diff --git a/CMakeLists.txt b/CMakeLists.txt index 7e2367df..31b2ef36 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -8,7 +8,9 @@ set(CMAKE_CXX_STANDARD 20) set(CMAKE_CXX_STANDARD_REQUIRED ON) set(CMAKE_CXX_EXTENSIONS OFF) -set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE) +if(NOT MINGW) + set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE) +endif() if(NOT WIN32 AND NOT CMAKE_CROSSCOMPILING) message(FATAL_ERROR "This CMake build currently supports Windows only. For cross-compilation from Linux, use the clang-cl toolchain.") diff --git a/CMakePresets.json b/CMakePresets.json index cb08b959..f1c4167c 100644 --- a/CMakePresets.json +++ b/CMakePresets.json @@ -14,7 +14,8 @@ "cacheVariables": { "PLATFORM_DEFINES": "_WINDOWS64", "PLATFORM_NAME": "Windows64", - "IGGY_LIBS": "iggy_w64.lib;iggyperfmon_w64.lib;iggyexpruntime_w64.lib" + "IGGY_LIBS": "iggy_w64.lib;iggyperfmon_w64.lib;iggyexpruntime_w64.lib", + "CMAKE_ASM_MASM_COMPILER": "C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.16.27023/bin/HostX64/x64/ml64.exe" } } ], diff --git a/Minecraft.Client/Assets/RmlUiAssets/HelpOptions.rcss b/Minecraft.Client/Assets/RmlUiAssets/HelpOptions.rcss new file mode 100644 index 00000000..1a8e7ae6 --- /dev/null +++ b/Minecraft.Client/Assets/RmlUiAssets/HelpOptions.rcss @@ -0,0 +1,55 @@ +body +{ + width: 100%; + height: 100%; + font-family: Minecraft Default; + font-size: 14px; + color: #ffffff; + background: transparent; + display: flex; + align-items: center; + justify-content: center; +} + +div#help_options_menu +{ + width: 624px; + height: auto; + margin: auto; + background: rgba(0, 0, 0, 0.65); + border-top: 2px solid #555555; + border-bottom: 2px solid #555555; + border-left: 2px solid #555555; + border-right: 2px solid #555555; + border-radius: 8px; + padding: 40px 48px; +} + +div#buttons +{ + width: 100%; + display: flex; + flex-direction: column; + align-items: center; +} + +button +{ + display: block; + width: 524px; + height: 64px; + margin-bottom: 11px; + color: #ffffff; + font-size: 25px; + text-align: center; + line-height: 64px; + cursor: pointer; + border: none; + background: transparent; + decorator: image(images/Button_Background.png); +} + +button:hover +{ + decorator: image(images/button_highlighted.png); +} diff --git a/Minecraft.Client/Assets/RmlUiAssets/HelpOptions.rml b/Minecraft.Client/Assets/RmlUiAssets/HelpOptions.rml new file mode 100644 index 00000000..6156124d --- /dev/null +++ b/Minecraft.Client/Assets/RmlUiAssets/HelpOptions.rml @@ -0,0 +1,17 @@ + + + Help & Options + + + +
+
+ + + + + +
+
+ + diff --git a/Minecraft.Client/Assets/RmlUiAssets/PauseMenu.rcss b/Minecraft.Client/Assets/RmlUiAssets/PauseMenu.rcss new file mode 100644 index 00000000..8d539ef6 --- /dev/null +++ b/Minecraft.Client/Assets/RmlUiAssets/PauseMenu.rcss @@ -0,0 +1,158 @@ +body +{ + width: 100%; + height: 100%; + font-family: Minecraft Default; + font-size: 14px; + color: #ffffff; + background: transparent; + display: flex; + align-items: center; + justify-content: center; +} + +div#pause_menu +{ + width: 624px; + height: auto; + margin: auto; + background: rgba(0, 0, 0, 0.65); + border-top: 2px solid #555555; + border-bottom: 2px solid #555555; + border-left: 2px solid #555555; + border-right: 2px solid #555555; + border-radius: 8px; + padding: 40px 48px; +} + +div#buttons +{ + width: 100%; + display: flex; + flex-direction: column; + align-items: center; +} + +button +{ + display: block; + width: 524px; + height: 64px; + margin-bottom: 11px; + color: #ffffff; + font-size: 25px; + text-align: center; + line-height: 64px; + cursor: pointer; + border: none; + background: transparent; + decorator: image(images/Button_Background.png); +} + +button:hover +{ + decorator: image(images/button_highlighted.png); +} + +div#exit_dialog +{ + display: none; + width: 100%; + margin-top: 16px; + text-align: center; +} + +p#exit_text +{ + font-size: 31px; + color: #ffffff; + margin-bottom: 23px; +} + +div#exit_buttons +{ + width: 100%; + display: flex; + justify-content: center; + gap: 10px; + flex-wrap: wrap; +} + +button#exit_save, +button#exit_nosave, +button#exit_cancel +{ + display: inline-block; + width: auto; + min-width: 233px; + height: 58px; + margin: 0; + padding-left: 29px; + padding-right: 29px; + color: #ffffff; + font-size: 22px; + text-align: center; + line-height: 58px; + cursor: pointer; + border: none; + background: transparent; + decorator: image(images/Button_Background2.png); +} + +button#exit_save:hover, +button#exit_nosave:hover, +button#exit_cancel:hover +{ + decorator: image(images/button_highlighted.png); +} + +div#save_dialog +{ + display: none; + width: 100%; + margin-top: 16px; + text-align: center; +} + +p#save_text +{ + font-size: 31px; + color: #ffffff; + margin-bottom: 23px; +} + +div#save_buttons +{ + width: 100%; + display: flex; + justify-content: center; + gap: 10px; +} + +button#save_cancel, +button#save_confirm +{ + display: inline-block; + width: auto; + min-width: 204px; + height: 58px; + margin: 0; + padding-left: 29px; + padding-right: 29px; + color: #ffffff; + font-size: 22px; + text-align: center; + line-height: 58px; + cursor: pointer; + border: none; + background: transparent; + decorator: image(images/Button_Background2.png); +} + +button#save_cancel:hover, +button#save_confirm:hover +{ + decorator: image(images/button_highlighted.png); +} + + diff --git a/Minecraft.Client/Assets/RmlUiAssets/PauseMenu.rml b/Minecraft.Client/Assets/RmlUiAssets/PauseMenu.rml new file mode 100644 index 00000000..97610aa2 --- /dev/null +++ b/Minecraft.Client/Assets/RmlUiAssets/PauseMenu.rml @@ -0,0 +1,32 @@ + + + Pause Menu + + + +
+
+ + + + + +
+
+

Exit Game?

+
+ + + +
+
+
+

Overwrite existing save?

+
+ + +
+
+
+ + diff --git a/Minecraft.Client/Assets/RmlUiAssets/SettingsMenu.rcss b/Minecraft.Client/Assets/RmlUiAssets/SettingsMenu.rcss new file mode 100644 index 00000000..911340ea --- /dev/null +++ b/Minecraft.Client/Assets/RmlUiAssets/SettingsMenu.rcss @@ -0,0 +1,104 @@ +body +{ + width: 100%; + height: 100%; + font-family: Minecraft Default; + font-size: 14px; + color: #ffffff; + background: transparent; + display: flex; + align-items: center; + justify-content: center; +} + +div#settings_menu +{ + width: 624px; + height: auto; + margin: auto; + background: rgba(0, 0, 0, 0.65); + border-top: 2px solid #555555; + border-bottom: 2px solid #555555; + border-left: 2px solid #555555; + border-right: 2px solid #555555; + border-radius: 8px; + padding: 40px 48px; +} + +div#buttons +{ + width: 100%; + display: flex; + flex-direction: column; + align-items: center; +} + +button +{ + display: block; + width: 524px; + height: 64px; + margin-bottom: 11px; + color: #ffffff; + font-size: 25px; + text-align: center; + line-height: 64px; + cursor: pointer; + border: none; + background: transparent; + decorator: image(images/Button_Background.png); +} + +button:hover +{ + decorator: image(images/button_highlighted.png); +} + +div#confirm_dialog +{ + display: none; + width: 100%; + margin-top: 16px; + text-align: center; +} + +p#confirm_text +{ + font-size: 31px; + color: #ffffff; + margin-bottom: 23px; +} + +div#confirm_buttons +{ + width: 100%; + display: flex; + justify-content: center; + gap: 10px; +} + +button#confirm_no, +button#confirm_yes +{ + display: inline-block; + width: auto; + min-width: 204px; + height: 58px; + margin: 0; + padding-left: 29px; + padding-right: 29px; + color: #ffffff; + font-size: 22px; + text-align: center; + line-height: 58px; + cursor: pointer; + border: none; + background: transparent; + decorator: image(images/Button_Background2.png); +} + +button#confirm_no:hover, +button#confirm_yes:hover +{ + decorator: image(images/button_highlighted.png); +} diff --git a/Minecraft.Client/Assets/RmlUiAssets/SettingsMenu.rml b/Minecraft.Client/Assets/RmlUiAssets/SettingsMenu.rml new file mode 100644 index 00000000..37b86491 --- /dev/null +++ b/Minecraft.Client/Assets/RmlUiAssets/SettingsMenu.rml @@ -0,0 +1,25 @@ + + + Settings + + + +
+
+ + + + + + +
+
+

Reset all settings to defaults?

+
+ + +
+
+
+ + diff --git a/Minecraft.Client/Assets/RmlUiAssets/fonts/MinecraftTen.ttf b/Minecraft.Client/Assets/RmlUiAssets/fonts/MinecraftTen.ttf new file mode 100644 index 00000000..4cd602f8 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/fonts/MinecraftTen.ttf differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/fonts/Mojangles.ttf b/Minecraft.Client/Assets/RmlUiAssets/fonts/Mojangles.ttf new file mode 100644 index 00000000..7f3d0dcf Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/fonts/Mojangles.ttf differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/fonts/chars.txt b/Minecraft.Client/Assets/RmlUiAssets/fonts/chars.txt new file mode 100644 index 00000000..f7ba12ae Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/fonts/chars.txt differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/AdventureIcon.png b/Minecraft.Client/Assets/RmlUiAssets/images/AdventureIcon.png new file mode 100644 index 00000000..60d306fe Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/AdventureIcon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Button_Background.png b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Background.png new file mode 100644 index 00000000..28a3269f Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Background.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Button_Background2.png b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Background2.png new file mode 100644 index 00000000..2bfa4f49 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Background2.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Button_Square.png b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Square.png new file mode 100644 index 00000000..08433912 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Square.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Button_Square_Highlighted.png b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Square_Highlighted.png new file mode 100644 index 00000000..c3620fe3 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Button_Square_Highlighted.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Change_Skin_Icon.png b/Minecraft.Client/Assets/RmlUiAssets/images/Change_Skin_Icon.png new file mode 100644 index 00000000..62d5012f Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Change_Skin_Icon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/CreateWorldIcon.png b/Minecraft.Client/Assets/RmlUiAssets/images/CreateWorldIcon.png new file mode 100644 index 00000000..7debd644 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/CreateWorldIcon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/CreativeIcon.png b/Minecraft.Client/Assets/RmlUiAssets/images/CreativeIcon.png new file mode 100644 index 00000000..2ef0df6d Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/CreativeIcon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/ModIcon.png b/Minecraft.Client/Assets/RmlUiAssets/images/ModIcon.png new file mode 100644 index 00000000..517b4304 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/ModIcon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/OriginalLogo.png b/Minecraft.Client/Assets/RmlUiAssets/images/OriginalLogo.png new file mode 100644 index 00000000..eacf7a03 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/OriginalLogo.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/SaveChest.png b/Minecraft.Client/Assets/RmlUiAssets/images/SaveChest.png new file mode 100644 index 00000000..04e46ee9 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/SaveChest.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/SliderHandlerBackground.png b/Minecraft.Client/Assets/RmlUiAssets/images/SliderHandlerBackground.png new file mode 100644 index 00000000..9b113b61 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/SliderHandlerBackground.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Slider_Handle.png b/Minecraft.Client/Assets/RmlUiAssets/images/Slider_Handle.png new file mode 100644 index 00000000..acfd1fe8 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Slider_Handle.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/SurvivalIcon.png b/Minecraft.Client/Assets/RmlUiAssets/images/SurvivalIcon.png new file mode 100644 index 00000000..6d045a6e Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/SurvivalIcon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Toggle_Switch_Off.png b/Minecraft.Client/Assets/RmlUiAssets/images/Toggle_Switch_Off.png new file mode 100644 index 00000000..f47fc6f3 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Toggle_Switch_Off.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Toggle_Switch_On.png b/Minecraft.Client/Assets/RmlUiAssets/images/Toggle_Switch_On.png new file mode 100644 index 00000000..64a7ea2d Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Toggle_Switch_On.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/Trash_Bin_Icon.png b/Minecraft.Client/Assets/RmlUiAssets/images/Trash_Bin_Icon.png new file mode 100644 index 00000000..543d8917 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/Trash_Bin_Icon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/button_highlighted.png b/Minecraft.Client/Assets/RmlUiAssets/images/button_highlighted.png new file mode 100644 index 00000000..af20c131 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/button_highlighted.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/down_icon.png b/Minecraft.Client/Assets/RmlUiAssets/images/down_icon.png new file mode 100644 index 00000000..4d52c5fa Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/down_icon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/frame_background.png b/Minecraft.Client/Assets/RmlUiAssets/images/frame_background.png new file mode 100644 index 00000000..d6dae3f3 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/frame_background.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/lceheadwer.png b/Minecraft.Client/Assets/RmlUiAssets/images/lceheadwer.png new file mode 100644 index 00000000..abf38ba0 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/lceheadwer.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/images/up_icon.png b/Minecraft.Client/Assets/RmlUiAssets/images/up_icon.png new file mode 100644 index 00000000..b119bcd2 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/images/up_icon.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/panorama/Background_Day.png b/Minecraft.Client/Assets/RmlUiAssets/panorama/Background_Day.png new file mode 100644 index 00000000..c680c099 Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/panorama/Background_Day.png differ diff --git a/Minecraft.Client/Assets/RmlUiAssets/panorama/Background_Night.png b/Minecraft.Client/Assets/RmlUiAssets/panorama/Background_Night.png new file mode 100644 index 00000000..b915b51a Binary files /dev/null and b/Minecraft.Client/Assets/RmlUiAssets/panorama/Background_Night.png differ diff --git a/Minecraft.Client/CMakeLists.txt b/Minecraft.Client/CMakeLists.txt index 27885cd6..bdcea8f7 100644 --- a/Minecraft.Client/CMakeLists.txt +++ b/Minecraft.Client/CMakeLists.txt @@ -20,8 +20,7 @@ if(PLATFORM_NAME STREQUAL "Windows64") set_target_properties(Minecraft.Client PROPERTIES WIN32_EXECUTABLE TRUE) endif() -target_include_directories(Minecraft.Client - PRIVATE +target_include_directories(Minecraft.Client PRIVATE "${CMAKE_BINARY_DIR}/generated/" # This is for the generated BuildVer.h "${CMAKE_CURRENT_SOURCE_DIR}" "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/Iggy/inc" @@ -29,6 +28,73 @@ target_include_directories(Minecraft.Client "${CMAKE_CURRENT_SOURCE_DIR}/Common/Libs/zlib/inc" "${CMAKE_SOURCE_DIR}/include/" ) + +if(PLATFORM_NAME STREQUAL "Windows64") + target_include_directories(Minecraft.Client PRIVATE + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/inc" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath" + "${CMAKE_SOURCE_DIR}/Minecraft.World/x64headers" + ) + + # Build RmlUi as a static library (skip samples, tests, backends) + set(BUILD_SHARED_LIBS OFF CACHE BOOL "" FORCE) + + # Download FreeType for RmlUi font engine + include(FetchContent) + set(CMAKE_POLICY_VERSION_MINIMUM 3.5) + set(FT_DISABLE_ZLIB ON CACHE BOOL "" FORCE) + set(FT_DISABLE_BZIP2 ON CACHE BOOL "" FORCE) + set(FT_DISABLE_PNG ON CACHE BOOL "" FORCE) + set(FT_DISABLE_HARFBUZZ ON CACHE BOOL "" FORCE) + set(FT_DISABLE_BROTLI ON CACHE BOOL "" FORCE) + FetchContent_Declare( + freetype + GIT_REPOSITORY https://gitlab.freedesktop.org/freetype/freetype.git + GIT_TAG VER-2-13-2 + GIT_SHALLOW TRUE + ) + FetchContent_MakeAvailable(freetype) + if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU" AND MINGW) + target_compile_definitions(freetype PRIVATE _INC_SETJMPEX) + endif() + if(TARGET freetype-interface AND NOT TARGET Freetype::Freetype) + add_library(Freetype::Freetype ALIAS freetype-interface) + endif() + set(RMLUI_SAMPLES OFF CACHE BOOL "" FORCE) + set(RMLUI_TESTS OFF CACHE BOOL "" FORCE) + set(RMLUI_LUA_BINDINGS OFF CACHE BOOL "" FORCE) + set(RMLUI_LOTTIE_PLUGIN OFF CACHE BOOL "" FORCE) + set(RMLUI_SVG_PLUGIN OFF CACHE BOOL "" FORCE) + set(RMLUI_FONT_ENGINE "freetype" CACHE STRING "" FORCE) + set(CMAKE_DISABLE_FIND_PACKAGE_Freetype ON CACHE BOOL "" FORCE) + add_subdirectory("${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/RmlUi" "${CMAKE_BINARY_DIR}/rmlui") + target_link_libraries(Minecraft.Client PRIVATE rmlui) + + # Add RmlUi include path only to files that include RmlUi headers, + # not to the whole target (avoids PCH macro conflicts with windows.h). + set(RMLUI_INCLUDE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/RmlUi/Include") + set(_RMLUI_SOURCES + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_HelpAndOptionsMenu.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_HelpAndOptionsMenu.h" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_PauseMenu.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_PauseMenu.h" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_SettingsMenu.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_SettingsMenu.h" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_SkinSelectMenu.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Common/UI/UIScene_SkinSelectMenu.h" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/RmlUi_Renderer_D3D11.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/RmlUi_Renderer_D3D11.h" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/RmlUi_SystemInterface_Win64.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/RmlUi_SystemInterface_Win64.h" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/RmlUi_FileInterface.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/RmlUi_FileInterface.h" + "${CMAKE_CURRENT_SOURCE_DIR}/RmlManager.cpp" + "${CMAKE_CURRENT_SOURCE_DIR}/RmlManager.h" + ) + set_property(SOURCE ${_RMLUI_SOURCES} APPEND PROPERTY INCLUDE_DIRECTORIES "${RMLUI_INCLUDE_DIR}") +endif() target_compile_definitions(Minecraft.Client PRIVATE ${MINECRAFT_SHARED_DEFINES} ) @@ -46,33 +112,28 @@ target_link_libraries(Minecraft.Client PRIVATE Minecraft.World d3d11 d3dcompiler + ws2_32 + $<$:legacy_stdio_definitions> + # Allow our override of CRT internal symbols (__acrt_iob_func) to take priority + $<$>:-Wl,--allow-multiple-definition> + # Group libs with circular deps so linker rescan resolves all symbols + $<$>:-Wl,--start-group> XInput9_1_0 - wsock32 - legacy_stdio_definitions $<$: # Debug 4J libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/4J_Input_d.lib" - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/4J_Storage_d.lib" - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/4J_Render_PC_d.lib" - # Debug Discord libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/Discord/libs/discord-rpc_d.lib" - # Debug ZSTD libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/zstd/libs/libzstd_static.lib" - # Debug ZLIB libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Common/Libs/zlib/libs/zlibstatic.lib" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/Debug/4J_Input_d.a" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/Debug/4J_Storage_d.a" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/Debug/4J_Render_PC_d.a" > $<$>: # Release 4J libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/4J_Input.lib" - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/4J_Storage.lib" - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/4J_Render_PC.lib" - # Release Discord libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/Discord/libs/discord-rpc.lib" - # Release ZSTD libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/zstd/libs/libzstd_static.lib" - # Release ZLIB libraries - "${CMAKE_CURRENT_SOURCE_DIR}/Common/Libs/zlib/libs/zlibstatic.lib" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/Release/4J_Input.a" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/Release/4J_Storage.a" + "${CMAKE_CURRENT_SOURCE_DIR}/Windows64/Libs/4JLibs/libs/Release/4J_Render_PC.a" > + $<$>:z> + $<$>:zstd> + $<$>:-Wl,--end-group> ) # Iggy libs diff --git a/Minecraft.Client/Common/Consoles_App.h b/Minecraft.Client/Common/Consoles_App.h index ced6b2a3..54d1f46b 100644 --- a/Minecraft.Client/Common/Consoles_App.h +++ b/Minecraft.Client/Common/Consoles_App.h @@ -875,6 +875,7 @@ private: bool m_bResetNether; + wstring m_currentSaveFolderName; // 4J Added: for hardcore world deletion on Win64 DWORD m_dwRequiredTexturePackID; #ifdef _XBOX_ONE vector m_vTMSPPData; @@ -889,6 +890,11 @@ private: #endif + // 4J Added - for hardcore world deletion on Win64 +public: + void SetCurrentSaveFolderName(const wstring& name) { m_currentSaveFolderName = name; } + const wstring& GetCurrentSaveFolderName() const { return m_currentSaveFolderName; } + // 4J-PB - language and locale functions public: diff --git a/Minecraft.Client/Common/UI/IUIScene_PauseMenu.cpp b/Minecraft.Client/Common/UI/IUIScene_PauseMenu.cpp index 01ee6c59..a56d6bba 100644 --- a/Minecraft.Client/Common/UI/IUIScene_PauseMenu.cpp +++ b/Minecraft.Client/Common/UI/IUIScene_PauseMenu.cpp @@ -1,17 +1,19 @@ +#include "stdafx.h" #include "IUIScene_PauseMenu.h" -#include "..\..\Minecraft.h" -#include "..\..\MinecraftServer.h" -#include "..\..\MultiPlayerLevel.h" -#include "..\..\ProgressRenderer.h" -#include "..\..\..\Minecraft.World\net.minecraft.world.level.h" -#include "..\..\..\Minecraft.World\net.minecraft.world.phys.h" -#include "..\..\TexturePackRepository.h" -#include "..\..\TexturePack.h" -#include "..\..\DLCTexturePack.h" -#include "..\..\..\Minecraft.World\StringHelpers.h" +#include "../../Minecraft.h" +#include "../../MinecraftServer.h" +#include "../../MultiPlayerLevel.h" +#include "../../ProgressRenderer.h" +#include "../../../Minecraft.World/net.minecraft.world.level.h" +#include "../../../Minecraft.World/net.minecraft.world.phys.h" +#include "../../TexturePackRepository.h" +#include "../../TexturePack.h" +#include "../../../Minecraft.World/StringHelpers.h" +#ifndef _XBOX #include "UI.h" -#include "../DLC/DLCPack.h" +#endif + int IUIScene_PauseMenu::ExitGameDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result) { @@ -40,43 +42,14 @@ int IUIScene_PauseMenu::ExitGameSaveDialogReturned(void *pParam,int iPad,C4JStor #endif // Exit with or without saving - // Decline means save in this dialog - if(result==C4JStorage::EMessage_ResultDecline || result==C4JStorage::EMessage_ResultThirdOption) + if(result==C4JStorage::EMessage_ResultAccept || result==C4JStorage::EMessage_ResultDecline || result==C4JStorage::EMessage_ResultThirdOption) { + if( result==C4JStorage::EMessage_ResultAccept ) // Cancel + { + return 0; + } if( result==C4JStorage::EMessage_ResultDecline ) // Save { - // 4J-PB - Is the player trying to save but they are using a trial texturepack ? - if(!Minecraft::GetInstance()->skins->isUsingDefaultSkin()) - { - TexturePack *tPack = Minecraft::GetInstance()->skins->getSelected(); - DLCTexturePack *pDLCTexPack=static_cast(tPack); - - DLCPack *pDLCPack=pDLCTexPack->getDLCInfoParentPack();//tPack->getDLCPack(); - if(!pDLCPack->hasPurchasedFile( DLCManager::e_DLCType_Texture, L"" )) - { -#ifdef _XBOX - // upsell - ULONGLONG ullOfferID_Full; - // get the dlc texture pack - DLCTexturePack *pDLCTexPack=(DLCTexturePack *)tPack; - - app.GetDLCFullOfferIDForPackID(pDLCTexPack->getDLCParentPackId(),&ullOfferID_Full); - - // tell sentient about the upsell of the full version of the skin pack - TelemetryManager->RecordUpsellPresented(iPad, eSet_UpsellID_Texture_DLC, ullOfferID_Full & 0xFFFFFFFF); -#endif - - UINT uiIDA[2]; - uiIDA[0]=IDS_CONFIRM_OK; - uiIDA[1]=IDS_CONFIRM_CANCEL; - - // Give the player a warning about the trial version of the texture pack - //ui.RequestAlertMessage(IDS_WARNING_DLC_TRIALTEXTUREPACK_TITLE, IDS_WARNING_DLC_TRIALTEXTUREPACK_TEXT, uiIDA, 2, ProfileManager.GetPrimaryPad() , &IUIScene_PauseMenu::WarningTrialTexturePackReturned, pParam); - - return S_OK; - } - } - // does the save exist? bool bSaveExists; StorageManager.DoesSaveExist(&bSaveExists); @@ -98,7 +71,7 @@ int IUIScene_PauseMenu::ExitGameSaveDialogReturned(void *pParam,int iPad,C4JStor MinecraftServer::getInstance()->setSaveOnExit( true ); } } - else + else // EMessage_ResultThirdOption - Don't Save { // been a few requests for a confirm on exit without saving UINT uiIDA[2]; @@ -212,141 +185,6 @@ int IUIScene_PauseMenu::ExitGameDeclineSaveReturned(void *pParam,int iPad,C4JSto -int IUIScene_PauseMenu::WarningTrialTexturePackReturned(void *pParam,int iPad,C4JStorage::EMessageResult result) -{ -#if defined(__PS3__) || defined(__ORBIS__) || defined(__PSVITA__) - if(result==C4JStorage::EMessage_ResultAccept) - { - if(!ProfileManager.IsSignedInLive(iPad)) - { - // you're not signed in to PSN! - - } - else - { - // 4J-PB - need to check this user can access the store - bool bContentRestricted; - ProfileManager.GetChatAndContentRestrictions(iPad,true,nullptr,&bContentRestricted,nullptr); - if(bContentRestricted) - { - UINT uiIDA[1]; - uiIDA[0]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_ONLINE_SERVICE_TITLE, IDS_CONTENT_RESTRICTION, uiIDA, 1, iPad); - } - else - { - // need to get info on the pack to see if the user has already downloaded it - TexturePack *tPack = Minecraft::GetInstance()->skins->getSelected(); - DLCTexturePack *pDLCTexPack=(DLCTexturePack *)tPack; - - // retrieve the store name for the skin pack - DLCPack *pDLCPack=pDLCTexPack->getDLCInfoParentPack();//tPack->getDLCPack(); - const char *pchPackName=wstringtofilename(pDLCPack->getName()); - app.DebugPrintf("Texture Pack - %s\n",pchPackName); - SONYDLC *pSONYDLCInfo=app.GetSONYDLCInfo((char *)pchPackName); - - if(pSONYDLCInfo!=nullptr) - { - char chName[42]; - char chSkuID[SCE_NP_COMMERCE2_SKU_ID_LEN]; - - memset(chSkuID,0,SCE_NP_COMMERCE2_SKU_ID_LEN); - // find the info on the skin pack - // we have to retrieve the skuid from the store info, it can't be hardcoded since Sony may change it. - // So we assume the first sku for the product is the one we want -#ifdef __ORBIS__ - sprintf(chName,"%s",pSONYDLCInfo->chDLCKeyname); -#else - sprintf(chName,"%s-%s",app.GetCommerceCategory(),pSONYDLCInfo->chDLCKeyname); -#endif - app.GetDLCSkuIDFromProductList(chName,chSkuID); - // 4J-PB - need to check for an empty store -#if defined __ORBIS__ || defined __PSVITA__ || defined __PS3__ - if(app.CheckForEmptyStore(iPad)==false) -#endif - { - if(app.DLCAlreadyPurchased(chSkuID)) - { - app.DownloadAlreadyPurchased(chSkuID); - } - else - { - app.Checkout(chSkuID); - } - } - } - } - } - } -#endif // - -#ifdef _XBOX_ONE - IUIScene_PauseMenu* pScene = (IUIScene_PauseMenu*)pParam; - - if(result==C4JStorage::EMessage_ResultAccept) - { - if(ProfileManager.IsSignedIn(iPad)) - { - if (ProfileManager.IsSignedInLive(iPad)) - { - TexturePack *tPack = Minecraft::GetInstance()->skins->getSelected(); - // get the dlc texture pack - DLCTexturePack *pDLCTexPack=(DLCTexturePack *)tPack; - - DLCPack *pDLCPack=pDLCTexPack->getDLCInfoParentPack(); - - DLC_INFO *pDLCInfo=app.GetDLCInfoForProductName((WCHAR *)pDLCPack->getName().c_str()); - - StorageManager.InstallOffer(1,(WCHAR *)pDLCInfo->wsProductId.c_str(),nullptr,nullptr); - - // the license change coming in when the offer has been installed will cause this scene to refresh - } - else - { - // 4J-JEV: Fix for XB1: #165863 - XR-074: Compliance: With no active network connection user is unable to convert from Trial to Full texture pack and is not messaged why. - UINT uiIDA[1] = { IDS_CONFIRM_OK }; - ui.RequestErrorMessage(IDS_PRO_NOTONLINE_TITLE, IDS_PRO_XBOXLIVE_NOTIFICATION, uiIDA, 1, iPad); - } - } - } - -#endif - -#ifdef _XBOX - IUIScene_PauseMenu* pScene = (IUIScene_PauseMenu*)pParam; - - //pScene->m_bIgnoreInput = false; - pScene->ShowScene( true ); - if(result==C4JStorage::EMessage_ResultAccept) - { - if(ProfileManager.IsSignedIn(iPad)) - { - ULONGLONG ullIndexA[1]; - - TexturePack *tPack = Minecraft::GetInstance()->skins->getSelected(); - // get the dlc texture pack - DLCTexturePack *pDLCTexPack=(DLCTexturePack *)tPack; - - // Need to get the parent packs id, since this may be one of many child packs with their own ids - app.GetDLCFullOfferIDForPackID(pDLCTexPack->getDLCParentPackId(),&ullIndexA[0]); - - // need to allow downloads here, or the player would need to quit the game to let the download of a texture pack happen. This might affect the network traffic, since the download could take all the bandwidth... - XBackgroundDownloadSetMode(XBACKGROUND_DOWNLOAD_MODE_ALWAYS_ALLOW); - - StorageManager.InstallOffer(1,ullIndexA,nullptr,nullptr); - } - } - else - { - TelemetryManager->RecordUpsellResponded(iPad, eSet_UpsellID_Texture_DLC, ( pScene->m_pDLCPack->getPurchaseOfferId() & 0xFFFFFFFF ), eSen_UpsellOutcome_Declined); - } -#endif - - - return 0; -} - - int IUIScene_PauseMenu::SaveWorldThreadProc( LPVOID lpParameter ) { bool bAutosave=static_cast(lpParameter); @@ -407,11 +245,73 @@ int IUIScene_PauseMenu::ExitWorldThreadProc( void* lpParameter ) return S_OK; } +#ifdef _WINDOWS64 +static bool Win64_DeleteSaveDirectory(const wchar_t* wPath) +{ + wchar_t wSearch[MAX_PATH]; + swprintf_s(wSearch, MAX_PATH, L"%s\\*", wPath); + WIN32_FIND_DATAW fd; + HANDLE hFind = FindFirstFileW(wSearch, &fd); + if (hFind != INVALID_HANDLE_VALUE) + { + do + { + if (wcscmp(fd.cFileName, L".") == 0 || wcscmp(fd.cFileName, L"..") == 0) continue; + wchar_t wChild[MAX_PATH]; + swprintf_s(wChild, MAX_PATH, L"%s\\%s", wPath, fd.cFileName); + if (fd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + Win64_DeleteSaveDirectory(wChild); + else + DeleteFileW(wChild); + } while (FindNextFileW(hFind, &fd)); + FindClose(hFind); + } + return RemoveDirectoryW(wPath) != 0; +} +#endif // _WINDOWS64 + // This function performs the meat of exiting from a level. It should be called from a thread other than the main thread. void IUIScene_PauseMenu::_ExitWorld(LPVOID lpParameter) { +#ifndef MINECRAFT_SERVER_BUILD Minecraft *pMinecraft=Minecraft::GetInstance(); + // 4J Added: Capture hardcore delete info before the server is destroyed +#ifdef _WINDOWS64 + bool shouldDeleteHardcoreWorld = false; + wstring hardcoreSaveFolderName; + if (MinecraftServer::getInstance() != nullptr && MinecraftServer::getInstance()->getDeleteWorldOnExit()) + { + shouldDeleteHardcoreWorld = true; + // Try 1: Use the save folder name stored by UIScene_LoadMenu::StartGameFromSave (works for existing saves) + hardcoreSaveFolderName = app.GetCurrentSaveFolderName(); + if (!hardcoreSaveFolderName.empty()) + { + app.DebugPrintf("Hardcore mode: save folder from app = '%ls'\n", hardcoreSaveFolderName.c_str()); + } + // Try 2: StorageManager (may work for new saves after first autosave) + if (hardcoreSaveFolderName.empty()) + { + char szSaveFolder[MAX_SAVEFILENAME_LENGTH] = {}; + StorageManager.GetSaveUniqueFilename(szSaveFolder); + if (szSaveFolder[0] != '\0') + { + wchar_t wSaveFolder[MAX_SAVEFILENAME_LENGTH] = {}; + mbstowcs(wSaveFolder, szSaveFolder, MAX_SAVEFILENAME_LENGTH - 1); + hardcoreSaveFolderName = wSaveFolder; + app.DebugPrintf("Hardcore mode: save folder from StorageManager = '%s'\n", szSaveFolder); + } + } + // Try 3: Stored during loadLevel + if (hardcoreSaveFolderName.empty()) + { + hardcoreSaveFolderName = MinecraftServer::getInstance()->getSaveFolderName(); + app.DebugPrintf("Hardcore mode: save folder from server = '%ls'\n", hardcoreSaveFolderName.c_str()); + } + MinecraftServer::getInstance()->setDeleteWorldOnExit(false); + } +#endif + int exitReasonStringId = pMinecraft->progressRenderer->getCurrentTitle(); int exitReasonTitleId = IDS_CONNECTION_LOST; @@ -622,6 +522,17 @@ void IUIScene_PauseMenu::_ExitWorld(LPVOID lpParameter) { Sleep(1); } + // 4J Added: Hardcore mode — delete world save data now that the server is fully stopped +#ifdef _WINDOWS64 + if (shouldDeleteHardcoreWorld && !hardcoreSaveFolderName.empty()) + { + wchar_t wFolderPath[MAX_PATH] = {}; + swprintf_s(wFolderPath, MAX_PATH, L"Windows64\\GameHDD\\%s", hardcoreSaveFolderName.c_str()); + app.DebugPrintf("Hardcore mode: Deleting world save folder '%ls'\n", wFolderPath); + Win64_DeleteSaveDirectory(wFolderPath); + } +#endif + pMinecraft->setLevel(nullptr,exitReasonStringId,nullptr,saveStats); TelemetryManager->Flush(); @@ -649,6 +560,7 @@ void IUIScene_PauseMenu::_ExitWorld(LPVOID lpParameter) // Make sure we don't think saving is disabled in the menus StorageManager.SetSaveDisabled(false); #endif +#endif } @@ -705,4 +617,4 @@ int IUIScene_PauseMenu::DisableAutosaveDialogReturned(void *pParam,int iPad,C4JS app.SetAction(iPad,eAppAction_SaveGame); } return 0; -} \ No newline at end of file +} diff --git a/Minecraft.Client/Common/UI/IUIScene_PauseMenu.h b/Minecraft.Client/Common/UI/IUIScene_PauseMenu.h index 7233df3a..7ce4e949 100644 --- a/Minecraft.Client/Common/UI/IUIScene_PauseMenu.h +++ b/Minecraft.Client/Common/UI/IUIScene_PauseMenu.h @@ -2,15 +2,11 @@ class IUIScene_PauseMenu { -protected: - DLCPack *m_pDLCPack; - public: static int ExitGameDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); static int ExitGameSaveDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); static int ExitGameAndSaveReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); static int ExitGameDeclineSaveReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); - static int WarningTrialTexturePackReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); static int SaveGameDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); static int EnableAutosaveDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); static int DisableAutosaveDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); diff --git a/Minecraft.Client/Common/UI/IUIScene_StartGame.cpp b/Minecraft.Client/Common/UI/IUIScene_StartGame.cpp index 18d71289..ab70cc91 100644 --- a/Minecraft.Client/Common/UI/IUIScene_StartGame.cpp +++ b/Minecraft.Client/Common/UI/IUIScene_StartGame.cpp @@ -12,8 +12,7 @@ IUIScene_StartGame::IUIScene_StartGame(int iPad, UILayer *parentLayer) : UIScene m_bShowTexturePackDescription = false; m_iSetTexturePackDescription = -1; - Minecraft *pMinecraft = Minecraft::GetInstance(); - m_currentTexturePackIndex = pMinecraft->skins->getTexturePackIndex(0); + m_currentTexturePackIndex = 0; } void IUIScene_StartGame::HandleDLCMountingComplete() diff --git a/Minecraft.Client/Common/UI/UIControl_EnchantmentButton.cpp b/Minecraft.Client/Common/UI/UIControl_EnchantmentButton.cpp index c71f47c2..b5199010 100644 --- a/Minecraft.Client/Common/UI/UIControl_EnchantmentButton.cpp +++ b/Minecraft.Client/Common/UI/UIControl_EnchantmentButton.cpp @@ -5,6 +5,7 @@ #include "..\..\MultiplayerLocalPlayer.h" #include "../../Font.h" #include "..\..\..\Minecraft.World\StringHelpers.h" +#include #include "..\..\Windows64\KeyboardMouseInput.h" diff --git a/Minecraft.Client/Common/UI/UIEnums.h b/Minecraft.Client/Common/UI/UIEnums.h index d3c60419..752755af 100644 --- a/Minecraft.Client/Common/UI/UIEnums.h +++ b/Minecraft.Client/Common/UI/UIEnums.h @@ -81,6 +81,7 @@ enum EUIScene eUIComponent_Chat, eUIScene_ReinstallMenu, eUIScene_SkinSelectMenu, + eUIScene_AchievementsMenu, eUIScene_TextEntry, eUIScene_InGameHostOptionsMenu, eUIScene_InGamePlayerOptionsMenu, diff --git a/Minecraft.Client/Common/UI/UIScene.h b/Minecraft.Client/Common/UI/UIScene.h index 57054e1c..39ffd506 100644 --- a/Minecraft.Client/Common/UI/UIScene.h +++ b/Minecraft.Client/Common/UI/UIScene.h @@ -167,7 +167,7 @@ public: // Returns true if this scene has focus for the pad passed in virtual bool hasFocus(int iPad) { return bHasFocus && iPad == m_iPad; } - void gainFocus(); + virtual void gainFocus(); void loseFocus(); virtual void updateTooltips(); diff --git a/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.cpp b/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.cpp index 71926a4f..5240a60f 100644 --- a/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.cpp +++ b/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.cpp @@ -1,235 +1,260 @@ +#include "stdafx.h" + +// Non-RmlUi includes that depend on Windows macros #include "UI.h" +#include "../../Minecraft.h" + +// RmlUi includes (Windows macros temporarily suppressed) +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + #include "UIScene_HelpAndOptionsMenu.h" -#include "..\..\Minecraft.h" +#include "RmlManager.h" +#include "../../Windows64/KeyboardMouseInput.h" +#include +#include -UIScene_HelpAndOptionsMenu::UIScene_HelpAndOptionsMenu(int iPad, void *initData, UILayer *parentLayer) : UIScene(iPad, parentLayer) +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") + +UIScene_HelpAndOptionsMenu::UIScene_HelpAndOptionsMenu(int iPad, void *initData, UILayer *parentLayer) + : UIScene(iPad, parentLayer) + , m_document(nullptr) { - // Setup all the Iggy references we need for this scene - initialiseMovie(); - - m_bNotInGame=(Minecraft::GetInstance()->level==nullptr); - - m_buttons[BUTTON_HAO_CHANGESKIN].init(IDS_CHANGE_SKIN,BUTTON_HAO_CHANGESKIN); - m_buttons[BUTTON_HAO_HOWTOPLAY].init(IDS_HOW_TO_PLAY,BUTTON_HAO_HOWTOPLAY); - m_buttons[BUTTON_HAO_CONTROLS].init(IDS_CONTROLS,BUTTON_HAO_CONTROLS); - m_buttons[BUTTON_HAO_SETTINGS].init(IDS_SETTINGS,BUTTON_HAO_SETTINGS); - m_buttons[BUTTON_HAO_CREDITS].init(IDS_CREDITS,BUTTON_HAO_CREDITS); - //m_buttons[BUTTON_HAO_REINSTALL].init(app.GetString(IDS_REINSTALL_CONTENT),BUTTON_HAO_REINSTALL); - m_buttons[BUTTON_HAO_DEBUG].init(IDS_DEBUG_SETTINGS,BUTTON_HAO_DEBUG); - - /* 4J-TomK - we should never remove a control before the other buttons controls are initialised! - (because vita touchboxes are rebuilt on remove since the remaining positions might change) */ - // We don't have a reinstall content, so remove the button - removeControl( &m_buttons[BUTTON_HAO_REINSTALL], false ); - -#ifndef _DEBUG - removeControl( &m_buttons[BUTTON_HAO_DEBUG], false); -#else - if(!app.DebugSettingsOn()) removeControl( &m_buttons[BUTTON_HAO_DEBUG], false); -#endif - -#ifdef _XBOX_ONE - // 4J-PB - in order to buy the skin packs, we need the signed offer ids for them, which we get in the availability info - // we need to retrieve this info though, so do it here - app.AddDLCRequest(e_Marketplace_Content); // content is skin packs, texture packs and mash-up packs - - // we also need to mount the local DLC so we can tell what's been purchased - app.StartInstallDLCProcess(iPad); -#endif - - - - // 4J-PB - do not need a storage device to see this menu - just need one when you choose to re-install them - bool bNotInGame=(Minecraft::GetInstance()->level==nullptr); - - // any content to be re-installed? - if(m_iPad==ProfileManager.GetPrimaryPad() && bNotInGame) + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx) { - // We should show the reinstall menu - app.DebugPrintf("Reinstall Menu required...\n"); + app.DebugPrintf("[RmlHelpAndOptions] No RmlUi context available\n"); + return; + } + + m_document = ctx->LoadDocument("HelpOptions.rml"); + if (m_document) + { + m_document->Show(); + + bool bNotInGame = (Minecraft::GetInstance()->level == nullptr); + + // Change Skin: hide if not full version + if (!ProfileManager.IsFullVersion()) + { + auto* el = m_document->GetElementById("change_skin"); + if (el) el->SetProperty("display", "none"); + } + + // Credits: hide in splitscreen + if (app.GetLocalPlayerCount() > 1) + { + auto* el = m_document->GetElementById("credits"); + if (el) el->SetProperty("display", "none"); + } + + // Register button click listeners + m_document->GetElementById("change_skin")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("how_to_play")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("controls")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("settings")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("credits")->AddEventListener(Rml::EventId::Click, this); + + // Focus first visible button + m_document->GetElementById("change_skin")->Focus(); } else { - removeControl( &m_buttons[BUTTON_HAO_REINSTALL], false); + app.DebugPrintf("[RmlHelpAndOptions] Failed to load HelpOptions.rml\n"); } - if(app.GetLocalPlayerCount()>1) - { - // no credits in splitscreen - removeControl( &m_buttons[BUTTON_HAO_CREDITS], false); + ui.HidePressStart(); -#if TO_BE_IMPLEMENTED - app.AdjustSplitscreenScene(m_hObj,&m_OriginalPosition,m_iPad,false); -#endif - if(ProfileManager.GetPrimaryPad()!=m_iPad) - { - removeControl( &m_buttons[BUTTON_HAO_REINSTALL], false); - } - } - - if(!ProfileManager.IsFullVersion() )//|| ProfileManager.IsGuest(m_iPad)) - { - removeControl( &m_buttons[BUTTON_HAO_CHANGESKIN], false); - } - - // 4J-TomK Moved horizontal resize check to the end to prevent horizontal scaling for buttons that might get removed anyways (debug options for example) - doHorizontalResizeCheck(); - - //StorageManager.TMSPP_GetUserQuotaInfo(C4JStorage::eGlobalStorage_TitleUser,iPad); - //StorageManager.WebServiceRequestGetFriends(iPad); + TelemetryManager->RecordMenuShown(m_iPad, eUIScene_HelpAndOptionsMenu, 0); } UIScene_HelpAndOptionsMenu::~UIScene_HelpAndOptionsMenu() { + if (m_document) + { + m_document->GetElementById("change_skin")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("how_to_play")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("controls")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("settings")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("credits")->RemoveEventListener(Rml::EventId::Click, this); + m_document->Close(); + m_document = nullptr; + } } -wstring UIScene_HelpAndOptionsMenu::getMoviePath() +void UIScene_HelpAndOptionsMenu::gainFocus() { - if(app.GetLocalPlayerCount() > 1) + if (!bHasFocus && stealsFocus()) { - return L"HelpAndOptionsMenuSplit"; + bHasFocus = true; + updateTooltips(); + updateComponents(); + if (!m_bFocussedOnce) + { + } + handleGainFocus(m_bFocussedOnce); + if (bHasFocus) + m_bFocussedOnce = true; + if (m_document) + m_document->Show(); } - else +} + +void UIScene_HelpAndOptionsMenu::handleLoseFocus() +{ + if (m_document) + m_document->Hide(); +} + +void UIScene_HelpAndOptionsMenu::reloadMovie(bool force) +{ +} + +void UIScene_HelpAndOptionsMenu::tick() +{ + m_hasTickedOnce = true; + + UIScene::tick(); + + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx || !bHasFocus) + return; + + ctx->ProcessMouseMove(g_KBMInput.GetMouseX(), g_KBMInput.GetMouseY(), 0); + + if (g_KBMInput.IsMouseButtonPressed(KeyboardMouseInput::MOUSE_LEFT)) + ctx->ProcessMouseButtonDown(0, 0); + if (g_KBMInput.IsMouseButtonReleased(KeyboardMouseInput::MOUSE_LEFT)) + ctx->ProcessMouseButtonUp(0, 0); + if (g_KBMInput.IsMouseButtonPressed(KeyboardMouseInput::MOUSE_RIGHT)) + ctx->ProcessMouseButtonDown(1, 0); + if (g_KBMInput.IsMouseButtonReleased(KeyboardMouseInput::MOUSE_RIGHT)) + ctx->ProcessMouseButtonUp(1, 0); + + int wheelDelta = g_KBMInput.GetMouseWheel(); + if (wheelDelta != 0) + ctx->ProcessMouseWheel(wheelDelta / -120, 0); + + ctx->Update(); +} + +void UIScene_HelpAndOptionsMenu::render(S32 width, S32 height, C4JRender::eViewportType viewport) +{ +} + +void UIScene_HelpAndOptionsMenu::handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled) +{ + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx) + return; + + if (!pressed) + return; + + switch (key) { - return L"HelpAndOptionsMenu"; + case ACTION_MENU_UP: + ctx->ProcessKeyDown(Rml::Input::KI_UP, 0); + handled = true; + break; + case ACTION_MENU_DOWN: + ctx->ProcessKeyDown(Rml::Input::KI_DOWN, 0); + handled = true; + break; + case ACTION_MENU_OK: + if (g_KBMInput.IsKeyPressed(VK_RETURN)) + { + ctx->ProcessKeyDown(Rml::Input::KI_RETURN, 0); + handled = true; + } + break; + case ACTION_MENU_CANCEL: + case ACTION_MENU_PAUSEMENU: + navigateBack(); + handled = true; + break; + default: + break; + } +} + +void UIScene_HelpAndOptionsMenu::ProcessEvent(Rml::Event& event) +{ + const Rml::String& id = event.GetCurrentElement()->GetId(); + + if (event == Rml::EventId::Click) + { + if (id == "change_skin") OnChangeSkin(); + else if (id == "how_to_play") OnHowToPlay(); + else if (id == "controls") OnControls(); + else if (id == "settings") OnSettings(); + else if (id == "credits") OnCredits(); } } void UIScene_HelpAndOptionsMenu::updateTooltips() { - ui.SetTooltips( m_iPad, IDS_TOOLTIPS_SELECT,IDS_TOOLTIPS_BACK); } void UIScene_HelpAndOptionsMenu::updateComponents() { - bool bNotInGame=(Minecraft::GetInstance()->level==nullptr); - if(bNotInGame) + bool bNotInGame = (Minecraft::GetInstance()->level == nullptr); + if (bNotInGame) { - m_parentLayer->showComponent(m_iPad,eUIComponent_Panorama,true); - m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,true); + m_parentLayer->showComponent(m_iPad, eUIComponent_Panorama, true); + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, true); } else { - m_parentLayer->showComponent(m_iPad,eUIComponent_Panorama,false); - - if( app.GetLocalPlayerCount() == 1 ) m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,true); - else m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,false); + m_parentLayer->showComponent(m_iPad, eUIComponent_Panorama, false); + if (app.GetLocalPlayerCount() == 1) + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, true); + else + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, false); } } -void UIScene_HelpAndOptionsMenu::handleReload() +void UIScene_HelpAndOptionsMenu::navigateBack() { -#ifndef _DEBUG // def _FINAL_BUILD // disable debug settings in release builds - removeControl( &m_buttons[BUTTON_HAO_DEBUG], false); -#else - if(!app.DebugSettingsOn()) removeControl( &m_buttons[BUTTON_HAO_DEBUG], false); -#endif - - // 4J-PB - do not need a storage device to see this menu - just need one when you choose to re-install them - bool bNotInGame=(Minecraft::GetInstance()->level==nullptr); - - // any content to be re-installed? - if(m_iPad==ProfileManager.GetPrimaryPad() && bNotInGame) - { - // We should show the reinstall menu - app.DebugPrintf("Reinstall Menu required...\n"); - } - else - { - removeControl( &m_buttons[BUTTON_HAO_REINSTALL], false); - } - - if(app.GetLocalPlayerCount()>1) - { - // no credits in splitscreen - removeControl( &m_buttons[BUTTON_HAO_CREDITS], false); - -#if TO_BE_IMPLEMENTED - app.AdjustSplitscreenScene(m_hObj,&m_OriginalPosition,m_iPad,false); -#endif - if(ProfileManager.GetPrimaryPad()!=m_iPad) - { - removeControl( &m_buttons[BUTTON_HAO_REINSTALL], false); - } - } - - if(!ProfileManager.IsFullVersion() )//|| ProfileManager.IsGuest(m_iPad)) - { -#if TO_BE_IMPLEMENTED - m_Buttons[BUTTON_HAO_CHANGESKIN].SetEnable(FALSE); - m_Buttons[BUTTON_HAO_CHANGESKIN].EnableInput(FALSE); - // set the focus to the second button - - XuiElementSetUserFocus(m_Buttons[BUTTON_HAO_HOWTOPLAY].m_hObj, m_iPad); -#endif - } - - if(!ProfileManager.IsFullVersion() )//|| ProfileManager.IsGuest(m_iPad)) - { - removeControl( &m_buttons[BUTTON_HAO_CHANGESKIN], false); - } - - doHorizontalResizeCheck(); + ui.PlayUISFX(eSFX_Back); + UIScene::navigateBack(); } -void UIScene_HelpAndOptionsMenu::handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled) -{ - //app.DebugPrintf("UIScene_DebugOverlay handling input for pad %d, key %d, down- %s, pressed- %s, released- %s\n", iPad, key, down?"TRUE":"FALSE", pressed?"TRUE":"FALSE", released?"TRUE":"FALSE"); - - ui.AnimateKeyPress(m_iPad, key, repeat, pressed, released); - - switch(key) - { - case ACTION_MENU_CANCEL: - if(pressed && !repeat) - { - navigateBack(); - } - break; - case ACTION_MENU_OK: -#ifdef __ORBIS__ - case ACTION_MENU_TOUCHPAD_PRESS: -#endif - //CD - Added for audio - if(pressed) - { - ui.PlayUISFX(eSFX_Press); - } - - case ACTION_MENU_UP: - case ACTION_MENU_DOWN: - sendInputToMovie(key, repeat, pressed, released); - break; - } -} - -void UIScene_HelpAndOptionsMenu::handlePress(F64 controlId, F64 childId) +void UIScene_HelpAndOptionsMenu::OnChangeSkin() { ui.PlayUISFX(eSFX_Press); - - switch(static_cast(controlId)) - { - case BUTTON_HAO_CHANGESKIN: - ui.NavigateToScene(m_iPad, eUIScene_SkinSelectMenu); - break; - case BUTTON_HAO_HOWTOPLAY: - ui.NavigateToScene(m_iPad, eUIScene_HowToPlayMenu); - break; - case BUTTON_HAO_CONTROLS: - ui.NavigateToScene(m_iPad, eUIScene_ControlsMenu); - break; - case BUTTON_HAO_SETTINGS: - ui.NavigateToScene(m_iPad, eUIScene_SettingsMenu); - break; - case BUTTON_HAO_CREDITS: - ui.NavigateToScene(m_iPad, eUIScene_Credits); - break; - case BUTTON_HAO_REINSTALL: - ui.NavigateToScene(m_iPad, eUIScene_ReinstallMenu); - break; - case BUTTON_HAO_DEBUG: - ui.NavigateToScene(m_iPad, eUIScene_DebugOptions); - break; - } + ui.NavigateToScene(m_iPad, eUIScene_SkinSelectMenu); } + +void UIScene_HelpAndOptionsMenu::OnHowToPlay() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_HowToPlayMenu); +} + +void UIScene_HelpAndOptionsMenu::OnControls() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_ControlsMenu); +} + +void UIScene_HelpAndOptionsMenu::OnSettings() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_SettingsMenu); +} + +void UIScene_HelpAndOptionsMenu::OnCredits() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_Credits); +} + + diff --git a/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.h b/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.h index 203011d3..a8c720d4 100644 --- a/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.h +++ b/Minecraft.Client/Common/UI/UIScene_HelpAndOptionsMenu.h @@ -1,50 +1,57 @@ #pragma once +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + #include "UIScene.h" +#include +#include -#define BUTTON_HAO_CHANGESKIN 0 -#define BUTTON_HAO_HOWTOPLAY 1 -#define BUTTON_HAO_CONTROLS 2 -#define BUTTON_HAO_SETTINGS 3 -#define BUTTON_HAO_CREDITS 4 -#define BUTTON_HAO_REINSTALL 5 -#define BUTTON_HAO_DEBUG 6 -#define BUTTONS_HAO_MAX BUTTON_HAO_DEBUG + 1 +// NOTE: pop_macro at end of file -class UIScene_HelpAndOptionsMenu : public UIScene +class UIScene_HelpAndOptionsMenu : public UIScene, public Rml::EventListener { -private: - UIControl_Button m_buttons[BUTTONS_HAO_MAX]; - UI_BEGIN_MAP_ELEMENTS_AND_NAMES(UIScene) - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_CHANGESKIN], "Button1") - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_HOWTOPLAY], "Button2") - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_CONTROLS], "Button3") - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_SETTINGS], "Button4") - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_CREDITS], "Button5") - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_REINSTALL], "Button6") - UI_MAP_ELEMENT( m_buttons[BUTTON_HAO_DEBUG], "Button7") - UI_END_MAP_ELEMENTS_AND_NAMES() - - bool m_bNotInGame; public: UIScene_HelpAndOptionsMenu(int iPad, void *initData, UILayer *parentLayer); virtual ~UIScene_HelpAndOptionsMenu(); - virtual EUIScene getSceneType() { return eUIScene_HelpAndOptionsMenu;} - - virtual void updateTooltips(); - virtual void updateComponents(); + virtual EUIScene getSceneType() { return eUIScene_HelpAndOptionsMenu; } -protected: - // TODO: This should be pure virtual in this class - virtual wstring getMoviePath(); + virtual void tick(); + virtual void render(S32 width, S32 height, C4JRender::eViewportType viewport); -public: - virtual void handleReload(); - - // INPUT virtual void handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled); + // Override focus/movie lifecycle for RmlUi (no Iggy movie) + virtual void gainFocus() override; + virtual void reloadMovie(bool force) override; + + // Rml::EventListener + void ProcessEvent(Rml::Event& event) override; + protected: - void handlePress(F64 controlId, F64 childId); -}; \ No newline at end of file + virtual wstring getMoviePath() { return L""; } + + virtual void handleLoseFocus() override; + virtual void updateTooltips() override; + virtual void updateComponents() override; + + void navigateBack(); + + void OnChangeSkin(); + void OnHowToPlay(); + void OnControls(); + void OnSettings(); + void OnCredits(); + +private: + Rml::ElementDocument* m_document; +}; + +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") diff --git a/Minecraft.Client/Common/UI/UIScene_LoadCreateJoinMenu.cpp b/Minecraft.Client/Common/UI/UIScene_LoadCreateJoinMenu.cpp index 6b1bbc6d..9e3ac8d3 100644 --- a/Minecraft.Client/Common/UI/UIScene_LoadCreateJoinMenu.cpp +++ b/Minecraft.Client/Common/UI/UIScene_LoadCreateJoinMenu.cpp @@ -286,11 +286,16 @@ UIScene_LoadCreateJoinMenu::UIScene_LoadCreateJoinMenu(int iPad, void* initData, #endif // block input if we're waiting for DLC to install, and wipe the saves list. The end of dlc mounting custom message will fill the list again +#if defined(_WINDOWS64) + // On Windows64, DLC is file-based and doesn't block loading; initialise immediately + Initialise(); +#else if (app.StartInstallDLCProcess(m_iPad) == true || app.DLCInstallPending()) // if we're waiting for DLC to mount, don't fill the save list. The custom message on end of dlc mounting will do that m_bIgnoreInput = true; else Initialise(); +#endif #ifdef __PSVITA__ if (CGameNetworkManager::usingAdhocMode() && SQRNetworkManager_AdHoc_Vita::GetAdhocStatus()) @@ -554,6 +559,7 @@ void UIScene_LoadCreateJoinMenu::handleGainFocus(bool navBack) m_bIgnoreInput = false; +#if !defined(_WINDOWS64) if (app.StartInstallDLCProcess(m_iPad) == false) m_bIgnoreInput = false; @@ -567,6 +573,7 @@ void UIScene_LoadCreateJoinMenu::handleGainFocus(bool navBack) m_controlSavesTimer.setVisible(true); } +#endif if (m_bMultiplayerAllowed) diff --git a/Minecraft.Client/Common/UI/UIScene_PauseMenu.cpp b/Minecraft.Client/Common/UI/UIScene_PauseMenu.cpp index 8e1bdc6a..9c9ba14a 100644 --- a/Minecraft.Client/Common/UI/UIScene_PauseMenu.cpp +++ b/Minecraft.Client/Common/UI/UIScene_PauseMenu.cpp @@ -1,736 +1,488 @@ +#include "stdafx.h" + +// Non-RmlUi includes that depend on Windows macros #include "UI.h" +#include "../../MinecraftServer.h" +#include "../../MultiPlayerLocalPlayer.h" + +// RmlUi includes (Windows macros temporarily suppressed) +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + #include "UIScene_PauseMenu.h" -#include "..\..\MinecraftServer.h" -#include "..\..\MultiplayerLocalPlayer.h" -#include "..\..\TexturePackRepository.h" -#include "..\..\TexturePack.h" -#include "..\..\DLCTexturePack.h" -#include "../DLC/DLCPack.h" -#include "../Tutorial/TutorialMode.h" -#include "..\..\..\Minecraft.World\StringHelpers.h" +#include "RmlManager.h" +#include "../../Windows64/KeyboardMouseInput.h" +#include +#include -UIScene_PauseMenu::UIScene_PauseMenu(int iPad, void *initData, UILayer *parentLayer) : UIScene(iPad, parentLayer) +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") + +UIScene_PauseMenu::UIScene_PauseMenu(int iPad, void *initData, UILayer *parentLayer) + : UIScene(iPad, parentLayer) + , m_document(nullptr) + , m_bIgnoreInput(false) + , m_bExitDialog(false) + , m_bSaveDialog(false) { - // Setup all the Iggy references we need for this scene - initialiseMovie(); - m_bIgnoreInput=false; - m_eAction=eAction_None; - - m_buttons[BUTTON_PAUSE_RESUMEGAME].init(app.GetString(IDS_RESUME_GAME),BUTTON_PAUSE_RESUMEGAME); - m_buttons[BUTTON_PAUSE_HELPANDOPTIONS].init(app.GetString(IDS_HELP_AND_OPTIONS),BUTTON_PAUSE_HELPANDOPTIONS); - m_buttons[BUTTON_PAUSE_LEADERBOARDS].init(app.GetString(IDS_LEADERBOARDS),BUTTON_PAUSE_LEADERBOARDS); -#ifdef _DURANGO - m_buttons[BUTTON_PAUSE_XBOXHELP].init(app.GetString(IDS_XBOX_HELP_APP), BUTTON_PAUSE_XBOXHELP); -#else - m_buttons[BUTTON_PAUSE_ACHIEVEMENTS].init(app.GetString(IDS_ACHIEVEMENTS),BUTTON_PAUSE_ACHIEVEMENTS); -#endif -#if defined(_XBOX_ONE) || defined(__ORBIS__) - m_bTrialTexturePack = false; - if(!Minecraft::GetInstance()->skins->isUsingDefaultSkin()) + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx) { - TexturePack *tPack = Minecraft::GetInstance()->skins->getSelected(); - DLCTexturePack *pDLCTexPack=(DLCTexturePack *)tPack; - - m_pDLCPack=pDLCTexPack->getDLCInfoParentPack();//tPack->getDLCPack(); - - if(!m_pDLCPack->hasPurchasedFile( DLCManager::e_DLCType_Texture, L"" )) - { - m_bTrialTexturePack = true; - } + app.DebugPrintf("[RmlPauseMenu] No RmlUi context available\n"); + return; } - // 4J-TomK - check for all possible labels being fed into BUTTON_PAUSE_SAVEGAME (Bug 163775) - // this has to be done before button initialisation! - wchar_t saveButtonLabels[2][256]; - swprintf( saveButtonLabels[0], 256, L"%ls", app.GetString( IDS_SAVE_GAME )); - swprintf( saveButtonLabels[1], 256, L"%ls", app.GetString( IDS_DISABLE_AUTOSAVE )); - m_buttons[BUTTON_PAUSE_SAVEGAME].setAllPossibleLabels(2,saveButtonLabels); - - if(app.GetGameHostOption(eGameHostOption_DisableSaving) || m_bTrialTexturePack) + m_document = ctx->LoadDocument("PauseMenu.rml"); + if (m_document) { - m_savesDisabled = true; - m_buttons[BUTTON_PAUSE_SAVEGAME].init(app.GetString(IDS_SAVE_GAME),BUTTON_PAUSE_SAVEGAME); + app.DebugPrintf("[RmlPauseMenu] Document loaded OK, showing...\n"); + m_document->Show(); + app.DebugPrintf("[RmlPauseMenu] Document shown, visible=%d, context docs=%d\n", + m_document->IsVisible(), ctx->GetNumDocuments()); + m_document->GetElementById("resume")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("help_options")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("achievements")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("save_game")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_game")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_save")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_nosave")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_cancel")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("save_cancel")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("save_confirm")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("resume")->Focus(); } else { - m_savesDisabled = false; - m_buttons[BUTTON_PAUSE_SAVEGAME].init(app.GetString(IDS_DISABLE_AUTOSAVE),BUTTON_PAUSE_SAVEGAME); - } -#else - m_buttons[BUTTON_PAUSE_SAVEGAME].init(app.GetString(IDS_SAVE_GAME),BUTTON_PAUSE_SAVEGAME); -#endif - m_buttons[BUTTON_PAUSE_EXITGAME].init(app.GetString(IDS_EXIT_GAME),BUTTON_PAUSE_EXITGAME); - - if(!ProfileManager.IsFullVersion()) - { - // hide the trial timer - ui.ShowTrialTimer(false); + app.DebugPrintf("[RmlPauseMenu] Failed to load PauseMenu.rml\n"); } - updateControlsVisibility(); - - doHorizontalResizeCheck(); - - // get rid of the quadrant display if it's on ui.HidePressStart(); -#if TO_BE_IMPLEMENTED - XuiSetTimer(m_hObj,IGNORE_KEYPRESS_TIMERID,IGNORE_KEYPRESS_TIME); -#endif + // parentLayer->addComponent(iPad,eUIComponent_Logo); - if( g_NetworkManager.IsLocalGame() && g_NetworkManager.GetPlayerCount() == 1 ) + if (g_NetworkManager.IsLocalGame() && g_NetworkManager.GetPlayerCount() == 1) { - app.SetXuiServerAction(ProfileManager.GetPrimaryPad(),eXuiServerAction_PauseServer,(void *)TRUE); + app.SetXuiServerAction(ProfileManager.GetPrimaryPad(), eXuiServerAction_PauseServer, (void*)TRUE); } TelemetryManager->RecordMenuShown(m_iPad, eUIScene_PauseMenu, 0); TelemetryManager->RecordPauseOrInactive(m_iPad); - - Minecraft *pMinecraft = Minecraft::GetInstance(); - if(pMinecraft != nullptr && pMinecraft->localgameModes[iPad] != nullptr ) - { - TutorialMode *gameMode = static_cast(pMinecraft->localgameModes[iPad]); - - // This just allows it to be shown - gameMode->getTutorial()->showTutorialPopup(false); - } - m_bErrorDialogRunning = false; } UIScene_PauseMenu::~UIScene_PauseMenu() { - Minecraft *pMinecraft = Minecraft::GetInstance(); - if(pMinecraft != nullptr && pMinecraft->localgameModes[m_iPad] != nullptr ) + if (m_document) { - TutorialMode *gameMode = static_cast(pMinecraft->localgameModes[m_iPad]); - - // This just allows it to be shown - gameMode->getTutorial()->showTutorialPopup(true); + m_document->GetElementById("resume")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("help_options")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("achievements")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("save_game")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_game")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_save")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_nosave")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("exit_cancel")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("save_cancel")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("save_confirm")->RemoveEventListener(Rml::EventId::Click, this); + m_document->Close(); + m_document = nullptr; } - m_parentLayer->showComponent(m_iPad,eUIComponent_Panorama,false); - m_parentLayer->showComponent(m_iPad,eUIComponent_MenuBackground,false); - m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,false); + m_parentLayer->showComponent(m_iPad, eUIComponent_Panorama, false); + m_parentLayer->showComponent(m_iPad, eUIComponent_MenuBackground, false); + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, false); } -wstring UIScene_PauseMenu::getMoviePath() +void UIScene_PauseMenu::gainFocus() { - if(app.GetLocalPlayerCount() > 1) + if (!bHasFocus && stealsFocus()) { - return L"PauseMenuSplit"; - } - else - { - return L"PauseMenu"; + bHasFocus = true; + updateTooltips(); + updateComponents(); + if (!m_bFocussedOnce) + { + // No Iggy set-focus call needed for RmlUi scene + } + handleGainFocus(m_bFocussedOnce); + if (bHasFocus) + m_bFocussedOnce = true; + if (m_document) + m_document->Show(); } } +void UIScene_PauseMenu::handleLoseFocus() +{ + if (m_document) + m_document->Hide(); +} + +void UIScene_PauseMenu::reloadMovie(bool force) +{ +} + void UIScene_PauseMenu::tick() { + // Signal the base class that we've ticked once, so it sets m_bCanHandleInput + // (the Iggy while-loop that normally sets this never runs without a loaded SWF) + m_hasTickedOnce = true; + UIScene::tick(); -#ifdef __PSVITA__ - // 4J-MGH - Need to check for installed DLC here, as we delay the installation of the key file on Vita - if(!app.DLCInstallProcessCompleted()) app.StartInstallDLCProcess(0); -#endif + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx || !bHasFocus) + return; + // Forward mouse position (for hover effects and correct click targeting) + ctx->ProcessMouseMove(g_KBMInput.GetMouseX(), g_KBMInput.GetMouseY(), 0); -#if defined _XBOX_ONE || defined __ORBIS__ - if(!m_bTrialTexturePack && m_savesDisabled != (app.GetGameHostOption(eGameHostOption_DisableSaving) != 0) && ProfileManager.GetPrimaryPad() == m_iPad ) - { - // We show the save button if saves are disabled as this lets us show a prompt to enable them (via purchasing a texture pack) - if( app.GetGameHostOption(eGameHostOption_DisableSaving) ) - { - m_savesDisabled = true; - m_buttons[BUTTON_PAUSE_SAVEGAME].setLabel( app.GetString(IDS_SAVE_GAME) ); - } - else - { - m_savesDisabled = false; - m_buttons[BUTTON_PAUSE_SAVEGAME].setLabel( app.GetString(IDS_DISABLE_AUTOSAVE) ); - } - } -#endif + // Forward mouse button presses/releases + if (g_KBMInput.IsMouseButtonPressed(KeyboardMouseInput::MOUSE_LEFT)) + ctx->ProcessMouseButtonDown(0, 0); + if (g_KBMInput.IsMouseButtonReleased(KeyboardMouseInput::MOUSE_LEFT)) + ctx->ProcessMouseButtonUp(0, 0); + if (g_KBMInput.IsMouseButtonPressed(KeyboardMouseInput::MOUSE_RIGHT)) + ctx->ProcessMouseButtonDown(1, 0); + if (g_KBMInput.IsMouseButtonReleased(KeyboardMouseInput::MOUSE_RIGHT)) + ctx->ProcessMouseButtonUp(1, 0); -#ifdef __ORBIS__ - // Process the error dialog (for a patch being available) - if(m_bErrorDialogRunning) - { - SceErrorDialogStatus stat = sceErrorDialogUpdateStatus(); - if( stat == SCE_ERROR_DIALOG_STATUS_FINISHED ) - { - sceErrorDialogTerminate(); - m_bErrorDialogRunning=false; - } - } -#endif + // Forward scroll wheel + int wheelDelta = g_KBMInput.GetMouseWheel(); + if (wheelDelta != 0) + ctx->ProcessMouseWheel(wheelDelta / -120, 0); + + ctx->Update(); } -void UIScene_PauseMenu::updateTooltips() +void UIScene_PauseMenu::render(S32 width, S32 height, C4JRender::eViewportType viewport) { - bool bUserisClientSide = ProfileManager.IsSignedInLive(m_iPad); - bool bIsisPrimaryHost=g_NetworkManager.IsHost() && (ProfileManager.GetPrimaryPad()==m_iPad); - -#ifdef _XBOX_ONE - bool bDisplayBanTip = !g_NetworkManager.IsLocalGame() && !bIsisPrimaryHost && !ProfileManager.IsGuest(m_iPad); -#endif - - int iY = -1; -#if defined __PS3__ || defined __ORBIS__ - if(m_iPad == ProfileManager.GetPrimaryPad() ) iY = IDS_TOOLTIPS_GAME_INVITES; -#endif - int iRB = -1; - int iX = -1; - - if(ProfileManager.IsFullVersion()) - { - if(StorageManager.GetSaveDisabled()) - { - iX = bIsisPrimaryHost?IDS_TOOLTIPS_SELECTDEVICE:-1; -#ifdef _XBOX_ONE - iRB = bDisplayBanTip?IDS_TOOLTIPS_BANLEVEL:-1; -#endif - if( CSocialManager::Instance()->IsTitleAllowedToPostImages() && CSocialManager::Instance()->AreAllUsersAllowedToPostImages() && bUserisClientSide ) - { -#ifndef __PS3__ - iY = IDS_TOOLTIPS_SHARE; -#endif - } - } - else - { - iX = bIsisPrimaryHost?IDS_TOOLTIPS_CHANGEDEVICE:-1; -#ifdef _XBOX_ONE - iRB = bDisplayBanTip?IDS_TOOLTIPS_BANLEVEL:-1; -#endif - if( CSocialManager::Instance()->IsTitleAllowedToPostImages() && CSocialManager::Instance()->AreAllUsersAllowedToPostImages() && bUserisClientSide) - { -#ifndef __PS3__ - iY = IDS_TOOLTIPS_SHARE; -#endif - } - } - } - ui.SetTooltips( m_iPad, IDS_TOOLTIPS_SELECT,IDS_TOOLTIPS_BACK,iX,iY, -1,-1,-1,iRB); -} - -void UIScene_PauseMenu::updateComponents() -{ - m_parentLayer->showComponent(m_iPad,eUIComponent_Panorama,false); - m_parentLayer->showComponent(m_iPad,eUIComponent_MenuBackground,true); - - if( app.GetLocalPlayerCount() == 1 ) m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,true); - else m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,false); -} - -void UIScene_PauseMenu::handlePreReload() -{ -#if defined _XBOX_ONE || defined __ORBIS__ - if(ProfileManager.GetPrimaryPad() == m_iPad) - { - // 4J-TomK - check for all possible labels being fed into BUTTON_PAUSE_SAVEGAME (Bug 163775) - // this has to be done before button initialisation! - wchar_t saveButtonLabels[2][256]; - swprintf( saveButtonLabels[0], 256, L"%ls", app.GetString( IDS_SAVE_GAME )); - swprintf( saveButtonLabels[1], 256, L"%ls", app.GetString( IDS_DISABLE_AUTOSAVE )); - m_buttons[BUTTON_PAUSE_SAVEGAME].setAllPossibleLabels(2,saveButtonLabels); - } -#endif -} - -void UIScene_PauseMenu::handleReload() -{ - updateTooltips(); - updateControlsVisibility(); - -#if defined _XBOX_ONE || defined __ORBIS__ - if(ProfileManager.GetPrimaryPad() == m_iPad) - { - // We show the save button if saves are disabled as this lets us show a prompt to enable them (via purchasing a texture pack) - if( app.GetGameHostOption(eGameHostOption_DisableSaving) || m_bTrialTexturePack ) - { - m_savesDisabled = true; - m_buttons[BUTTON_PAUSE_SAVEGAME].setLabel( app.GetString(IDS_SAVE_GAME) ); - } - else - { - m_savesDisabled = false; - m_buttons[BUTTON_PAUSE_SAVEGAME].setLabel( app.GetString(IDS_DISABLE_AUTOSAVE) ); - } - } -#endif - - doHorizontalResizeCheck(); -} - -void UIScene_PauseMenu::updateControlsVisibility() -{ - // are we the primary player? - // 4J-PB - fix for 7844 & 7845 - - // TCR # 128: XLA Pause Menu: When in a multiplayer game as a client the Pause Menu does not have a Leaderboards option. - // TCR # 128: XLA Pause Menu: When in a multiplayer game as a client the Pause Menu does not have an Achievements option. - if(ProfileManager.GetPrimaryPad()==m_iPad) // && g_NetworkManager.IsHost()) - { - // are we in splitscreen? - // how many local players do we have? - if( app.GetLocalPlayerCount()>1 ) - { - // Hide the BUTTON_PAUSE_LEADERBOARDS and BUTTON_PAUSE_ACHIEVEMENTS - removeControl( &m_buttons[BUTTON_PAUSE_LEADERBOARDS], false ); -#ifndef _XBOX_ONE - removeControl( &m_buttons[BUTTON_PAUSE_ACHIEVEMENTS], false ); -#endif - } - - if( !g_NetworkManager.IsHost() ) - { - // Hide the BUTTON_PAUSE_SAVEGAME - removeControl( &m_buttons[BUTTON_PAUSE_SAVEGAME], false ); - } - } - else - { - // Hide the BUTTON_PAUSE_LEADERBOARDS, BUTTON_PAUSE_ACHIEVEMENTS and BUTTON_PAUSE_SAVEGAME - removeControl( &m_buttons[BUTTON_PAUSE_LEADERBOARDS], false ); -#ifndef _XBOX_ONE - removeControl( &m_buttons[BUTTON_PAUSE_ACHIEVEMENTS], false ); -#endif - removeControl( &m_buttons[BUTTON_PAUSE_SAVEGAME], false ); - } - - // is saving disabled? - if(StorageManager.GetSaveDisabled()) - { -#ifdef _XBOX - // disable save button - m_buttons[BUTTON_PAUSE_SAVEGAME].setEnable(false); -#endif - } - -#if defined(__PS3__) || defined (__PSVITA__) || defined(__ORBIS__) - // We don't have a way to display trophies/achievements, so remove the button, and we're allowed to not have it on Xbox One - removeControl( &m_buttons[BUTTON_PAUSE_ACHIEVEMENTS], false ); -#endif - } void UIScene_PauseMenu::handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled) { - if(m_bIgnoreInput) - { + if (m_bIgnoreInput) return; + + if (pressed) + { + if (iPad == ProfileManager.GetPrimaryPad() && g_NetworkManager.IsLocalGame()) + { + app.SetXuiServerAction(ProfileManager.GetPrimaryPad(), eXuiServerAction_PauseServer, (void*)FALSE); + } } - //app.DebugPrintf("UIScene_DebugOverlay handling input for pad %d, key %d, down- %s, pressed- %s, released- %s\n", iPad, key, down?"TRUE":"FALSE", pressed?"TRUE":"FALSE", released?"TRUE":"FALSE"); - ui.AnimateKeyPress(iPad, key, repeat, pressed, released); + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx) + return; -#ifdef _XBOX_ONE - bool bIsisPrimaryHost=g_NetworkManager.IsHost() && (ProfileManager.GetPrimaryPad()==iPad); - bool bDisplayBanTip = !g_NetworkManager.IsLocalGame() && !bIsisPrimaryHost && !ProfileManager.IsGuest(iPad); -#endif + // Translate game actions to RmlUi key events. + // Mouse clicks are forwarded in tick() via ProcessMouseButtonDown/Up, + // so here we only handle keyboard-originated actions. + if (!pressed) + return; - switch(key) + switch (key) { -#ifdef _DURANGO - case ACTION_MENU_GTC_RESUME: -#endif -#if defined(__PS3__) // not for Orbis - we want to use the pause menu (touchpad press) to select a menu item - case ACTION_MENU_PAUSEMENU: -#endif - case ACTION_MENU_CANCEL: - if(pressed) - { - if( iPad == ProfileManager.GetPrimaryPad() && g_NetworkManager.IsLocalGame() ) - { - app.SetXuiServerAction(ProfileManager.GetPrimaryPad(),eXuiServerAction_PauseServer,(void *)FALSE); - } - - ui.PlayUISFX(eSFX_Back); - navigateBack(); - if(!ProfileManager.IsFullVersion()) - { - ui.ShowTrialTimer(true); - } - } + case ACTION_MENU_UP: + ctx->ProcessKeyDown(Rml::Input::KI_UP, 0); + handled = true; + break; + case ACTION_MENU_DOWN: + ctx->ProcessKeyDown(Rml::Input::KI_DOWN, 0); + handled = true; break; case ACTION_MENU_OK: -#ifdef __ORBIS__ - case ACTION_MENU_TOUCHPAD_PRESS: -#endif - case ACTION_MENU_UP: - case ACTION_MENU_DOWN: - if(pressed) + // Only forward as keyboard Enter if actually coming from keyboard, + // not from mouse click (mouse is handled in tick()). + if (g_KBMInput.IsKeyPressed(VK_RETURN)) { - sendInputToMovie(key, repeat, pressed, released); + ctx->ProcessKeyDown(Rml::Input::KI_RETURN, 0); + handled = true; } break; - -#ifdef _XBOX_ONE - case ACTION_MENU_RIGHT_SCROLL: - if( bDisplayBanTip ) + case ACTION_MENU_CANCEL: + case ACTION_MENU_PAUSEMENU: + if (m_bExitDialog) { - UINT uiIDA[2]; - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_ACTION_BAN_LEVEL_TITLE, IDS_ACTION_BAN_LEVEL_DESCRIPTION, uiIDA, 2, iPad,&UIScene_PauseMenu::BanGameDialogReturned,(LPVOID)GetCallbackUniqueId() ); - - //rfHandled = TRUE; + OnExitCancel(); + handled = true; + } + else if (m_bSaveDialog) + { + OnSaveCancel(); + handled = true; + } + else + { + navigateBack(); + handled = true; } break; -#endif - } -} - -void UIScene_PauseMenu::handlePress(F64 controlId, F64 childId) -{ - ui.PlayUISFX(eSFX_Press); - - if(m_bIgnoreInput) return; - - switch(static_cast(controlId)) - { - case BUTTON_PAUSE_RESUMEGAME: - if( m_iPad == ProfileManager.GetPrimaryPad() && g_NetworkManager.IsLocalGame() ) - { - app.SetXuiServerAction(ProfileManager.GetPrimaryPad(),eXuiServerAction_PauseServer,(void *)FALSE); - } - navigateBack(); - break; - case BUTTON_PAUSE_LEADERBOARDS: - { - UINT uiIDA[1]; - uiIDA[0]=IDS_OK; - - //4J Gordon: Being used for the leaderboards proper now - // guests can't look at leaderboards - if(ProfileManager.IsGuest(m_iPad)) - { - ui.RequestAlertMessage(IDS_PRO_GUESTPROFILE_TITLE, IDS_PRO_GUESTPROFILE_TEXT, uiIDA, 1, ProfileManager.GetPrimaryPad()); - } - else if(!ProfileManager.IsSignedInLive(m_iPad)) - { - UINT uiIDA[1] = { IDS_OK }; - ui.RequestErrorMessage(IDS_PRO_NOTONLINE_TITLE, IDS_PRO_XBOXLIVE_NOTIFICATION, uiIDA, 1, m_iPad); - } - else - { - bool bContentRestricted=false; - - if(!bContentRestricted) - ui.NavigateToScene(m_iPad, eUIScene_LeaderboardsMenu); - } - } - break; - case BUTTON_PAUSE_HELPANDOPTIONS: - ui.NavigateToScene(m_iPad,eUIScene_HelpAndOptionsMenu); - break; - case BUTTON_PAUSE_SAVEGAME: - PerformActionSaveGame(); - break; - case BUTTON_PAUSE_EXITGAME: - { - Minecraft *pMinecraft = Minecraft::GetInstance(); - // Check if it's the trial version - if(ProfileManager.IsFullVersion()) - { - UINT uiIDA[3]; - - // is it the primary player exiting? - if(m_iPad==ProfileManager.GetPrimaryPad()) - { - int playTime = -1; - if( pMinecraft->localplayers[m_iPad] != nullptr ) - { - playTime = static_cast(pMinecraft->localplayers[m_iPad]->getSessionTimer()); - } - -#if defined(_XBOX_ONE) || defined(__ORBIS__) - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - - if(g_NetworkManager.IsHost() && StorageManager.GetSaveDisabled()) - { - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_EXIT_GAME_SAVE; - uiIDA[2]=IDS_EXIT_GAME_NO_SAVE; - - if(g_NetworkManager.GetPlayerCount()>1) - { - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME_CONFIRM_DISCONNECT_SAVE, uiIDA, 3, m_iPad,&UIScene_PauseMenu::ExitGameSaveDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - else - { - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME, uiIDA, 3, m_iPad,&UIScene_PauseMenu::ExitGameSaveDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - } - else if(g_NetworkManager.IsHost() && g_NetworkManager.GetPlayerCount()>1) - { - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME_CONFIRM_DISCONNECT, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::ExitGameDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - else - { - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::ExitGameDialogReturned, (LPVOID)GetCallbackUniqueId()); - } -#else - if(StorageManager.GetSaveDisabled()) - { - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME_PROGRESS_LOST, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::ExitGameDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - else - { - if( g_NetworkManager.IsHost() ) - { - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_EXIT_GAME_SAVE; - uiIDA[2]=IDS_EXIT_GAME_NO_SAVE; - - if(g_NetworkManager.GetPlayerCount()>1) - { - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME_CONFIRM_DISCONNECT_SAVE, uiIDA, 3, m_iPad,&UIScene_PauseMenu::ExitGameSaveDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - else - { - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME, uiIDA, 3, m_iPad,&UIScene_PauseMenu::ExitGameSaveDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - } - else - { - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::ExitGameDialogReturned, (LPVOID)GetCallbackUniqueId()); - } - } -#endif - } - else - { - int playTime = -1; - if( pMinecraft->localplayers[m_iPad] != nullptr ) - { - playTime = static_cast(pMinecraft->localplayers[m_iPad]->getSessionTimer()); - } - - TelemetryManager->RecordLevelExit(m_iPad, eSen_LevelExitStatus_Exited); - - - // just exit the player - app.SetAction(m_iPad,eAppAction_ExitPlayer); - } - } - else - { - // is it the primary player exiting? - if(m_iPad==ProfileManager.GetPrimaryPad()) - { - int playTime = -1; - if( pMinecraft->localplayers[m_iPad] != nullptr ) - { - playTime = static_cast(pMinecraft->localplayers[m_iPad]->getSessionTimer()); - } - - // adjust the trial time played - ui.ReduceTrialTimerValue(); - - // exit the level - UINT uiIDA[2]; - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_EXIT_GAME, IDS_CONFIRM_EXIT_GAME_PROGRESS_LOST, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::ExitGameDialogReturned, (LPVOID)GetCallbackUniqueId()); - - } - else - { - int playTime = -1; - if( pMinecraft->localplayers[m_iPad] != nullptr ) - { - playTime = static_cast(pMinecraft->localplayers[m_iPad]->getSessionTimer()); - } - - TelemetryManager->RecordLevelExit(m_iPad, eSen_LevelExitStatus_Exited); - - // just exit the player - app.SetAction(m_iPad,eAppAction_ExitPlayer); - } - } - } + default: break; } } -void UIScene_PauseMenu::PerformActionSaveGame() +void UIScene_PauseMenu::ProcessEvent(Rml::Event& event) { - // 4J-PB - Is the player trying to save but they are using a trial texturepack ? - if(!Minecraft::GetInstance()->skins->isUsingDefaultSkin()) + if (m_bIgnoreInput) + return; + + const Rml::String& id = event.GetCurrentElement()->GetId(); + + if (event == Rml::EventId::Click) { - TexturePack *tPack = Minecraft::GetInstance()->skins->getSelected(); - DLCTexturePack *pDLCTexPack=static_cast(tPack); - - m_pDLCPack=pDLCTexPack->getDLCInfoParentPack();//tPack->getDLCPack(); - - if(!m_pDLCPack->hasPurchasedFile( DLCManager::e_DLCType_Texture, L"" )) - { - return; - } - else - { - m_bTrialTexturePack = false; - } - } - - // does the save exist? - bool bSaveExists; - C4JStorage::ESaveGameState result=StorageManager.DoesSaveExist(&bSaveExists); - -#ifdef _XBOX - if(result == C4JStorage::ELoadGame_DeviceRemoved) - { - // this will be a tester trying to be clever - UINT uiIDA[2]; - uiIDA[0]=IDS_SELECTANEWDEVICE; - uiIDA[1]=IDS_NODEVICE_DECLINE; - - ui.RequestAlertMessage(IDS_STORAGEDEVICEPROBLEM_TITLE, IDS_FAILED_TO_LOADSAVE_TEXT, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::DeviceRemovedDialogReturned,(LPVOID)GetCallbackUniqueId()); - } - else -#endif - { -#if defined(_XBOX_ONE) || defined(__ORBIS__) - if(!m_savesDisabled) - { - UINT uiIDA[2]; - uiIDA[0]=IDS_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_TITLE_DISABLE_AUTOSAVE, IDS_CONFIRM_DISABLE_AUTOSAVE, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::DisableAutosaveDialogReturned,(LPVOID)GetCallbackUniqueId()); - } - else -#endif - // we need to ask if they are sure they want to overwrite the existing game - if(bSaveExists) - { - UINT uiIDA[2]; - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_TITLE_SAVE_GAME, IDS_CONFIRM_SAVE_GAME, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::SaveGameDialogReturned,(LPVOID)GetCallbackUniqueId()); - } - else - { -#if defined(_XBOX_ONE) || defined(__ORBIS__) - UINT uiIDA[2]; - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_TITLE_ENABLE_AUTOSAVE, IDS_CONFIRM_ENABLE_AUTOSAVE, uiIDA, 2, m_iPad,&IUIScene_PauseMenu::EnableAutosaveDialogReturned,(LPVOID)GetCallbackUniqueId()); -#else - // flag a app action of save game - app.SetAction(m_iPad,eAppAction_SaveGame); -#endif - } + if (id == "resume") OnResume(); + else if (id == "help_options") OnHelpAndOptions(); + else if (id == "achievements") OnAchievements(); + else if (id == "save_game") OnSaveGame(); + else if (id == "exit_game") OnExitGame(); + else if (id == "exit_save") OnExitSave(); + else if (id == "exit_nosave") OnExitNoSave(); + else if (id == "exit_cancel") OnExitCancel(); + else if (id == "save_confirm") OnSaveConfirm(); + else if (id == "save_cancel") OnSaveCancel(); } } void UIScene_PauseMenu::ShowScene(bool show) { - app.DebugPrintf("UIScene_PauseMenu::ShowScene is not implemented\n"); + if (m_document) + { + if (show) + m_document->Show(); + else + m_document->Hide(); + } } -void UIScene_PauseMenu::HandleDLCInstalled() +void UIScene_PauseMenu::navigateBack() { - // mounted DLC may have changed - if(app.StartInstallDLCProcess(m_iPad)==false) + if (m_iPad == ProfileManager.GetPrimaryPad() && g_NetworkManager.IsLocalGame()) { - // not doing a mount, so re-enable input - //m_bIgnoreInput=false; - app.DebugPrintf("UIScene_PauseMenu::HandleDLCInstalled - m_bIgnoreInput false\n"); + app.SetXuiServerAction(ProfileManager.GetPrimaryPad(), eXuiServerAction_PauseServer, (void*)FALSE); + } + + ui.PlayUISFX(eSFX_Back); + UIScene::navigateBack(); +} + +void UIScene_PauseMenu::OnResume() +{ + ui.PlayUISFX(eSFX_Back); + navigateBack(); +} + +void UIScene_PauseMenu::OnHelpAndOptions() +{ + ui.NavigateToScene(m_iPad, eUIScene_HelpAndOptionsMenu); +} + +void UIScene_PauseMenu::OnAchievements() +{ + ui.NavigateToScene(m_iPad, eUIScene_AchievementsMenu); +} + +void UIScene_PauseMenu::ShowSaveDialog(bool show) +{ + if (!m_document) + return; + + auto* saveDialog = m_document->GetElementById("save_dialog"); + auto* buttons = m_document->GetElementById("buttons"); + + if (show) + { + if (buttons) buttons->SetProperty("display", "none"); + if (saveDialog) saveDialog->SetProperty("display", "block"); + + auto* saveText = m_document->GetElementById("save_text"); + + if (!ProfileManager.IsFullVersion()) + { + if (saveText) saveText->SetInnerRML("You need to unlock the full game to save."); + } + else + { + if (saveText) saveText->SetInnerRML("Overwrite existing save?"); + } + + m_document->GetElementById("save_cancel")->Focus(); + m_bSaveDialog = true; } else { - // 4J-PB - Somehow, on th edisc build, we get in here, but don't call HandleDLCMountingComplete, so input locks up - //m_bIgnoreInput=true; - app.DebugPrintf("UIScene_PauseMenu::HandleDLCInstalled - m_bIgnoreInput true\n"); + if (saveDialog) saveDialog->SetProperty("display", "none"); + if (buttons) buttons->SetProperty("display", "block"); + + m_document->GetElementById("save_game")->Focus(); + m_bSaveDialog = false; } - // this will send a CustomMessage_DLCMountingComplete when done } - -void UIScene_PauseMenu::HandleDLCMountingComplete() -{ - // check if we should display the save option - - //m_bIgnoreInput=false; - app.DebugPrintf("UIScene_PauseMenu::HandleDLCMountingComplete - m_bIgnoreInput false \n"); - - // if(ProfileManager.IsFullVersion()) - // { - // bool bIsisPrimaryHost=g_NetworkManager.IsHost() && (ProfileManager.GetPrimaryPad()==m_iPad); - // - // if(bIsisPrimaryHost) - // { - // m_buttons[BUTTON_PAUSE_SAVEGAME].setEnable(true); - // } - // } -} - -int UIScene_PauseMenu::UnlockFullSaveReturned(void *pParam,int iPad,C4JStorage::EMessageResult result) +void UIScene_PauseMenu::OnSaveGame() { - Minecraft *pMinecraft=Minecraft::GetInstance(); + if (m_iPad != ProfileManager.GetPrimaryPad()) + return; - if(result==C4JStorage::EMessage_ResultAccept) + if (!ProfileManager.IsFullVersion()) { - if(ProfileManager.IsSignedInLive(pMinecraft->player->GetXboxPad())) + ShowSaveDialog(true); + return; + } + + bool bSaveExists; + StorageManager.DoesSaveExist(&bSaveExists); + + if (bSaveExists) + { + ShowSaveDialog(true); + } + else + { + app.SetAction(m_iPad, eAppAction_SaveGame); + } +} + +void UIScene_PauseMenu::OnSaveConfirm() +{ + ShowSaveDialog(false); + + if (!ProfileManager.IsFullVersion()) + { + ProfileManager.DisplayFullVersionPurchase(false, m_iPad, eSen_UpsellID_Full_Version_Of_Game); + } + else + { + app.SetAction(m_iPad, eAppAction_SaveGame); + } +} + +void UIScene_PauseMenu::OnSaveCancel() +{ + ShowSaveDialog(false); +} + +void UIScene_PauseMenu::ShowExitDialog(bool show) +{ + if (!m_document) + return; + + auto* exitDialog = m_document->GetElementById("exit_dialog"); + auto* buttons = m_document->GetElementById("buttons"); + + if (show) + { + Minecraft* pMinecraft = Minecraft::GetInstance(); + + if (buttons) buttons->SetProperty("display", "none"); + if (exitDialog) exitDialog->SetProperty("display", "block"); + + // Configure dialog for current state + auto* exitSave = m_document->GetElementById("exit_save"); + auto* exitNosave = m_document->GetElementById("exit_nosave"); + auto* exitCancel = m_document->GetElementById("exit_cancel"); + auto* exitText = m_document->GetElementById("exit_text"); + + if (ProfileManager.IsFullVersion() && m_iPad == ProfileManager.GetPrimaryPad()) { - // 4J-PB - need to check this user can access the store -#if defined(__PS3__) || defined(__PSVITA__) - bool bContentRestricted; - ProfileManager.GetChatAndContentRestrictions(ProfileManager.GetPrimaryPad(),true,nullptr,&bContentRestricted,nullptr); - if(bContentRestricted) + // Full version primary pad: show Save & Exit + Exit without Saving + Cancel + if (exitSave) exitSave->SetProperty("display", "inline-block"); + if (exitNosave) exitNosave->SetProperty("display", "inline-block"); + + if (g_NetworkManager.GetPlayerCount() > 1) { - UINT uiIDA[1]; - uiIDA[0]=IDS_CONFIRM_OK; - ui.RequestAlertMessage(IDS_ONLINE_SERVICE_TITLE, IDS_CONTENT_RESTRICTION, uiIDA, 1, ProfileManager.GetPrimaryPad()); + if (exitText) exitText->SetInnerRML("Exit Game (disconnect other players)?"); } else -#endif { - ProfileManager.DisplayFullVersionPurchase(false,pMinecraft->player->GetXboxPad(),eSen_UpsellID_Full_Version_Of_Game); + if (exitText) exitText->SetInnerRML("Exit Game?"); } + if (exitCancel) exitCancel->Focus(); } + else if (!ProfileManager.IsFullVersion() && m_iPad == ProfileManager.GetPrimaryPad()) + { + // Trial version: simple OK/Cancel, "progress will be lost" + if (exitSave) exitSave->SetProperty("display", "none"); + if (exitNosave) exitNosave->SetProperty("display", "inline-block"); + if (exitNosave) exitNosave->SetInnerRML("Yes"); + if (exitText) exitText->SetInnerRML("Exit Game (progress will be lost)?"); + if (exitCancel) exitCancel->Focus(); + } + + m_bExitDialog = true; } else { - //SentientManager.RecordUpsellResponded(iPad, eSen_UpsellID_Full_Version_Of_Game, app.m_dwOfferID, eSen_UpsellOutcome_Declined); - } + if (exitDialog) exitDialog->SetProperty("display", "none"); + if (buttons) buttons->SetProperty("display", "block"); - return 0; + // Restore default button state + auto* exitSave = m_document->GetElementById("exit_save"); + auto* exitNosave = m_document->GetElementById("exit_nosave"); + if (exitSave) exitSave->SetProperty("display", "inline-block"); + if (exitNosave) exitNosave->SetInnerRML("Exit without Saving"); + if (exitNosave) exitNosave->SetProperty("display", "inline-block"); + + m_document->GetElementById("exit_game")->Focus(); + m_bExitDialog = false; + } } -int UIScene_PauseMenu::SaveGame_SignInReturned(void *pParam,bool bContinue, int iPad) +void UIScene_PauseMenu::OnExitGame() { - UIScene_PauseMenu* pClass = static_cast(ui.GetSceneFromCallbackId((size_t)pParam)); - if(pClass) pClass->SetIgnoreInput(false); + Minecraft* pMinecraft = Minecraft::GetInstance(); - if(bContinue==true) + // Non-primary pad: exit directly without dialog + if (m_iPad != ProfileManager.GetPrimaryPad()) { - if(pClass) pClass->PerformActionSaveGame(); + if (pMinecraft->localplayers[m_iPad] != nullptr) + TelemetryManager->RecordLevelExit(m_iPad, eSen_LevelExitStatus_Exited); + app.SetAction(m_iPad, eAppAction_ExitPlayer); + return; } - return 0; + // Primary pad: show inline confirmation dialog (no Iggy MessageBox) + ShowExitDialog(true); } -#ifdef _XBOX_ONE -int UIScene_PauseMenu::BanGameDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result) +void UIScene_PauseMenu::OnExitSave() { - // results switched for this dialog - if(result==C4JStorage::EMessage_ResultDecline) + ShowExitDialog(false); + SetIgnoreInput(true); + + Minecraft* pMinecraft = Minecraft::GetInstance(); + if (pMinecraft->localplayers[m_iPad] != nullptr) { - app.SetAction(iPad,eAppAction_BanLevel); + int playTime = static_cast(pMinecraft->localplayers[m_iPad]->getSessionTimer()); + TelemetryManager->RecordLevelExit(m_iPad, eSen_LevelExitStatus_Exited); } - return 0; -} -#endif -void UIScene_PauseMenu::SetIgnoreInput(bool ignoreInput) + MinecraftServer::getInstance()->setSaveOnExit(true); + app.SetAction(m_iPad, eAppAction_ExitWorld); +} + +void UIScene_PauseMenu::OnExitNoSave() { - m_bIgnoreInput = ignoreInput; -} \ No newline at end of file + ShowExitDialog(false); + SetIgnoreInput(true); + + Minecraft* pMinecraft = Minecraft::GetInstance(); + if (pMinecraft->localplayers[m_iPad] != nullptr) + { + int playTime = static_cast(pMinecraft->localplayers[m_iPad]->getSessionTimer()); + TelemetryManager->RecordLevelExit(m_iPad, eSen_LevelExitStatus_Exited); + } + + app.SetAction(m_iPad, eAppAction_ExitWorld); +} + +void UIScene_PauseMenu::OnExitCancel() +{ + ShowExitDialog(false); +} diff --git a/Minecraft.Client/Common/UI/UIScene_PauseMenu.h b/Minecraft.Client/Common/UI/UIScene_PauseMenu.h index 854b6aa5..9d67f593 100644 --- a/Minecraft.Client/Common/UI/UIScene_PauseMenu.h +++ b/Minecraft.Client/Common/UI/UIScene_PauseMenu.h @@ -1,103 +1,70 @@ #pragma once +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + #include "UIScene.h" #include "IUIScene_PauseMenu.h" +#include +#include -#define BUTTON_PAUSE_RESUMEGAME 0 -#define BUTTON_PAUSE_HELPANDOPTIONS 1 -#define BUTTON_PAUSE_LEADERBOARDS 2 +// NOTE: pop_macro at end of file -#ifdef _XBOX_ONE -#define BUTTON_PAUSE_XBOXHELP 3 -#else -#define BUTTON_PAUSE_ACHIEVEMENTS 3 -#endif - -#define BUTTON_PAUSE_SAVEGAME 4 -#define BUTTON_PAUSE_EXITGAME 5 -#define BUTTONS_PAUSE_MAX BUTTON_PAUSE_EXITGAME + 1 - -class UIScene_PauseMenu : public UIScene, public IUIScene_PauseMenu +class UIScene_PauseMenu : public UIScene, public IUIScene_PauseMenu, public Rml::EventListener { -private: - bool m_savesDisabled; - bool m_bTrialTexturePack; - bool m_bErrorDialogRunning; - - enum eActions - { - eAction_None=0, -#if defined(__PS3__) || defined(__PSVITA__) || defined(__ORBIS__) - eAction_ViewLeaderboardsPSN, - eAction_ViewInvitesPSN, - eAction_SaveGamePSN, - eAction_BuyTexturePackPSN -#endif - - }; - eActions m_eAction; - - UIControl_Button m_buttons[BUTTONS_PAUSE_MAX]; - UI_BEGIN_MAP_ELEMENTS_AND_NAMES(UIScene) - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_RESUMEGAME], "Button1") - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_HELPANDOPTIONS], "Button2") - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_LEADERBOARDS], "Button3") -#ifdef _DURANGO - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_XBOXHELP], "Button4") -#else - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_ACHIEVEMENTS], "Button4") -#endif - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_SAVEGAME], "Button5") - UI_MAP_ELEMENT( m_buttons[BUTTON_PAUSE_EXITGAME], "Button6") - UI_END_MAP_ELEMENTS_AND_NAMES() - - virtual void HandleDLCMountingComplete(); - virtual void HandleDLCInstalled(); -#ifdef _XBOX_ONE - virtual void HandleDLCLicenseChange(); -#endif - static int UnlockFullSaveReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); - static int SaveGame_SignInReturned(void *pParam,bool bContinue, int iPad); - public: UIScene_PauseMenu(int iPad, void *initData, UILayer *parentLayer); virtual ~UIScene_PauseMenu(); - virtual EUIScene getSceneType() { return eUIScene_PauseMenu;} + virtual EUIScene getSceneType() { return eUIScene_PauseMenu; } virtual void tick(); + virtual void render(S32 width, S32 height, C4JRender::eViewportType viewport); - virtual void updateTooltips(); - virtual void updateComponents(); - virtual void handlePreReload(); - virtual void handleReload(); - -protected: - void updateControlsVisibility(); - - // TODO: This should be pure virtual in this class - virtual wstring getMoviePath(); - -public: - // INPUT virtual void handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled); + // Override focus/movie lifecycle for RmlUi (no Iggy movie) + virtual void gainFocus() override; + virtual void reloadMovie(bool force) override; + + // Rml::EventListener + void ProcessEvent(Rml::Event& event) override; + protected: - void handlePress(F64 controlId, F64 childId); - virtual void ShowScene(bool show); - virtual void SetIgnoreInput(bool ignoreInput); - bool m_bIgnoreInput; + virtual wstring getMoviePath() { return L""; } + + virtual void handleLoseFocus() override; + + void ShowScene(bool show) override; + void SetIgnoreInput(bool ignoreInput) override { m_bIgnoreInput = ignoreInput; } + + void navigateBack(); + + void OnResume(); + void OnHelpAndOptions(); + void OnAchievements(); + void OnSaveGame(); + void OnExitGame(); + void OnExitSave(); + void OnExitNoSave(); + void OnExitCancel(); + void OnSaveConfirm(); + void OnSaveCancel(); + + void ShowExitDialog(bool show); + void ShowSaveDialog(bool show); private: - void PerformActionSaveGame(); - -protected: -#ifdef _XBOX_ONE - static int BanGameDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); - virtual long long getDefaultGtcButtons() { return _360_GTC_BACK | _360_GTC_PLAY; } -#endif - -#ifdef __ORBIS__ - bool CheckForPatch(); -#endif + Rml::ElementDocument* m_document; + bool m_bIgnoreInput; + bool m_bExitDialog; + bool m_bSaveDialog; }; + +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") diff --git a/Minecraft.Client/Common/UI/UIScene_SettingsMenu.cpp b/Minecraft.Client/Common/UI/UIScene_SettingsMenu.cpp index 497af6d1..4e70c087 100644 --- a/Minecraft.Client/Common/UI/UIScene_SettingsMenu.cpp +++ b/Minecraft.Client/Common/UI/UIScene_SettingsMenu.cpp @@ -1,168 +1,316 @@ +#include "stdafx.h" + +// Non-RmlUi includes that depend on Windows macros #include "UI.h" +#include "../../Minecraft.h" + +// RmlUi includes (Windows macros temporarily suppressed) +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + #include "UIScene_SettingsMenu.h" -#include "..\..\Minecraft.h" +#include "RmlManager.h" +#include "../../Windows64/KeyboardMouseInput.h" +#include +#include -UIScene_SettingsMenu::UIScene_SettingsMenu(int iPad, void *initData, UILayer *parentLayer) : UIScene(iPad, parentLayer) +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") + +UIScene_SettingsMenu::UIScene_SettingsMenu(int iPad, void *initData, UILayer *parentLayer) + : UIScene(iPad, parentLayer) + , m_document(nullptr) + , m_bConfirmDialog(false) { - // Setup all the Iggy references we need for this scene - initialiseMovie(); - - bool bNotInGame=(Minecraft::GetInstance()->level==nullptr); - - m_buttons[BUTTON_ALL_OPTIONS].init(IDS_OPTIONS,BUTTON_ALL_OPTIONS); - m_buttons[BUTTON_ALL_AUDIO].init(IDS_AUDIO,BUTTON_ALL_AUDIO); - m_buttons[BUTTON_ALL_CONTROL].init(IDS_CONTROL,BUTTON_ALL_CONTROL); - m_buttons[BUTTON_ALL_GRAPHICS].init(IDS_GRAPHICS,BUTTON_ALL_GRAPHICS); - m_buttons[BUTTON_ALL_UI].init(IDS_USER_INTERFACE,BUTTON_ALL_UI); - m_buttons[BUTTON_ALL_RESETTODEFAULTS].init(IDS_RESET_TO_DEFAULTS,BUTTON_ALL_RESETTODEFAULTS); - - if(ProfileManager.GetPrimaryPad()!=m_iPad) + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx) { - removeControl( &m_buttons[BUTTON_ALL_AUDIO], bNotInGame); - removeControl( &m_buttons[BUTTON_ALL_GRAPHICS], bNotInGame); + app.DebugPrintf("[RmlSettingsMenu] No RmlUi context available\n"); + return; } - doHorizontalResizeCheck(); - - if(app.GetLocalPlayerCount()>1) + m_document = ctx->LoadDocument("SettingsMenu.rml"); + if (m_document) { -#if TO_BE_IMPLEMENTED - app.AdjustSplitscreenScene(m_hObj,&m_OriginalPosition,m_iPad,false); -#endif + m_document->Show(); + + // Apply conditional visibility matching original Iggy scene behavior + bool bNotInGame = (Minecraft::GetInstance()->level == nullptr); + + // Audio and Graphics: hidden if not primary pad + if (ProfileManager.GetPrimaryPad() != m_iPad) + { + auto* elAudio = m_document->GetElementById("audio"); + if (elAudio) elAudio->SetProperty("display", "none"); + auto* elGraphics = m_document->GetElementById("graphics"); + if (elGraphics) elGraphics->SetProperty("display", "none"); + } + + // Register button click listeners + m_document->GetElementById("options")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("audio")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("controls")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("graphics")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("ui")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("reset_defaults")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("confirm_yes")->AddEventListener(Rml::EventId::Click, this); + m_document->GetElementById("confirm_no")->AddEventListener(Rml::EventId::Click, this); + + // Focus first visible button + m_document->GetElementById("options")->Focus(); } + else + { + app.DebugPrintf("[RmlSettingsMenu] Failed to load SettingsMenu.rml\n"); + } + + updateComponents(); + + ui.HidePressStart(); + TelemetryManager->RecordMenuShown(m_iPad, eUIScene_SettingsMenu, 0); } UIScene_SettingsMenu::~UIScene_SettingsMenu() { -} - -wstring UIScene_SettingsMenu::getMoviePath() -{ - if(app.GetLocalPlayerCount() > 1) + if (m_document) { - return L"SettingsMenuSplit"; - } - else - { - return L"SettingsMenu"; + m_document->GetElementById("options")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("audio")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("controls")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("graphics")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("ui")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("reset_defaults")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("confirm_yes")->RemoveEventListener(Rml::EventId::Click, this); + m_document->GetElementById("confirm_no")->RemoveEventListener(Rml::EventId::Click, this); + m_document->Close(); + m_document = nullptr; } } -void UIScene_SettingsMenu::handleReload() +void UIScene_SettingsMenu::gainFocus() { - bool bNotInGame=(Minecraft::GetInstance()->level==nullptr); - if(ProfileManager.GetPrimaryPad()!=m_iPad) + if (!bHasFocus && stealsFocus()) { - removeControl( &m_buttons[BUTTON_ALL_AUDIO], bNotInGame); - removeControl( &m_buttons[BUTTON_ALL_GRAPHICS], bNotInGame); + bHasFocus = true; + updateTooltips(); + updateComponents(); + if (!m_bFocussedOnce) + { + } + handleGainFocus(m_bFocussedOnce); + if (bHasFocus) + m_bFocussedOnce = true; + if (m_document) + m_document->Show(); } +} - doHorizontalResizeCheck(); +void UIScene_SettingsMenu::handleLoseFocus() +{ + if (m_document) + m_document->Hide(); } void UIScene_SettingsMenu::updateTooltips() { - ui.SetTooltips( m_iPad, IDS_TOOLTIPS_SELECT,IDS_TOOLTIPS_BACK); } void UIScene_SettingsMenu::updateComponents() { - bool bNotInGame=(Minecraft::GetInstance()->level==nullptr); - if(bNotInGame) + bool bNotInGame = (Minecraft::GetInstance()->level == nullptr); + + if (bNotInGame) { - m_parentLayer->showComponent(m_iPad,eUIComponent_Panorama,true); - m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,true); + m_parentLayer->showComponent(m_iPad, eUIComponent_Panorama, true); + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, true); } else { - m_parentLayer->showComponent(m_iPad,eUIComponent_Panorama,false); - - if( app.GetLocalPlayerCount() == 1 ) m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,true); - else m_parentLayer->showComponent(m_iPad,eUIComponent_Logo,false); + m_parentLayer->showComponent(m_iPad, eUIComponent_Panorama, false); + if (app.GetLocalPlayerCount() == 1) + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, true); + else + m_parentLayer->showComponent(m_iPad, eUIComponent_Logo, false); } } +void UIScene_SettingsMenu::reloadMovie(bool force) +{ +} + +void UIScene_SettingsMenu::tick() +{ + m_hasTickedOnce = true; + + UIScene::tick(); + + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx || !bHasFocus) + return; + + ctx->ProcessMouseMove(g_KBMInput.GetMouseX(), g_KBMInput.GetMouseY(), 0); + + if (g_KBMInput.IsMouseButtonPressed(KeyboardMouseInput::MOUSE_LEFT)) + ctx->ProcessMouseButtonDown(0, 0); + if (g_KBMInput.IsMouseButtonReleased(KeyboardMouseInput::MOUSE_LEFT)) + ctx->ProcessMouseButtonUp(0, 0); + if (g_KBMInput.IsMouseButtonPressed(KeyboardMouseInput::MOUSE_RIGHT)) + ctx->ProcessMouseButtonDown(1, 0); + if (g_KBMInput.IsMouseButtonReleased(KeyboardMouseInput::MOUSE_RIGHT)) + ctx->ProcessMouseButtonUp(1, 0); + + int wheelDelta = g_KBMInput.GetMouseWheel(); + if (wheelDelta != 0) + ctx->ProcessMouseWheel(wheelDelta / -120, 0); + + ctx->Update(); +} + +void UIScene_SettingsMenu::render(S32 width, S32 height, C4JRender::eViewportType viewport) +{ +} + void UIScene_SettingsMenu::handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled) { - //app.DebugPrintf("UIScene_DebugOverlay handling input for pad %d, key %d, down- %s, pressed- %s, released- %s\n", iPad, key, down?"TRUE":"FALSE", pressed?"TRUE":"FALSE", released?"TRUE":"FALSE"); - ui.AnimateKeyPress(m_iPad, key, repeat, pressed, released); + Rml::Context* ctx = RmlManager::Get().GetContext(); + if (!ctx) + return; - switch(key) + if (!pressed) + return; + + switch (key) { - case ACTION_MENU_CANCEL: - if(pressed) - { - // if the profile data has been changed, then force a profile write - // It seems we're allowed to break the 5 minute rule if it's the result of a user action - - app.CheckGameSettingsChanged(true,iPad); - navigateBack(); - } + case ACTION_MENU_UP: + ctx->ProcessKeyDown(Rml::Input::KI_UP, 0); + handled = true; + break; + case ACTION_MENU_DOWN: + ctx->ProcessKeyDown(Rml::Input::KI_DOWN, 0); + handled = true; break; case ACTION_MENU_OK: -#ifdef __ORBIS__ - case ACTION_MENU_TOUCHPAD_PRESS: -#endif - sendInputToMovie(key, repeat, pressed, released); - break; - case ACTION_MENU_UP: - case ACTION_MENU_DOWN: - sendInputToMovie(key, repeat, pressed, released); - break; - } -} - -void UIScene_SettingsMenu::handlePress(F64 controlId, F64 childId) -{ - //CD - Added for audio - ui.PlayUISFX(eSFX_Press); - - switch(static_cast(controlId)) - { - case BUTTON_ALL_OPTIONS: - ui.NavigateToScene(m_iPad, eUIScene_SettingsOptionsMenu); - break; - case BUTTON_ALL_AUDIO: - ui.NavigateToScene(m_iPad, eUIScene_SettingsAudioMenu); - break; - case BUTTON_ALL_CONTROL: - ui.NavigateToScene(m_iPad, eUIScene_SettingsControlMenu); - break; - case BUTTON_ALL_GRAPHICS: - ui.NavigateToScene(m_iPad, eUIScene_SettingsGraphicsMenu); - break; - case BUTTON_ALL_UI: - ui.NavigateToScene(m_iPad, eUIScene_SettingsUIMenu); - break; - case BUTTON_ALL_RESETTODEFAULTS: + if (g_KBMInput.IsKeyPressed(VK_RETURN)) { - // check they really want to do this - UINT uiIDA[2]; - uiIDA[0]=IDS_CONFIRM_CANCEL; - uiIDA[1]=IDS_CONFIRM_OK; - - ui.RequestAlertMessage(IDS_DEFAULTS_TITLE, IDS_DEFAULTS_TEXT, uiIDA, 2, m_iPad,&UIScene_SettingsMenu::ResetDefaultsDialogReturned,this); - } + ctx->ProcessKeyDown(Rml::Input::KI_RETURN, 0); + handled = true; + } + break; + case ACTION_MENU_CANCEL: + case ACTION_MENU_PAUSEMENU: + if (m_bConfirmDialog) + { + OnConfirmNo(); + handled = true; + } + else + { + app.CheckGameSettingsChanged(true, iPad); + navigateBack(); + handled = true; + } + break; + default: break; } } -int UIScene_SettingsMenu::ResetDefaultsDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result) +void UIScene_SettingsMenu::ProcessEvent(Rml::Event& event) { - UIScene_SettingsMenu* pClass = static_cast(pParam); + const Rml::String& id = event.GetCurrentElement()->GetId(); - // results switched for this dialog - if(result==C4JStorage::EMessage_ResultDecline) + if (event == Rml::EventId::Click) { -#if (defined __PS3__ || defined __ORBIS__ || defined _DURANGO || defined __PSVITA__) - app.SetDefaultOptions(StorageManager.GetDashboardProfileSettings(pClass->m_iPad),pClass->m_iPad); -#else - app.SetDefaultOptions(ProfileManager.GetDashboardProfileSettings(pClass->m_iPad),pClass->m_iPad); -#endif - // if the profile data has been changed, then force a profile write - // It seems we're allowed to break the 5 minute rule if it's the result of a user action - app.CheckGameSettingsChanged(true,iPad); + if (id == "options") OnOptions(); + else if (id == "audio") OnAudio(); + else if (id == "controls") OnControls(); + else if (id == "graphics") OnGraphics(); + else if (id == "ui") OnUI(); + else if (id == "reset_defaults") OnResetToDefaults(); + else if (id == "confirm_yes") OnConfirmYes(); + else if (id == "confirm_no") OnConfirmNo(); } - return 0; +} + +void UIScene_SettingsMenu::navigateBack() +{ + ui.PlayUISFX(eSFX_Back); + UIScene::navigateBack(); +} + +void UIScene_SettingsMenu::OnOptions() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_SettingsOptionsMenu); +} + +void UIScene_SettingsMenu::OnAudio() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_SettingsAudioMenu); +} + +void UIScene_SettingsMenu::OnControls() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_SettingsControlMenu); +} + +void UIScene_SettingsMenu::OnGraphics() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_SettingsGraphicsMenu); +} + +void UIScene_SettingsMenu::OnUI() +{ + ui.PlayUISFX(eSFX_Press); + ui.NavigateToScene(m_iPad, eUIScene_SettingsUIMenu); +} + +void UIScene_SettingsMenu::ShowConfirmDialog(bool show) +{ + if (!m_document) + return; + + auto* confirmDialog = m_document->GetElementById("confirm_dialog"); + auto* buttons = m_document->GetElementById("buttons"); + + if (show) + { + if (buttons) buttons->SetProperty("display", "none"); + if (confirmDialog) confirmDialog->SetProperty("display", "block"); + m_document->GetElementById("confirm_no")->Focus(); + m_bConfirmDialog = true; + } + else + { + if (confirmDialog) confirmDialog->SetProperty("display", "none"); + if (buttons) buttons->SetProperty("display", "block"); + m_document->GetElementById("reset_defaults")->Focus(); + m_bConfirmDialog = false; + } +} + +void UIScene_SettingsMenu::OnResetToDefaults() +{ + ShowConfirmDialog(true); +} + +void UIScene_SettingsMenu::OnConfirmYes() +{ + app.SetDefaultOptions(ProfileManager.GetDashboardProfileSettings(m_iPad), m_iPad); + app.CheckGameSettingsChanged(true, m_iPad); + navigateBack(); +} + +void UIScene_SettingsMenu::OnConfirmNo() +{ + ShowConfirmDialog(false); } diff --git a/Minecraft.Client/Common/UI/UIScene_SettingsMenu.h b/Minecraft.Client/Common/UI/UIScene_SettingsMenu.h index 7f5fe169..f63ebb32 100644 --- a/Minecraft.Client/Common/UI/UIScene_SettingsMenu.h +++ b/Minecraft.Client/Common/UI/UIScene_SettingsMenu.h @@ -1,47 +1,63 @@ #pragma once +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + #include "UIScene.h" +#include +#include -#define BUTTON_ALL_OPTIONS 0 -#define BUTTON_ALL_AUDIO 1 -#define BUTTON_ALL_CONTROL 2 -#define BUTTON_ALL_GRAPHICS 4 -#define BUTTON_ALL_UI 5 -#define BUTTON_ALL_RESETTODEFAULTS 6 -#define BUTTONS_ALL_MAX BUTTON_ALL_RESETTODEFAULTS + 1 +// NOTE: pop_macro at end of file -class UIScene_SettingsMenu : public UIScene +class UIScene_SettingsMenu : public UIScene, public Rml::EventListener { -private: - UIControl_Button m_buttons[BUTTONS_ALL_MAX]; - UI_BEGIN_MAP_ELEMENTS_AND_NAMES(UIScene) - UI_MAP_ELEMENT( m_buttons[BUTTON_ALL_OPTIONS], "Button1") - UI_MAP_ELEMENT( m_buttons[BUTTON_ALL_AUDIO], "Button2") - UI_MAP_ELEMENT( m_buttons[BUTTON_ALL_CONTROL], "Button3") - UI_MAP_ELEMENT( m_buttons[BUTTON_ALL_GRAPHICS], "Button4") - UI_MAP_ELEMENT( m_buttons[BUTTON_ALL_UI], "Button5") - UI_MAP_ELEMENT( m_buttons[BUTTON_ALL_RESETTODEFAULTS], "Button6") - UI_END_MAP_ELEMENTS_AND_NAMES() public: UIScene_SettingsMenu(int iPad, void *initData, UILayer *parentLayer); virtual ~UIScene_SettingsMenu(); - virtual EUIScene getSceneType() { return eUIScene_SettingsMenu;} - - virtual void updateTooltips(); - virtual void updateComponents(); - virtual void handleReload(); + virtual EUIScene getSceneType() { return eUIScene_SettingsMenu; } -protected: - // TODO: This should be pure virtual in this class - virtual wstring getMoviePath(); + virtual void tick(); + virtual void render(S32 width, S32 height, C4JRender::eViewportType viewport); -public: - // INPUT virtual void handleInput(int iPad, int key, bool repeat, bool pressed, bool released, bool &handled); -protected: - void handlePress(F64 controlId, F64 childId); + // Override focus/movie lifecycle for RmlUi (no Iggy movie) + virtual void gainFocus() override; + virtual void reloadMovie(bool force) override; - static int ResetDefaultsDialogReturned(void *pParam,int iPad,C4JStorage::EMessageResult result); -}; \ No newline at end of file + // Rml::EventListener + void ProcessEvent(Rml::Event& event) override; + +protected: + virtual wstring getMoviePath() { return L""; } + + virtual void handleLoseFocus() override; + virtual void updateTooltips() override; + virtual void updateComponents() override; + + void navigateBack(); + + void OnOptions(); + void OnAudio(); + void OnControls(); + void OnGraphics(); + void OnUI(); + void OnResetToDefaults(); + void OnConfirmYes(); + void OnConfirmNo(); + + void ShowConfirmDialog(bool show); + +private: + Rml::ElementDocument* m_document; + bool m_bConfirmDialog; +}; + +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") diff --git a/Minecraft.Client/Common/UI/UIStructs.h b/Minecraft.Client/Common/UI/UIStructs.h index aad77d3f..78869b1a 100644 --- a/Minecraft.Client/Common/UI/UIStructs.h +++ b/Minecraft.Client/Common/UI/UIStructs.h @@ -367,19 +367,19 @@ typedef struct _LaunchMoreOptionsMenuInitData _LaunchMoreOptionsMenuInitData() { - memset(this,0,sizeof(_LaunchMoreOptionsMenuInitData)); bOnlineGame = true; + bInviteOnly = false; bAllowFriendsOfFriends = true; + bGenerateOptions = false; + bStructures = false; + bFlatWorld = false; + bBonusChest = false; bPVP = true; + bTrust = false; bFireSpreads = true; bTNT = true; - iPad = -1; - worldSize = 3; - seed = L""; - bDisableSaving = false; - newWorldSize = e_worldSize_Unknown; - newWorldSizeOverwriteEdges = false; - + bHostPrivileges = false; + bResetNether = false; bMobGriefing = true; bKeepInventory = false; bDoMobSpawning = true; @@ -387,6 +387,15 @@ typedef struct _LaunchMoreOptionsMenuInitData bDoTileDrops = true; bNaturalRegeneration = true; bDoDaylightCycle = true; + bOnlineSettingChangedBySystem = false; + iPad = -1; + dwTexturePack = 0; + seed = L""; + worldSize = 3; + bDisableSaving = false; + currentWorldSize = e_worldSize_Unknown; + newWorldSize = e_worldSize_Unknown; + newWorldSizeOverwriteEdges = false; } } LaunchMoreOptionsMenuInitData; diff --git a/Minecraft.Client/GameRenderer.cpp b/Minecraft.Client/GameRenderer.cpp index 41894eb6..d853eae9 100644 --- a/Minecraft.Client/GameRenderer.cpp +++ b/Minecraft.Client/GameRenderer.cpp @@ -240,7 +240,7 @@ void GameRenderer::tick(bool first) // 4J - add bFirst PIXEndNamedEvent(); darkenWorldAmountO = darkenWorldAmount; - int idx = BossMobGuiInfo::getIndexFromDimension((int)mc->level->dimension); + int idx = BossMobGuiInfo::getIndexFromDimension((int)(intptr_t)mc->level->dimension); if (BossMobGuiInfo::darkenWorld[idx]) { darkenWorldAmount += 1.0f / (static_cast(SharedConstants::TICKS_PER_SECOND) * 1); diff --git a/Minecraft.Client/MinecraftServer.h b/Minecraft.Client/MinecraftServer.h index cf9c7bec..7674cecd 100644 --- a/Minecraft.Client/MinecraftServer.h +++ b/Minecraft.Client/MinecraftServer.h @@ -265,6 +265,8 @@ private: private: // 4J Added bool m_saveOnExit; + bool m_deleteWorldOnExit; // 4J Added - for hardcore mode world deletion + wstring m_saveFolderName; // 4J Added - stored for hardcore world deletion bool m_suspending; public: @@ -278,6 +280,9 @@ public: void chunkPacketManagement_PostTick(); void setSaveOnExit(bool save) { m_saveOnExit = save; s_bSaveOnExitAnswered = true; } + void setDeleteWorldOnExit(bool del) { m_deleteWorldOnExit = del; } + bool getDeleteWorldOnExit() const { return m_deleteWorldOnExit; } + const wstring& getSaveFolderName() const { return m_saveFolderName; } void Suspend(); bool IsSuspending(); diff --git a/Minecraft.Client/RmlManager.cpp b/Minecraft.Client/RmlManager.cpp new file mode 100644 index 00000000..4e5aa17a --- /dev/null +++ b/Minecraft.Client/RmlManager.cpp @@ -0,0 +1,156 @@ +#include "stdafx.h" +#include "RmlManager.h" + +extern ID3D11Device* g_pd3dDevice; +extern ID3D11DeviceContext* g_pImmediateContext; +extern ID3D11RenderTargetView* g_pRenderTargetView; + +RmlManager& RmlManager::Get() +{ + static RmlManager instance; + return instance; +} + +RmlManager::RmlManager() + : m_initialised(false) + , m_context(nullptr) +{ +} + +RmlManager::~RmlManager() +{ + Shutdown(); +} + +bool RmlManager::Init(int viewport_width, int viewport_height) +{ + if (m_initialised) + return true; + + m_file_interface = std::make_unique(); + m_system_interface = std::make_unique(); + m_renderer = std::make_unique(g_pd3dDevice, g_pImmediateContext); + + if (!m_renderer->IsInitialised()) + { + app.DebugPrintf("[RmlManager] Renderer failed to initialise\n"); + return false; + } + + // Set asset path for RmlUi files (relative to exe directory) + char exePath[MAX_PATH]; + GetModuleFileNameA(nullptr, exePath, MAX_PATH); + char* lastSlash = strrchr(exePath, '\\'); + if (lastSlash) + *(lastSlash + 1) = '\0'; + Rml::String assetPath = Rml::String(exePath) + "Assets/RmlUiAssets"; + m_file_interface->SetAssetPath(assetPath); + + Rml::SetRenderInterface(m_renderer.get()); + Rml::SetSystemInterface(m_system_interface.get()); + Rml::SetFileInterface(m_file_interface.get()); + + if (!Rml::Initialise()) + { + app.DebugPrintf("[RmlManager] RmlUi Core failed to initialise\n"); + return false; + } + + // Load default fonts + Rml::LoadFontFace("C:/Windows/Fonts/segoeui.ttf", false); + Rml::LoadFontFace("C:/Windows/Fonts/segoeuib.ttf", false); + Rml::LoadFontFace("C:/Windows/Fonts/segoeuii.ttf", false); + Rml::LoadFontFace("C:/Windows/Fonts/segoeuiz.ttf", false); + + // Minecraft's UI fonts (from tryashtar/minecraft-ttf) + if (!Rml::LoadFontFace(assetPath + "/fonts/Mojangles.ttf", "Minecraft Default", Rml::Style::FontStyle::Normal, Rml::Style::FontWeight::Normal, false)) + app.DebugPrintf("[RmlManager] FAILED loading Mojangles.ttf\n"); + if (!Rml::LoadFontFace(assetPath + "/fonts/MinecraftTen.ttf", "Minecraft Default", Rml::Style::FontStyle::Normal, Rml::Style::FontWeight::Bold, true)) + app.DebugPrintf("[RmlManager] FAILED loading MinecraftTen.ttf\n"); + + m_context = Rml::CreateContext("main", Rml::Vector2i(viewport_width, viewport_height)); + if (!m_context) + { + app.DebugPrintf("[RmlManager] Failed to create RmlUi context\n"); + Rml::Shutdown(); + return false; + } + app.DebugPrintf("[RmlManager] Context created (%dx%d)\n", viewport_width, viewport_height); + + m_renderer->SetViewport(viewport_width, viewport_height); + m_initialised = true; + app.DebugPrintf("[RmlManager] Initialised (%dx%d)\n", viewport_width, viewport_height); + return true; +} + +void RmlManager::Shutdown() +{ + if (!m_initialised) + return; + + if (m_context) + { + m_context->UnloadAllDocuments(); + Rml::RemoveContext(m_context->GetName()); + m_context = nullptr; + } + + Rml::Shutdown(); + + m_renderer.reset(); + m_system_interface.reset(); + m_file_interface.reset(); + + m_initialised = false; + app.DebugPrintf("[RmlManager] Shutdown\n"); +} + +void RmlManager::Update() +{ + if (!m_initialised || !m_context) + return; + m_context->Update(); +} + +void RmlManager::Render() +{ + if (!m_initialised || !m_context) + return; + + // Ensure we render to the backbuffer (gamma post-process may have restored a + // different RTV). Use the singleton backbuffer RTV that the rest of the engine + // binds at frame start. + if (g_pRenderTargetView) + { + ID3D11DepthStencilView* nullDSV = nullptr; + g_pImmediateContext->OMSetRenderTargets(1, &g_pRenderTargetView, nullDSV); + } + + // Clear any SRV still bound from gamma post-process so it doesn't + // interfere with RmlUi's own texture binding. + ID3D11ShaderResourceView* nullSRV[1] = { nullptr }; + g_pImmediateContext->PSSetShaderResources(0, 1, nullSRV); + + m_renderer->BeginFrame(); + + int docCount = 0; + for (int i = 0; i < m_context->GetNumDocuments(); i++) + { + auto* doc = m_context->GetDocument(i); + if (doc && doc->IsVisible()) + docCount++; + } + if (docCount == 0) + app.DebugPrintf("[RmlManager] WARNING: No visible documents in context\n"); + + m_context->Render(); + m_renderer->EndFrame(); +} + +void RmlManager::OnResize(int width, int height) +{ + if (!m_initialised || !m_context) + return; + m_context->SetDimensions(Rml::Vector2i(width, height)); + m_renderer->SetViewport(width, height); +} diff --git a/Minecraft.Client/RmlManager.h b/Minecraft.Client/RmlManager.h new file mode 100644 index 00000000..e9af36da --- /dev/null +++ b/Minecraft.Client/RmlManager.h @@ -0,0 +1,52 @@ +#pragma once + +#pragma push_macro("byte") +#pragma push_macro("GetNextSibling") +#pragma push_macro("GetFirstChild") +#undef byte +#undef GetNextSibling +#undef GetFirstChild + +#include +#include +#include +#include "RmlUi_Renderer_D3D11.h" +#include "RmlUi_SystemInterface_Win64.h" +#include "RmlUi_FileInterface.h" + +// NOTE: pop_macro at end of file + +class RmlManager +{ +public: + static RmlManager& Get(); + + bool Init(int viewport_width, int viewport_height); + void Shutdown(); + + void Update(); + void Render(); + + Rml::Context* GetContext() const { return m_context; } + RenderInterface_D3D11* GetRenderInterface() const { return m_renderer.get(); } + + void OnResize(int width, int height); + +private: + RmlManager(); + ~RmlManager(); + RmlManager(const RmlManager&) = delete; + RmlManager& operator=(const RmlManager&) = delete; + + bool m_initialised; + + std::unique_ptr m_renderer; + std::unique_ptr m_system_interface; + std::unique_ptr m_file_interface; + + Rml::Context* m_context; +}; + +#pragma pop_macro("GetFirstChild") +#pragma pop_macro("GetNextSibling") +#pragma pop_macro("byte") diff --git a/Minecraft.Client/Tesselator.h b/Minecraft.Client/Tesselator.h index 2edc4a24..d0be867a 100644 --- a/Minecraft.Client/Tesselator.h +++ b/Minecraft.Client/Tesselator.h @@ -1,5 +1,7 @@ #pragma once +#include + class ChunkRebuildData; class Tesselator { diff --git a/Minecraft.Client/TileRenderer.cpp b/Minecraft.Client/TileRenderer.cpp index a5ee810f..4356437d 100644 --- a/Minecraft.Client/TileRenderer.cpp +++ b/Minecraft.Client/TileRenderer.cpp @@ -156,7 +156,7 @@ TileRenderer::TileRenderer( LevelSource* level, int xMin, int yMin, int zMin, un this->zMin2 = zMin-2; this->tileIds = tileIds; cache = s_tlsCache; - std::memset(cache, 0, TILE_RENDERER_CACHE_SIZE * sizeof(unsigned int)); + memset(cache, 0, TILE_RENDERER_CACHE_SIZE * sizeof(unsigned int)); } TileRenderer::~TileRenderer() = default; diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXColors.h b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXColors.h new file mode 100644 index 00000000..18614505 --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXColors.h @@ -0,0 +1,311 @@ +//------------------------------------------------------------------------------------- +// DirectXColors.h -- C++ Color Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +#include "DirectXMath.h" + +namespace DirectX +{ + + namespace Colors + { + // Standard colors (Red/Green/Blue/Alpha) in sRGB colorspace + XMGLOBALCONST XMVECTORF32 AliceBlue = { { { 0.941176534f, 0.972549081f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 AntiqueWhite = { { { 0.980392218f, 0.921568692f, 0.843137324f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Aqua = { { { 0.f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Aquamarine = { { { 0.498039246f, 1.f, 0.831372619f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Azure = { { { 0.941176534f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Beige = { { { 0.960784376f, 0.960784376f, 0.862745166f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Bisque = { { { 1.f, 0.894117713f, 0.768627524f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Black = { { { 0.f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 BlanchedAlmond = { { { 1.f, 0.921568692f, 0.803921640f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Blue = { { { 0.f, 0.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 BlueViolet = { { { 0.541176498f, 0.168627456f, 0.886274576f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Brown = { { { 0.647058845f, 0.164705887f, 0.164705887f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 BurlyWood = { { { 0.870588303f, 0.721568644f, 0.529411793f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 CadetBlue = { { { 0.372549027f, 0.619607866f, 0.627451003f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Chartreuse = { { { 0.498039246f, 1.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Chocolate = { { { 0.823529482f, 0.411764741f, 0.117647067f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Coral = { { { 1.f, 0.498039246f, 0.313725501f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 CornflowerBlue = { { { 0.392156899f, 0.584313750f, 0.929411829f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Cornsilk = { { { 1.f, 0.972549081f, 0.862745166f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Crimson = { { { 0.862745166f, 0.078431375f, 0.235294133f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Cyan = { { { 0.f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkBlue = { { { 0.f, 0.f, 0.545098066f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkCyan = { { { 0.f, 0.545098066f, 0.545098066f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkGoldenrod = { { { 0.721568644f, 0.525490224f, 0.043137256f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkGray = { { { 0.662745118f, 0.662745118f, 0.662745118f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkGreen = { { { 0.f, 0.392156899f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkKhaki = { { { 0.741176486f, 0.717647076f, 0.419607878f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkMagenta = { { { 0.545098066f, 0.f, 0.545098066f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkOliveGreen = { { { 0.333333343f, 0.419607878f, 0.184313729f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkOrange = { { { 1.f, 0.549019635f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkOrchid = { { { 0.600000024f, 0.196078449f, 0.800000072f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkRed = { { { 0.545098066f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSalmon = { { { 0.913725555f, 0.588235319f, 0.478431404f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSeaGreen = { { { 0.560784340f, 0.737254918f, 0.545098066f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSlateBlue = { { { 0.282352954f, 0.239215702f, 0.545098066f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSlateGray = { { { 0.184313729f, 0.309803933f, 0.309803933f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkTurquoise = { { { 0.f, 0.807843208f, 0.819607913f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkViolet = { { { 0.580392182f, 0.f, 0.827451050f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DeepPink = { { { 1.f, 0.078431375f, 0.576470613f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DeepSkyBlue = { { { 0.f, 0.749019623f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DimGray = { { { 0.411764741f, 0.411764741f, 0.411764741f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DodgerBlue = { { { 0.117647067f, 0.564705908f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Firebrick = { { { 0.698039234f, 0.133333340f, 0.133333340f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 FloralWhite = { { { 1.f, 0.980392218f, 0.941176534f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 ForestGreen = { { { 0.133333340f, 0.545098066f, 0.133333340f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Fuchsia = { { { 1.f, 0.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Gainsboro = { { { 0.862745166f, 0.862745166f, 0.862745166f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 GhostWhite = { { { 0.972549081f, 0.972549081f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Gold = { { { 1.f, 0.843137324f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Goldenrod = { { { 0.854902029f, 0.647058845f, 0.125490203f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Gray = { { { 0.501960814f, 0.501960814f, 0.501960814f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Green = { { { 0.f, 0.501960814f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 GreenYellow = { { { 0.678431392f, 1.f, 0.184313729f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Honeydew = { { { 0.941176534f, 1.f, 0.941176534f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 HotPink = { { { 1.f, 0.411764741f, 0.705882370f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 IndianRed = { { { 0.803921640f, 0.360784322f, 0.360784322f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Indigo = { { { 0.294117659f, 0.f, 0.509803951f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Ivory = { { { 1.f, 1.f, 0.941176534f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Khaki = { { { 0.941176534f, 0.901960850f, 0.549019635f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Lavender = { { { 0.901960850f, 0.901960850f, 0.980392218f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LavenderBlush = { { { 1.f, 0.941176534f, 0.960784376f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LawnGreen = { { { 0.486274540f, 0.988235354f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LemonChiffon = { { { 1.f, 0.980392218f, 0.803921640f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightBlue = { { { 0.678431392f, 0.847058892f, 0.901960850f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightCoral = { { { 0.941176534f, 0.501960814f, 0.501960814f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightCyan = { { { 0.878431439f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightGoldenrodYellow = { { { 0.980392218f, 0.980392218f, 0.823529482f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightGray = { { { 0.827451050f, 0.827451050f, 0.827451050f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightGreen = { { { 0.564705908f, 0.933333397f, 0.564705908f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightPink = { { { 1.f, 0.713725507f, 0.756862819f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSalmon = { { { 1.f, 0.627451003f, 0.478431404f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSeaGreen = { { { 0.125490203f, 0.698039234f, 0.666666687f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSkyBlue = { { { 0.529411793f, 0.807843208f, 0.980392218f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSlateGray = { { { 0.466666698f, 0.533333361f, 0.600000024f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSteelBlue = { { { 0.690196097f, 0.768627524f, 0.870588303f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightYellow = { { { 1.f, 1.f, 0.878431439f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Lime = { { { 0.f, 1.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LimeGreen = { { { 0.196078449f, 0.803921640f, 0.196078449f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Linen = { { { 0.980392218f, 0.941176534f, 0.901960850f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Magenta = { { { 1.f, 0.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Maroon = { { { 0.501960814f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumAquamarine = { { { 0.400000036f, 0.803921640f, 0.666666687f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumBlue = { { { 0.f, 0.f, 0.803921640f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumOrchid = { { { 0.729411781f, 0.333333343f, 0.827451050f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumPurple = { { { 0.576470613f, 0.439215720f, 0.858823597f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumSeaGreen = { { { 0.235294133f, 0.701960802f, 0.443137288f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumSlateBlue = { { { 0.482352972f, 0.407843173f, 0.933333397f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumSpringGreen = { { { 0.f, 0.980392218f, 0.603921592f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumTurquoise = { { { 0.282352954f, 0.819607913f, 0.800000072f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumVioletRed = { { { 0.780392230f, 0.082352944f, 0.521568656f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MidnightBlue = { { { 0.098039225f, 0.098039225f, 0.439215720f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MintCream = { { { 0.960784376f, 1.f, 0.980392218f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MistyRose = { { { 1.f, 0.894117713f, 0.882353008f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Moccasin = { { { 1.f, 0.894117713f, 0.709803939f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 NavajoWhite = { { { 1.f, 0.870588303f, 0.678431392f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Navy = { { { 0.f, 0.f, 0.501960814f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 OldLace = { { { 0.992156923f, 0.960784376f, 0.901960850f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Olive = { { { 0.501960814f, 0.501960814f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 OliveDrab = { { { 0.419607878f, 0.556862772f, 0.137254909f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Orange = { { { 1.f, 0.647058845f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 OrangeRed = { { { 1.f, 0.270588249f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Orchid = { { { 0.854902029f, 0.439215720f, 0.839215755f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleGoldenrod = { { { 0.933333397f, 0.909803987f, 0.666666687f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleGreen = { { { 0.596078455f, 0.984313786f, 0.596078455f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleTurquoise = { { { 0.686274529f, 0.933333397f, 0.933333397f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleVioletRed = { { { 0.858823597f, 0.439215720f, 0.576470613f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PapayaWhip = { { { 1.f, 0.937254965f, 0.835294187f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PeachPuff = { { { 1.f, 0.854902029f, 0.725490212f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Peru = { { { 0.803921640f, 0.521568656f, 0.247058839f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Pink = { { { 1.f, 0.752941251f, 0.796078503f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Plum = { { { 0.866666734f, 0.627451003f, 0.866666734f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PowderBlue = { { { 0.690196097f, 0.878431439f, 0.901960850f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Purple = { { { 0.501960814f, 0.f, 0.501960814f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Red = { { { 1.f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 RosyBrown = { { { 0.737254918f, 0.560784340f, 0.560784340f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 RoyalBlue = { { { 0.254901975f, 0.411764741f, 0.882353008f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SaddleBrown = { { { 0.545098066f, 0.270588249f, 0.074509807f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Salmon = { { { 0.980392218f, 0.501960814f, 0.447058856f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SandyBrown = { { { 0.956862807f, 0.643137276f, 0.376470625f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SeaGreen = { { { 0.180392161f, 0.545098066f, 0.341176480f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SeaShell = { { { 1.f, 0.960784376f, 0.933333397f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Sienna = { { { 0.627451003f, 0.321568638f, 0.176470593f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Silver = { { { 0.752941251f, 0.752941251f, 0.752941251f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SkyBlue = { { { 0.529411793f, 0.807843208f, 0.921568692f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SlateBlue = { { { 0.415686309f, 0.352941185f, 0.803921640f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SlateGray = { { { 0.439215720f, 0.501960814f, 0.564705908f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Snow = { { { 1.f, 0.980392218f, 0.980392218f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SpringGreen = { { { 0.f, 1.f, 0.498039246f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SteelBlue = { { { 0.274509817f, 0.509803951f, 0.705882370f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Tan = { { { 0.823529482f, 0.705882370f, 0.549019635f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Teal = { { { 0.f, 0.501960814f, 0.501960814f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Thistle = { { { 0.847058892f, 0.749019623f, 0.847058892f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Tomato = { { { 1.f, 0.388235331f, 0.278431386f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Transparent = { { { 0.f, 0.f, 0.f, 0.f } } }; + XMGLOBALCONST XMVECTORF32 Turquoise = { { { 0.250980407f, 0.878431439f, 0.815686345f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Violet = { { { 0.933333397f, 0.509803951f, 0.933333397f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Wheat = { { { 0.960784376f, 0.870588303f, 0.701960802f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 White = { { { 1.f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 WhiteSmoke = { { { 0.960784376f, 0.960784376f, 0.960784376f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Yellow = { { { 1.f, 1.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 YellowGreen = { { { 0.603921592f, 0.803921640f, 0.196078449f, 1.f } } }; + + } // namespace Colors + + namespace ColorsLinear + { + // Standard colors (Red/Green/Blue/Alpha) in linear colorspace + XMGLOBALCONST XMVECTORF32 AliceBlue = { { { 0.871367335f, 0.938685894f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 AntiqueWhite = { { { 0.955973506f, 0.830770075f, 0.679542601f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Aqua = { { { 0.f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Aquamarine = { { { 0.212230787f, 1.f, 0.658374965f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Azure = { { { 0.871367335f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Beige = { { { 0.913098991f, 0.913098991f, 0.715693772f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Bisque = { { { 1.f, 0.775822461f, 0.552011609f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Black = { { { 0.f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 BlanchedAlmond = { { { 1.f, 0.830770075f, 0.610495746f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Blue = { { { 0.f, 0.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 BlueViolet = { { { 0.254152179f, 0.024157630f, 0.760524750f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Brown = { { { 0.376262218f, 0.023153365f, 0.023153365f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 BurlyWood = { { { 0.730461001f, 0.479320228f, 0.242281199f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 CadetBlue = { { { 0.114435382f, 0.341914445f, 0.351532698f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Chartreuse = { { { 0.212230787f, 1.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Chocolate = { { { 0.644479871f, 0.141263321f, 0.012983031f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Coral = { { { 1.f, 0.212230787f, 0.080219828f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 CornflowerBlue = { { { 0.127437726f, 0.300543845f, 0.846873462f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Cornsilk = { { { 1.f, 0.938685894f, 0.715693772f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Crimson = { { { 0.715693772f, 0.006995410f, 0.045186214f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Cyan = { { { 0.f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkBlue = { { { 0.f, 0.f, 0.258182913f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkCyan = { { { 0.f, 0.258182913f, 0.258182913f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkGoldenrod = { { { 0.479320228f, 0.238397658f, 0.003346536f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkGray = { { { 0.396755308f, 0.396755308f, 0.396755308f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkGreen = { { { 0.f, 0.127437726f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkKhaki = { { { 0.508881450f, 0.473531544f, 0.147027299f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkMagenta = { { { 0.258182913f, 0.f, 0.258182913f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkOliveGreen = { { { 0.090841733f, 0.147027299f, 0.028426038f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkOrange = { { { 1.f, 0.262250721f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkOrchid = { { { 0.318546832f, 0.031896040f, 0.603827536f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkRed = { { { 0.258182913f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSalmon = { { { 0.814846814f, 0.304987371f, 0.194617867f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSeaGreen = { { { 0.274677366f, 0.502886593f, 0.258182913f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSlateBlue = { { { 0.064803280f, 0.046665095f, 0.258182913f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkSlateGray = { { { 0.028426038f, 0.078187428f, 0.078187428f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkTurquoise = { { { 0.f, 0.617206752f, 0.637597024f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DarkViolet = { { { 0.296138316f, 0.f, 0.651405811f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DeepPink = { { { 1.f, 0.006995410f, 0.291770697f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DeepSkyBlue = { { { 0.f, 0.520995677f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DimGray = { { { 0.141263321f, 0.141263321f, 0.141263321f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 DodgerBlue = { { { 0.012983031f, 0.278894335f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Firebrick = { { { 0.445201248f, 0.015996292f, 0.015996292f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 FloralWhite = { { { 1.f, 0.955973506f, 0.871367335f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 ForestGreen = { { { 0.015996292f, 0.258182913f, 0.015996292f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Fuchsia = { { { 1.f, 0.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Gainsboro = { { { 0.715693772f, 0.715693772f, 0.715693772f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 GhostWhite = { { { 0.938685894f, 0.938685894f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Gold = { { { 1.f, 0.679542601f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Goldenrod = { { { 0.701102138f, 0.376262218f, 0.014443844f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Gray = { { { 0.215860531f, 0.215860531f, 0.215860531f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Green = { { { 0.f, 0.215860531f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 GreenYellow = { { { 0.417885154f, 1.f, 0.028426038f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Honeydew = { { { 0.871367335f, 1.f, 0.871367335f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 HotPink = { { { 1.f, 0.141263321f, 0.456411064f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 IndianRed = { { { 0.610495746f, 0.107023112f, 0.107023112f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Indigo = { { { 0.070360109f, 0.f, 0.223227978f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Ivory = { { { 1.f, 1.f, 0.871367335f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Khaki = { { { 0.871367335f, 0.791298151f, 0.262250721f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Lavender = { { { 0.791298151f, 0.791298151f, 0.955973506f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LavenderBlush = { { { 1.f, 0.871367335f, 0.913098991f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LawnGreen = { { { 0.201556295f, 0.973445475f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LemonChiffon = { { { 1.f, 0.955973506f, 0.610495746f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightBlue = { { { 0.417885154f, 0.686685443f, 0.791298151f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightCoral = { { { 0.871367335f, 0.215860531f, 0.215860531f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightCyan = { { { 0.745404482f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightGoldenrodYellow = { { { 0.955973506f, 0.955973506f, 0.644479871f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightGray = { { { 0.651405811f, 0.651405811f, 0.651405811f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightGreen = { { { 0.278894335f, 0.854992807f, 0.278894335f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightPink = { { { 1.f, 0.467783839f, 0.533276618f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSalmon = { { { 1.f, 0.351532698f, 0.194617867f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSeaGreen = { { { 0.014443844f, 0.445201248f, 0.401977867f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSkyBlue = { { { 0.242281199f, 0.617206752f, 0.955973506f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSlateGray = { { { 0.184475034f, 0.246201396f, 0.318546832f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightSteelBlue = { { { 0.434153706f, 0.552011609f, 0.730461001f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LightYellow = { { { 1.f, 1.f, 0.745404482f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Lime = { { { 0.f, 1.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 LimeGreen = { { { 0.031896040f, 0.610495746f, 0.031896040f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Linen = { { { 0.955973506f, 0.871367335f, 0.791298151f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Magenta = { { { 1.f, 0.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Maroon = { { { 0.215860531f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumAquamarine = { { { 0.132868364f, 0.610495746f, 0.401977867f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumBlue = { { { 0.f, 0.f, 0.610495746f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumOrchid = { { { 0.491020888f, 0.090841733f, 0.651405811f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumPurple = { { { 0.291770697f, 0.162029430f, 0.708376050f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumSeaGreen = { { { 0.045186214f, 0.450785846f, 0.165132239f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumSlateBlue = { { { 0.198069349f, 0.138431653f, 0.854992807f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumSpringGreen = { { { 0.f, 0.955973506f, 0.323143244f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumTurquoise = { { { 0.064803280f, 0.637597024f, 0.603827536f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MediumVioletRed = { { { 0.571125031f, 0.007499032f, 0.234550655f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MidnightBlue = { { { 0.009721218f, 0.009721218f, 0.162029430f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MintCream = { { { 0.913098991f, 1.f, 0.955973506f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 MistyRose = { { { 1.f, 0.775822461f, 0.752942443f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Moccasin = { { { 1.f, 0.775822461f, 0.462077051f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 NavajoWhite = { { { 1.f, 0.730461001f, 0.417885154f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Navy = { { { 0.f, 0.f, 0.215860531f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 OldLace = { { { 0.982250869f, 0.913098991f, 0.791298151f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Olive = { { { 0.215860531f, 0.215860531f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 OliveDrab = { { { 0.147027299f, 0.270497859f, 0.016807375f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Orange = { { { 1.f, 0.376262218f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 OrangeRed = { { { 1.f, 0.059511241f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Orchid = { { { 0.701102138f, 0.162029430f, 0.672443330f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleGoldenrod = { { { 0.854992807f, 0.806952477f, 0.401977867f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleGreen = { { { 0.313988745f, 0.964686573f, 0.313988745f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleTurquoise = { { { 0.428690553f, 0.854992807f, 0.854992807f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PaleVioletRed = { { { 0.708376050f, 0.162029430f, 0.291770697f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PapayaWhip = { { { 1.f, 0.863157392f, 0.665387452f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PeachPuff = { { { 1.f, 0.701102138f, 0.485149980f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Peru = { { { 0.610495746f, 0.234550655f, 0.049706575f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Pink = { { { 1.f, 0.527115345f, 0.597202003f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Plum = { { { 0.723055363f, 0.351532698f, 0.723055363f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 PowderBlue = { { { 0.434153706f, 0.745404482f, 0.791298151f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Purple = { { { 0.215860531f, 0.f, 0.215860531f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Red = { { { 1.f, 0.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 RosyBrown = { { { 0.502886593f, 0.274677366f, 0.274677366f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 RoyalBlue = { { { 0.052860655f, 0.141263321f, 0.752942443f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SaddleBrown = { { { 0.258182913f, 0.059511241f, 0.006512091f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Salmon = { { { 0.955973506f, 0.215860531f, 0.168269455f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SandyBrown = { { { 0.904661357f, 0.371237785f, 0.116970696f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SeaGreen = { { { 0.027320892f, 0.258182913f, 0.095307484f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SeaShell = { { { 1.f, 0.913098991f, 0.854992807f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Sienna = { { { 0.351532698f, 0.084376216f, 0.026241222f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Silver = { { { 0.527115345f, 0.527115345f, 0.527115345f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SkyBlue = { { { 0.242281199f, 0.617206752f, 0.830770075f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SlateBlue = { { { 0.144128501f, 0.102241747f, 0.610495746f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SlateGray = { { { 0.162029430f, 0.215860531f, 0.278894335f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Snow = { { { 1.f, 0.955973506f, 0.955973506f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SpringGreen = { { { 0.f, 1.f, 0.212230787f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 SteelBlue = { { { 0.061246071f, 0.223227978f, 0.456411064f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Tan = { { { 0.644479871f, 0.456411064f, 0.262250721f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Teal = { { { 0.f, 0.215860531f, 0.215860531f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Thistle = { { { 0.686685443f, 0.520995677f, 0.686685443f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Tomato = { { { 1.f, 0.124771863f, 0.063010029f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Transparent = { { { 0.f, 0.f, 0.f, 0.f } } }; + XMGLOBALCONST XMVECTORF32 Turquoise = { { { 0.051269468f, 0.745404482f, 0.630757332f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Violet = { { { 0.854992807f, 0.223227978f, 0.854992807f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Wheat = { { { 0.913098991f, 0.730461001f, 0.450785846f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 White = { { { 1.f, 1.f, 1.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 WhiteSmoke = { { { 0.913098991f, 0.913098991f, 0.913098991f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 Yellow = { { { 1.f, 1.f, 0.f, 1.f } } }; + XMGLOBALCONST XMVECTORF32 YellowGreen = { { { 0.323143244f, 0.610495746f, 0.031896040f, 1.f } } }; + + } // namespace ColorsLinear + +} // namespace DirectX diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMath.h b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMath.h new file mode 100644 index 00000000..ae3cce1d --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMath.h @@ -0,0 +1,2385 @@ +//------------------------------------------------------------------------------------- +// DirectXMath.h -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +#ifndef __cplusplus +#error DirectX Math requires C++ +#endif + +#define DIRECTX_MATH_VERSION 321 + +#if defined(_MSC_VER) && (_MSC_VER < 1910) +#error DirectX Math requires Visual C++ 2017 or later. +#endif + +#if defined(_MSC_VER) && !defined(_M_ARM) && !defined(_M_ARM64) && !defined(_M_HYBRID_X86_ARM64) && !defined(_M_ARM64EC) && (!_MANAGED) && (!_M_CEE) && (!defined(_M_IX86_FP) || (_M_IX86_FP > 1)) && !defined(_XM_NO_INTRINSICS_) && !defined(_XM_VECTORCALL_) +#define _XM_VECTORCALL_ 1 +#endif + +#if _XM_VECTORCALL_ +#define XM_CALLCONV __vectorcall +#elif defined(__GNUC__) +#define XM_CALLCONV +#else +#define XM_CALLCONV __fastcall +#endif + +#ifndef XM_DEPRECATED +#if (__cplusplus >= 201402L) +#define XM_DEPRECATED [[deprecated]] +#elif defined(__GNUC__) +#define XM_DEPRECATED __attribute__ ((deprecated)) +#else +#define XM_DEPRECATED __declspec(deprecated("This is deprecated and will be removed in a future version.")) +#endif +#endif + +#if !defined(_XM_AVX2_INTRINSICS_) && defined(__AVX2__) && !defined(_XM_NO_INTRINSICS_) +#define _XM_AVX2_INTRINSICS_ +#endif + +#if !defined(_XM_FMA3_INTRINSICS_) && defined(_XM_AVX2_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) +#define _XM_FMA3_INTRINSICS_ +#endif + +#if !defined(_XM_F16C_INTRINSICS_) && defined(_XM_AVX2_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) +#define _XM_F16C_INTRINSICS_ +#endif + +#if !defined(_XM_F16C_INTRINSICS_) && defined(__F16C__) && !defined(_XM_NO_INTRINSICS_) +#define _XM_F16C_INTRINSICS_ +#endif + +#if defined(_XM_FMA3_INTRINSICS_) && !defined(_XM_AVX_INTRINSICS_) +#define _XM_AVX_INTRINSICS_ +#endif + +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_AVX_INTRINSICS_) +#define _XM_AVX_INTRINSICS_ +#endif + +#if !defined(_XM_AVX_INTRINSICS_) && defined(__AVX__) && !defined(_XM_NO_INTRINSICS_) +#define _XM_AVX_INTRINSICS_ +#endif + +#if defined(_XM_AVX_INTRINSICS_) && !defined(_XM_SSE4_INTRINSICS_) +#define _XM_SSE4_INTRINSICS_ +#endif + +#if defined(_XM_SSE4_INTRINSICS_) && !defined(_XM_SSE3_INTRINSICS_) +#define _XM_SSE3_INTRINSICS_ +#endif + +#if defined(_XM_SSE3_INTRINSICS_) && !defined(_XM_SSE_INTRINSICS_) +#define _XM_SSE_INTRINSICS_ +#endif + +#if !defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) +#if (defined(_M_IX86) || defined(_M_X64) || __i386__ || __x86_64__ || __powerpc64__) && !defined(_M_HYBRID_X86_ARM64) && !defined(_M_ARM64EC) +#define _XM_SSE_INTRINSICS_ +#elif defined(_M_ARM) || defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __arm__ || __aarch64__ +#define _XM_ARM_NEON_INTRINSICS_ +#elif !defined(_XM_NO_INTRINSICS_) +#error DirectX Math does not support this target +#endif +#endif // !_XM_ARM_NEON_INTRINSICS_ && !_XM_SSE_INTRINSICS_ && !_XM_NO_INTRINSICS_ + +#if defined(_XM_SSE_INTRINSICS_) && defined(_MSC_VER) && (_MSC_VER >= 1920) && !defined(__clang__) && !defined(_XM_SVML_INTRINSICS_) && !defined(_XM_DISABLE_INTEL_SVML_) +#define _XM_SVML_INTRINSICS_ +#endif + +#if !defined(_XM_NO_XMVECTOR_OVERLOADS_) && (defined(__clang__) || defined(__GNUC__)) && !defined(_XM_NO_INTRINSICS_) +#define _XM_NO_XMVECTOR_OVERLOADS_ +#endif + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable:4514 4820) +// C4514/4820: Off by default noise +#endif +#include +#include +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +#ifndef _XM_NO_INTRINSICS_ + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4987) +// C4987: Off by default noise +#endif +#if defined(_MSC_VER) || defined(__MINGW32__) +#include +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +#if (defined(__clang__) || defined(__GNUC__)) && (__x86_64__ || __i386__) && !defined(__MINGW32__) && !defined(_MSC_VER) +#include +#endif + +#ifdef _XM_SSE_INTRINSICS_ +#include +#include + +#ifdef _XM_SSE3_INTRINSICS_ +#include +#endif + +#ifdef _XM_SSE4_INTRINSICS_ +#include +#endif + +#ifdef _XM_AVX_INTRINSICS_ +#include +#endif + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && (defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC)) +#include +#else +#include +#endif +#endif +#endif // !_XM_NO_INTRINSICS_ + +#include "sal.h" +#include + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4005 4668) +// C4005/4668: Old header issue +#endif +#include +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +#if (__cplusplus >= 201703L) +#define XM_ALIGNED_DATA(x) alignas(x) +#define XM_ALIGNED_STRUCT(x) struct alignas(x) +#elif defined(__GNUC__) +#define XM_ALIGNED_DATA(x) __attribute__ ((aligned(x))) +#define XM_ALIGNED_STRUCT(x) struct __attribute__ ((aligned(x))) +#else +#define XM_ALIGNED_DATA(x) __declspec(align(x)) +#define XM_ALIGNED_STRUCT(x) __declspec(align(x)) struct +#endif + +#if (__cplusplus >= 202002L) +#include +#endif + +/**************************************************************************** + * + * Conditional intrinsics + * + ****************************************************************************/ + +#if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + +#if defined(_XM_NO_MOVNT_) +#define XM_STREAM_PS( p, a ) _mm_store_ps((p), (a)) +#define XM256_STREAM_PS( p, a ) _mm256_store_ps((p), (a)) +#define XM_SFENCE() +#else +#define XM_STREAM_PS( p, a ) _mm_stream_ps((p), (a)) +#define XM256_STREAM_PS( p, a ) _mm256_stream_ps((p), (a)) +#define XM_SFENCE() _mm_sfence() +#endif + +#if defined(_XM_FMA3_INTRINSICS_) +#define XM_FMADD_PS( a, b, c ) _mm_fmadd_ps((a), (b), (c)) +#define XM_FNMADD_PS( a, b, c ) _mm_fnmadd_ps((a), (b), (c)) +#else +#define XM_FMADD_PS( a, b, c ) _mm_add_ps(_mm_mul_ps((a), (b)), (c)) +#define XM_FNMADD_PS( a, b, c ) _mm_sub_ps((c), _mm_mul_ps((a), (b))) +#endif + +#if defined(_XM_AVX_INTRINSICS_) && defined(_XM_FAVOR_INTEL_) +#define XM_PERMUTE_PS( v, c ) _mm_permute_ps((v), c ) +#else +#define XM_PERMUTE_PS( v, c ) _mm_shuffle_ps((v), (v), c ) +#endif + +#if (defined(__GNUC__) && !defined(__clang__) && (__GNUC__ < 11)) || defined(__powerpc64__) +#define XM_LOADU_SI16( p ) _mm_cvtsi32_si128(*reinterpret_cast(p)) +#else +#define XM_LOADU_SI16( p ) _mm_loadu_si16(p) +#endif + +#endif // _XM_SSE_INTRINSICS_ && !_XM_NO_INTRINSICS_ + +#if defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + +#if defined(__clang__) || defined(__GNUC__) +#define XM_PREFETCH( a ) __builtin_prefetch(a) +#elif defined(_MSC_VER) +#define XM_PREFETCH( a ) __prefetch(a) +#else +#define XM_PREFETCH( a ) +#endif + +#endif // _XM_ARM_NEON_INTRINSICS_ && !_XM_NO_INTRINSICS_ + +namespace DirectX +{ + + /**************************************************************************** + * + * Constant definitions + * + ****************************************************************************/ + +#if defined(__XNAMATH_H__) && defined(XM_PI) +#undef XM_PI +#undef XM_2PI +#undef XM_1DIVPI +#undef XM_1DIV2PI +#undef XM_PIDIV2 +#undef XM_PIDIV4 +#undef XM_SELECT_0 +#undef XM_SELECT_1 +#undef XM_PERMUTE_0X +#undef XM_PERMUTE_0Y +#undef XM_PERMUTE_0Z +#undef XM_PERMUTE_0W +#undef XM_PERMUTE_1X +#undef XM_PERMUTE_1Y +#undef XM_PERMUTE_1Z +#undef XM_PERMUTE_1W +#undef XM_CRMASK_CR6 +#undef XM_CRMASK_CR6TRUE +#undef XM_CRMASK_CR6FALSE +#undef XM_CRMASK_CR6BOUNDS +#undef XM_CACHE_LINE_SIZE +#endif + + constexpr float XM_PI = 3.141592654f; + constexpr float XM_2PI = 6.283185307f; + constexpr float XM_1DIVPI = 0.318309886f; + constexpr float XM_1DIV2PI = 0.159154943f; + constexpr float XM_PIDIV2 = 1.570796327f; + constexpr float XM_PIDIV4 = 0.785398163f; + + constexpr uint32_t XM_SELECT_0 = 0x00000000; + constexpr uint32_t XM_SELECT_1 = 0xFFFFFFFF; + + constexpr uint32_t XM_PERMUTE_0X = 0; + constexpr uint32_t XM_PERMUTE_0Y = 1; + constexpr uint32_t XM_PERMUTE_0Z = 2; + constexpr uint32_t XM_PERMUTE_0W = 3; + constexpr uint32_t XM_PERMUTE_1X = 4; + constexpr uint32_t XM_PERMUTE_1Y = 5; + constexpr uint32_t XM_PERMUTE_1Z = 6; + constexpr uint32_t XM_PERMUTE_1W = 7; + + constexpr uint32_t XM_SWIZZLE_X = 0; + constexpr uint32_t XM_SWIZZLE_Y = 1; + constexpr uint32_t XM_SWIZZLE_Z = 2; + constexpr uint32_t XM_SWIZZLE_W = 3; + + constexpr uint32_t XM_CRMASK_CR6 = 0x000000F0; + constexpr uint32_t XM_CRMASK_CR6TRUE = 0x00000080; + constexpr uint32_t XM_CRMASK_CR6FALSE = 0x00000020; + constexpr uint32_t XM_CRMASK_CR6BOUNDS = XM_CRMASK_CR6FALSE; + +#if defined(_M_ARM) || defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __arm__ || __aarch64__ + constexpr size_t XM_CACHE_LINE_SIZE = 128; +#else + constexpr size_t XM_CACHE_LINE_SIZE = 64; +#endif + + + /**************************************************************************** + * + * Macros + * + ****************************************************************************/ + +#if defined(__XNAMATH_H__) && defined(XMComparisonAllTrue) +#undef XMComparisonAllTrue +#undef XMComparisonAnyTrue +#undef XMComparisonAllFalse +#undef XMComparisonAnyFalse +#undef XMComparisonMixed +#undef XMComparisonAllInBounds +#undef XMComparisonAnyOutOfBounds +#endif + + // Unit conversion + + constexpr float XMConvertToRadians(float fDegrees) noexcept { return fDegrees * (XM_PI / 180.0f); } + constexpr float XMConvertToDegrees(float fRadians) noexcept { return fRadians * (180.0f / XM_PI); } + + // Condition register evaluation proceeding a recording (R) comparison + + constexpr bool XMComparisonAllTrue(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6TRUE) == XM_CRMASK_CR6TRUE; } + constexpr bool XMComparisonAnyTrue(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6FALSE) != XM_CRMASK_CR6FALSE; } + constexpr bool XMComparisonAllFalse(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6FALSE) == XM_CRMASK_CR6FALSE; } + constexpr bool XMComparisonAnyFalse(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6TRUE) != XM_CRMASK_CR6TRUE; } + constexpr bool XMComparisonMixed(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6) == 0; } + constexpr bool XMComparisonAllInBounds(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6BOUNDS) == XM_CRMASK_CR6BOUNDS; } + constexpr bool XMComparisonAnyOutOfBounds(uint32_t CR) noexcept { return (CR & XM_CRMASK_CR6BOUNDS) != XM_CRMASK_CR6BOUNDS; } + + + /**************************************************************************** + * + * Data types + * + ****************************************************************************/ + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable:4068 4201 4365 4324 4820) + // C4068: ignore unknown pragmas + // C4201: nonstandard extension used : nameless struct/union + // C4365: Off by default noise + // C4324/4820: padding warnings +#endif + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 25000, "FXMVECTOR is 16 bytes") +#endif + +//------------------------------------------------------------------------------ +#if defined(_XM_NO_INTRINSICS_) + struct __vector4 + { + union + { + float vector4_f32[4]; + uint32_t vector4_u32[4]; + }; + }; +#endif // _XM_NO_INTRINSICS_ + + //------------------------------------------------------------------------------ + // Vector intrinsic: Four 32 bit floating point components aligned on a 16 byte + // boundary and mapped to hardware vector registers +#if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + using XMVECTOR = __m128; +#elif defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + using XMVECTOR = float32x4_t; +#else + using XMVECTOR = __vector4; +#endif + + // Fix-up for (1st-3rd) XMVECTOR parameters that are pass-in-register for x86, ARM, ARM64, and vector call; by reference otherwise +#if ( defined(_M_IX86) || defined(_M_ARM) || defined(_M_ARM64) || _XM_VECTORCALL_ || __i386__ || __arm__ || __aarch64__ ) && !defined(_XM_NO_INTRINSICS_) + typedef const XMVECTOR FXMVECTOR; +#else + typedef const XMVECTOR& FXMVECTOR; +#endif + + // Fix-up for (4th) XMVECTOR parameter to pass in-register for ARM, ARM64, and vector call; by reference otherwise +#if ( defined(_M_ARM) || defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || _XM_VECTORCALL_ || __arm__ || __aarch64__ ) && !defined(_XM_NO_INTRINSICS_) + typedef const XMVECTOR GXMVECTOR; +#else + typedef const XMVECTOR& GXMVECTOR; +#endif + + // Fix-up for (5th & 6th) XMVECTOR parameter to pass in-register for ARM64 and vector call; by reference otherwise +#if ( defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || _XM_VECTORCALL_ || __aarch64__ ) && !defined(_XM_NO_INTRINSICS_) + typedef const XMVECTOR HXMVECTOR; +#else + typedef const XMVECTOR& HXMVECTOR; +#endif + + // Fix-up for (7th+) XMVECTOR parameters to pass by reference + typedef const XMVECTOR& CXMVECTOR; + + //------------------------------------------------------------------------------ + // Conversion types for constants + XM_ALIGNED_STRUCT(16) XMVECTORF32 + { + union + { + float f[4]; + XMVECTOR v; + }; + + inline operator XMVECTOR() const noexcept { return v; } + inline operator const float* () const noexcept { return f; } + #ifdef _XM_NO_INTRINSICS_ + #elif defined(_XM_SSE_INTRINSICS_) + inline operator __m128i() const noexcept { return _mm_castps_si128(v); } + inline operator __m128d() const noexcept { return _mm_castps_pd(v); } + #elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(__GNUC__) || defined(_ARM64_DISTINCT_NEON_TYPES)) + inline operator int32x4_t() const noexcept { return vreinterpretq_s32_f32(v); } + inline operator uint32x4_t() const noexcept { return vreinterpretq_u32_f32(v); } + #endif + }; + + XM_ALIGNED_STRUCT(16) XMVECTORI32 + { + union + { + int32_t i[4]; + XMVECTOR v; + }; + + inline operator XMVECTOR() const noexcept { return v; } + #ifdef _XM_NO_INTRINSICS_ + #elif defined(_XM_SSE_INTRINSICS_) + inline operator __m128i() const noexcept { return _mm_castps_si128(v); } + inline operator __m128d() const noexcept { return _mm_castps_pd(v); } + #elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(__GNUC__) || defined(_ARM64_DISTINCT_NEON_TYPES)) + inline operator int32x4_t() const noexcept { return vreinterpretq_s32_f32(v); } + inline operator uint32x4_t() const noexcept { return vreinterpretq_u32_f32(v); } + #endif + }; + + XM_ALIGNED_STRUCT(16) XMVECTORU8 + { + union + { + uint8_t u[16]; + XMVECTOR v; + }; + + inline operator XMVECTOR() const noexcept { return v; } + #ifdef _XM_NO_INTRINSICS_ + #elif defined(_XM_SSE_INTRINSICS_) + inline operator __m128i() const noexcept { return _mm_castps_si128(v); } + inline operator __m128d() const noexcept { return _mm_castps_pd(v); } + #elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(__GNUC__) || defined(_ARM64_DISTINCT_NEON_TYPES)) + inline operator int32x4_t() const noexcept { return vreinterpretq_s32_f32(v); } + inline operator uint32x4_t() const noexcept { return vreinterpretq_u32_f32(v); } + #endif + }; + + XM_ALIGNED_STRUCT(16) XMVECTORU32 + { + union + { + uint32_t u[4]; + XMVECTOR v; + }; + + inline operator XMVECTOR() const noexcept { return v; } + #ifdef _XM_NO_INTRINSICS_ + #elif defined(_XM_SSE_INTRINSICS_) + inline operator __m128i() const noexcept { return _mm_castps_si128(v); } + inline operator __m128d() const noexcept { return _mm_castps_pd(v); } + #elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(__GNUC__) || defined(_ARM64_DISTINCT_NEON_TYPES)) + inline operator int32x4_t() const noexcept { return vreinterpretq_s32_f32(v); } + inline operator uint32x4_t() const noexcept { return vreinterpretq_u32_f32(v); } + #endif + }; + + //------------------------------------------------------------------------------ + // Vector operators + +#ifndef _XM_NO_XMVECTOR_OVERLOADS_ + XMVECTOR XM_CALLCONV operator+ (FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV operator- (FXMVECTOR V) noexcept; + + XMVECTOR& XM_CALLCONV operator+= (XMVECTOR& V1, FXMVECTOR V2) noexcept; + XMVECTOR& XM_CALLCONV operator-= (XMVECTOR& V1, FXMVECTOR V2) noexcept; + XMVECTOR& XM_CALLCONV operator*= (XMVECTOR& V1, FXMVECTOR V2) noexcept; + XMVECTOR& XM_CALLCONV operator/= (XMVECTOR& V1, FXMVECTOR V2) noexcept; + + XMVECTOR& operator*= (XMVECTOR& V, float S) noexcept; + XMVECTOR& operator/= (XMVECTOR& V, float S) noexcept; + + XMVECTOR XM_CALLCONV operator+ (FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV operator- (FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV operator* (FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV operator/ (FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV operator* (FXMVECTOR V, float S) noexcept; + XMVECTOR XM_CALLCONV operator* (float S, FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV operator/ (FXMVECTOR V, float S) noexcept; +#endif /* !_XM_NO_XMVECTOR_OVERLOADS_ */ + + //------------------------------------------------------------------------------ + // Matrix type: Sixteen 32 bit floating point components aligned on a + // 16 byte boundary and mapped to four hardware vector registers + + struct XMMATRIX; + + // Fix-up for (1st) XMMATRIX parameter to pass in-register for ARM64 and vector call; by reference otherwise +#if ( defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || _XM_VECTORCALL_ || __aarch64__ ) && !defined(_XM_NO_INTRINSICS_) + typedef const XMMATRIX FXMMATRIX; +#else + typedef const XMMATRIX& FXMMATRIX; +#endif + + // Fix-up for (2nd+) XMMATRIX parameters to pass by reference + typedef const XMMATRIX& CXMMATRIX; + +#ifdef _XM_NO_INTRINSICS_ + struct XMMATRIX + #else + XM_ALIGNED_STRUCT(16) XMMATRIX + #endif + { + #ifdef _XM_NO_INTRINSICS_ + union + { + XMVECTOR r[4]; + struct + { + float _11, _12, _13, _14; + float _21, _22, _23, _24; + float _31, _32, _33, _34; + float _41, _42, _43, _44; + }; + float m[4][4]; + }; + #else + XMVECTOR r[4]; + #endif + + XMMATRIX() = default; + + XMMATRIX(const XMMATRIX&) = default; + + #if defined(_MSC_VER) && (_MSC_FULL_VER < 191426431) + XMMATRIX& operator= (const XMMATRIX& M) noexcept { r[0] = M.r[0]; r[1] = M.r[1]; r[2] = M.r[2]; r[3] = M.r[3]; return *this; } + #else + XMMATRIX& operator=(const XMMATRIX&) = default; + + XMMATRIX(XMMATRIX&&) = default; + XMMATRIX& operator=(XMMATRIX&&) = default; + #endif + + constexpr XMMATRIX(FXMVECTOR R0, FXMVECTOR R1, FXMVECTOR R2, CXMVECTOR R3) noexcept : r{ R0,R1,R2,R3 } {} + XMMATRIX(float m00, float m01, float m02, float m03, + float m10, float m11, float m12, float m13, + float m20, float m21, float m22, float m23, + float m30, float m31, float m32, float m33) noexcept; + explicit XMMATRIX(_In_reads_(16) const float* pArray) noexcept; + + #ifdef _XM_NO_INTRINSICS_ + float operator() (size_t Row, size_t Column) const noexcept { return m[Row][Column]; } + float& operator() (size_t Row, size_t Column) noexcept { return m[Row][Column]; } + #endif + + XMMATRIX operator+ () const noexcept { return *this; } + XMMATRIX operator- () const noexcept; + + XMMATRIX& XM_CALLCONV operator+= (FXMMATRIX M) noexcept; + XMMATRIX& XM_CALLCONV operator-= (FXMMATRIX M) noexcept; + XMMATRIX& XM_CALLCONV operator*= (FXMMATRIX M) noexcept; + XMMATRIX& operator*= (float S) noexcept; + XMMATRIX& operator/= (float S) noexcept; + + XMMATRIX XM_CALLCONV operator+ (FXMMATRIX M) const noexcept; + XMMATRIX XM_CALLCONV operator- (FXMMATRIX M) const noexcept; + XMMATRIX XM_CALLCONV operator* (FXMMATRIX M) const noexcept; + XMMATRIX operator* (float S) const noexcept; + XMMATRIX operator/ (float S) const noexcept; + + friend XMMATRIX XM_CALLCONV operator* (float S, FXMMATRIX M) noexcept; + }; + + //------------------------------------------------------------------------------ + // 2D Vector; 32 bit floating point components + struct XMFLOAT2 + { + float x; + float y; + + XMFLOAT2() = default; + + XMFLOAT2(const XMFLOAT2&) = default; + XMFLOAT2& operator=(const XMFLOAT2&) = default; + + XMFLOAT2(XMFLOAT2&&) = default; + XMFLOAT2& operator=(XMFLOAT2&&) = default; + + constexpr XMFLOAT2(float _x, float _y) noexcept : x(_x), y(_y) {} + explicit XMFLOAT2(_In_reads_(2) const float* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT2&) const = default; + auto operator <=> (const XMFLOAT2&) const = default; + #endif + }; + + // 2D Vector; 32 bit floating point components aligned on a 16 byte boundary + XM_ALIGNED_STRUCT(16) XMFLOAT2A : public XMFLOAT2 + { + using XMFLOAT2::XMFLOAT2; + }; + + //------------------------------------------------------------------------------ + // 2D Vector; 32 bit signed integer components + struct XMINT2 + { + int32_t x; + int32_t y; + + XMINT2() = default; + + XMINT2(const XMINT2&) = default; + XMINT2& operator=(const XMINT2&) = default; + + XMINT2(XMINT2&&) = default; + XMINT2& operator=(XMINT2&&) = default; + + constexpr XMINT2(int32_t _x, int32_t _y) noexcept : x(_x), y(_y) {} + explicit XMINT2(_In_reads_(2) const int32_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMINT2&) const = default; + auto operator <=> (const XMINT2&) const = default; + #endif + }; + + // 2D Vector; 32 bit unsigned integer components + struct XMUINT2 + { + uint32_t x; + uint32_t y; + + XMUINT2() = default; + + XMUINT2(const XMUINT2&) = default; + XMUINT2& operator=(const XMUINT2&) = default; + + XMUINT2(XMUINT2&&) = default; + XMUINT2& operator=(XMUINT2&&) = default; + + constexpr XMUINT2(uint32_t _x, uint32_t _y) noexcept : x(_x), y(_y) {} + explicit XMUINT2(_In_reads_(2) const uint32_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMUINT2&) const = default; + auto operator <=> (const XMUINT2&) const = default; + #endif + }; + + //------------------------------------------------------------------------------ + // 3D Vector; 32 bit floating point components + struct XMFLOAT3 + { + float x; + float y; + float z; + + XMFLOAT3() = default; + + XMFLOAT3(const XMFLOAT3&) = default; + XMFLOAT3& operator=(const XMFLOAT3&) = default; + + XMFLOAT3(XMFLOAT3&&) = default; + XMFLOAT3& operator=(XMFLOAT3&&) = default; + + constexpr XMFLOAT3(float _x, float _y, float _z) noexcept : x(_x), y(_y), z(_z) {} + explicit XMFLOAT3(_In_reads_(3) const float* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT3&) const = default; + auto operator <=> (const XMFLOAT3&) const = default; + #endif + }; + + // 3D Vector; 32 bit floating point components aligned on a 16 byte boundary + XM_ALIGNED_STRUCT(16) XMFLOAT3A : public XMFLOAT3 + { + using XMFLOAT3::XMFLOAT3; + }; + + //------------------------------------------------------------------------------ + // 3D Vector; 32 bit signed integer components + struct XMINT3 + { + int32_t x; + int32_t y; + int32_t z; + + XMINT3() = default; + + XMINT3(const XMINT3&) = default; + XMINT3& operator=(const XMINT3&) = default; + + XMINT3(XMINT3&&) = default; + XMINT3& operator=(XMINT3&&) = default; + + constexpr XMINT3(int32_t _x, int32_t _y, int32_t _z) noexcept : x(_x), y(_y), z(_z) {} + explicit XMINT3(_In_reads_(3) const int32_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMINT3&) const = default; + auto operator <=> (const XMINT3&) const = default; + #endif + }; + + // 3D Vector; 32 bit unsigned integer components + struct XMUINT3 + { + uint32_t x; + uint32_t y; + uint32_t z; + + XMUINT3() = default; + + XMUINT3(const XMUINT3&) = default; + XMUINT3& operator=(const XMUINT3&) = default; + + XMUINT3(XMUINT3&&) = default; + XMUINT3& operator=(XMUINT3&&) = default; + + constexpr XMUINT3(uint32_t _x, uint32_t _y, uint32_t _z) noexcept : x(_x), y(_y), z(_z) {} + explicit XMUINT3(_In_reads_(3) const uint32_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMUINT3&) const = default; + auto operator <=> (const XMUINT3&) const = default; + #endif + }; + + //------------------------------------------------------------------------------ + // 4D Vector; 32 bit floating point components + struct XMFLOAT4 + { + float x; + float y; + float z; + float w; + + XMFLOAT4() = default; + + XMFLOAT4(const XMFLOAT4&) = default; + XMFLOAT4& operator=(const XMFLOAT4&) = default; + + XMFLOAT4(XMFLOAT4&&) = default; + XMFLOAT4& operator=(XMFLOAT4&&) = default; + + constexpr XMFLOAT4(float _x, float _y, float _z, float _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMFLOAT4(_In_reads_(4) const float* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT4&) const = default; + auto operator <=> (const XMFLOAT4&) const = default; + #endif + }; + + // 4D Vector; 32 bit floating point components aligned on a 16 byte boundary + XM_ALIGNED_STRUCT(16) XMFLOAT4A : public XMFLOAT4 + { + using XMFLOAT4::XMFLOAT4; + }; + + //------------------------------------------------------------------------------ + // 4D Vector; 32 bit signed integer components + struct XMINT4 + { + int32_t x; + int32_t y; + int32_t z; + int32_t w; + + XMINT4() = default; + + XMINT4(const XMINT4&) = default; + XMINT4& operator=(const XMINT4&) = default; + + XMINT4(XMINT4&&) = default; + XMINT4& operator=(XMINT4&&) = default; + + constexpr XMINT4(int32_t _x, int32_t _y, int32_t _z, int32_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMINT4(_In_reads_(4) const int32_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMINT4&) const = default; + auto operator <=> (const XMINT4&) const = default; + #endif + }; + + // 4D Vector; 32 bit unsigned integer components + struct XMUINT4 + { + uint32_t x; + uint32_t y; + uint32_t z; + uint32_t w; + + XMUINT4() = default; + + XMUINT4(const XMUINT4&) = default; + XMUINT4& operator=(const XMUINT4&) = default; + + XMUINT4(XMUINT4&&) = default; + XMUINT4& operator=(XMUINT4&&) = default; + + constexpr XMUINT4(uint32_t _x, uint32_t _y, uint32_t _z, uint32_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMUINT4(_In_reads_(4) const uint32_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + + #if (__cplusplus >= 202002L) + bool operator == (const XMUINT4&) const = default; + auto operator <=> (const XMUINT4&) const = default; + #endif + }; + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wgnu-anonymous-struct" +#pragma clang diagnostic ignored "-Wnested-anon-types" +#pragma clang diagnostic ignored "-Wunknown-warning-option" +#pragma clang diagnostic ignored "-Wunsafe-buffer-usage" +#endif + + //------------------------------------------------------------------------------ + // 3x3 Matrix: 32 bit floating point components + struct XMFLOAT3X3 + { + union + { + struct + { + float _11, _12, _13; + float _21, _22, _23; + float _31, _32, _33; + }; + float m[3][3]; + }; + + XMFLOAT3X3() = default; + + XMFLOAT3X3(const XMFLOAT3X3&) = default; + XMFLOAT3X3& operator=(const XMFLOAT3X3&) = default; + + XMFLOAT3X3(XMFLOAT3X3&&) = default; + XMFLOAT3X3& operator=(XMFLOAT3X3&&) = default; + + constexpr XMFLOAT3X3(float m00, float m01, float m02, + float m10, float m11, float m12, + float m20, float m21, float m22) noexcept + : _11(m00), _12(m01), _13(m02), + _21(m10), _22(m11), _23(m12), + _31(m20), _32(m21), _33(m22) + {} + explicit XMFLOAT3X3(_In_reads_(9) const float* pArray) noexcept; + + float operator() (size_t Row, size_t Column) const noexcept { return m[Row][Column]; } + float& operator() (size_t Row, size_t Column) noexcept { return m[Row][Column]; } + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT3X3& M) const noexcept + { + return _11 == M._11 && _12 == M._12 && _13 == M._13 + && _21 == M._21 && _22 == M._22 && _23 == M._23 + && _31 == M._31 && _32 == M._32 && _33 == M._33; + } + + auto operator <=> (const XMFLOAT3X3& M) const noexcept + { + if (auto cmp = _11 <=> M._11; cmp != 0) return cmp; + if (auto cmp = _12 <=> M._12; cmp != 0) return cmp; + if (auto cmp = _13 <=> M._13; cmp != 0) return cmp; + if (auto cmp = _21 <=> M._21; cmp != 0) return cmp; + if (auto cmp = _22 <=> M._22; cmp != 0) return cmp; + if (auto cmp = _23 <=> M._23; cmp != 0) return cmp; + if (auto cmp = _31 <=> M._31; cmp != 0) return cmp; + if (auto cmp = _32 <=> M._32; cmp != 0) return cmp; + return _33 <=> M._33; + } + #endif + }; + + //------------------------------------------------------------------------------ + // 4x3 Row-major Matrix: 32 bit floating point components + struct XMFLOAT4X3 + { + union + { + struct + { + float _11, _12, _13; + float _21, _22, _23; + float _31, _32, _33; + float _41, _42, _43; + }; + float m[4][3]; + float f[12]; + }; + + XMFLOAT4X3() = default; + + XMFLOAT4X3(const XMFLOAT4X3&) = default; + XMFLOAT4X3& operator=(const XMFLOAT4X3&) = default; + + XMFLOAT4X3(XMFLOAT4X3&&) = default; + XMFLOAT4X3& operator=(XMFLOAT4X3&&) = default; + + constexpr XMFLOAT4X3(float m00, float m01, float m02, + float m10, float m11, float m12, + float m20, float m21, float m22, + float m30, float m31, float m32) noexcept + : _11(m00), _12(m01), _13(m02), + _21(m10), _22(m11), _23(m12), + _31(m20), _32(m21), _33(m22), + _41(m30), _42(m31), _43(m32) + {} + explicit XMFLOAT4X3(_In_reads_(12) const float* pArray) noexcept; + + float operator() (size_t Row, size_t Column) const noexcept { return m[Row][Column]; } + float& operator() (size_t Row, size_t Column) noexcept { return m[Row][Column]; } + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT4X3& M) const noexcept + { + return _11 == M._11 && _12 == M._12 && _13 == M._13 + && _21 == M._21 && _22 == M._22 && _23 == M._23 + && _31 == M._31 && _32 == M._32 && _33 == M._33 + && _41 == M._41 && _42 == M._42 && _43 == M._43; + } + + auto operator <=> (const XMFLOAT4X3& M) const noexcept + { + if (auto cmp = _11 <=> M._11; cmp != 0) return cmp; + if (auto cmp = _12 <=> M._12; cmp != 0) return cmp; + if (auto cmp = _13 <=> M._13; cmp != 0) return cmp; + if (auto cmp = _21 <=> M._21; cmp != 0) return cmp; + if (auto cmp = _22 <=> M._22; cmp != 0) return cmp; + if (auto cmp = _23 <=> M._23; cmp != 0) return cmp; + if (auto cmp = _31 <=> M._31; cmp != 0) return cmp; + if (auto cmp = _32 <=> M._32; cmp != 0) return cmp; + if (auto cmp = _33 <=> M._33; cmp != 0) return cmp; + if (auto cmp = _41 <=> M._41; cmp != 0) return cmp; + if (auto cmp = _42 <=> M._42; cmp != 0) return cmp; + return _43 <=> M._43; + } + #endif + }; + + // 4x3 Row-major Matrix: 32 bit floating point components aligned on a 16 byte boundary + XM_ALIGNED_STRUCT(16) XMFLOAT4X3A : public XMFLOAT4X3 + { + using XMFLOAT4X3::XMFLOAT4X3; + }; + + //------------------------------------------------------------------------------ + // 3x4 Column-major Matrix: 32 bit floating point components + struct XMFLOAT3X4 + { + union + { + struct + { + float _11, _12, _13, _14; + float _21, _22, _23, _24; + float _31, _32, _33, _34; + }; + float m[3][4]; + float f[12]; + }; + + XMFLOAT3X4() = default; + + XMFLOAT3X4(const XMFLOAT3X4&) = default; + XMFLOAT3X4& operator=(const XMFLOAT3X4&) = default; + + XMFLOAT3X4(XMFLOAT3X4&&) = default; + XMFLOAT3X4& operator=(XMFLOAT3X4&&) = default; + + constexpr XMFLOAT3X4(float m00, float m01, float m02, float m03, + float m10, float m11, float m12, float m13, + float m20, float m21, float m22, float m23) noexcept + : _11(m00), _12(m01), _13(m02), _14(m03), + _21(m10), _22(m11), _23(m12), _24(m13), + _31(m20), _32(m21), _33(m22), _34(m23) + {} + explicit XMFLOAT3X4(_In_reads_(12) const float* pArray) noexcept; + + float operator() (size_t Row, size_t Column) const noexcept { return m[Row][Column]; } + float& operator() (size_t Row, size_t Column) noexcept { return m[Row][Column]; } + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT3X4& M) const noexcept + { + return _11 == M._11 && _12 == M._12 && _13 == M._13 && _14 == M._14 + && _21 == M._21 && _22 == M._22 && _23 == M._23 && _24 == M._24 + && _31 == M._31 && _32 == M._32 && _33 == M._33 && _34 == M._34; + } + + auto operator <=> (const XMFLOAT3X4& M) const noexcept + { + if (auto cmp = _11 <=> M._11; cmp != 0) return cmp; + if (auto cmp = _12 <=> M._12; cmp != 0) return cmp; + if (auto cmp = _13 <=> M._13; cmp != 0) return cmp; + if (auto cmp = _14 <=> M._14; cmp != 0) return cmp; + if (auto cmp = _21 <=> M._21; cmp != 0) return cmp; + if (auto cmp = _22 <=> M._22; cmp != 0) return cmp; + if (auto cmp = _23 <=> M._23; cmp != 0) return cmp; + if (auto cmp = _24 <=> M._24; cmp != 0) return cmp; + if (auto cmp = _31 <=> M._31; cmp != 0) return cmp; + if (auto cmp = _32 <=> M._32; cmp != 0) return cmp; + if (auto cmp = _33 <=> M._33; cmp != 0) return cmp; + return _34 <=> M._34; + } + #endif + }; + + // 3x4 Column-major Matrix: 32 bit floating point components aligned on a 16 byte boundary + XM_ALIGNED_STRUCT(16) XMFLOAT3X4A : public XMFLOAT3X4 + { + using XMFLOAT3X4::XMFLOAT3X4; + }; + + //------------------------------------------------------------------------------ + // 4x4 Matrix: 32 bit floating point components + struct XMFLOAT4X4 + { + union + { + struct + { + float _11, _12, _13, _14; + float _21, _22, _23, _24; + float _31, _32, _33, _34; + float _41, _42, _43, _44; + }; + float m[4][4]; + }; + + XMFLOAT4X4() = default; + + XMFLOAT4X4(const XMFLOAT4X4&) = default; + XMFLOAT4X4& operator=(const XMFLOAT4X4&) = default; + + XMFLOAT4X4(XMFLOAT4X4&&) = default; + XMFLOAT4X4& operator=(XMFLOAT4X4&&) = default; + + constexpr XMFLOAT4X4(float m00, float m01, float m02, float m03, + float m10, float m11, float m12, float m13, + float m20, float m21, float m22, float m23, + float m30, float m31, float m32, float m33) noexcept + : _11(m00), _12(m01), _13(m02), _14(m03), + _21(m10), _22(m11), _23(m12), _24(m13), + _31(m20), _32(m21), _33(m22), _34(m23), + _41(m30), _42(m31), _43(m32), _44(m33) + {} + explicit XMFLOAT4X4(_In_reads_(16) const float* pArray) noexcept; + + float operator() (size_t Row, size_t Column) const noexcept { return m[Row][Column]; } + float& operator() (size_t Row, size_t Column) noexcept { return m[Row][Column]; } + + #if (__cplusplus >= 202002L) + bool operator == (const XMFLOAT4X4& M) const noexcept + { + return _11 == M._11 && _12 == M._12 && _13 == M._13 && _14 == M._14 + && _21 == M._21 && _22 == M._22 && _23 == M._23 && _24 == M._24 + && _31 == M._31 && _32 == M._32 && _33 == M._33 && _34 == M._34 + && _41 == M._41 && _42 == M._42 && _43 == M._43 && _44 == M._44; + } + + auto operator <=> (const XMFLOAT4X4& M) const noexcept + { + if (auto cmp = _11 <=> M._11; cmp != 0) return cmp; + if (auto cmp = _12 <=> M._12; cmp != 0) return cmp; + if (auto cmp = _13 <=> M._13; cmp != 0) return cmp; + if (auto cmp = _14 <=> M._14; cmp != 0) return cmp; + if (auto cmp = _21 <=> M._21; cmp != 0) return cmp; + if (auto cmp = _22 <=> M._22; cmp != 0) return cmp; + if (auto cmp = _23 <=> M._23; cmp != 0) return cmp; + if (auto cmp = _24 <=> M._24; cmp != 0) return cmp; + if (auto cmp = _31 <=> M._31; cmp != 0) return cmp; + if (auto cmp = _32 <=> M._32; cmp != 0) return cmp; + if (auto cmp = _33 <=> M._33; cmp != 0) return cmp; + if (auto cmp = _34 <=> M._34; cmp != 0) return cmp; + if (auto cmp = _41 <=> M._41; cmp != 0) return cmp; + if (auto cmp = _42 <=> M._42; cmp != 0) return cmp; + if (auto cmp = _43 <=> M._43; cmp != 0) return cmp; + return _44 <=> M._44; + } + #endif + }; + + // 4x4 Matrix: 32 bit floating point components aligned on a 16 byte boundary + XM_ALIGNED_STRUCT(16) XMFLOAT4X4A : public XMFLOAT4X4 + { + using XMFLOAT4X4::XMFLOAT4X4; + }; + + //////////////////////////////////////////////////////////////////////////////// + +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +/**************************************************************************** + * + * Data conversion operations + * + ****************************************************************************/ + + XMVECTOR XM_CALLCONV XMConvertVectorIntToFloat(FXMVECTOR VInt, uint32_t DivExponent) noexcept; + XMVECTOR XM_CALLCONV XMConvertVectorFloatToInt(FXMVECTOR VFloat, uint32_t MulExponent) noexcept; + XMVECTOR XM_CALLCONV XMConvertVectorUIntToFloat(FXMVECTOR VUInt, uint32_t DivExponent) noexcept; + XMVECTOR XM_CALLCONV XMConvertVectorFloatToUInt(FXMVECTOR VFloat, uint32_t MulExponent) noexcept; + +#if defined(__XNAMATH_H__) && defined(XMVectorSetBinaryConstant) +#undef XMVectorSetBinaryConstant +#undef XMVectorSplatConstant +#undef XMVectorSplatConstantInt +#endif + + XMVECTOR XM_CALLCONV XMVectorSetBinaryConstant(uint32_t C0, uint32_t C1, uint32_t C2, uint32_t C3) noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatConstant(int32_t IntConstant, uint32_t DivExponent) noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatConstantInt(int32_t IntConstant) noexcept; + + /**************************************************************************** + * + * Load operations + * + ****************************************************************************/ + + XMVECTOR XM_CALLCONV XMLoadInt(_In_ const uint32_t* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat(_In_ const float* pSource) noexcept; + + XMVECTOR XM_CALLCONV XMLoadInt2(_In_reads_(2) const uint32_t* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadInt2A(_In_reads_(2) const uint32_t* PSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat2(_In_ const XMFLOAT2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat2A(_In_ const XMFLOAT2A* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadSInt2(_In_ const XMINT2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUInt2(_In_ const XMUINT2* pSource) noexcept; + + XMVECTOR XM_CALLCONV XMLoadInt3(_In_reads_(3) const uint32_t* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadInt3A(_In_reads_(3) const uint32_t* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat3(_In_ const XMFLOAT3* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat3A(_In_ const XMFLOAT3A* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadSInt3(_In_ const XMINT3* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUInt3(_In_ const XMUINT3* pSource) noexcept; + + XMVECTOR XM_CALLCONV XMLoadInt4(_In_reads_(4) const uint32_t* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadInt4A(_In_reads_(4) const uint32_t* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat4(_In_ const XMFLOAT4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat4A(_In_ const XMFLOAT4A* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadSInt4(_In_ const XMINT4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUInt4(_In_ const XMUINT4* pSource) noexcept; + + XMMATRIX XM_CALLCONV XMLoadFloat3x3(_In_ const XMFLOAT3X3* pSource) noexcept; + XMMATRIX XM_CALLCONV XMLoadFloat4x3(_In_ const XMFLOAT4X3* pSource) noexcept; + XMMATRIX XM_CALLCONV XMLoadFloat4x3A(_In_ const XMFLOAT4X3A* pSource) noexcept; + XMMATRIX XM_CALLCONV XMLoadFloat3x4(_In_ const XMFLOAT3X4* pSource) noexcept; + XMMATRIX XM_CALLCONV XMLoadFloat3x4A(_In_ const XMFLOAT3X4A* pSource) noexcept; + XMMATRIX XM_CALLCONV XMLoadFloat4x4(_In_ const XMFLOAT4X4* pSource) noexcept; + XMMATRIX XM_CALLCONV XMLoadFloat4x4A(_In_ const XMFLOAT4X4A* pSource) noexcept; + + /**************************************************************************** + * + * Store operations + * + ****************************************************************************/ + + void XM_CALLCONV XMStoreInt(_Out_ uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat(_Out_ float* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreInt2(_Out_writes_(2) uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreInt2A(_Out_writes_(2) uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat2(_Out_ XMFLOAT2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat2A(_Out_ XMFLOAT2A* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreSInt2(_Out_ XMINT2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUInt2(_Out_ XMUINT2* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreInt3(_Out_writes_(3) uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreInt3A(_Out_writes_(3) uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat3(_Out_ XMFLOAT3* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat3A(_Out_ XMFLOAT3A* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreSInt3(_Out_ XMINT3* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUInt3(_Out_ XMUINT3* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreInt4(_Out_writes_(4) uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreInt4A(_Out_writes_(4) uint32_t* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat4(_Out_ XMFLOAT4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat4A(_Out_ XMFLOAT4A* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreSInt4(_Out_ XMINT4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUInt4(_Out_ XMUINT4* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreFloat3x3(_Out_ XMFLOAT3X3* pDestination, _In_ FXMMATRIX M) noexcept; + void XM_CALLCONV XMStoreFloat4x3(_Out_ XMFLOAT4X3* pDestination, _In_ FXMMATRIX M) noexcept; + void XM_CALLCONV XMStoreFloat4x3A(_Out_ XMFLOAT4X3A* pDestination, _In_ FXMMATRIX M) noexcept; + void XM_CALLCONV XMStoreFloat3x4(_Out_ XMFLOAT3X4* pDestination, _In_ FXMMATRIX M) noexcept; + void XM_CALLCONV XMStoreFloat3x4A(_Out_ XMFLOAT3X4A* pDestination, _In_ FXMMATRIX M) noexcept; + void XM_CALLCONV XMStoreFloat4x4(_Out_ XMFLOAT4X4* pDestination, _In_ FXMMATRIX M) noexcept; + void XM_CALLCONV XMStoreFloat4x4A(_Out_ XMFLOAT4X4A* pDestination, _In_ FXMMATRIX M) noexcept; + + /**************************************************************************** + * + * General vector operations + * + ****************************************************************************/ + + XMVECTOR XM_CALLCONV XMVectorZero() noexcept; + XMVECTOR XM_CALLCONV XMVectorSet(float x, float y, float z, float w) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetInt(uint32_t x, uint32_t y, uint32_t z, uint32_t w) noexcept; + XMVECTOR XM_CALLCONV XMVectorReplicate(float Value) noexcept; + XMVECTOR XM_CALLCONV XMVectorReplicatePtr(_In_ const float* pValue) noexcept; + XMVECTOR XM_CALLCONV XMVectorReplicateInt(uint32_t Value) noexcept; + XMVECTOR XM_CALLCONV XMVectorReplicateIntPtr(_In_ const uint32_t* pValue) noexcept; + XMVECTOR XM_CALLCONV XMVectorTrueInt() noexcept; + XMVECTOR XM_CALLCONV XMVectorFalseInt() noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatX(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatY(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatZ(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatW(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatOne() noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatInfinity() noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatQNaN() noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatEpsilon() noexcept; + XMVECTOR XM_CALLCONV XMVectorSplatSignMask() noexcept; + + float XM_CALLCONV XMVectorGetByIndex(FXMVECTOR V, size_t i) noexcept; + float XM_CALLCONV XMVectorGetX(FXMVECTOR V) noexcept; + float XM_CALLCONV XMVectorGetY(FXMVECTOR V) noexcept; + float XM_CALLCONV XMVectorGetZ(FXMVECTOR V) noexcept; + float XM_CALLCONV XMVectorGetW(FXMVECTOR V) noexcept; + + void XM_CALLCONV XMVectorGetByIndexPtr(_Out_ float* f, _In_ FXMVECTOR V, _In_ size_t i) noexcept; + void XM_CALLCONV XMVectorGetXPtr(_Out_ float* x, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorGetYPtr(_Out_ float* y, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorGetZPtr(_Out_ float* z, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorGetWPtr(_Out_ float* w, _In_ FXMVECTOR V) noexcept; + + uint32_t XM_CALLCONV XMVectorGetIntByIndex(FXMVECTOR V, size_t i) noexcept; + uint32_t XM_CALLCONV XMVectorGetIntX(FXMVECTOR V) noexcept; + uint32_t XM_CALLCONV XMVectorGetIntY(FXMVECTOR V) noexcept; + uint32_t XM_CALLCONV XMVectorGetIntZ(FXMVECTOR V) noexcept; + uint32_t XM_CALLCONV XMVectorGetIntW(FXMVECTOR V) noexcept; + + void XM_CALLCONV XMVectorGetIntByIndexPtr(_Out_ uint32_t* x, _In_ FXMVECTOR V, _In_ size_t i) noexcept; + void XM_CALLCONV XMVectorGetIntXPtr(_Out_ uint32_t* x, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorGetIntYPtr(_Out_ uint32_t* y, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorGetIntZPtr(_Out_ uint32_t* z, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorGetIntWPtr(_Out_ uint32_t* w, _In_ FXMVECTOR V) noexcept; + + XMVECTOR XM_CALLCONV XMVectorSetByIndex(FXMVECTOR V, float f, size_t i) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetX(FXMVECTOR V, float x) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetY(FXMVECTOR V, float y) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetZ(FXMVECTOR V, float z) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetW(FXMVECTOR V, float w) noexcept; + + XMVECTOR XM_CALLCONV XMVectorSetByIndexPtr(_In_ FXMVECTOR V, _In_ const float* f, _In_ size_t i) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetXPtr(_In_ FXMVECTOR V, _In_ const float* x) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetYPtr(_In_ FXMVECTOR V, _In_ const float* y) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetZPtr(_In_ FXMVECTOR V, _In_ const float* z) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetWPtr(_In_ FXMVECTOR V, _In_ const float* w) noexcept; + + XMVECTOR XM_CALLCONV XMVectorSetIntByIndex(FXMVECTOR V, uint32_t x, size_t i) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntX(FXMVECTOR V, uint32_t x) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntY(FXMVECTOR V, uint32_t y) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntZ(FXMVECTOR V, uint32_t z) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntW(FXMVECTOR V, uint32_t w) noexcept; + + XMVECTOR XM_CALLCONV XMVectorSetIntByIndexPtr(_In_ FXMVECTOR V, _In_ const uint32_t* x, _In_ size_t i) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntXPtr(_In_ FXMVECTOR V, _In_ const uint32_t* x) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntYPtr(_In_ FXMVECTOR V, _In_ const uint32_t* y) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntZPtr(_In_ FXMVECTOR V, _In_ const uint32_t* z) noexcept; + XMVECTOR XM_CALLCONV XMVectorSetIntWPtr(_In_ FXMVECTOR V, _In_ const uint32_t* w) noexcept; + +#if defined(__XNAMATH_H__) && defined(XMVectorSwizzle) +#undef XMVectorSwizzle +#endif + + XMVECTOR XM_CALLCONV XMVectorSwizzle(FXMVECTOR V, uint32_t E0, uint32_t E1, uint32_t E2, uint32_t E3) noexcept; + XMVECTOR XM_CALLCONV XMVectorPermute(FXMVECTOR V1, FXMVECTOR V2, uint32_t PermuteX, uint32_t PermuteY, uint32_t PermuteZ, uint32_t PermuteW) noexcept; + XMVECTOR XM_CALLCONV XMVectorSelectControl(uint32_t VectorIndex0, uint32_t VectorIndex1, uint32_t VectorIndex2, uint32_t VectorIndex3) noexcept; + XMVECTOR XM_CALLCONV XMVectorSelect(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR Control) noexcept; + XMVECTOR XM_CALLCONV XMVectorMergeXY(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorMergeZW(FXMVECTOR V1, FXMVECTOR V2) noexcept; + +#if defined(__XNAMATH_H__) && defined(XMVectorShiftLeft) +#undef XMVectorShiftLeft +#undef XMVectorRotateLeft +#undef XMVectorRotateRight +#undef XMVectorInsert +#endif + + XMVECTOR XM_CALLCONV XMVectorShiftLeft(FXMVECTOR V1, FXMVECTOR V2, uint32_t Elements) noexcept; + XMVECTOR XM_CALLCONV XMVectorRotateLeft(FXMVECTOR V, uint32_t Elements) noexcept; + XMVECTOR XM_CALLCONV XMVectorRotateRight(FXMVECTOR V, uint32_t Elements) noexcept; + XMVECTOR XM_CALLCONV XMVectorInsert(FXMVECTOR VD, FXMVECTOR VS, uint32_t VSLeftRotateElements, + uint32_t Select0, uint32_t Select1, uint32_t Select2, uint32_t Select3) noexcept; + + XMVECTOR XM_CALLCONV XMVectorEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorEqualR(_Out_ uint32_t* pCR, _In_ FXMVECTOR V1, _In_ FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorEqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorEqualIntR(_Out_ uint32_t* pCR, _In_ FXMVECTOR V, _In_ FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorNearEqual(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR Epsilon) noexcept; + XMVECTOR XM_CALLCONV XMVectorNotEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorNotEqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorGreater(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorGreaterR(_Out_ uint32_t* pCR, _In_ FXMVECTOR V1, _In_ FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorGreaterOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorGreaterOrEqualR(_Out_ uint32_t* pCR, _In_ FXMVECTOR V1, _In_ FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorLess(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorLessOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorInBounds(FXMVECTOR V, FXMVECTOR Bounds) noexcept; + XMVECTOR XM_CALLCONV XMVectorInBoundsR(_Out_ uint32_t* pCR, _In_ FXMVECTOR V, _In_ FXMVECTOR Bounds) noexcept; + + XMVECTOR XM_CALLCONV XMVectorIsNaN(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorIsInfinite(FXMVECTOR V) noexcept; + + XMVECTOR XM_CALLCONV XMVectorMin(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorMax(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorRound(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorTruncate(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorFloor(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorCeiling(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorClamp(FXMVECTOR V, FXMVECTOR Min, FXMVECTOR Max) noexcept; + XMVECTOR XM_CALLCONV XMVectorSaturate(FXMVECTOR V) noexcept; + + XMVECTOR XM_CALLCONV XMVectorAndInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorAndCInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorOrInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorNorInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorXorInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + + XMVECTOR XM_CALLCONV XMVectorNegate(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorAdd(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorSum(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorAddAngles(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorSubtract(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorSubtractAngles(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorMultiply(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorMultiplyAdd(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR V3) noexcept; + XMVECTOR XM_CALLCONV XMVectorDivide(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorNegativeMultiplySubtract(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR V3) noexcept; + XMVECTOR XM_CALLCONV XMVectorScale(FXMVECTOR V, float ScaleFactor) noexcept; + XMVECTOR XM_CALLCONV XMVectorReciprocalEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorReciprocal(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSqrtEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSqrt(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorReciprocalSqrtEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorReciprocalSqrt(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorExp2(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorExp10(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorExpE(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorExp(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorLog2(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorLog10(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorLogE(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorLog(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorPow(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorAbs(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorMod(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVectorModAngles(FXMVECTOR Angles) noexcept; + XMVECTOR XM_CALLCONV XMVectorSin(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSinEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorCos(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorCosEst(FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorSinCos(_Out_ XMVECTOR* pSin, _Out_ XMVECTOR* pCos, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMVectorSinCosEst(_Out_ XMVECTOR* pSin, _Out_ XMVECTOR* pCos, _In_ FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorTan(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorTanEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorSinH(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorCosH(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorTanH(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorASin(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorASinEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorACos(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorACosEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorATan(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorATanEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVectorATan2(FXMVECTOR Y, FXMVECTOR X) noexcept; + XMVECTOR XM_CALLCONV XMVectorATan2Est(FXMVECTOR Y, FXMVECTOR X) noexcept; + XMVECTOR XM_CALLCONV XMVectorLerp(FXMVECTOR V0, FXMVECTOR V1, float t) noexcept; + XMVECTOR XM_CALLCONV XMVectorLerpV(FXMVECTOR V0, FXMVECTOR V1, FXMVECTOR T) noexcept; + XMVECTOR XM_CALLCONV XMVectorHermite(FXMVECTOR Position0, FXMVECTOR Tangent0, FXMVECTOR Position1, GXMVECTOR Tangent1, float t) noexcept; + XMVECTOR XM_CALLCONV XMVectorHermiteV(FXMVECTOR Position0, FXMVECTOR Tangent0, FXMVECTOR Position1, GXMVECTOR Tangent1, HXMVECTOR T) noexcept; + XMVECTOR XM_CALLCONV XMVectorCatmullRom(FXMVECTOR Position0, FXMVECTOR Position1, FXMVECTOR Position2, GXMVECTOR Position3, float t) noexcept; + XMVECTOR XM_CALLCONV XMVectorCatmullRomV(FXMVECTOR Position0, FXMVECTOR Position1, FXMVECTOR Position2, GXMVECTOR Position3, HXMVECTOR T) noexcept; + XMVECTOR XM_CALLCONV XMVectorBaryCentric(FXMVECTOR Position0, FXMVECTOR Position1, FXMVECTOR Position2, float f, float g) noexcept; + XMVECTOR XM_CALLCONV XMVectorBaryCentricV(FXMVECTOR Position0, FXMVECTOR Position1, FXMVECTOR Position2, GXMVECTOR F, HXMVECTOR G) noexcept; + + /**************************************************************************** + * + * 2D vector operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMVector2Equal(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector2EqualR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2EqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector2EqualIntR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2NearEqual(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR Epsilon) noexcept; + bool XM_CALLCONV XMVector2NotEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2NotEqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2Greater(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector2GreaterR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2GreaterOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector2GreaterOrEqualR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2Less(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2LessOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector2InBounds(FXMVECTOR V, FXMVECTOR Bounds) noexcept; + + bool XM_CALLCONV XMVector2IsNaN(FXMVECTOR V) noexcept; + bool XM_CALLCONV XMVector2IsInfinite(FXMVECTOR V) noexcept; + + XMVECTOR XM_CALLCONV XMVector2Dot(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector2Cross(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector2LengthSq(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2ReciprocalLengthEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2ReciprocalLength(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2LengthEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2Length(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2NormalizeEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2Normalize(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2ClampLength(FXMVECTOR V, float LengthMin, float LengthMax) noexcept; + XMVECTOR XM_CALLCONV XMVector2ClampLengthV(FXMVECTOR V, FXMVECTOR LengthMin, FXMVECTOR LengthMax) noexcept; + XMVECTOR XM_CALLCONV XMVector2Reflect(FXMVECTOR Incident, FXMVECTOR Normal) noexcept; + XMVECTOR XM_CALLCONV XMVector2Refract(FXMVECTOR Incident, FXMVECTOR Normal, float RefractionIndex) noexcept; + XMVECTOR XM_CALLCONV XMVector2RefractV(FXMVECTOR Incident, FXMVECTOR Normal, FXMVECTOR RefractionIndex) noexcept; + XMVECTOR XM_CALLCONV XMVector2Orthogonal(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector2AngleBetweenNormalsEst(FXMVECTOR N1, FXMVECTOR N2) noexcept; + XMVECTOR XM_CALLCONV XMVector2AngleBetweenNormals(FXMVECTOR N1, FXMVECTOR N2) noexcept; + XMVECTOR XM_CALLCONV XMVector2AngleBetweenVectors(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector2LinePointDistance(FXMVECTOR LinePoint1, FXMVECTOR LinePoint2, FXMVECTOR Point) noexcept; + XMVECTOR XM_CALLCONV XMVector2IntersectLine(FXMVECTOR Line1Point1, FXMVECTOR Line1Point2, FXMVECTOR Line2Point1, GXMVECTOR Line2Point2) noexcept; + XMVECTOR XM_CALLCONV XMVector2Transform(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT4* XM_CALLCONV XMVector2TransformStream(_Out_writes_bytes_(sizeof(XMFLOAT4) + OutputStride * (VectorCount - 1)) XMFLOAT4* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT2) + InputStride * (VectorCount - 1)) const XMFLOAT2* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + XMVECTOR XM_CALLCONV XMVector2TransformCoord(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT2* XM_CALLCONV XMVector2TransformCoordStream(_Out_writes_bytes_(sizeof(XMFLOAT2) + OutputStride * (VectorCount - 1)) XMFLOAT2* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT2) + InputStride * (VectorCount - 1)) const XMFLOAT2* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + XMVECTOR XM_CALLCONV XMVector2TransformNormal(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT2* XM_CALLCONV XMVector2TransformNormalStream(_Out_writes_bytes_(sizeof(XMFLOAT2) + OutputStride * (VectorCount - 1)) XMFLOAT2* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT2) + InputStride * (VectorCount - 1)) const XMFLOAT2* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + + /**************************************************************************** + * + * 3D vector operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMVector3Equal(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector3EqualR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3EqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector3EqualIntR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3NearEqual(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR Epsilon) noexcept; + bool XM_CALLCONV XMVector3NotEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3NotEqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3Greater(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector3GreaterR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3GreaterOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector3GreaterOrEqualR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3Less(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3LessOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector3InBounds(FXMVECTOR V, FXMVECTOR Bounds) noexcept; + + bool XM_CALLCONV XMVector3IsNaN(FXMVECTOR V) noexcept; + bool XM_CALLCONV XMVector3IsInfinite(FXMVECTOR V) noexcept; + + XMVECTOR XM_CALLCONV XMVector3Dot(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector3Cross(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector3LengthSq(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3ReciprocalLengthEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3ReciprocalLength(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3LengthEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3Length(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3NormalizeEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3Normalize(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3ClampLength(FXMVECTOR V, float LengthMin, float LengthMax) noexcept; + XMVECTOR XM_CALLCONV XMVector3ClampLengthV(FXMVECTOR V, FXMVECTOR LengthMin, FXMVECTOR LengthMax) noexcept; + XMVECTOR XM_CALLCONV XMVector3Reflect(FXMVECTOR Incident, FXMVECTOR Normal) noexcept; + XMVECTOR XM_CALLCONV XMVector3Refract(FXMVECTOR Incident, FXMVECTOR Normal, float RefractionIndex) noexcept; + XMVECTOR XM_CALLCONV XMVector3RefractV(FXMVECTOR Incident, FXMVECTOR Normal, FXMVECTOR RefractionIndex) noexcept; + XMVECTOR XM_CALLCONV XMVector3Orthogonal(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector3AngleBetweenNormalsEst(FXMVECTOR N1, FXMVECTOR N2) noexcept; + XMVECTOR XM_CALLCONV XMVector3AngleBetweenNormals(FXMVECTOR N1, FXMVECTOR N2) noexcept; + XMVECTOR XM_CALLCONV XMVector3AngleBetweenVectors(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector3LinePointDistance(FXMVECTOR LinePoint1, FXMVECTOR LinePoint2, FXMVECTOR Point) noexcept; + void XM_CALLCONV XMVector3ComponentsFromNormal(_Out_ XMVECTOR* pParallel, _Out_ XMVECTOR* pPerpendicular, _In_ FXMVECTOR V, _In_ FXMVECTOR Normal) noexcept; + XMVECTOR XM_CALLCONV XMVector3Rotate(FXMVECTOR V, FXMVECTOR RotationQuaternion) noexcept; + XMVECTOR XM_CALLCONV XMVector3InverseRotate(FXMVECTOR V, FXMVECTOR RotationQuaternion) noexcept; + XMVECTOR XM_CALLCONV XMVector3Transform(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT4* XM_CALLCONV XMVector3TransformStream(_Out_writes_bytes_(sizeof(XMFLOAT4) + OutputStride * (VectorCount - 1)) XMFLOAT4* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT3) + InputStride * (VectorCount - 1)) const XMFLOAT3* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + XMVECTOR XM_CALLCONV XMVector3TransformCoord(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT3* XM_CALLCONV XMVector3TransformCoordStream(_Out_writes_bytes_(sizeof(XMFLOAT3) + OutputStride * (VectorCount - 1)) XMFLOAT3* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT3) + InputStride * (VectorCount - 1)) const XMFLOAT3* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + XMVECTOR XM_CALLCONV XMVector3TransformNormal(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT3* XM_CALLCONV XMVector3TransformNormalStream(_Out_writes_bytes_(sizeof(XMFLOAT3) + OutputStride * (VectorCount - 1)) XMFLOAT3* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT3) + InputStride * (VectorCount - 1)) const XMFLOAT3* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + XMVECTOR XM_CALLCONV XMVector3Project(FXMVECTOR V, float ViewportX, float ViewportY, float ViewportWidth, float ViewportHeight, float ViewportMinZ, float ViewportMaxZ, + FXMMATRIX Projection, CXMMATRIX View, CXMMATRIX World) noexcept; + XMFLOAT3* XM_CALLCONV XMVector3ProjectStream(_Out_writes_bytes_(sizeof(XMFLOAT3) + OutputStride * (VectorCount - 1)) XMFLOAT3* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT3) + InputStride * (VectorCount - 1)) const XMFLOAT3* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, + _In_ float ViewportX, _In_ float ViewportY, _In_ float ViewportWidth, _In_ float ViewportHeight, _In_ float ViewportMinZ, _In_ float ViewportMaxZ, + _In_ FXMMATRIX Projection, _In_ CXMMATRIX View, _In_ CXMMATRIX World) noexcept; + XMVECTOR XM_CALLCONV XMVector3Unproject(FXMVECTOR V, float ViewportX, float ViewportY, float ViewportWidth, float ViewportHeight, float ViewportMinZ, float ViewportMaxZ, + FXMMATRIX Projection, CXMMATRIX View, CXMMATRIX World) noexcept; + XMFLOAT3* XM_CALLCONV XMVector3UnprojectStream(_Out_writes_bytes_(sizeof(XMFLOAT3) + OutputStride * (VectorCount - 1)) XMFLOAT3* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT3) + InputStride * (VectorCount - 1)) const XMFLOAT3* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, + _In_ float ViewportX, _In_ float ViewportY, _In_ float ViewportWidth, _In_ float ViewportHeight, _In_ float ViewportMinZ, _In_ float ViewportMaxZ, + _In_ FXMMATRIX Projection, _In_ CXMMATRIX View, _In_ CXMMATRIX World) noexcept; + + /**************************************************************************** + * + * 4D vector operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMVector4Equal(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector4EqualR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4EqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector4EqualIntR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4NearEqual(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR Epsilon) noexcept; + bool XM_CALLCONV XMVector4NotEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4NotEqualInt(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4Greater(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector4GreaterR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4GreaterOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + uint32_t XM_CALLCONV XMVector4GreaterOrEqualR(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4Less(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4LessOrEqual(FXMVECTOR V1, FXMVECTOR V2) noexcept; + bool XM_CALLCONV XMVector4InBounds(FXMVECTOR V, FXMVECTOR Bounds) noexcept; + + bool XM_CALLCONV XMVector4IsNaN(FXMVECTOR V) noexcept; + bool XM_CALLCONV XMVector4IsInfinite(FXMVECTOR V) noexcept; + + XMVECTOR XM_CALLCONV XMVector4Dot(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector4Cross(FXMVECTOR V1, FXMVECTOR V2, FXMVECTOR V3) noexcept; + XMVECTOR XM_CALLCONV XMVector4LengthSq(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4ReciprocalLengthEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4ReciprocalLength(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4LengthEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4Length(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4NormalizeEst(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4Normalize(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4ClampLength(FXMVECTOR V, float LengthMin, float LengthMax) noexcept; + XMVECTOR XM_CALLCONV XMVector4ClampLengthV(FXMVECTOR V, FXMVECTOR LengthMin, FXMVECTOR LengthMax) noexcept; + XMVECTOR XM_CALLCONV XMVector4Reflect(FXMVECTOR Incident, FXMVECTOR Normal) noexcept; + XMVECTOR XM_CALLCONV XMVector4Refract(FXMVECTOR Incident, FXMVECTOR Normal, float RefractionIndex) noexcept; + XMVECTOR XM_CALLCONV XMVector4RefractV(FXMVECTOR Incident, FXMVECTOR Normal, FXMVECTOR RefractionIndex) noexcept; + XMVECTOR XM_CALLCONV XMVector4Orthogonal(FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMVector4AngleBetweenNormalsEst(FXMVECTOR N1, FXMVECTOR N2) noexcept; + XMVECTOR XM_CALLCONV XMVector4AngleBetweenNormals(FXMVECTOR N1, FXMVECTOR N2) noexcept; + XMVECTOR XM_CALLCONV XMVector4AngleBetweenVectors(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMVector4Transform(FXMVECTOR V, FXMMATRIX M) noexcept; + XMFLOAT4* XM_CALLCONV XMVector4TransformStream(_Out_writes_bytes_(sizeof(XMFLOAT4) + OutputStride * (VectorCount - 1)) XMFLOAT4* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT4) + InputStride * (VectorCount - 1)) const XMFLOAT4* pInputStream, + _In_ size_t InputStride, _In_ size_t VectorCount, _In_ FXMMATRIX M) noexcept; + + /**************************************************************************** + * + * Matrix operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMMatrixIsNaN(FXMMATRIX M) noexcept; + bool XM_CALLCONV XMMatrixIsInfinite(FXMMATRIX M) noexcept; + bool XM_CALLCONV XMMatrixIsIdentity(FXMMATRIX M) noexcept; + + XMMATRIX XM_CALLCONV XMMatrixMultiply(FXMMATRIX M1, CXMMATRIX M2) noexcept; + XMMATRIX XM_CALLCONV XMMatrixMultiplyTranspose(FXMMATRIX M1, CXMMATRIX M2) noexcept; + XMMATRIX XM_CALLCONV XMMatrixTranspose(FXMMATRIX M) noexcept; + XMMATRIX XM_CALLCONV XMMatrixInverse(_Out_opt_ XMVECTOR* pDeterminant, _In_ FXMMATRIX M) noexcept; + XMMATRIX XM_CALLCONV XMMatrixVectorTensorProduct(FXMVECTOR V1, FXMVECTOR V2) noexcept; + XMVECTOR XM_CALLCONV XMMatrixDeterminant(FXMMATRIX M) noexcept; + + _Success_(return) + bool XM_CALLCONV XMMatrixDecompose(_Out_ XMVECTOR* outScale, _Out_ XMVECTOR* outRotQuat, _Out_ XMVECTOR* outTrans, _In_ FXMMATRIX M) noexcept; + + XMMATRIX XM_CALLCONV XMMatrixIdentity() noexcept; + XMMATRIX XM_CALLCONV XMMatrixSet(float m00, float m01, float m02, float m03, + float m10, float m11, float m12, float m13, + float m20, float m21, float m22, float m23, + float m30, float m31, float m32, float m33) noexcept; + XMMATRIX XM_CALLCONV XMMatrixTranslation(float OffsetX, float OffsetY, float OffsetZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixTranslationFromVector(FXMVECTOR Offset) noexcept; + XMMATRIX XM_CALLCONV XMMatrixScaling(float ScaleX, float ScaleY, float ScaleZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixScalingFromVector(FXMVECTOR Scale) noexcept; + XMMATRIX XM_CALLCONV XMMatrixRotationX(float Angle) noexcept; + XMMATRIX XM_CALLCONV XMMatrixRotationY(float Angle) noexcept; + XMMATRIX XM_CALLCONV XMMatrixRotationZ(float Angle) noexcept; + + // Rotates about y-axis (Yaw), then x-axis (Pitch), then z-axis (Roll) + XMMATRIX XM_CALLCONV XMMatrixRotationRollPitchYaw(float Pitch, float Yaw, float Roll) noexcept; + + // Rotates about y-axis (Angles.y), then x-axis (Angles.x), then z-axis (Angles.z) + XMMATRIX XM_CALLCONV XMMatrixRotationRollPitchYawFromVector(FXMVECTOR Angles) noexcept; + + XMMATRIX XM_CALLCONV XMMatrixRotationNormal(FXMVECTOR NormalAxis, float Angle) noexcept; + XMMATRIX XM_CALLCONV XMMatrixRotationAxis(FXMVECTOR Axis, float Angle) noexcept; + XMMATRIX XM_CALLCONV XMMatrixRotationQuaternion(FXMVECTOR Quaternion) noexcept; + XMMATRIX XM_CALLCONV XMMatrixTransformation2D(FXMVECTOR ScalingOrigin, float ScalingOrientation, FXMVECTOR Scaling, + FXMVECTOR RotationOrigin, float Rotation, GXMVECTOR Translation) noexcept; + XMMATRIX XM_CALLCONV XMMatrixTransformation(FXMVECTOR ScalingOrigin, FXMVECTOR ScalingOrientationQuaternion, FXMVECTOR Scaling, + GXMVECTOR RotationOrigin, HXMVECTOR RotationQuaternion, HXMVECTOR Translation) noexcept; + XMMATRIX XM_CALLCONV XMMatrixAffineTransformation2D(FXMVECTOR Scaling, FXMVECTOR RotationOrigin, float Rotation, FXMVECTOR Translation) noexcept; + XMMATRIX XM_CALLCONV XMMatrixAffineTransformation(FXMVECTOR Scaling, FXMVECTOR RotationOrigin, FXMVECTOR RotationQuaternion, GXMVECTOR Translation) noexcept; + XMMATRIX XM_CALLCONV XMMatrixReflect(FXMVECTOR ReflectionPlane) noexcept; + XMMATRIX XM_CALLCONV XMMatrixShadow(FXMVECTOR ShadowPlane, FXMVECTOR LightPosition) noexcept; + + XMMATRIX XM_CALLCONV XMMatrixLookAtLH(FXMVECTOR EyePosition, FXMVECTOR FocusPosition, FXMVECTOR UpDirection) noexcept; + XMMATRIX XM_CALLCONV XMMatrixLookAtRH(FXMVECTOR EyePosition, FXMVECTOR FocusPosition, FXMVECTOR UpDirection) noexcept; + XMMATRIX XM_CALLCONV XMMatrixLookToLH(FXMVECTOR EyePosition, FXMVECTOR EyeDirection, FXMVECTOR UpDirection) noexcept; + XMMATRIX XM_CALLCONV XMMatrixLookToRH(FXMVECTOR EyePosition, FXMVECTOR EyeDirection, FXMVECTOR UpDirection) noexcept; + XMMATRIX XM_CALLCONV XMMatrixPerspectiveLH(float ViewWidth, float ViewHeight, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixPerspectiveRH(float ViewWidth, float ViewHeight, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixPerspectiveFovLH(float FovAngleY, float AspectRatio, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixPerspectiveFovRH(float FovAngleY, float AspectRatio, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixPerspectiveOffCenterLH(float ViewLeft, float ViewRight, float ViewBottom, float ViewTop, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixPerspectiveOffCenterRH(float ViewLeft, float ViewRight, float ViewBottom, float ViewTop, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixOrthographicLH(float ViewWidth, float ViewHeight, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixOrthographicRH(float ViewWidth, float ViewHeight, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixOrthographicOffCenterLH(float ViewLeft, float ViewRight, float ViewBottom, float ViewTop, float NearZ, float FarZ) noexcept; + XMMATRIX XM_CALLCONV XMMatrixOrthographicOffCenterRH(float ViewLeft, float ViewRight, float ViewBottom, float ViewTop, float NearZ, float FarZ) noexcept; + + + /**************************************************************************** + * + * Quaternion operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMQuaternionEqual(FXMVECTOR Q1, FXMVECTOR Q2) noexcept; + bool XM_CALLCONV XMQuaternionNotEqual(FXMVECTOR Q1, FXMVECTOR Q2) noexcept; + + bool XM_CALLCONV XMQuaternionIsNaN(FXMVECTOR Q) noexcept; + bool XM_CALLCONV XMQuaternionIsInfinite(FXMVECTOR Q) noexcept; + bool XM_CALLCONV XMQuaternionIsIdentity(FXMVECTOR Q) noexcept; + + XMVECTOR XM_CALLCONV XMQuaternionDot(FXMVECTOR Q1, FXMVECTOR Q2) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionMultiply(FXMVECTOR Q1, FXMVECTOR Q2) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionLengthSq(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionReciprocalLength(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionLength(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionNormalizeEst(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionNormalize(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionConjugate(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionInverse(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionLn(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionExp(FXMVECTOR Q) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionSlerp(FXMVECTOR Q0, FXMVECTOR Q1, float t) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionSlerpV(FXMVECTOR Q0, FXMVECTOR Q1, FXMVECTOR T) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionSquad(FXMVECTOR Q0, FXMVECTOR Q1, FXMVECTOR Q2, GXMVECTOR Q3, float t) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionSquadV(FXMVECTOR Q0, FXMVECTOR Q1, FXMVECTOR Q2, GXMVECTOR Q3, HXMVECTOR T) noexcept; + void XM_CALLCONV XMQuaternionSquadSetup(_Out_ XMVECTOR* pA, _Out_ XMVECTOR* pB, _Out_ XMVECTOR* pC, _In_ FXMVECTOR Q0, _In_ FXMVECTOR Q1, _In_ FXMVECTOR Q2, _In_ GXMVECTOR Q3) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionBaryCentric(FXMVECTOR Q0, FXMVECTOR Q1, FXMVECTOR Q2, float f, float g) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionBaryCentricV(FXMVECTOR Q0, FXMVECTOR Q1, FXMVECTOR Q2, GXMVECTOR F, HXMVECTOR G) noexcept; + + XMVECTOR XM_CALLCONV XMQuaternionIdentity() noexcept; + + // Rotates about y-axis (Yaw), then x-axis (Pitch), then z-axis (Roll) + XMVECTOR XM_CALLCONV XMQuaternionRotationRollPitchYaw(float Pitch, float Yaw, float Roll) noexcept; + + // Rotates about y-axis (Angles.y), then x-axis (Angles.x), then z-axis (Angles.z) + XMVECTOR XM_CALLCONV XMQuaternionRotationRollPitchYawFromVector(FXMVECTOR Angles) noexcept; + + XMVECTOR XM_CALLCONV XMQuaternionRotationNormal(FXMVECTOR NormalAxis, float Angle) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionRotationAxis(FXMVECTOR Axis, float Angle) noexcept; + XMVECTOR XM_CALLCONV XMQuaternionRotationMatrix(FXMMATRIX M) noexcept; + + void XM_CALLCONV XMQuaternionToAxisAngle(_Out_ XMVECTOR* pAxis, _Out_ float* pAngle, _In_ FXMVECTOR Q) noexcept; + + /**************************************************************************** + * + * Plane operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMPlaneEqual(FXMVECTOR P1, FXMVECTOR P2) noexcept; + bool XM_CALLCONV XMPlaneNearEqual(FXMVECTOR P1, FXMVECTOR P2, FXMVECTOR Epsilon) noexcept; + bool XM_CALLCONV XMPlaneNotEqual(FXMVECTOR P1, FXMVECTOR P2) noexcept; + + bool XM_CALLCONV XMPlaneIsNaN(FXMVECTOR P) noexcept; + bool XM_CALLCONV XMPlaneIsInfinite(FXMVECTOR P) noexcept; + + XMVECTOR XM_CALLCONV XMPlaneDot(FXMVECTOR P, FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMPlaneDotCoord(FXMVECTOR P, FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMPlaneDotNormal(FXMVECTOR P, FXMVECTOR V) noexcept; + XMVECTOR XM_CALLCONV XMPlaneNormalizeEst(FXMVECTOR P) noexcept; + XMVECTOR XM_CALLCONV XMPlaneNormalize(FXMVECTOR P) noexcept; + XMVECTOR XM_CALLCONV XMPlaneIntersectLine(FXMVECTOR P, FXMVECTOR LinePoint1, FXMVECTOR LinePoint2) noexcept; + void XM_CALLCONV XMPlaneIntersectPlane(_Out_ XMVECTOR* pLinePoint1, _Out_ XMVECTOR* pLinePoint2, _In_ FXMVECTOR P1, _In_ FXMVECTOR P2) noexcept; + + // Transforms a plane given an inverse transpose matrix + XMVECTOR XM_CALLCONV XMPlaneTransform(FXMVECTOR P, FXMMATRIX ITM) noexcept; + + // Transforms an array of planes given an inverse transpose matrix + XMFLOAT4* XM_CALLCONV XMPlaneTransformStream(_Out_writes_bytes_(sizeof(XMFLOAT4) + OutputStride * (PlaneCount - 1)) XMFLOAT4* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(XMFLOAT4) + InputStride * (PlaneCount - 1)) const XMFLOAT4* pInputStream, + _In_ size_t InputStride, _In_ size_t PlaneCount, _In_ FXMMATRIX ITM) noexcept; + + XMVECTOR XM_CALLCONV XMPlaneFromPointNormal(FXMVECTOR Point, FXMVECTOR Normal) noexcept; + XMVECTOR XM_CALLCONV XMPlaneFromPoints(FXMVECTOR Point1, FXMVECTOR Point2, FXMVECTOR Point3) noexcept; + + /**************************************************************************** + * + * Color operations + * + ****************************************************************************/ + + bool XM_CALLCONV XMColorEqual(FXMVECTOR C1, FXMVECTOR C2) noexcept; + bool XM_CALLCONV XMColorNotEqual(FXMVECTOR C1, FXMVECTOR C2) noexcept; + bool XM_CALLCONV XMColorGreater(FXMVECTOR C1, FXMVECTOR C2) noexcept; + bool XM_CALLCONV XMColorGreaterOrEqual(FXMVECTOR C1, FXMVECTOR C2) noexcept; + bool XM_CALLCONV XMColorLess(FXMVECTOR C1, FXMVECTOR C2) noexcept; + bool XM_CALLCONV XMColorLessOrEqual(FXMVECTOR C1, FXMVECTOR C2) noexcept; + + bool XM_CALLCONV XMColorIsNaN(FXMVECTOR C) noexcept; + bool XM_CALLCONV XMColorIsInfinite(FXMVECTOR C) noexcept; + + XMVECTOR XM_CALLCONV XMColorNegative(FXMVECTOR C) noexcept; + XMVECTOR XM_CALLCONV XMColorModulate(FXMVECTOR C1, FXMVECTOR C2) noexcept; + XMVECTOR XM_CALLCONV XMColorAdjustSaturation(FXMVECTOR C, float Saturation) noexcept; + XMVECTOR XM_CALLCONV XMColorAdjustContrast(FXMVECTOR C, float Contrast) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToHSL(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorHSLToRGB(FXMVECTOR hsl) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToHSV(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorHSVToRGB(FXMVECTOR hsv) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToYUV(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorYUVToRGB(FXMVECTOR yuv) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToYUV_HD(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorYUVToRGB_HD(FXMVECTOR yuv) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToYUV_UHD(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorYUVToRGB_UHD(FXMVECTOR yuv) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToXYZ(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorXYZToRGB(FXMVECTOR xyz) noexcept; + + XMVECTOR XM_CALLCONV XMColorXYZToSRGB(FXMVECTOR xyz) noexcept; + XMVECTOR XM_CALLCONV XMColorSRGBToXYZ(FXMVECTOR srgb) noexcept; + + XMVECTOR XM_CALLCONV XMColorRGBToSRGB(FXMVECTOR rgb) noexcept; + XMVECTOR XM_CALLCONV XMColorSRGBToRGB(FXMVECTOR srgb) noexcept; + + + /**************************************************************************** + * + * Miscellaneous operations + * + ****************************************************************************/ + + bool XMVerifyCPUSupport() noexcept; + + XMVECTOR XM_CALLCONV XMFresnelTerm(FXMVECTOR CosIncidentAngle, FXMVECTOR RefractionIndex) noexcept; + + bool XMScalarNearEqual(float S1, float S2, float Epsilon) noexcept; + float XMScalarModAngle(float Value) noexcept; + + float XMScalarSin(float Value) noexcept; + float XMScalarSinEst(float Value) noexcept; + + float XMScalarCos(float Value) noexcept; + float XMScalarCosEst(float Value) noexcept; + + void XMScalarSinCos(_Out_ float* pSin, _Out_ float* pCos, float Value) noexcept; + void XMScalarSinCosEst(_Out_ float* pSin, _Out_ float* pCos, float Value) noexcept; + + float XMScalarASin(float Value) noexcept; + float XMScalarASinEst(float Value) noexcept; + + float XMScalarACos(float Value) noexcept; + float XMScalarACosEst(float Value) noexcept; + + /**************************************************************************** + * + * Templates + * + ****************************************************************************/ + +#if defined(__XNAMATH_H__) && defined(XMMin) +#undef XMMin +#undef XMMax +#endif + + template inline T XMMin(T a, T b) noexcept { return (a < b) ? a : b; } + template inline T XMMax(T a, T b) noexcept { return (a > b) ? a : b; } + + //------------------------------------------------------------------------------ + +#if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + +// PermuteHelper internal template (SSE only) + namespace MathInternal + { + // Slow path fallback for permutes that do not map to a single SSE shuffle opcode. + template struct PermuteHelper + { + static XMVECTOR XM_CALLCONV Permute(FXMVECTOR v1, FXMVECTOR v2) noexcept + { + static const XMVECTORU32 selectMask = + { { { + WhichX ? 0xFFFFFFFF : 0, + WhichY ? 0xFFFFFFFF : 0, + WhichZ ? 0xFFFFFFFF : 0, + WhichW ? 0xFFFFFFFF : 0, + } } }; + + XMVECTOR shuffled1 = XM_PERMUTE_PS(v1, Shuffle); + XMVECTOR shuffled2 = XM_PERMUTE_PS(v2, Shuffle); + + XMVECTOR masked1 = _mm_andnot_ps(selectMask, shuffled1); + XMVECTOR masked2 = _mm_and_ps(selectMask, shuffled2); + + return _mm_or_ps(masked1, masked2); + } + }; + + // Fast path for permutes that only read from the first vector. + template struct PermuteHelper + { + static XMVECTOR XM_CALLCONV Permute(FXMVECTOR v1, FXMVECTOR) noexcept { return XM_PERMUTE_PS(v1, Shuffle); } + }; + + // Fast path for permutes that only read from the second vector. + template struct PermuteHelper + { + static XMVECTOR XM_CALLCONV Permute(FXMVECTOR, FXMVECTOR v2) noexcept { return XM_PERMUTE_PS(v2, Shuffle); } + }; + + // Fast path for permutes that read XY from the first vector, ZW from the second. + template struct PermuteHelper + { + static XMVECTOR XM_CALLCONV Permute(FXMVECTOR v1, FXMVECTOR v2) noexcept { return _mm_shuffle_ps(v1, v2, Shuffle); } + }; + + // Fast path for permutes that read XY from the second vector, ZW from the first. + template struct PermuteHelper + { + static XMVECTOR XM_CALLCONV Permute(FXMVECTOR v1, FXMVECTOR v2) noexcept { return _mm_shuffle_ps(v2, v1, Shuffle); } + }; + } + +#endif // _XM_SSE_INTRINSICS_ && !_XM_NO_INTRINSICS_ + + // General permute template + template + inline XMVECTOR XM_CALLCONV XMVectorPermute(FXMVECTOR V1, FXMVECTOR V2) noexcept + { + static_assert(PermuteX <= 7, "PermuteX template parameter out of range"); + static_assert(PermuteY <= 7, "PermuteY template parameter out of range"); + static_assert(PermuteZ <= 7, "PermuteZ template parameter out of range"); + static_assert(PermuteW <= 7, "PermuteW template parameter out of range"); + + #if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + constexpr uint32_t Shuffle = _MM_SHUFFLE(PermuteW & 3, PermuteZ & 3, PermuteY & 3, PermuteX & 3); + + constexpr bool WhichX = PermuteX > 3; + constexpr bool WhichY = PermuteY > 3; + constexpr bool WhichZ = PermuteZ > 3; + constexpr bool WhichW = PermuteW > 3; + + return MathInternal::PermuteHelper::Permute(V1, V2); + #else + + return XMVectorPermute(V1, V2, PermuteX, PermuteY, PermuteZ, PermuteW); + + #endif + } + + // Special-case permute templates + template<> constexpr XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 2, 3>(FXMVECTOR V1, FXMVECTOR) noexcept { return V1; } + template<> constexpr XMVECTOR XM_CALLCONV XMVectorPermute<4, 5, 6, 7>(FXMVECTOR, FXMVECTOR V2) noexcept { return V2; } + +#if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 4, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_movelh_ps(V1, V2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<6, 7, 2, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_movehl_ps(V1, V2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 4, 1, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_unpacklo_ps(V1, V2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 6, 3, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_unpackhi_ps(V1, V2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 3, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_castpd_ps(_mm_unpackhi_pd(_mm_castps_pd(V1), _mm_castps_pd(V2))); } +#endif + +#if defined(_XM_SSE4_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 1, 2, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x1); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 5, 2, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 5, 2, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x3); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 6, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x4); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 1, 6, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x5); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 5, 6, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x6); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 5, 6, 3>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x7); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 2, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x8); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 1, 2, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0x9); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 5, 2, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0xA); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 5, 2, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0xB); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0xC); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<4, 1, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0xD); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 5, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return _mm_blend_ps(V1, V2, 0xE); } +#endif + +#if defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + + // If the indices are all in the range 0-3 or 4-7, then use XMVectorSwizzle instead + // The mirror cases are not spelled out here as the programmer can always swap the arguments + // (i.e. prefer permutes where the X element comes from the V1 vector instead of the V2 vector) + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 4, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_low_f32(V1), vget_low_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 0, 4, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_low_f32(V1)), vget_low_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 5, 4>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_low_f32(V1), vrev64_f32(vget_low_f32(V2))); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 0, 5, 4>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_low_f32(V1)), vrev64_f32(vget_low_f32(V2))); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 3, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_high_f32(V1), vget_high_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<3, 2, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_high_f32(V1)), vget_high_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 3, 7, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_high_f32(V1), vrev64_f32(vget_high_f32(V2))); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<3, 2, 7, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_high_f32(V1)), vrev64_f32(vget_high_f32(V2))); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_low_f32(V1), vget_high_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 0, 6, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_low_f32(V1)), vget_high_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 1, 7, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_low_f32(V1), vrev64_f32(vget_high_f32(V2))); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 0, 7, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_low_f32(V1)), vrev64_f32(vget_high_f32(V2))); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<3, 2, 4, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_high_f32(V1)), vget_low_f32(V2)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 3, 5, 4>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vget_high_f32(V1), vrev64_f32(vget_low_f32(V2))); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<3, 2, 5, 4>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vcombine_f32(vrev64_f32(vget_high_f32(V1)), vrev64_f32(vget_low_f32(V2))); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 4, 2, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vtrnq_f32(V1, V2).val[0]; } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 5, 3, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vtrnq_f32(V1, V2).val[1]; } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 4, 1, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vzipq_f32(V1, V2).val[0]; } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 6, 3, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vzipq_f32(V1, V2).val[1]; } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<0, 2, 4, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vuzpq_f32(V1, V2).val[0]; } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 3, 5, 7>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vuzpq_f32(V1, V2).val[1]; } + + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<1, 2, 3, 4>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vextq_f32(V1, V2, 1); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<2, 3, 4, 5>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vextq_f32(V1, V2, 2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorPermute<3, 4, 5, 6>(FXMVECTOR V1, FXMVECTOR V2) noexcept { return vextq_f32(V1, V2, 3); } + +#endif // _XM_ARM_NEON_INTRINSICS_ && !_XM_NO_INTRINSICS_ + + //------------------------------------------------------------------------------ + + // General swizzle template + template + inline XMVECTOR XM_CALLCONV XMVectorSwizzle(FXMVECTOR V) noexcept + { + static_assert(SwizzleX <= 3, "SwizzleX template parameter out of range"); + static_assert(SwizzleY <= 3, "SwizzleY template parameter out of range"); + static_assert(SwizzleZ <= 3, "SwizzleZ template parameter out of range"); + static_assert(SwizzleW <= 3, "SwizzleW template parameter out of range"); + + #if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + return XM_PERMUTE_PS(V, _MM_SHUFFLE(SwizzleW, SwizzleZ, SwizzleY, SwizzleX)); + #else + + return XMVectorSwizzle(V, SwizzleX, SwizzleY, SwizzleZ, SwizzleW); + + #endif + } + + // Specialized swizzles + template<> constexpr XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 1, 2, 3>(FXMVECTOR V) noexcept { return V; } + +#if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 1, 0, 1>(FXMVECTOR V) noexcept { return _mm_movelh_ps(V, V); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 3, 2, 3>(FXMVECTOR V) noexcept { return _mm_movehl_ps(V, V); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 0, 1, 1>(FXMVECTOR V) noexcept { return _mm_unpacklo_ps(V, V); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 2, 3, 3>(FXMVECTOR V) noexcept { return _mm_unpackhi_ps(V, V); } +#endif + +#if defined(_XM_SSE3_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 0, 2, 2>(FXMVECTOR V) noexcept { return _mm_moveldup_ps(V); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 1, 3, 3>(FXMVECTOR V) noexcept { return _mm_movehdup_ps(V); } +#endif + +#if defined(_XM_AVX2_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) && defined(_XM_FAVOR_INTEL_) + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 0, 0, 0>(FXMVECTOR V) noexcept { return _mm_broadcastss_ps(V); } +#endif + +#if defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 0, 0, 0>(FXMVECTOR V) noexcept { return vdupq_lane_f32(vget_low_f32(V), 0); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 1, 1, 1>(FXMVECTOR V) noexcept { return vdupq_lane_f32(vget_low_f32(V), 1); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 2, 2, 2>(FXMVECTOR V) noexcept { return vdupq_lane_f32(vget_high_f32(V), 0); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<3, 3, 3, 3>(FXMVECTOR V) noexcept { return vdupq_lane_f32(vget_high_f32(V), 1); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 0, 3, 2>(FXMVECTOR V) noexcept { return vrev64q_f32(V); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 1, 0, 1>(FXMVECTOR V) noexcept { float32x2_t vt = vget_low_f32(V); return vcombine_f32(vt, vt); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 3, 2, 3>(FXMVECTOR V) noexcept { float32x2_t vt = vget_high_f32(V); return vcombine_f32(vt, vt); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 0, 1, 0>(FXMVECTOR V) noexcept { float32x2_t vt = vrev64_f32(vget_low_f32(V)); return vcombine_f32(vt, vt); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<3, 2, 3, 2>(FXMVECTOR V) noexcept { float32x2_t vt = vrev64_f32(vget_high_f32(V)); return vcombine_f32(vt, vt); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 1, 3, 2>(FXMVECTOR V) noexcept { return vcombine_f32(vget_low_f32(V), vrev64_f32(vget_high_f32(V))); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 0, 2, 3>(FXMVECTOR V) noexcept { return vcombine_f32(vrev64_f32(vget_low_f32(V)), vget_high_f32(V)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 3, 1, 0>(FXMVECTOR V) noexcept { return vcombine_f32(vget_high_f32(V), vrev64_f32(vget_low_f32(V))); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<3, 2, 0, 1>(FXMVECTOR V) noexcept { return vcombine_f32(vrev64_f32(vget_high_f32(V)), vget_low_f32(V)); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<3, 2, 1, 0>(FXMVECTOR V) noexcept { return vcombine_f32(vrev64_f32(vget_high_f32(V)), vrev64_f32(vget_low_f32(V))); } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 0, 2, 2>(FXMVECTOR V) noexcept { return vtrnq_f32(V, V).val[0]; } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 1, 3, 3>(FXMVECTOR V) noexcept { return vtrnq_f32(V, V).val[1]; } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 0, 1, 1>(FXMVECTOR V) noexcept { return vzipq_f32(V, V).val[0]; } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 2, 3, 3>(FXMVECTOR V) noexcept { return vzipq_f32(V, V).val[1]; } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<0, 2, 0, 2>(FXMVECTOR V) noexcept { return vuzpq_f32(V, V).val[0]; } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 3, 1, 3>(FXMVECTOR V) noexcept { return vuzpq_f32(V, V).val[1]; } + + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<1, 2, 3, 0>(FXMVECTOR V) noexcept { return vextq_f32(V, V, 1); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<2, 3, 0, 1>(FXMVECTOR V) noexcept { return vextq_f32(V, V, 2); } + template<> inline XMVECTOR XM_CALLCONV XMVectorSwizzle<3, 0, 1, 2>(FXMVECTOR V) noexcept { return vextq_f32(V, V, 3); } + +#endif // _XM_ARM_NEON_INTRINSICS_ && !_XM_NO_INTRINSICS_ + + //------------------------------------------------------------------------------ + + template + inline XMVECTOR XM_CALLCONV XMVectorShiftLeft(FXMVECTOR V1, FXMVECTOR V2) noexcept + { + static_assert(Elements < 4, "Elements template parameter out of range"); + return XMVectorPermute(V1, V2); + } + + template + inline XMVECTOR XM_CALLCONV XMVectorRotateLeft(FXMVECTOR V) noexcept + { + static_assert(Elements < 4, "Elements template parameter out of range"); + return XMVectorSwizzle(V); + } + + template + inline XMVECTOR XM_CALLCONV XMVectorRotateRight(FXMVECTOR V) noexcept + { + static_assert(Elements < 4, "Elements template parameter out of range"); + return XMVectorSwizzle<(4 - Elements) & 3, (5 - Elements) & 3, (6 - Elements) & 3, (7 - Elements) & 3>(V); + } + + template + inline XMVECTOR XM_CALLCONV XMVectorInsert(FXMVECTOR VD, FXMVECTOR VS) noexcept + { + XMVECTOR Control = XMVectorSelectControl(Select0 & 1, Select1 & 1, Select2 & 1, Select3 & 1); + return XMVectorSelect(VD, XMVectorRotateLeft(VS), Control); + } + + /**************************************************************************** + * + * Globals + * + ****************************************************************************/ + + // The purpose of the following global constants is to prevent redundant + // reloading of the constants when they are referenced by more than one + // separate inline math routine called within the same function. Declaring + // a constant locally within a routine is sufficient to prevent redundant + // reloads of that constant when that single routine is called multiple + // times in a function, but if the constant is used (and declared) in a + // separate math routine it would be reloaded. + +#ifndef XMGLOBALCONST +#if __cplusplus >= 201703L +#define XMGLOBALCONST inline constexpr +#elif defined(__GNUC__) && !defined(__MINGW32__) +#define XMGLOBALCONST extern const __attribute__((weak)) +#else +#define XMGLOBALCONST extern const __declspec(selectany) +#endif +#endif + + XMGLOBALCONST XMVECTORF32 g_XMSinCoefficients0 = { { { -0.16666667f, +0.0083333310f, -0.00019840874f, +2.7525562e-06f } } }; + XMGLOBALCONST XMVECTORF32 g_XMSinCoefficients1 = { { { -2.3889859e-08f, -0.16665852f /*Est1*/, +0.0083139502f /*Est2*/, -0.00018524670f /*Est3*/ } } }; + XMGLOBALCONST XMVECTORF32 g_XMCosCoefficients0 = { { { -0.5f, +0.041666638f, -0.0013888378f, +2.4760495e-05f } } }; + XMGLOBALCONST XMVECTORF32 g_XMCosCoefficients1 = { { { -2.6051615e-07f, -0.49992746f /*Est1*/, +0.041493919f /*Est2*/, -0.0012712436f /*Est3*/ } } }; + XMGLOBALCONST XMVECTORF32 g_XMTanCoefficients0 = { { { 1.0f, 0.333333333f, 0.133333333f, 5.396825397e-2f } } }; + XMGLOBALCONST XMVECTORF32 g_XMTanCoefficients1 = { { { 2.186948854e-2f, 8.863235530e-3f, 3.592128167e-3f, 1.455834485e-3f } } }; + XMGLOBALCONST XMVECTORF32 g_XMTanCoefficients2 = { { { 5.900274264e-4f, 2.391290764e-4f, 9.691537707e-5f, 3.927832950e-5f } } }; + XMGLOBALCONST XMVECTORF32 g_XMArcCoefficients0 = { { { +1.5707963050f, -0.2145988016f, +0.0889789874f, -0.0501743046f } } }; + XMGLOBALCONST XMVECTORF32 g_XMArcCoefficients1 = { { { +0.0308918810f, -0.0170881256f, +0.0066700901f, -0.0012624911f } } }; + XMGLOBALCONST XMVECTORF32 g_XMATanCoefficients0 = { { { -0.3333314528f, +0.1999355085f, -0.1420889944f, +0.1065626393f } } }; + XMGLOBALCONST XMVECTORF32 g_XMATanCoefficients1 = { { { -0.0752896400f, +0.0429096138f, -0.0161657367f, +0.0028662257f } } }; + XMGLOBALCONST XMVECTORF32 g_XMATanEstCoefficients0 = { { { +0.999866f, +0.999866f, +0.999866f, +0.999866f } } }; + XMGLOBALCONST XMVECTORF32 g_XMATanEstCoefficients1 = { { { -0.3302995f, +0.180141f, -0.085133f, +0.0208351f } } }; + XMGLOBALCONST XMVECTORF32 g_XMTanEstCoefficients = { { { 2.484f, -1.954923183e-1f, 2.467401101f, XM_1DIVPI } } }; + XMGLOBALCONST XMVECTORF32 g_XMArcEstCoefficients = { { { +1.5707288f, -0.2121144f, +0.0742610f, -0.0187293f } } }; + XMGLOBALCONST XMVECTORF32 g_XMPiConstants0 = { { { XM_PI, XM_2PI, XM_1DIVPI, XM_1DIV2PI } } }; + XMGLOBALCONST XMVECTORF32 g_XMIdentityR0 = { { { 1.0f, 0.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMIdentityR1 = { { { 0.0f, 1.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMIdentityR2 = { { { 0.0f, 0.0f, 1.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMIdentityR3 = { { { 0.0f, 0.0f, 0.0f, 1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegIdentityR0 = { { { -1.0f, 0.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegIdentityR1 = { { { 0.0f, -1.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegIdentityR2 = { { { 0.0f, 0.0f, -1.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegIdentityR3 = { { { 0.0f, 0.0f, 0.0f, -1.0f } } }; + XMGLOBALCONST XMVECTORU32 g_XMNegativeZero = { { { 0x80000000, 0x80000000, 0x80000000, 0x80000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMNegate3 = { { { 0x80000000, 0x80000000, 0x80000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskXY = { { { 0xFFFFFFFF, 0xFFFFFFFF, 0x00000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMMask3 = { { { 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskX = { { { 0xFFFFFFFF, 0x00000000, 0x00000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskY = { { { 0x00000000, 0xFFFFFFFF, 0x00000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskZ = { { { 0x00000000, 0x00000000, 0xFFFFFFFF, 0x00000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskW = { { { 0x00000000, 0x00000000, 0x00000000, 0xFFFFFFFF } } }; + XMGLOBALCONST XMVECTORF32 g_XMOne = { { { 1.0f, 1.0f, 1.0f, 1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMOne3 = { { { 1.0f, 1.0f, 1.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMZero = { { { 0.0f, 0.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMTwo = { { { 2.f, 2.f, 2.f, 2.f } } }; + XMGLOBALCONST XMVECTORF32 g_XMFour = { { { 4.f, 4.f, 4.f, 4.f } } }; + XMGLOBALCONST XMVECTORF32 g_XMSix = { { { 6.f, 6.f, 6.f, 6.f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegativeOne = { { { -1.0f, -1.0f, -1.0f, -1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMOneHalf = { { { 0.5f, 0.5f, 0.5f, 0.5f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegativeOneHalf = { { { -0.5f, -0.5f, -0.5f, -0.5f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegativeTwoPi = { { { -XM_2PI, -XM_2PI, -XM_2PI, -XM_2PI } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegativePi = { { { -XM_PI, -XM_PI, -XM_PI, -XM_PI } } }; + XMGLOBALCONST XMVECTORF32 g_XMHalfPi = { { { XM_PIDIV2, XM_PIDIV2, XM_PIDIV2, XM_PIDIV2 } } }; + XMGLOBALCONST XMVECTORF32 g_XMPi = { { { XM_PI, XM_PI, XM_PI, XM_PI } } }; + XMGLOBALCONST XMVECTORF32 g_XMReciprocalPi = { { { XM_1DIVPI, XM_1DIVPI, XM_1DIVPI, XM_1DIVPI } } }; + XMGLOBALCONST XMVECTORF32 g_XMTwoPi = { { { XM_2PI, XM_2PI, XM_2PI, XM_2PI } } }; + XMGLOBALCONST XMVECTORF32 g_XMReciprocalTwoPi = { { { XM_1DIV2PI, XM_1DIV2PI, XM_1DIV2PI, XM_1DIV2PI } } }; + XMGLOBALCONST XMVECTORF32 g_XMEpsilon = { { { 1.192092896e-7f, 1.192092896e-7f, 1.192092896e-7f, 1.192092896e-7f } } }; + XMGLOBALCONST XMVECTORI32 g_XMInfinity = { { { 0x7F800000, 0x7F800000, 0x7F800000, 0x7F800000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMQNaN = { { { 0x7FC00000, 0x7FC00000, 0x7FC00000, 0x7FC00000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMQNaNTest = { { { 0x007FFFFF, 0x007FFFFF, 0x007FFFFF, 0x007FFFFF } } }; + XMGLOBALCONST XMVECTORI32 g_XMAbsMask = { { { 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF } } }; + XMGLOBALCONST XMVECTORI32 g_XMFltMin = { { { 0x00800000, 0x00800000, 0x00800000, 0x00800000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMFltMax = { { { 0x7F7FFFFF, 0x7F7FFFFF, 0x7F7FFFFF, 0x7F7FFFFF } } }; + XMGLOBALCONST XMVECTORU32 g_XMNegOneMask = { { { 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskA8R8G8B8 = { { { 0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipA8R8G8B8 = { { { 0x00000000, 0x00000000, 0x00000000, 0x80000000 } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixAA8R8G8B8 = { { { 0.0f, 0.0f, 0.0f, float(0x80000000U) } } }; + XMGLOBALCONST XMVECTORF32 g_XMNormalizeA8R8G8B8 = { { { 1.0f / (255.0f * float(0x10000)), 1.0f / (255.0f * float(0x100)), 1.0f / 255.0f, 1.0f / (255.0f * float(0x1000000)) } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskA2B10G10R10 = { { { 0x000003FF, 0x000FFC00, 0x3FF00000, 0xC0000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipA2B10G10R10 = { { { 0x00000200, 0x00080000, 0x20000000, 0x80000000 } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixAA2B10G10R10 = { { { -512.0f, -512.0f * float(0x400), -512.0f * float(0x100000), float(0x80000000U) } } }; + XMGLOBALCONST XMVECTORF32 g_XMNormalizeA2B10G10R10 = { { { 1.0f / 511.0f, 1.0f / (511.0f * float(0x400)), 1.0f / (511.0f * float(0x100000)), 1.0f / (3.0f * float(0x40000000)) } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskX16Y16 = { { { 0x0000FFFF, 0xFFFF0000, 0x00000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMFlipX16Y16 = { { { 0x00008000, 0x00000000, 0x00000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixX16Y16 = { { { -32768.0f, 0.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNormalizeX16Y16 = { { { 1.0f / 32767.0f, 1.0f / (32767.0f * 65536.0f), 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskX16Y16Z16W16 = { { { 0x0000FFFF, 0x0000FFFF, 0xFFFF0000, 0xFFFF0000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMFlipX16Y16Z16W16 = { { { 0x00008000, 0x00008000, 0x00000000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixX16Y16Z16W16 = { { { -32768.0f, -32768.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNormalizeX16Y16Z16W16 = { { { 1.0f / 32767.0f, 1.0f / 32767.0f, 1.0f / (32767.0f * 65536.0f), 1.0f / (32767.0f * 65536.0f) } } }; + XMGLOBALCONST XMVECTORF32 g_XMNoFraction = { { { 8388608.0f, 8388608.0f, 8388608.0f, 8388608.0f } } }; + XMGLOBALCONST XMVECTORI32 g_XMMaskByte = { { { 0x000000FF, 0x000000FF, 0x000000FF, 0x000000FF } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegateX = { { { -1.0f, 1.0f, 1.0f, 1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegateY = { { { 1.0f, -1.0f, 1.0f, 1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegateZ = { { { 1.0f, 1.0f, -1.0f, 1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMNegateW = { { { 1.0f, 1.0f, 1.0f, -1.0f } } }; + XMGLOBALCONST XMVECTORU32 g_XMSelect0101 = { { { XM_SELECT_0, XM_SELECT_1, XM_SELECT_0, XM_SELECT_1 } } }; + XMGLOBALCONST XMVECTORU32 g_XMSelect1010 = { { { XM_SELECT_1, XM_SELECT_0, XM_SELECT_1, XM_SELECT_0 } } }; + XMGLOBALCONST XMVECTORI32 g_XMOneHalfMinusEpsilon = { { { 0x3EFFFFFD, 0x3EFFFFFD, 0x3EFFFFFD, 0x3EFFFFFD } } }; + XMGLOBALCONST XMVECTORU32 g_XMSelect1000 = { { { XM_SELECT_1, XM_SELECT_0, XM_SELECT_0, XM_SELECT_0 } } }; + XMGLOBALCONST XMVECTORU32 g_XMSelect1100 = { { { XM_SELECT_1, XM_SELECT_1, XM_SELECT_0, XM_SELECT_0 } } }; + XMGLOBALCONST XMVECTORU32 g_XMSelect1110 = { { { XM_SELECT_1, XM_SELECT_1, XM_SELECT_1, XM_SELECT_0 } } }; + XMGLOBALCONST XMVECTORU32 g_XMSelect1011 = { { { XM_SELECT_1, XM_SELECT_0, XM_SELECT_1, XM_SELECT_1 } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixupY16 = { { { 1.0f, 1.0f / 65536.0f, 0.0f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixupY16W16 = { { { 1.0f, 1.0f, 1.0f / 65536.0f, 1.0f / 65536.0f } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipY = { { { 0, 0x80000000, 0, 0 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipZ = { { { 0, 0, 0x80000000, 0 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipW = { { { 0, 0, 0, 0x80000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipYZ = { { { 0, 0x80000000, 0x80000000, 0 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipZW = { { { 0, 0, 0x80000000, 0x80000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMFlipYW = { { { 0, 0x80000000, 0, 0x80000000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMMaskDec4 = { { { 0x3FF, 0x3FF << 10, 0x3FF << 20, static_cast(0xC0000000) } } }; + XMGLOBALCONST XMVECTORI32 g_XMXorDec4 = { { { 0x200, 0x200 << 10, 0x200 << 20, 0 } } }; + XMGLOBALCONST XMVECTORF32 g_XMAddUDec4 = { { { 0, 0, 0, 32768.0f * 65536.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMAddDec4 = { { { -512.0f, -512.0f * 1024.0f, -512.0f * 1024.0f * 1024.0f, 0 } } }; + XMGLOBALCONST XMVECTORF32 g_XMMulDec4 = { { { 1.0f, 1.0f / 1024.0f, 1.0f / (1024.0f * 1024.0f), 1.0f / (1024.0f * 1024.0f * 1024.0f) } } }; + XMGLOBALCONST XMVECTORU32 g_XMMaskByte4 = { { { 0xFF, 0xFF00, 0xFF0000, 0xFF000000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMXorByte4 = { { { 0x80, 0x8000, 0x800000, 0x00000000 } } }; + XMGLOBALCONST XMVECTORF32 g_XMAddByte4 = { { { -128.0f, -128.0f * 256.0f, -128.0f * 65536.0f, 0 } } }; + XMGLOBALCONST XMVECTORF32 g_XMFixUnsigned = { { { 32768.0f * 65536.0f, 32768.0f * 65536.0f, 32768.0f * 65536.0f, 32768.0f * 65536.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMMaxInt = { { { 65536.0f * 32768.0f - 128.0f, 65536.0f * 32768.0f - 128.0f, 65536.0f * 32768.0f - 128.0f, 65536.0f * 32768.0f - 128.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMMaxUInt = { { { 65536.0f * 65536.0f - 256.0f, 65536.0f * 65536.0f - 256.0f, 65536.0f * 65536.0f - 256.0f, 65536.0f * 65536.0f - 256.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMUnsignedFix = { { { 32768.0f * 65536.0f, 32768.0f * 65536.0f, 32768.0f * 65536.0f, 32768.0f * 65536.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMsrgbScale = { { { 12.92f, 12.92f, 12.92f, 1.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMsrgbA = { { { 0.055f, 0.055f, 0.055f, 0.0f } } }; + XMGLOBALCONST XMVECTORF32 g_XMsrgbA1 = { { { 1.055f, 1.055f, 1.055f, 1.0f } } }; + XMGLOBALCONST XMVECTORI32 g_XMExponentBias = { { { 127, 127, 127, 127 } } }; + XMGLOBALCONST XMVECTORI32 g_XMSubnormalExponent = { { { -126, -126, -126, -126 } } }; + XMGLOBALCONST XMVECTORI32 g_XMNumTrailing = { { { 23, 23, 23, 23 } } }; + XMGLOBALCONST XMVECTORI32 g_XMMinNormal = { { { 0x00800000, 0x00800000, 0x00800000, 0x00800000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMNegInfinity = { { { 0xFF800000, 0xFF800000, 0xFF800000, 0xFF800000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMNegQNaN = { { { 0xFFC00000, 0xFFC00000, 0xFFC00000, 0xFFC00000 } } }; + XMGLOBALCONST XMVECTORI32 g_XMBin128 = { { { 0x43000000, 0x43000000, 0x43000000, 0x43000000 } } }; + XMGLOBALCONST XMVECTORU32 g_XMBinNeg150 = { { { 0xC3160000, 0xC3160000, 0xC3160000, 0xC3160000 } } }; + XMGLOBALCONST XMVECTORI32 g_XM253 = { { { 253, 253, 253, 253 } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst1 = { { { -6.93147182e-1f, -6.93147182e-1f, -6.93147182e-1f, -6.93147182e-1f } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst2 = { { { +2.40226462e-1f, +2.40226462e-1f, +2.40226462e-1f, +2.40226462e-1f } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst3 = { { { -5.55036440e-2f, -5.55036440e-2f, -5.55036440e-2f, -5.55036440e-2f } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst4 = { { { +9.61597636e-3f, +9.61597636e-3f, +9.61597636e-3f, +9.61597636e-3f } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst5 = { { { -1.32823968e-3f, -1.32823968e-3f, -1.32823968e-3f, -1.32823968e-3f } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst6 = { { { +1.47491097e-4f, +1.47491097e-4f, +1.47491097e-4f, +1.47491097e-4f } } }; + XMGLOBALCONST XMVECTORF32 g_XMExpEst7 = { { { -1.08635004e-5f, -1.08635004e-5f, -1.08635004e-5f, -1.08635004e-5f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst0 = { { { +1.442693f, +1.442693f, +1.442693f, +1.442693f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst1 = { { { -0.721242f, -0.721242f, -0.721242f, -0.721242f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst2 = { { { +0.479384f, +0.479384f, +0.479384f, +0.479384f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst3 = { { { -0.350295f, -0.350295f, -0.350295f, -0.350295f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst4 = { { { +0.248590f, +0.248590f, +0.248590f, +0.248590f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst5 = { { { -0.145700f, -0.145700f, -0.145700f, -0.145700f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst6 = { { { +0.057148f, +0.057148f, +0.057148f, +0.057148f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLogEst7 = { { { -0.010578f, -0.010578f, -0.010578f, -0.010578f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLgE = { { { +1.442695f, +1.442695f, +1.442695f, +1.442695f } } }; + XMGLOBALCONST XMVECTORF32 g_XMInvLgE = { { { +6.93147182e-1f, +6.93147182e-1f, +6.93147182e-1f, +6.93147182e-1f } } }; + XMGLOBALCONST XMVECTORF32 g_XMLg10 = { { { +3.321928f, +3.321928f, +3.321928f, +3.321928f } } }; + XMGLOBALCONST XMVECTORF32 g_XMInvLg10 = { { { +3.010299956e-1f, +3.010299956e-1f, +3.010299956e-1f, +3.010299956e-1f } } }; + XMGLOBALCONST XMVECTORF32 g_UByteMax = { { { 255.0f, 255.0f, 255.0f, 255.0f } } }; + XMGLOBALCONST XMVECTORF32 g_ByteMin = { { { -127.0f, -127.0f, -127.0f, -127.0f } } }; + XMGLOBALCONST XMVECTORF32 g_ByteMax = { { { 127.0f, 127.0f, 127.0f, 127.0f } } }; + XMGLOBALCONST XMVECTORF32 g_ShortMin = { { { -32767.0f, -32767.0f, -32767.0f, -32767.0f } } }; + XMGLOBALCONST XMVECTORF32 g_ShortMax = { { { 32767.0f, 32767.0f, 32767.0f, 32767.0f } } }; + XMGLOBALCONST XMVECTORF32 g_UShortMax = { { { 65535.0f, 65535.0f, 65535.0f, 65535.0f } } }; + + /**************************************************************************** + * + * Implementation + * + ****************************************************************************/ + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable:4068 4214 4204 4365 4616 4640 6001 6101) + // C4068/4616: ignore unknown pragmas + // C4214/4204: nonstandard extension used + // C4365/4640: Off by default noise + // C6001/6101: False positives +#endif + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 25000, "FXMVECTOR is 16 bytes") +#pragma prefast(disable : 26495, "Union initialization confuses /analyze") +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wfloat-equal" +#pragma clang diagnostic ignored "-Wundefined-reinterpret-cast" +#pragma clang diagnostic ignored "-Wunknown-warning-option" +#pragma clang diagnostic ignored "-Wunsafe-buffer-usage" +#endif + +//------------------------------------------------------------------------------ + + inline XMVECTOR XM_CALLCONV XMVectorSetBinaryConstant(uint32_t C0, uint32_t C1, uint32_t C2, uint32_t C3) noexcept + { + #if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = (0 - (C0 & 1)) & 0x3F800000; + vResult.u[1] = (0 - (C1 & 1)) & 0x3F800000; + vResult.u[2] = (0 - (C2 & 1)) & 0x3F800000; + vResult.u[3] = (0 - (C3 & 1)) & 0x3F800000; + return vResult.v; + #elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = (0 - (C0 & 1)) & 0x3F800000; + vResult.u[1] = (0 - (C1 & 1)) & 0x3F800000; + vResult.u[2] = (0 - (C2 & 1)) & 0x3F800000; + vResult.u[3] = (0 - (C3 & 1)) & 0x3F800000; + return vResult.v; + #else // XM_SSE_INTRINSICS_ + static const XMVECTORU32 g_vMask1 = { { { 1, 1, 1, 1 } } }; + // Move the parms to a vector + __m128i vTemp = _mm_set_epi32(static_cast(C3), static_cast(C2), static_cast(C1), static_cast(C0)); + // Mask off the low bits + vTemp = _mm_and_si128(vTemp, g_vMask1); + // 0xFFFFFFFF on true bits + vTemp = _mm_cmpeq_epi32(vTemp, g_vMask1); + // 0xFFFFFFFF -> 1.0f, 0x00000000 -> 0.0f + vTemp = _mm_and_si128(vTemp, g_XMOne); + return _mm_castsi128_ps(vTemp); + #endif + } + + //------------------------------------------------------------------------------ + + inline XMVECTOR XM_CALLCONV XMVectorSplatConstant(int32_t IntConstant, uint32_t DivExponent) noexcept + { + assert(IntConstant >= -16 && IntConstant <= 15); + assert(DivExponent < 32); + #if defined(_XM_NO_INTRINSICS_) + + using DirectX::XMConvertVectorIntToFloat; + + XMVECTORI32 V = { { { IntConstant, IntConstant, IntConstant, IntConstant } } }; + return XMConvertVectorIntToFloat(V.v, DivExponent); + + #elif defined(_XM_ARM_NEON_INTRINSICS_) + // Splat the int + int32x4_t vScale = vdupq_n_s32(IntConstant); + // Convert to a float + XMVECTOR vResult = vcvtq_f32_s32(vScale); + // Convert DivExponent into 1.0f/(1<(&vScale)[0]); + return vResult; + #else // XM_SSE_INTRINSICS_ + // Splat the int + __m128i vScale = _mm_set1_epi32(IntConstant); + // Convert to a float + XMVECTOR vResult = _mm_cvtepi32_ps(vScale); + // Convert DivExponent into 1.0f/(1<(uScale)); + // Multiply by the reciprocal (Perform a right shift by DivExponent) + vResult = _mm_mul_ps(vResult, _mm_castsi128_ps(vScale)); + return vResult; + #endif + } + + //------------------------------------------------------------------------------ + + inline XMVECTOR XM_CALLCONV XMVectorSplatConstantInt(int32_t IntConstant) noexcept + { + assert(IntConstant >= -16 && IntConstant <= 15); + #if defined(_XM_NO_INTRINSICS_) + + XMVECTORI32 V = { { { IntConstant, IntConstant, IntConstant, IntConstant } } }; + return V.v; + + #elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t V = vdupq_n_s32(IntConstant); + return reinterpret_cast(&V)[0]; + #else // XM_SSE_INTRINSICS_ + __m128i V = _mm_set1_epi32(IntConstant); + return _mm_castsi128_ps(V); + #endif + } + +#include "DirectXMathConvert.inl" +#include "DirectXMathVector.inl" +#include "DirectXMathMatrix.inl" +#include "DirectXMathMisc.inl" + +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +} // namespace DirectX + diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathConvert.inl b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathConvert.inl new file mode 100644 index 00000000..e3a6bdc6 --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathConvert.inl @@ -0,0 +1,2219 @@ +//------------------------------------------------------------------------------------- +// DirectXMathConvert.inl -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +/**************************************************************************** + * + * Data conversion + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable:4701) +// C4701: false positives +#endif + +inline XMVECTOR XM_CALLCONV XMConvertVectorIntToFloat +( + FXMVECTOR VInt, + uint32_t DivExponent +) noexcept +{ + assert(DivExponent < 32); +#if defined(_XM_NO_INTRINSICS_) + float fScale = 1.0f / static_cast(1U << DivExponent); + uint32_t ElementIndex = 0; + XMVECTOR Result; + do + { + auto iTemp = static_cast(VInt.vector4_u32[ElementIndex]); + Result.vector4_f32[ElementIndex] = static_cast(iTemp)* fScale; + } + while (++ElementIndex < 4); + return Result; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fScale = 1.0f / static_cast(1U << DivExponent); + float32x4_t vResult = vcvtq_f32_s32(vreinterpretq_s32_f32(VInt)); + return vmulq_n_f32(vResult, fScale); +#else // _XM_SSE_INTRINSICS_ + // Convert to floats + XMVECTOR vResult = _mm_cvtepi32_ps(_mm_castps_si128(VInt)); + // Convert DivExponent into 1.0f/(1<(uScale)); + vResult = _mm_mul_ps(vResult, _mm_castsi128_ps(vScale)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMConvertVectorFloatToInt +( + FXMVECTOR VFloat, + uint32_t MulExponent +) noexcept +{ + assert(MulExponent < 32); +#if defined(_XM_NO_INTRINSICS_) + // Get the scalar factor. + auto fScale = static_cast(1U << MulExponent); + uint32_t ElementIndex = 0; + XMVECTOR Result; + do + { + int32_t iResult; + float fTemp = VFloat.vector4_f32[ElementIndex] * fScale; + if (fTemp <= -(65536.0f * 32768.0f)) + { + iResult = (-0x7FFFFFFF) - 1; + } + else if (fTemp > (65536.0f * 32768.0f) - 128.0f) + { + iResult = 0x7FFFFFFF; + } + else + { + iResult = static_cast(fTemp); + } + Result.vector4_u32[ElementIndex] = static_cast(iResult); + } + while (++ElementIndex < 4); + return Result; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmulq_n_f32(VFloat, static_cast(1U << MulExponent)); + // In case of positive overflow, detect it + uint32x4_t vOverflow = vcgtq_f32(vResult, g_XMMaxInt); + // Float to int conversion + int32x4_t vResulti = vcvtq_s32_f32(vResult); + // If there was positive overflow, set to 0x7FFFFFFF + vResult = vreinterpretq_f32_u32(vandq_u32(vOverflow, g_XMAbsMask)); + vOverflow = vbicq_u32(vreinterpretq_u32_s32(vResulti), vOverflow); + vOverflow = vorrq_u32(vOverflow, vreinterpretq_u32_f32(vResult)); + return vreinterpretq_f32_u32(vOverflow); +#else // _XM_SSE_INTRINSICS_ + XMVECTOR vResult = _mm_set_ps1(static_cast(1U << MulExponent)); + vResult = _mm_mul_ps(vResult, VFloat); + // In case of positive overflow, detect it + XMVECTOR vOverflow = _mm_cmpgt_ps(vResult, g_XMMaxInt); + // Float to int conversion + __m128i vResulti = _mm_cvttps_epi32(vResult); + // If there was positive overflow, set to 0x7FFFFFFF + vResult = _mm_and_ps(vOverflow, g_XMAbsMask); + vOverflow = _mm_andnot_ps(vOverflow, _mm_castsi128_ps(vResulti)); + vOverflow = _mm_or_ps(vOverflow, vResult); + return vOverflow; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMConvertVectorUIntToFloat +( + FXMVECTOR VUInt, + uint32_t DivExponent +) noexcept +{ + assert(DivExponent < 32); +#if defined(_XM_NO_INTRINSICS_) + float fScale = 1.0f / static_cast(1U << DivExponent); + uint32_t ElementIndex = 0; + XMVECTOR Result; + do + { + Result.vector4_f32[ElementIndex] = static_cast(VUInt.vector4_u32[ElementIndex])* fScale; + } + while (++ElementIndex < 4); + return Result; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fScale = 1.0f / static_cast(1U << DivExponent); + float32x4_t vResult = vcvtq_f32_u32(vreinterpretq_u32_f32(VUInt)); + return vmulq_n_f32(vResult, fScale); +#else // _XM_SSE_INTRINSICS_ + // For the values that are higher than 0x7FFFFFFF, a fixup is needed + // Determine which ones need the fix. + XMVECTOR vMask = _mm_and_ps(VUInt, g_XMNegativeZero); + // Force all values positive + XMVECTOR vResult = _mm_xor_ps(VUInt, vMask); + // Convert to floats + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Convert 0x80000000 -> 0xFFFFFFFF + __m128i iMask = _mm_srai_epi32(_mm_castps_si128(vMask), 31); + // For only the ones that are too big, add the fixup + vMask = _mm_and_ps(_mm_castsi128_ps(iMask), g_XMFixUnsigned); + vResult = _mm_add_ps(vResult, vMask); + // Convert DivExponent into 1.0f/(1<(uScale)); + vResult = _mm_mul_ps(vResult, _mm_castsi128_ps(iMask)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMConvertVectorFloatToUInt +( + FXMVECTOR VFloat, + uint32_t MulExponent +) noexcept +{ + assert(MulExponent < 32); +#if defined(_XM_NO_INTRINSICS_) + // Get the scalar factor. + auto fScale = static_cast(1U << MulExponent); + uint32_t ElementIndex = 0; + XMVECTOR Result; + do + { + uint32_t uResult; + float fTemp = VFloat.vector4_f32[ElementIndex] * fScale; + if (fTemp <= 0.0f) + { + uResult = 0; + } + else if (fTemp >= (65536.0f * 65536.0f)) + { + uResult = 0xFFFFFFFFU; + } + else + { + uResult = static_cast(fTemp); + } + Result.vector4_u32[ElementIndex] = uResult; + } + while (++ElementIndex < 4); + return Result; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmulq_n_f32(VFloat, static_cast(1U << MulExponent)); + // In case of overflow, detect it + uint32x4_t vOverflow = vcgtq_f32(vResult, g_XMMaxUInt); + // Float to int conversion + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + // If there was overflow, set to 0xFFFFFFFFU + vResult = vreinterpretq_f32_u32(vbicq_u32(vResulti, vOverflow)); + vOverflow = vorrq_u32(vOverflow, vreinterpretq_u32_f32(vResult)); + return vreinterpretq_f32_u32(vOverflow); +#else // _XM_SSE_INTRINSICS_ + XMVECTOR vResult = _mm_set_ps1(static_cast(1U << MulExponent)); + vResult = _mm_mul_ps(vResult, VFloat); + // Clamp to >=0 + vResult = _mm_max_ps(vResult, g_XMZero); + // Any numbers that are too big, set to 0xFFFFFFFFU + XMVECTOR vOverflow = _mm_cmpgt_ps(vResult, g_XMMaxUInt); + XMVECTOR vValue = g_XMUnsignedFix; + // Too large for a signed integer? + XMVECTOR vMask = _mm_cmpge_ps(vResult, vValue); + // Zero for number's lower than 0x80000000, 32768.0f*65536.0f otherwise + vValue = _mm_and_ps(vValue, vMask); + // Perform fixup only on numbers too large (Keeps low bit precision) + vResult = _mm_sub_ps(vResult, vValue); + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Convert from signed to unsigned pnly if greater than 0x80000000 + vMask = _mm_and_ps(vMask, g_XMNegativeZero); + vResult = _mm_xor_ps(_mm_castsi128_ps(vResulti), vMask); + // On those that are too large, set to 0xFFFFFFFF + vResult = _mm_or_ps(vResult, vOverflow); + return vResult; +#endif +} + +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +/**************************************************************************** + * + * Vector and matrix load operations + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt(const uint32_t* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = *pSource; + V.vector4_u32[1] = 0; + V.vector4_u32[2] = 0; + V.vector4_u32[3] = 0; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t zero = vdupq_n_u32(0); + return vreinterpretq_f32_u32(vld1q_lane_u32(pSource, zero, 0)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_load_ss(reinterpret_cast(pSource)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat(const float* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = *pSource; + V.vector4_f32[1] = 0.f; + V.vector4_f32[2] = 0.f; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t zero = vdupq_n_f32(0); + return vld1q_lane_f32(pSource, zero, 0); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_load_ss(pSource); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt2(const uint32_t* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = pSource[0]; + V.vector4_u32[1] = pSource[1]; + V.vector4_u32[2] = 0; + V.vector4_u32[3] = 0; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t x = vld1_u32(pSource); + uint32x2_t zero = vdup_n_u32(0); + return vreinterpretq_f32_u32(vcombine_u32(x, zero)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt2A(const uint32_t* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = pSource[0]; + V.vector4_u32[1] = pSource[1]; + V.vector4_u32[2] = 0; + V.vector4_u32[3] = 0; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + uint32x2_t x = vld1_u32_ex(pSource, 64); +#else + uint32x2_t x = vld1_u32(pSource); +#endif + uint32x2_t zero = vdup_n_u32(0); + return vreinterpretq_f32_u32(vcombine_u32(x, zero)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat2(const XMFLOAT2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = pSource->x; + V.vector4_f32[1] = pSource->y; + V.vector4_f32[2] = 0.f; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t x = vld1_f32(reinterpret_cast(pSource)); + float32x2_t zero = vdup_n_f32(0); + return vcombine_f32(x, zero); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat2A(const XMFLOAT2A* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = pSource->x; + V.vector4_f32[1] = pSource->y; + V.vector4_f32[2] = 0.f; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + float32x2_t x = vld1_f32_ex(reinterpret_cast(pSource), 64); +#else + float32x2_t x = vld1_f32(reinterpret_cast(pSource)); +#endif + float32x2_t zero = vdup_n_f32(0); + return vcombine_f32(x, zero); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadSInt2(const XMINT2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = static_cast(pSource->x); + V.vector4_f32[1] = static_cast(pSource->y); + V.vector4_f32[2] = 0.f; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x2_t x = vld1_s32(reinterpret_cast(pSource)); + float32x2_t v = vcvt_f32_s32(x); + float32x2_t zero = vdup_n_f32(0); + return vcombine_f32(v, zero); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 V = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + return _mm_cvtepi32_ps(_mm_castps_si128(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUInt2(const XMUINT2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = static_cast(pSource->x); + V.vector4_f32[1] = static_cast(pSource->y); + V.vector4_f32[2] = 0.f; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t x = vld1_u32(reinterpret_cast(pSource)); + float32x2_t v = vcvt_f32_u32(x); + float32x2_t zero = vdup_n_f32(0); + return vcombine_f32(v, zero); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 V = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + // For the values that are higher than 0x7FFFFFFF, a fixup is needed + // Determine which ones need the fix. + XMVECTOR vMask = _mm_and_ps(V, g_XMNegativeZero); + // Force all values positive + XMVECTOR vResult = _mm_xor_ps(V, vMask); + // Convert to floats + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Convert 0x80000000 -> 0xFFFFFFFF + __m128i iMask = _mm_srai_epi32(_mm_castps_si128(vMask), 31); + // For only the ones that are too big, add the fixup + vMask = _mm_and_ps(_mm_castsi128_ps(iMask), g_XMFixUnsigned); + vResult = _mm_add_ps(vResult, vMask); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt3(const uint32_t* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = pSource[0]; + V.vector4_u32[1] = pSource[1]; + V.vector4_u32[2] = pSource[2]; + V.vector4_u32[3] = 0; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t x = vld1_u32(pSource); + uint32x2_t zero = vdup_n_u32(0); + uint32x2_t y = vld1_lane_u32(pSource + 2, zero, 0); + return vreinterpretq_f32_u32(vcombine_u32(x, y)); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(reinterpret_cast(pSource + 2)); + return _mm_insert_ps(xy, z, 0x20); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(reinterpret_cast(pSource + 2)); + return _mm_movelh_ps(xy, z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt3A(const uint32_t* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = pSource[0]; + V.vector4_u32[1] = pSource[1]; + V.vector4_u32[2] = pSource[2]; + V.vector4_u32[3] = 0; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Reads an extra integer which is zero'd +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + uint32x4_t V = vld1q_u32_ex(pSource, 128); +#else + uint32x4_t V = vld1q_u32(pSource); +#endif + return vreinterpretq_f32_u32(vsetq_lane_u32(0, V, 3)); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(reinterpret_cast(pSource + 2)); + return _mm_insert_ps(xy, z, 0x20); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(reinterpret_cast(pSource + 2)); + return _mm_movelh_ps(xy, z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat3(const XMFLOAT3* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = pSource->x; + V.vector4_f32[1] = pSource->y; + V.vector4_f32[2] = pSource->z; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t x = vld1_f32(reinterpret_cast(pSource)); + float32x2_t zero = vdup_n_f32(0); + float32x2_t y = vld1_lane_f32(reinterpret_cast(pSource) + 2, zero, 0); + return vcombine_f32(x, y); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(&pSource->z); + return _mm_insert_ps(xy, z, 0x20); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(&pSource->z); + return _mm_movelh_ps(xy, z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat3A(const XMFLOAT3A* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = pSource->x; + V.vector4_f32[1] = pSource->y; + V.vector4_f32[2] = pSource->z; + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Reads an extra float which is zero'd +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + float32x4_t V = vld1q_f32_ex(reinterpret_cast(pSource), 128); +#else + float32x4_t V = vld1q_f32(reinterpret_cast(pSource)); +#endif + return vsetq_lane_f32(0, V, 3); +#elif defined(_XM_SSE4_INTRINSICS_) + // Reads an extra float which is zero'd + __m128 V = _mm_load_ps(&pSource->x); + return _mm_blend_ps(_mm_setzero_ps(), V, 0x7); +#elif defined(_XM_SSE_INTRINSICS_) + // Reads an extra float which is zero'd + __m128 V = _mm_load_ps(&pSource->x); + return _mm_and_ps(V, g_XMMask3); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadSInt3(const XMINT3* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR V; + V.vector4_f32[0] = static_cast(pSource->x); + V.vector4_f32[1] = static_cast(pSource->y); + V.vector4_f32[2] = static_cast(pSource->z); + V.vector4_f32[3] = 0.f; + return V; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x2_t x = vld1_s32(reinterpret_cast(pSource)); + int32x2_t zero = vdup_n_s32(0); + int32x2_t y = vld1_lane_s32(reinterpret_cast(pSource) + 2, zero, 0); + int32x4_t v = vcombine_s32(x, y); + return vcvtq_f32_s32(v); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(reinterpret_cast(&pSource->z)); + __m128 V = _mm_movelh_ps(xy, z); + return _mm_cvtepi32_ps(_mm_castps_si128(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUInt3(const XMUINT3* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = static_cast(pSource->x); + V.vector4_f32[1] = static_cast(pSource->y); + V.vector4_f32[2] = static_cast(pSource->z); + V.vector4_f32[3] = 0.f; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t x = vld1_u32(reinterpret_cast(pSource)); + uint32x2_t zero = vdup_n_u32(0); + uint32x2_t y = vld1_lane_u32(reinterpret_cast(pSource) + 2, zero, 0); + uint32x4_t v = vcombine_u32(x, y); + return vcvtq_f32_u32(v); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pSource))); + __m128 z = _mm_load_ss(reinterpret_cast(&pSource->z)); + __m128 V = _mm_movelh_ps(xy, z); + // For the values that are higher than 0x7FFFFFFF, a fixup is needed + // Determine which ones need the fix. + XMVECTOR vMask = _mm_and_ps(V, g_XMNegativeZero); + // Force all values positive + XMVECTOR vResult = _mm_xor_ps(V, vMask); + // Convert to floats + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Convert 0x80000000 -> 0xFFFFFFFF + __m128i iMask = _mm_srai_epi32(_mm_castps_si128(vMask), 31); + // For only the ones that are too big, add the fixup + vMask = _mm_and_ps(_mm_castsi128_ps(iMask), g_XMFixUnsigned); + vResult = _mm_add_ps(vResult, vMask); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt4(const uint32_t* pSource) noexcept +{ + assert(pSource); + +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = pSource[0]; + V.vector4_u32[1] = pSource[1]; + V.vector4_u32[2] = pSource[2]; + V.vector4_u32[3] = pSource[3]; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vld1q_u32(pSource)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_loadu_si128(reinterpret_cast(pSource)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadInt4A(const uint32_t* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_u32[0] = pSource[0]; + V.vector4_u32[1] = pSource[1]; + V.vector4_u32[2] = pSource[2]; + V.vector4_u32[3] = pSource[3]; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + return vld1q_u32_ex(pSource, 128); +#else + return vreinterpretq_f32_u32(vld1q_u32(pSource)); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_load_si128(reinterpret_cast(pSource)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat4(const XMFLOAT4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = pSource->x; + V.vector4_f32[1] = pSource->y; + V.vector4_f32[2] = pSource->z; + V.vector4_f32[3] = pSource->w; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vld1q_f32(reinterpret_cast(pSource)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_loadu_ps(&pSource->x); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat4A(const XMFLOAT4A* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = pSource->x; + V.vector4_f32[1] = pSource->y; + V.vector4_f32[2] = pSource->z; + V.vector4_f32[3] = pSource->w; + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + return vld1q_f32_ex(reinterpret_cast(pSource), 128); +#else + return vld1q_f32(reinterpret_cast(pSource)); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_load_ps(&pSource->x); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadSInt4(const XMINT4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR V; + V.vector4_f32[0] = static_cast(pSource->x); + V.vector4_f32[1] = static_cast(pSource->y); + V.vector4_f32[2] = static_cast(pSource->z); + V.vector4_f32[3] = static_cast(pSource->w); + return V; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t v = vld1q_s32(reinterpret_cast(pSource)); + return vcvtq_f32_s32(v); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_loadu_si128(reinterpret_cast(pSource)); + return _mm_cvtepi32_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUInt4(const XMUINT4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR V; + V.vector4_f32[0] = static_cast(pSource->x); + V.vector4_f32[1] = static_cast(pSource->y); + V.vector4_f32[2] = static_cast(pSource->z); + V.vector4_f32[3] = static_cast(pSource->w); + return V; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t v = vld1q_u32(reinterpret_cast(pSource)); + return vcvtq_f32_u32(v); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_loadu_si128(reinterpret_cast(pSource)); + // For the values that are higher than 0x7FFFFFFF, a fixup is needed + // Determine which ones need the fix. + XMVECTOR vMask = _mm_and_ps(_mm_castsi128_ps(V), g_XMNegativeZero); + // Force all values positive + XMVECTOR vResult = _mm_xor_ps(_mm_castsi128_ps(V), vMask); + // Convert to floats + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Convert 0x80000000 -> 0xFFFFFFFF + __m128i iMask = _mm_srai_epi32(_mm_castps_si128(vMask), 31); + // For only the ones that are too big, add the fixup + vMask = _mm_and_ps(_mm_castsi128_ps(iMask), g_XMFixUnsigned); + vResult = _mm_add_ps(vResult, vMask); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat3x3(const XMFLOAT3X3* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[0][1]; + M.r[0].vector4_f32[2] = pSource->m[0][2]; + M.r[0].vector4_f32[3] = 0.0f; + + M.r[1].vector4_f32[0] = pSource->m[1][0]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[1][2]; + M.r[1].vector4_f32[3] = 0.0f; + + M.r[2].vector4_f32[0] = pSource->m[2][0]; + M.r[2].vector4_f32[1] = pSource->m[2][1]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = 0.0f; + M.r[3].vector4_f32[0] = 0.0f; + M.r[3].vector4_f32[1] = 0.0f; + M.r[3].vector4_f32[2] = 0.0f; + M.r[3].vector4_f32[3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t v0 = vld1q_f32(&pSource->m[0][0]); + float32x4_t v1 = vld1q_f32(&pSource->m[1][1]); + float32x2_t v2 = vcreate_f32(static_cast(*reinterpret_cast(&pSource->m[2][2]))); + float32x4_t T = vextq_f32(v0, v1, 3); + + XMMATRIX M; + M.r[0] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(v0), g_XMMask3)); + M.r[1] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T), g_XMMask3)); + M.r[2] = vcombine_f32(vget_high_f32(v1), v2); + M.r[3] = g_XMIdentityR3; + return M; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 Z = _mm_setzero_ps(); + + __m128 V1 = _mm_loadu_ps(&pSource->m[0][0]); + __m128 V2 = _mm_loadu_ps(&pSource->m[1][1]); + __m128 V3 = _mm_load_ss(&pSource->m[2][2]); + + __m128 T1 = _mm_unpackhi_ps(V1, Z); + __m128 T2 = _mm_unpacklo_ps(V2, Z); + __m128 T3 = _mm_shuffle_ps(V3, T2, _MM_SHUFFLE(0, 1, 0, 0)); + __m128 T4 = _mm_movehl_ps(T2, T3); + __m128 T5 = _mm_movehl_ps(Z, T1); + + XMMATRIX M; + M.r[0] = _mm_movelh_ps(V1, T1); + M.r[1] = _mm_add_ps(T4, T5); + M.r[2] = _mm_shuffle_ps(V2, V3, _MM_SHUFFLE(1, 0, 3, 2)); + M.r[3] = g_XMIdentityR3; + return M; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat4x3(const XMFLOAT4X3* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[0][1]; + M.r[0].vector4_f32[2] = pSource->m[0][2]; + M.r[0].vector4_f32[3] = 0.0f; + + M.r[1].vector4_f32[0] = pSource->m[1][0]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[1][2]; + M.r[1].vector4_f32[3] = 0.0f; + + M.r[2].vector4_f32[0] = pSource->m[2][0]; + M.r[2].vector4_f32[1] = pSource->m[2][1]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = 0.0f; + + M.r[3].vector4_f32[0] = pSource->m[3][0]; + M.r[3].vector4_f32[1] = pSource->m[3][1]; + M.r[3].vector4_f32[2] = pSource->m[3][2]; + M.r[3].vector4_f32[3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t v0 = vld1q_f32(&pSource->m[0][0]); + float32x4_t v1 = vld1q_f32(&pSource->m[1][1]); + float32x4_t v2 = vld1q_f32(&pSource->m[2][2]); + + float32x4_t T1 = vextq_f32(v0, v1, 3); + float32x4_t T2 = vcombine_f32(vget_high_f32(v1), vget_low_f32(v2)); + float32x4_t T3 = vextq_f32(v2, v2, 1); + + XMMATRIX M; + M.r[0] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(v0), g_XMMask3)); + M.r[1] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T1), g_XMMask3)); + M.r[2] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T2), g_XMMask3)); + M.r[3] = vsetq_lane_f32(1.f, T3, 3); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + // Use unaligned load instructions to + // load the 12 floats + // vTemp1 = x1,y1,z1,x2 + XMVECTOR vTemp1 = _mm_loadu_ps(&pSource->m[0][0]); + // vTemp2 = y2,z2,x3,y3 + XMVECTOR vTemp2 = _mm_loadu_ps(&pSource->m[1][1]); + // vTemp4 = z3,x4,y4,z4 + XMVECTOR vTemp4 = _mm_loadu_ps(&pSource->m[2][2]); + // vTemp3 = x3,y3,z3,z3 + XMVECTOR vTemp3 = _mm_shuffle_ps(vTemp2, vTemp4, _MM_SHUFFLE(0, 0, 3, 2)); + // vTemp2 = y2,z2,x2,x2 + vTemp2 = _mm_shuffle_ps(vTemp2, vTemp1, _MM_SHUFFLE(3, 3, 1, 0)); + // vTemp2 = x2,y2,z2,z2 + vTemp2 = XM_PERMUTE_PS(vTemp2, _MM_SHUFFLE(1, 1, 0, 2)); + // vTemp1 = x1,y1,z1,0 + // vTemp2 = x2,y2,z2,0 + // vTemp3 = x3,y3,z3,0 +#ifdef _XM_SSE4_INTRINSICS_ + XMVECTOR zero = _mm_setzero_ps(); + vTemp1 = _mm_blend_ps(zero, vTemp1, 0x7); + vTemp2 = _mm_blend_ps(zero, vTemp2, 0x7); + vTemp3 = _mm_blend_ps(zero, vTemp3, 0x7); +#else + vTemp1 = _mm_and_ps(vTemp1, g_XMMask3); + vTemp2 = _mm_and_ps(vTemp2, g_XMMask3); + vTemp3 = _mm_and_ps(vTemp3, g_XMMask3); +#endif + // vTemp4i = x4,y4,z4,0 + __m128i vTemp4i = _mm_srli_si128(_mm_castps_si128(vTemp4), 32 / 8); + // vTemp4i = x4,y4,z4,1.0f + vTemp4i = _mm_or_si128(vTemp4i, g_XMIdentityR3); + XMMATRIX M(vTemp1, + vTemp2, + vTemp3, + _mm_castsi128_ps(vTemp4i)); + return M; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat4x3A(const XMFLOAT4X3A* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[0][1]; + M.r[0].vector4_f32[2] = pSource->m[0][2]; + M.r[0].vector4_f32[3] = 0.0f; + + M.r[1].vector4_f32[0] = pSource->m[1][0]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[1][2]; + M.r[1].vector4_f32[3] = 0.0f; + + M.r[2].vector4_f32[0] = pSource->m[2][0]; + M.r[2].vector4_f32[1] = pSource->m[2][1]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = 0.0f; + + M.r[3].vector4_f32[0] = pSource->m[3][0]; + M.r[3].vector4_f32[1] = pSource->m[3][1]; + M.r[3].vector4_f32[2] = pSource->m[3][2]; + M.r[3].vector4_f32[3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + float32x4_t v0 = vld1q_f32_ex(&pSource->m[0][0], 128); + float32x4_t v1 = vld1q_f32_ex(&pSource->m[1][1], 128); + float32x4_t v2 = vld1q_f32_ex(&pSource->m[2][2], 128); +#else + float32x4_t v0 = vld1q_f32(&pSource->m[0][0]); + float32x4_t v1 = vld1q_f32(&pSource->m[1][1]); + float32x4_t v2 = vld1q_f32(&pSource->m[2][2]); +#endif + + float32x4_t T1 = vextq_f32(v0, v1, 3); + float32x4_t T2 = vcombine_f32(vget_high_f32(v1), vget_low_f32(v2)); + float32x4_t T3 = vextq_f32(v2, v2, 1); + + XMMATRIX M; + M.r[0] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(v0), g_XMMask3)); + M.r[1] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T1), g_XMMask3)); + M.r[2] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T2), g_XMMask3)); + M.r[3] = vsetq_lane_f32(1.f, T3, 3); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + // Use aligned load instructions to + // load the 12 floats + // vTemp1 = x1,y1,z1,x2 + XMVECTOR vTemp1 = _mm_load_ps(&pSource->m[0][0]); + // vTemp2 = y2,z2,x3,y3 + XMVECTOR vTemp2 = _mm_load_ps(&pSource->m[1][1]); + // vTemp4 = z3,x4,y4,z4 + XMVECTOR vTemp4 = _mm_load_ps(&pSource->m[2][2]); + // vTemp3 = x3,y3,z3,z3 + XMVECTOR vTemp3 = _mm_shuffle_ps(vTemp2, vTemp4, _MM_SHUFFLE(0, 0, 3, 2)); + // vTemp2 = y2,z2,x2,x2 + vTemp2 = _mm_shuffle_ps(vTemp2, vTemp1, _MM_SHUFFLE(3, 3, 1, 0)); + // vTemp2 = x2,y2,z2,z2 + vTemp2 = XM_PERMUTE_PS(vTemp2, _MM_SHUFFLE(1, 1, 0, 2)); + // vTemp1 = x1,y1,z1,0 + // vTemp2 = x2,y2,z2,0 + // vTemp3 = x3,y3,z3,0 +#ifdef _XM_SSE4_INTRINSICS_ + XMVECTOR zero = _mm_setzero_ps(); + vTemp1 = _mm_blend_ps(zero, vTemp1, 0x7); + vTemp2 = _mm_blend_ps(zero, vTemp2, 0x7); + vTemp3 = _mm_blend_ps(zero, vTemp3, 0x7); +#else + vTemp1 = _mm_and_ps(vTemp1, g_XMMask3); + vTemp2 = _mm_and_ps(vTemp2, g_XMMask3); + vTemp3 = _mm_and_ps(vTemp3, g_XMMask3); +#endif + // vTemp4i = x4,y4,z4,0 + __m128i vTemp4i = _mm_srli_si128(_mm_castps_si128(vTemp4), 32 / 8); + // vTemp4i = x4,y4,z4,1.0f + vTemp4i = _mm_or_si128(vTemp4i, g_XMIdentityR3); + XMMATRIX M(vTemp1, + vTemp2, + vTemp3, + _mm_castsi128_ps(vTemp4i)); + return M; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat3x4(const XMFLOAT3X4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[1][0]; + M.r[0].vector4_f32[2] = pSource->m[2][0]; + M.r[0].vector4_f32[3] = 0.0f; + + M.r[1].vector4_f32[0] = pSource->m[0][1]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[2][1]; + M.r[1].vector4_f32[3] = 0.0f; + + M.r[2].vector4_f32[0] = pSource->m[0][2]; + M.r[2].vector4_f32[1] = pSource->m[1][2]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = 0.0f; + + M.r[3].vector4_f32[0] = pSource->m[0][3]; + M.r[3].vector4_f32[1] = pSource->m[1][3]; + M.r[3].vector4_f32[2] = pSource->m[2][3]; + M.r[3].vector4_f32[3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2x4_t vTemp0 = vld4_f32(&pSource->_11); + float32x4_t vTemp1 = vld1q_f32(&pSource->_31); + + float32x2_t l = vget_low_f32(vTemp1); + float32x4_t T0 = vcombine_f32(vTemp0.val[0], l); + float32x2_t rl = vrev64_f32(l); + float32x4_t T1 = vcombine_f32(vTemp0.val[1], rl); + + float32x2_t h = vget_high_f32(vTemp1); + float32x4_t T2 = vcombine_f32(vTemp0.val[2], h); + float32x2_t rh = vrev64_f32(h); + float32x4_t T3 = vcombine_f32(vTemp0.val[3], rh); + + XMMATRIX M = {}; + M.r[0] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T0), g_XMMask3)); + M.r[1] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T1), g_XMMask3)); + M.r[2] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T2), g_XMMask3)); + M.r[3] = vsetq_lane_f32(1.f, T3, 3); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + M.r[0] = _mm_loadu_ps(&pSource->_11); + M.r[1] = _mm_loadu_ps(&pSource->_21); + M.r[2] = _mm_loadu_ps(&pSource->_31); + M.r[3] = g_XMIdentityR3; + + // x.x,x.y,y.x,y.y + XMVECTOR vTemp1 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(1, 0, 1, 0)); + // x.z,x.w,y.z,y.w + XMVECTOR vTemp3 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + // z.x,z.y,w.x,w.y + XMVECTOR vTemp2 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(1, 0, 1, 0)); + // z.z,z.w,w.z,w.w + XMVECTOR vTemp4 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + XMMATRIX mResult; + + // x.x,y.x,z.x,w.x + mResult.r[0] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + // x.y,y.y,z.y,w.y + mResult.r[1] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + // x.z,y.z,z.z,w.z + mResult.r[2] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + // x.w,y.w,z.w,w.w + mResult.r[3] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(3, 1, 3, 1)); + return mResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat3x4A(const XMFLOAT3X4A* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[1][0]; + M.r[0].vector4_f32[2] = pSource->m[2][0]; + M.r[0].vector4_f32[3] = 0.0f; + + M.r[1].vector4_f32[0] = pSource->m[0][1]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[2][1]; + M.r[1].vector4_f32[3] = 0.0f; + + M.r[2].vector4_f32[0] = pSource->m[0][2]; + M.r[2].vector4_f32[1] = pSource->m[1][2]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = 0.0f; + + M.r[3].vector4_f32[0] = pSource->m[0][3]; + M.r[3].vector4_f32[1] = pSource->m[1][3]; + M.r[3].vector4_f32[2] = pSource->m[2][3]; + M.r[3].vector4_f32[3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + float32x2x4_t vTemp0 = vld4_f32_ex(&pSource->_11, 128); + float32x4_t vTemp1 = vld1q_f32_ex(&pSource->_31, 128); +#else + float32x2x4_t vTemp0 = vld4_f32(&pSource->_11); + float32x4_t vTemp1 = vld1q_f32(&pSource->_31); +#endif + + float32x2_t l = vget_low_f32(vTemp1); + float32x4_t T0 = vcombine_f32(vTemp0.val[0], l); + float32x2_t rl = vrev64_f32(l); + float32x4_t T1 = vcombine_f32(vTemp0.val[1], rl); + + float32x2_t h = vget_high_f32(vTemp1); + float32x4_t T2 = vcombine_f32(vTemp0.val[2], h); + float32x2_t rh = vrev64_f32(h); + float32x4_t T3 = vcombine_f32(vTemp0.val[3], rh); + + XMMATRIX M = {}; + M.r[0] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T0), g_XMMask3)); + M.r[1] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T1), g_XMMask3)); + M.r[2] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T2), g_XMMask3)); + M.r[3] = vsetq_lane_f32(1.f, T3, 3); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + M.r[0] = _mm_load_ps(&pSource->_11); + M.r[1] = _mm_load_ps(&pSource->_21); + M.r[2] = _mm_load_ps(&pSource->_31); + M.r[3] = g_XMIdentityR3; + + // x.x,x.y,y.x,y.y + XMVECTOR vTemp1 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(1, 0, 1, 0)); + // x.z,x.w,y.z,y.w + XMVECTOR vTemp3 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + // z.x,z.y,w.x,w.y + XMVECTOR vTemp2 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(1, 0, 1, 0)); + // z.z,z.w,w.z,w.w + XMVECTOR vTemp4 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + XMMATRIX mResult; + + // x.x,y.x,z.x,w.x + mResult.r[0] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + // x.y,y.y,z.y,w.y + mResult.r[1] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + // x.z,y.z,z.z,w.z + mResult.r[2] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + // x.w,y.w,z.w,w.w + mResult.r[3] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(3, 1, 3, 1)); + return mResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat4x4(const XMFLOAT4X4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[0][1]; + M.r[0].vector4_f32[2] = pSource->m[0][2]; + M.r[0].vector4_f32[3] = pSource->m[0][3]; + + M.r[1].vector4_f32[0] = pSource->m[1][0]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[1][2]; + M.r[1].vector4_f32[3] = pSource->m[1][3]; + + M.r[2].vector4_f32[0] = pSource->m[2][0]; + M.r[2].vector4_f32[1] = pSource->m[2][1]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = pSource->m[2][3]; + + M.r[3].vector4_f32[0] = pSource->m[3][0]; + M.r[3].vector4_f32[1] = pSource->m[3][1]; + M.r[3].vector4_f32[2] = pSource->m[3][2]; + M.r[3].vector4_f32[3] = pSource->m[3][3]; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX M; + M.r[0] = vld1q_f32(reinterpret_cast(&pSource->_11)); + M.r[1] = vld1q_f32(reinterpret_cast(&pSource->_21)); + M.r[2] = vld1q_f32(reinterpret_cast(&pSource->_31)); + M.r[3] = vld1q_f32(reinterpret_cast(&pSource->_41)); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + M.r[0] = _mm_loadu_ps(&pSource->_11); + M.r[1] = _mm_loadu_ps(&pSource->_21); + M.r[2] = _mm_loadu_ps(&pSource->_31); + M.r[3] = _mm_loadu_ps(&pSource->_41); + return M; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMLoadFloat4x4A(const XMFLOAT4X4A* pSource) noexcept +{ + assert(pSource); + assert((reinterpret_cast(pSource) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.r[0].vector4_f32[0] = pSource->m[0][0]; + M.r[0].vector4_f32[1] = pSource->m[0][1]; + M.r[0].vector4_f32[2] = pSource->m[0][2]; + M.r[0].vector4_f32[3] = pSource->m[0][3]; + + M.r[1].vector4_f32[0] = pSource->m[1][0]; + M.r[1].vector4_f32[1] = pSource->m[1][1]; + M.r[1].vector4_f32[2] = pSource->m[1][2]; + M.r[1].vector4_f32[3] = pSource->m[1][3]; + + M.r[2].vector4_f32[0] = pSource->m[2][0]; + M.r[2].vector4_f32[1] = pSource->m[2][1]; + M.r[2].vector4_f32[2] = pSource->m[2][2]; + M.r[2].vector4_f32[3] = pSource->m[2][3]; + + M.r[3].vector4_f32[0] = pSource->m[3][0]; + M.r[3].vector4_f32[1] = pSource->m[3][1]; + M.r[3].vector4_f32[2] = pSource->m[3][2]; + M.r[3].vector4_f32[3] = pSource->m[3][3]; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX M; +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + M.r[0] = vld1q_f32_ex(reinterpret_cast(&pSource->_11), 128); + M.r[1] = vld1q_f32_ex(reinterpret_cast(&pSource->_21), 128); + M.r[2] = vld1q_f32_ex(reinterpret_cast(&pSource->_31), 128); + M.r[3] = vld1q_f32_ex(reinterpret_cast(&pSource->_41), 128); +#else + M.r[0] = vld1q_f32(reinterpret_cast(&pSource->_11)); + M.r[1] = vld1q_f32(reinterpret_cast(&pSource->_21)); + M.r[2] = vld1q_f32(reinterpret_cast(&pSource->_31)); + M.r[3] = vld1q_f32(reinterpret_cast(&pSource->_41)); +#endif + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + M.r[0] = _mm_load_ps(&pSource->_11); + M.r[1] = _mm_load_ps(&pSource->_21); + M.r[2] = _mm_load_ps(&pSource->_31); + M.r[3] = _mm_load_ps(&pSource->_41); + return M; +#endif +} + +/**************************************************************************** + * + * Vector and matrix store operations + * + ****************************************************************************/ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + *pDestination = XMVectorGetIntX(V); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_u32(pDestination, *reinterpret_cast(&V), 0); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_ss(reinterpret_cast(pDestination), V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat +( + float* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + *pDestination = XMVectorGetX(V); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_f32(pDestination, V, 0); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_ss(pDestination, V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt2 +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination[0] = V.vector4_u32[0]; + pDestination[1] = V.vector4_u32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t VL = vget_low_u32(vreinterpretq_u32_f32(V)); + vst1_u32(pDestination, VL); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt2A +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + pDestination[0] = V.vector4_u32[0]; + pDestination[1] = V.vector4_u32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t VL = vget_low_u32(vreinterpretq_u32_f32(V)); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1_u32_ex(pDestination, VL, 64); +#else + vst1_u32(pDestination, VL); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat2 +( + XMFLOAT2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = V.vector4_f32[0]; + pDestination->y = V.vector4_f32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + vst1_f32(reinterpret_cast(pDestination), VL); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat2A +( + XMFLOAT2A* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = V.vector4_f32[0]; + pDestination->y = V.vector4_f32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1_f32_ex(reinterpret_cast(pDestination), VL, 64); +#else + vst1_f32(reinterpret_cast(pDestination), VL); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreSInt2 +( + XMINT2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = static_cast(V.vector4_f32[0]); + pDestination->y = static_cast(V.vector4_f32[1]); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t v = vget_low_f32(V); + int32x2_t iv = vcvt_s32_f32(v); + vst1_s32(reinterpret_cast(pDestination), iv); +#elif defined(_XM_SSE_INTRINSICS_) + // In case of positive overflow, detect it + XMVECTOR vOverflow = _mm_cmpgt_ps(V, g_XMMaxInt); + // Float to int conversion + __m128i vResulti = _mm_cvttps_epi32(V); + // If there was positive overflow, set to 0x7FFFFFFF + XMVECTOR vResult = _mm_and_ps(vOverflow, g_XMAbsMask); + vOverflow = _mm_andnot_ps(vOverflow, _mm_castsi128_ps(vResulti)); + vOverflow = _mm_or_ps(vOverflow, vResult); + // Write two ints + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(vOverflow)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUInt2 +( + XMUINT2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = static_cast(V.vector4_f32[0]); + pDestination->y = static_cast(V.vector4_f32[1]); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t v = vget_low_f32(V); + uint32x2_t iv = vcvt_u32_f32(v); + vst1_u32(reinterpret_cast(pDestination), iv); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to >=0 + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + // Any numbers that are too big, set to 0xFFFFFFFFU + XMVECTOR vOverflow = _mm_cmpgt_ps(vResult, g_XMMaxUInt); + XMVECTOR vValue = g_XMUnsignedFix; + // Too large for a signed integer? + XMVECTOR vMask = _mm_cmpge_ps(vResult, vValue); + // Zero for number's lower than 0x80000000, 32768.0f*65536.0f otherwise + vValue = _mm_and_ps(vValue, vMask); + // Perform fixup only on numbers too large (Keeps low bit precision) + vResult = _mm_sub_ps(vResult, vValue); + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Convert from signed to unsigned pnly if greater than 0x80000000 + vMask = _mm_and_ps(vMask, g_XMNegativeZero); + vResult = _mm_xor_ps(_mm_castsi128_ps(vResulti), vMask); + // On those that are too large, set to 0xFFFFFFFF + vResult = _mm_or_ps(vResult, vOverflow); + // Write two uints + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(vResult)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt3 +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination[0] = V.vector4_u32[0]; + pDestination[1] = V.vector4_u32[1]; + pDestination[2] = V.vector4_u32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t VL = vget_low_u32(vreinterpretq_u32_f32(V)); + vst1_u32(pDestination, VL); + vst1q_lane_u32(pDestination + 2, *reinterpret_cast(&V), 2); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); + __m128 z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(reinterpret_cast(&pDestination[2]), z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt3A +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + pDestination[0] = V.vector4_u32[0]; + pDestination[1] = V.vector4_u32[1]; + pDestination[2] = V.vector4_u32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t VL = vget_low_u32(vreinterpretq_u32_f32(V)); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1_u32_ex(pDestination, VL, 64); +#else + vst1_u32(pDestination, VL); +#endif + vst1q_lane_u32(pDestination + 2, *reinterpret_cast(&V), 2); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); + __m128 z = _mm_movehl_ps(V, V); + _mm_store_ss(reinterpret_cast(&pDestination[2]), z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3 +( + XMFLOAT3* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = V.vector4_f32[0]; + pDestination->y = V.vector4_f32[1]; + pDestination->z = V.vector4_f32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + vst1_f32(reinterpret_cast(pDestination), VL); + vst1q_lane_f32(reinterpret_cast(pDestination) + 2, V, 2); +#elif defined(_XM_SSE4_INTRINSICS_) + * reinterpret_cast(&pDestination->x) = _mm_extract_ps(V, 0); + *reinterpret_cast(&pDestination->y) = _mm_extract_ps(V, 1); + *reinterpret_cast(&pDestination->z) = _mm_extract_ps(V, 2); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); + __m128 z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(&pDestination->z, z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3A +( + XMFLOAT3A* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = V.vector4_f32[0]; + pDestination->y = V.vector4_f32[1]; + pDestination->z = V.vector4_f32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1_f32_ex(reinterpret_cast(pDestination), VL, 64); +#else + vst1_f32(reinterpret_cast(pDestination), VL); +#endif + vst1q_lane_f32(reinterpret_cast(pDestination) + 2, V, 2); +#elif defined(_XM_SSE4_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); + *reinterpret_cast(&pDestination->z) = _mm_extract_ps(V, 2); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(V)); + __m128 z = _mm_movehl_ps(V, V); + _mm_store_ss(&pDestination->z, z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreSInt3 +( + XMINT3* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = static_cast(V.vector4_f32[0]); + pDestination->y = static_cast(V.vector4_f32[1]); + pDestination->z = static_cast(V.vector4_f32[2]); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t v = vcvtq_s32_f32(V); + int32x2_t vL = vget_low_s32(v); + vst1_s32(reinterpret_cast(pDestination), vL); + vst1q_lane_s32(reinterpret_cast(pDestination) + 2, v, 2); +#elif defined(_XM_SSE_INTRINSICS_) + // In case of positive overflow, detect it + XMVECTOR vOverflow = _mm_cmpgt_ps(V, g_XMMaxInt); + // Float to int conversion + __m128i vResulti = _mm_cvttps_epi32(V); + // If there was positive overflow, set to 0x7FFFFFFF + XMVECTOR vResult = _mm_and_ps(vOverflow, g_XMAbsMask); + vOverflow = _mm_andnot_ps(vOverflow, _mm_castsi128_ps(vResulti)); + vOverflow = _mm_or_ps(vOverflow, vResult); + // Write 3 uints + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(vOverflow)); + __m128 z = XM_PERMUTE_PS(vOverflow, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(reinterpret_cast(&pDestination->z), z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUInt3 +( + XMUINT3* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = static_cast(V.vector4_f32[0]); + pDestination->y = static_cast(V.vector4_f32[1]); + pDestination->z = static_cast(V.vector4_f32[2]); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t v = vcvtq_u32_f32(V); + uint32x2_t vL = vget_low_u32(v); + vst1_u32(reinterpret_cast(pDestination), vL); + vst1q_lane_u32(reinterpret_cast(pDestination) + 2, v, 2); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to >=0 + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + // Any numbers that are too big, set to 0xFFFFFFFFU + XMVECTOR vOverflow = _mm_cmpgt_ps(vResult, g_XMMaxUInt); + XMVECTOR vValue = g_XMUnsignedFix; + // Too large for a signed integer? + XMVECTOR vMask = _mm_cmpge_ps(vResult, vValue); + // Zero for number's lower than 0x80000000, 32768.0f*65536.0f otherwise + vValue = _mm_and_ps(vValue, vMask); + // Perform fixup only on numbers too large (Keeps low bit precision) + vResult = _mm_sub_ps(vResult, vValue); + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Convert from signed to unsigned pnly if greater than 0x80000000 + vMask = _mm_and_ps(vMask, g_XMNegativeZero); + vResult = _mm_xor_ps(_mm_castsi128_ps(vResulti), vMask); + // On those that are too large, set to 0xFFFFFFFF + vResult = _mm_or_ps(vResult, vOverflow); + // Write 3 uints + _mm_store_sd(reinterpret_cast(pDestination), _mm_castps_pd(vResult)); + __m128 z = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(reinterpret_cast(&pDestination->z), z); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt4 +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination[0] = V.vector4_u32[0]; + pDestination[1] = V.vector4_u32[1]; + pDestination[2] = V.vector4_u32[2]; + pDestination[3] = V.vector4_u32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_u32(pDestination, vreinterpretq_u32_f32(V)); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_storeu_si128(reinterpret_cast<__m128i*>(pDestination), _mm_castps_si128(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreInt4A +( + uint32_t* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + pDestination[0] = V.vector4_u32[0]; + pDestination[1] = V.vector4_u32[1]; + pDestination[2] = V.vector4_u32[2]; + pDestination[3] = V.vector4_u32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1q_u32_ex(pDestination, V, 128); +#else + vst1q_u32(pDestination, vreinterpretq_u32_f32(V)); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_si128(reinterpret_cast<__m128i*>(pDestination), _mm_castps_si128(V)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat4 +( + XMFLOAT4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = V.vector4_f32[0]; + pDestination->y = V.vector4_f32[1]; + pDestination->z = V.vector4_f32[2]; + pDestination->w = V.vector4_f32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_f32(reinterpret_cast(pDestination), V); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_storeu_ps(&pDestination->x, V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat4A +( + XMFLOAT4A* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = V.vector4_f32[0]; + pDestination->y = V.vector4_f32[1]; + pDestination->z = V.vector4_f32[2]; + pDestination->w = V.vector4_f32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1q_f32_ex(reinterpret_cast(pDestination), V, 128); +#else + vst1q_f32(reinterpret_cast(pDestination), V); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_ps(&pDestination->x, V); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreSInt4 +( + XMINT4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = static_cast(V.vector4_f32[0]); + pDestination->y = static_cast(V.vector4_f32[1]); + pDestination->z = static_cast(V.vector4_f32[2]); + pDestination->w = static_cast(V.vector4_f32[3]); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t v = vcvtq_s32_f32(V); + vst1q_s32(reinterpret_cast(pDestination), v); +#elif defined(_XM_SSE_INTRINSICS_) + // In case of positive overflow, detect it + XMVECTOR vOverflow = _mm_cmpgt_ps(V, g_XMMaxInt); + // Float to int conversion + __m128i vResulti = _mm_cvttps_epi32(V); + // If there was positive overflow, set to 0x7FFFFFFF + XMVECTOR vResult = _mm_and_ps(vOverflow, g_XMAbsMask); + vOverflow = _mm_andnot_ps(vOverflow, _mm_castsi128_ps(vResulti)); + vOverflow = _mm_or_ps(vOverflow, vResult); + _mm_storeu_si128(reinterpret_cast<__m128i*>(pDestination), _mm_castps_si128(vOverflow)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUInt4 +( + XMUINT4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + pDestination->x = static_cast(V.vector4_f32[0]); + pDestination->y = static_cast(V.vector4_f32[1]); + pDestination->z = static_cast(V.vector4_f32[2]); + pDestination->w = static_cast(V.vector4_f32[3]); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t v = vcvtq_u32_f32(V); + vst1q_u32(reinterpret_cast(pDestination), v); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to >=0 + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + // Any numbers that are too big, set to 0xFFFFFFFFU + XMVECTOR vOverflow = _mm_cmpgt_ps(vResult, g_XMMaxUInt); + XMVECTOR vValue = g_XMUnsignedFix; + // Too large for a signed integer? + XMVECTOR vMask = _mm_cmpge_ps(vResult, vValue); + // Zero for number's lower than 0x80000000, 32768.0f*65536.0f otherwise + vValue = _mm_and_ps(vValue, vMask); + // Perform fixup only on numbers too large (Keeps low bit precision) + vResult = _mm_sub_ps(vResult, vValue); + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Convert from signed to unsigned pnly if greater than 0x80000000 + vMask = _mm_and_ps(vMask, g_XMNegativeZero); + vResult = _mm_xor_ps(_mm_castsi128_ps(vResulti), vMask); + // On those that are too large, set to 0xFFFFFFFF + vResult = _mm_or_ps(vResult, vOverflow); + _mm_storeu_si128(reinterpret_cast<__m128i*>(pDestination), _mm_castps_si128(vResult)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3x3 +( + XMFLOAT3X3* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[0].vector4_f32[1]; + pDestination->m[0][2] = M.r[0].vector4_f32[2]; + + pDestination->m[1][0] = M.r[1].vector4_f32[0]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[1].vector4_f32[2]; + + pDestination->m[2][0] = M.r[2].vector4_f32[0]; + pDestination->m[2][1] = M.r[2].vector4_f32[1]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t T1 = vextq_f32(M.r[0], M.r[1], 1); + float32x4_t T2 = vbslq_f32(g_XMMask3, M.r[0], T1); + vst1q_f32(&pDestination->m[0][0], T2); + + T1 = vextq_f32(M.r[1], M.r[1], 1); + T2 = vcombine_f32(vget_low_f32(T1), vget_low_f32(M.r[2])); + vst1q_f32(&pDestination->m[1][1], T2); + + vst1q_lane_f32(&pDestination->m[2][2], M.r[2], 2); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp1 = M.r[0]; + XMVECTOR vTemp2 = M.r[1]; + XMVECTOR vTemp3 = M.r[2]; + XMVECTOR vWork = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(0, 0, 2, 2)); + vTemp1 = _mm_shuffle_ps(vTemp1, vWork, _MM_SHUFFLE(2, 0, 1, 0)); + _mm_storeu_ps(&pDestination->m[0][0], vTemp1); + vTemp2 = _mm_shuffle_ps(vTemp2, vTemp3, _MM_SHUFFLE(1, 0, 2, 1)); + _mm_storeu_ps(&pDestination->m[1][1], vTemp2); + vTemp3 = XM_PERMUTE_PS(vTemp3, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(&pDestination->m[2][2], vTemp3); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat4x3 +( + XMFLOAT4X3* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[0].vector4_f32[1]; + pDestination->m[0][2] = M.r[0].vector4_f32[2]; + + pDestination->m[1][0] = M.r[1].vector4_f32[0]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[1].vector4_f32[2]; + + pDestination->m[2][0] = M.r[2].vector4_f32[0]; + pDestination->m[2][1] = M.r[2].vector4_f32[1]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + + pDestination->m[3][0] = M.r[3].vector4_f32[0]; + pDestination->m[3][1] = M.r[3].vector4_f32[1]; + pDestination->m[3][2] = M.r[3].vector4_f32[2]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t T1 = vextq_f32(M.r[0], M.r[1], 1); + float32x4_t T2 = vbslq_f32(g_XMMask3, M.r[0], T1); + vst1q_f32(&pDestination->m[0][0], T2); + + T1 = vextq_f32(M.r[1], M.r[1], 1); + T2 = vcombine_f32(vget_low_f32(T1), vget_low_f32(M.r[2])); + vst1q_f32(&pDestination->m[1][1], T2); + + T1 = vdupq_lane_f32(vget_high_f32(M.r[2]), 0); + T2 = vextq_f32(T1, M.r[3], 3); + vst1q_f32(&pDestination->m[2][2], T2); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp1 = M.r[0]; + XMVECTOR vTemp2 = M.r[1]; + XMVECTOR vTemp3 = M.r[2]; + XMVECTOR vTemp4 = M.r[3]; + XMVECTOR vTemp2x = _mm_shuffle_ps(vTemp2, vTemp3, _MM_SHUFFLE(1, 0, 2, 1)); + vTemp2 = _mm_shuffle_ps(vTemp2, vTemp1, _MM_SHUFFLE(2, 2, 0, 0)); + vTemp1 = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(0, 2, 1, 0)); + vTemp3 = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(0, 0, 2, 2)); + vTemp3 = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 1, 2, 0)); + _mm_storeu_ps(&pDestination->m[0][0], vTemp1); + _mm_storeu_ps(&pDestination->m[1][1], vTemp2x); + _mm_storeu_ps(&pDestination->m[2][2], vTemp3); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat4x3A +( + XMFLOAT4X3A* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[0].vector4_f32[1]; + pDestination->m[0][2] = M.r[0].vector4_f32[2]; + + pDestination->m[1][0] = M.r[1].vector4_f32[0]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[1].vector4_f32[2]; + + pDestination->m[2][0] = M.r[2].vector4_f32[0]; + pDestination->m[2][1] = M.r[2].vector4_f32[1]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + + pDestination->m[3][0] = M.r[3].vector4_f32[0]; + pDestination->m[3][1] = M.r[3].vector4_f32[1]; + pDestination->m[3][2] = M.r[3].vector4_f32[2]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + float32x4_t T1 = vextq_f32(M.r[0], M.r[1], 1); + float32x4_t T2 = vbslq_f32(g_XMMask3, M.r[0], T1); + vst1q_f32_ex(&pDestination->m[0][0], T2, 128); + + T1 = vextq_f32(M.r[1], M.r[1], 1); + T2 = vcombine_f32(vget_low_f32(T1), vget_low_f32(M.r[2])); + vst1q_f32_ex(&pDestination->m[1][1], T2, 128); + + T1 = vdupq_lane_f32(vget_high_f32(M.r[2]), 0); + T2 = vextq_f32(T1, M.r[3], 3); + vst1q_f32_ex(&pDestination->m[2][2], T2, 128); +#else + float32x4_t T1 = vextq_f32(M.r[0], M.r[1], 1); + float32x4_t T2 = vbslq_f32(g_XMMask3, M.r[0], T1); + vst1q_f32(&pDestination->m[0][0], T2); + + T1 = vextq_f32(M.r[1], M.r[1], 1); + T2 = vcombine_f32(vget_low_f32(T1), vget_low_f32(M.r[2])); + vst1q_f32(&pDestination->m[1][1], T2); + + T1 = vdupq_lane_f32(vget_high_f32(M.r[2]), 0); + T2 = vextq_f32(T1, M.r[3], 3); + vst1q_f32(&pDestination->m[2][2], T2); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + // x1,y1,z1,w1 + XMVECTOR vTemp1 = M.r[0]; + // x2,y2,z2,w2 + XMVECTOR vTemp2 = M.r[1]; + // x3,y3,z3,w3 + XMVECTOR vTemp3 = M.r[2]; + // x4,y4,z4,w4 + XMVECTOR vTemp4 = M.r[3]; + // z1,z1,x2,y2 + XMVECTOR vTemp = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(1, 0, 2, 2)); + // y2,z2,x3,y3 (Final) + vTemp2 = _mm_shuffle_ps(vTemp2, vTemp3, _MM_SHUFFLE(1, 0, 2, 1)); + // x1,y1,z1,x2 (Final) + vTemp1 = _mm_shuffle_ps(vTemp1, vTemp, _MM_SHUFFLE(2, 0, 1, 0)); + // z3,z3,x4,x4 + vTemp3 = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(0, 0, 2, 2)); + // z3,x4,y4,z4 (Final) + vTemp3 = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 1, 2, 0)); + // Store in 3 operations + _mm_store_ps(&pDestination->m[0][0], vTemp1); + _mm_store_ps(&pDestination->m[1][1], vTemp2); + _mm_store_ps(&pDestination->m[2][2], vTemp3); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3x4 +( + XMFLOAT3X4* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[1].vector4_f32[0]; + pDestination->m[0][2] = M.r[2].vector4_f32[0]; + pDestination->m[0][3] = M.r[3].vector4_f32[0]; + + pDestination->m[1][0] = M.r[0].vector4_f32[1]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[2].vector4_f32[1]; + pDestination->m[1][3] = M.r[3].vector4_f32[1]; + + pDestination->m[2][0] = M.r[0].vector4_f32[2]; + pDestination->m[2][1] = M.r[1].vector4_f32[2]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + pDestination->m[2][3] = M.r[3].vector4_f32[2]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4x2_t P0 = vzipq_f32(M.r[0], M.r[2]); + float32x4x2_t P1 = vzipq_f32(M.r[1], M.r[3]); + + float32x4x2_t T0 = vzipq_f32(P0.val[0], P1.val[0]); + float32x4x2_t T1 = vzipq_f32(P0.val[1], P1.val[1]); + + vst1q_f32(&pDestination->m[0][0], T0.val[0]); + vst1q_f32(&pDestination->m[1][0], T0.val[1]); + vst1q_f32(&pDestination->m[2][0], T1.val[0]); +#elif defined(_XM_SSE_INTRINSICS_) + // x.x,x.y,y.x,y.y + XMVECTOR vTemp1 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(1, 0, 1, 0)); + // x.z,x.w,y.z,y.w + XMVECTOR vTemp3 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + // z.x,z.y,w.x,w.y + XMVECTOR vTemp2 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(1, 0, 1, 0)); + // z.z,z.w,w.z,w.w + XMVECTOR vTemp4 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + + // x.x,y.x,z.x,w.x + XMVECTOR r0 = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + // x.y,y.y,z.y,w.y + XMVECTOR r1 = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + // x.z,y.z,z.z,w.z + XMVECTOR r2 = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + + _mm_storeu_ps(&pDestination->m[0][0], r0); + _mm_storeu_ps(&pDestination->m[1][0], r1); + _mm_storeu_ps(&pDestination->m[2][0], r2); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3x4A +( + XMFLOAT3X4A* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[1].vector4_f32[0]; + pDestination->m[0][2] = M.r[2].vector4_f32[0]; + pDestination->m[0][3] = M.r[3].vector4_f32[0]; + + pDestination->m[1][0] = M.r[0].vector4_f32[1]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[2].vector4_f32[1]; + pDestination->m[1][3] = M.r[3].vector4_f32[1]; + + pDestination->m[2][0] = M.r[0].vector4_f32[2]; + pDestination->m[2][1] = M.r[1].vector4_f32[2]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + pDestination->m[2][3] = M.r[3].vector4_f32[2]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4x2_t P0 = vzipq_f32(M.r[0], M.r[2]); + float32x4x2_t P1 = vzipq_f32(M.r[1], M.r[3]); + + float32x4x2_t T0 = vzipq_f32(P0.val[0], P1.val[0]); + float32x4x2_t T1 = vzipq_f32(P0.val[1], P1.val[1]); + +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1q_f32_ex(&pDestination->m[0][0], T0.val[0], 128); + vst1q_f32_ex(&pDestination->m[1][0], T0.val[1], 128); + vst1q_f32_ex(&pDestination->m[2][0], T1.val[0], 128); +#else + vst1q_f32(&pDestination->m[0][0], T0.val[0]); + vst1q_f32(&pDestination->m[1][0], T0.val[1]); + vst1q_f32(&pDestination->m[2][0], T1.val[0]); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + // x.x,x.y,y.x,y.y + XMVECTOR vTemp1 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(1, 0, 1, 0)); + // x.z,x.w,y.z,y.w + XMVECTOR vTemp3 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + // z.x,z.y,w.x,w.y + XMVECTOR vTemp2 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(1, 0, 1, 0)); + // z.z,z.w,w.z,w.w + XMVECTOR vTemp4 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + + // x.x,y.x,z.x,w.x + XMVECTOR r0 = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + // x.y,y.y,z.y,w.y + XMVECTOR r1 = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + // x.z,y.z,z.z,w.z + XMVECTOR r2 = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + + _mm_store_ps(&pDestination->m[0][0], r0); + _mm_store_ps(&pDestination->m[1][0], r1); + _mm_store_ps(&pDestination->m[2][0], r2); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat4x4 +( + XMFLOAT4X4* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[0].vector4_f32[1]; + pDestination->m[0][2] = M.r[0].vector4_f32[2]; + pDestination->m[0][3] = M.r[0].vector4_f32[3]; + + pDestination->m[1][0] = M.r[1].vector4_f32[0]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[1].vector4_f32[2]; + pDestination->m[1][3] = M.r[1].vector4_f32[3]; + + pDestination->m[2][0] = M.r[2].vector4_f32[0]; + pDestination->m[2][1] = M.r[2].vector4_f32[1]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + pDestination->m[2][3] = M.r[2].vector4_f32[3]; + + pDestination->m[3][0] = M.r[3].vector4_f32[0]; + pDestination->m[3][1] = M.r[3].vector4_f32[1]; + pDestination->m[3][2] = M.r[3].vector4_f32[2]; + pDestination->m[3][3] = M.r[3].vector4_f32[3]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_f32(reinterpret_cast(&pDestination->_11), M.r[0]); + vst1q_f32(reinterpret_cast(&pDestination->_21), M.r[1]); + vst1q_f32(reinterpret_cast(&pDestination->_31), M.r[2]); + vst1q_f32(reinterpret_cast(&pDestination->_41), M.r[3]); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_storeu_ps(&pDestination->_11, M.r[0]); + _mm_storeu_ps(&pDestination->_21, M.r[1]); + _mm_storeu_ps(&pDestination->_31, M.r[2]); + _mm_storeu_ps(&pDestination->_41, M.r[3]); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat4x4A +( + XMFLOAT4X4A* pDestination, + FXMMATRIX M +) noexcept +{ + assert(pDestination); + assert((reinterpret_cast(pDestination) & 0xF) == 0); +#if defined(_XM_NO_INTRINSICS_) + + pDestination->m[0][0] = M.r[0].vector4_f32[0]; + pDestination->m[0][1] = M.r[0].vector4_f32[1]; + pDestination->m[0][2] = M.r[0].vector4_f32[2]; + pDestination->m[0][3] = M.r[0].vector4_f32[3]; + + pDestination->m[1][0] = M.r[1].vector4_f32[0]; + pDestination->m[1][1] = M.r[1].vector4_f32[1]; + pDestination->m[1][2] = M.r[1].vector4_f32[2]; + pDestination->m[1][3] = M.r[1].vector4_f32[3]; + + pDestination->m[2][0] = M.r[2].vector4_f32[0]; + pDestination->m[2][1] = M.r[2].vector4_f32[1]; + pDestination->m[2][2] = M.r[2].vector4_f32[2]; + pDestination->m[2][3] = M.r[2].vector4_f32[3]; + + pDestination->m[3][0] = M.r[3].vector4_f32[0]; + pDestination->m[3][1] = M.r[3].vector4_f32[1]; + pDestination->m[3][2] = M.r[3].vector4_f32[2]; + pDestination->m[3][3] = M.r[3].vector4_f32[3]; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + vst1q_f32_ex(reinterpret_cast(&pDestination->_11), M.r[0], 128); + vst1q_f32_ex(reinterpret_cast(&pDestination->_21), M.r[1], 128); + vst1q_f32_ex(reinterpret_cast(&pDestination->_31), M.r[2], 128); + vst1q_f32_ex(reinterpret_cast(&pDestination->_41), M.r[3], 128); +#else + vst1q_f32(reinterpret_cast(&pDestination->_11), M.r[0]); + vst1q_f32(reinterpret_cast(&pDestination->_21), M.r[1]); + vst1q_f32(reinterpret_cast(&pDestination->_31), M.r[2]); + vst1q_f32(reinterpret_cast(&pDestination->_41), M.r[3]); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_ps(&pDestination->_11, M.r[0]); + _mm_store_ps(&pDestination->_21, M.r[1]); + _mm_store_ps(&pDestination->_31, M.r[2]); + _mm_store_ps(&pDestination->_41, M.r[3]); +#endif +} + diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathMatrix.inl b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathMatrix.inl new file mode 100644 index 00000000..5e96a331 --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathMatrix.inl @@ -0,0 +1,3554 @@ +//------------------------------------------------------------------------------------- +// DirectXMathMatrix.inl -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +/**************************************************************************** + * + * Matrix + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(push) +#pragma float_control(precise, on) +#endif + +// Return true if any entry in the matrix is NaN +inline bool XM_CALLCONV XMMatrixIsNaN(FXMMATRIX M) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + size_t i = 16; + auto pWork = reinterpret_cast(&M.m[0][0]); + do + { + // Fetch value into integer unit + uint32_t uTest = pWork[0]; + // Remove sign + uTest &= 0x7FFFFFFFU; + // NaN is 0x7F800001 through 0x7FFFFFFF inclusive + uTest -= 0x7F800001U; + if (uTest < 0x007FFFFFU) + { + break; // NaN found + } + ++pWork; // Next entry + } + while (--i); + return (i != 0); // i == 0 if nothing matched +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Load in registers + float32x4_t vX = M.r[0]; + float32x4_t vY = M.r[1]; + float32x4_t vZ = M.r[2]; + float32x4_t vW = M.r[3]; + // Test themselves to check for NaN + uint32x4_t xmask = vmvnq_u32(vceqq_f32(vX, vX)); + uint32x4_t ymask = vmvnq_u32(vceqq_f32(vY, vY)); + uint32x4_t zmask = vmvnq_u32(vceqq_f32(vZ, vZ)); + uint32x4_t wmask = vmvnq_u32(vceqq_f32(vW, vW)); + // Or all the results + xmask = vorrq_u32(xmask, zmask); + ymask = vorrq_u32(ymask, wmask); + xmask = vorrq_u32(xmask, ymask); + // If any tested true, return true + uint8x8x2_t vTemp = vzip_u8( + vget_low_u8(vreinterpretq_u8_u32(xmask)), + vget_high_u8(vreinterpretq_u8_u32(xmask))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + return (r != 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Load in registers + XMVECTOR vX = M.r[0]; + XMVECTOR vY = M.r[1]; + XMVECTOR vZ = M.r[2]; + XMVECTOR vW = M.r[3]; + // Test themselves to check for NaN + vX = _mm_cmpneq_ps(vX, vX); + vY = _mm_cmpneq_ps(vY, vY); + vZ = _mm_cmpneq_ps(vZ, vZ); + vW = _mm_cmpneq_ps(vW, vW); + // Or all the results + vX = _mm_or_ps(vX, vZ); + vY = _mm_or_ps(vY, vW); + vX = _mm_or_ps(vX, vY); + // If any tested true, return true + return (_mm_movemask_ps(vX) != 0); +#else +#endif +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(pop) +#endif + +//------------------------------------------------------------------------------ + +// Return true if any entry in the matrix is +/-INF +inline bool XM_CALLCONV XMMatrixIsInfinite(FXMMATRIX M) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + size_t i = 16; + auto pWork = reinterpret_cast(&M.m[0][0]); + do + { + // Fetch value into integer unit + uint32_t uTest = pWork[0]; + // Remove sign + uTest &= 0x7FFFFFFFU; + // INF is 0x7F800000 + if (uTest == 0x7F800000U) + { + break; // INF found + } + ++pWork; // Next entry + } + while (--i); + return (i != 0); // i == 0 if nothing matched +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Load in registers + float32x4_t vX = M.r[0]; + float32x4_t vY = M.r[1]; + float32x4_t vZ = M.r[2]; + float32x4_t vW = M.r[3]; + // Mask off the sign bits + vX = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(vX), g_XMAbsMask)); + vY = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(vY), g_XMAbsMask)); + vZ = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(vZ), g_XMAbsMask)); + vW = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(vW), g_XMAbsMask)); + // Compare to infinity + uint32x4_t xmask = vceqq_f32(vX, g_XMInfinity); + uint32x4_t ymask = vceqq_f32(vY, g_XMInfinity); + uint32x4_t zmask = vceqq_f32(vZ, g_XMInfinity); + uint32x4_t wmask = vceqq_f32(vW, g_XMInfinity); + // Or the answers together + xmask = vorrq_u32(xmask, zmask); + ymask = vorrq_u32(ymask, wmask); + xmask = vorrq_u32(xmask, ymask); + // If any tested true, return true + uint8x8x2_t vTemp = vzip_u8( + vget_low_u8(vreinterpretq_u8_u32(xmask)), + vget_high_u8(vreinterpretq_u8_u32(xmask))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + return (r != 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Mask off the sign bits + XMVECTOR vTemp1 = _mm_and_ps(M.r[0], g_XMAbsMask); + XMVECTOR vTemp2 = _mm_and_ps(M.r[1], g_XMAbsMask); + XMVECTOR vTemp3 = _mm_and_ps(M.r[2], g_XMAbsMask); + XMVECTOR vTemp4 = _mm_and_ps(M.r[3], g_XMAbsMask); + // Compare to infinity + vTemp1 = _mm_cmpeq_ps(vTemp1, g_XMInfinity); + vTemp2 = _mm_cmpeq_ps(vTemp2, g_XMInfinity); + vTemp3 = _mm_cmpeq_ps(vTemp3, g_XMInfinity); + vTemp4 = _mm_cmpeq_ps(vTemp4, g_XMInfinity); + // Or the answers together + vTemp1 = _mm_or_ps(vTemp1, vTemp2); + vTemp3 = _mm_or_ps(vTemp3, vTemp4); + vTemp1 = _mm_or_ps(vTemp1, vTemp3); + // If any are infinity, the signs are true. + return (_mm_movemask_ps(vTemp1) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +// Return true if the XMMatrix is equal to identity +inline bool XM_CALLCONV XMMatrixIsIdentity(FXMMATRIX M) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + // Use the integer pipeline to reduce branching to a minimum + auto pWork = reinterpret_cast(&M.m[0][0]); + // Convert 1.0f to zero and or them together + uint32_t uOne = pWork[0] ^ 0x3F800000U; + // Or all the 0.0f entries together + uint32_t uZero = pWork[1]; + uZero |= pWork[2]; + uZero |= pWork[3]; + // 2nd row + uZero |= pWork[4]; + uOne |= pWork[5] ^ 0x3F800000U; + uZero |= pWork[6]; + uZero |= pWork[7]; + // 3rd row + uZero |= pWork[8]; + uZero |= pWork[9]; + uOne |= pWork[10] ^ 0x3F800000U; + uZero |= pWork[11]; + // 4th row + uZero |= pWork[12]; + uZero |= pWork[13]; + uZero |= pWork[14]; + uOne |= pWork[15] ^ 0x3F800000U; + // If all zero entries are zero, the uZero==0 + uZero &= 0x7FFFFFFF; // Allow -0.0f + // If all 1.0f entries are 1.0f, then uOne==0 + uOne |= uZero; + return (uOne == 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t xmask = vceqq_f32(M.r[0], g_XMIdentityR0); + uint32x4_t ymask = vceqq_f32(M.r[1], g_XMIdentityR1); + uint32x4_t zmask = vceqq_f32(M.r[2], g_XMIdentityR2); + uint32x4_t wmask = vceqq_f32(M.r[3], g_XMIdentityR3); + xmask = vandq_u32(xmask, zmask); + ymask = vandq_u32(ymask, wmask); + xmask = vandq_u32(xmask, ymask); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(xmask)), vget_high_u8(vreinterpretq_u8_u32(xmask))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + return (r == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp1 = _mm_cmpeq_ps(M.r[0], g_XMIdentityR0); + XMVECTOR vTemp2 = _mm_cmpeq_ps(M.r[1], g_XMIdentityR1); + XMVECTOR vTemp3 = _mm_cmpeq_ps(M.r[2], g_XMIdentityR2); + XMVECTOR vTemp4 = _mm_cmpeq_ps(M.r[3], g_XMIdentityR3); + vTemp1 = _mm_and_ps(vTemp1, vTemp2); + vTemp3 = _mm_and_ps(vTemp3, vTemp4); + vTemp1 = _mm_and_ps(vTemp1, vTemp3); + return (_mm_movemask_ps(vTemp1) == 0x0f); +#endif +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ +// Perform a 4x4 matrix multiply by a 4x4 matrix +inline XMMATRIX XM_CALLCONV XMMatrixMultiply +( + FXMMATRIX M1, + CXMMATRIX M2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMMATRIX mResult; + // Cache the invariants in registers + float x = M1.m[0][0]; + float y = M1.m[0][1]; + float z = M1.m[0][2]; + float w = M1.m[0][3]; + // Perform the operation on the first row + mResult.m[0][0] = (M2.m[0][0] * x) + (M2.m[1][0] * y) + (M2.m[2][0] * z) + (M2.m[3][0] * w); + mResult.m[0][1] = (M2.m[0][1] * x) + (M2.m[1][1] * y) + (M2.m[2][1] * z) + (M2.m[3][1] * w); + mResult.m[0][2] = (M2.m[0][2] * x) + (M2.m[1][2] * y) + (M2.m[2][2] * z) + (M2.m[3][2] * w); + mResult.m[0][3] = (M2.m[0][3] * x) + (M2.m[1][3] * y) + (M2.m[2][3] * z) + (M2.m[3][3] * w); + // Repeat for all the other rows + x = M1.m[1][0]; + y = M1.m[1][1]; + z = M1.m[1][2]; + w = M1.m[1][3]; + mResult.m[1][0] = (M2.m[0][0] * x) + (M2.m[1][0] * y) + (M2.m[2][0] * z) + (M2.m[3][0] * w); + mResult.m[1][1] = (M2.m[0][1] * x) + (M2.m[1][1] * y) + (M2.m[2][1] * z) + (M2.m[3][1] * w); + mResult.m[1][2] = (M2.m[0][2] * x) + (M2.m[1][2] * y) + (M2.m[2][2] * z) + (M2.m[3][2] * w); + mResult.m[1][3] = (M2.m[0][3] * x) + (M2.m[1][3] * y) + (M2.m[2][3] * z) + (M2.m[3][3] * w); + x = M1.m[2][0]; + y = M1.m[2][1]; + z = M1.m[2][2]; + w = M1.m[2][3]; + mResult.m[2][0] = (M2.m[0][0] * x) + (M2.m[1][0] * y) + (M2.m[2][0] * z) + (M2.m[3][0] * w); + mResult.m[2][1] = (M2.m[0][1] * x) + (M2.m[1][1] * y) + (M2.m[2][1] * z) + (M2.m[3][1] * w); + mResult.m[2][2] = (M2.m[0][2] * x) + (M2.m[1][2] * y) + (M2.m[2][2] * z) + (M2.m[3][2] * w); + mResult.m[2][3] = (M2.m[0][3] * x) + (M2.m[1][3] * y) + (M2.m[2][3] * z) + (M2.m[3][3] * w); + x = M1.m[3][0]; + y = M1.m[3][1]; + z = M1.m[3][2]; + w = M1.m[3][3]; + mResult.m[3][0] = (M2.m[0][0] * x) + (M2.m[1][0] * y) + (M2.m[2][0] * z) + (M2.m[3][0] * w); + mResult.m[3][1] = (M2.m[0][1] * x) + (M2.m[1][1] * y) + (M2.m[2][1] * z) + (M2.m[3][1] * w); + mResult.m[3][2] = (M2.m[0][2] * x) + (M2.m[1][2] * y) + (M2.m[2][2] * z) + (M2.m[3][2] * w); + mResult.m[3][3] = (M2.m[0][3] * x) + (M2.m[1][3] * y) + (M2.m[2][3] * z) + (M2.m[3][3] * w); + return mResult; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX mResult; + float32x2_t VL = vget_low_f32(M1.r[0]); + float32x2_t VH = vget_high_f32(M1.r[0]); + // Perform the operation on the first row + float32x4_t vX = vmulq_lane_f32(M2.r[0], VL, 0); + float32x4_t vY = vmulq_lane_f32(M2.r[1], VL, 1); + float32x4_t vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + float32x4_t vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + mResult.r[0] = vaddq_f32(vZ, vW); + // Repeat for the other 3 rows + VL = vget_low_f32(M1.r[1]); + VH = vget_high_f32(M1.r[1]); + vX = vmulq_lane_f32(M2.r[0], VL, 0); + vY = vmulq_lane_f32(M2.r[1], VL, 1); + vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + mResult.r[1] = vaddq_f32(vZ, vW); + VL = vget_low_f32(M1.r[2]); + VH = vget_high_f32(M1.r[2]); + vX = vmulq_lane_f32(M2.r[0], VL, 0); + vY = vmulq_lane_f32(M2.r[1], VL, 1); + vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + mResult.r[2] = vaddq_f32(vZ, vW); + VL = vget_low_f32(M1.r[3]); + VH = vget_high_f32(M1.r[3]); + vX = vmulq_lane_f32(M2.r[0], VL, 0); + vY = vmulq_lane_f32(M2.r[1], VL, 1); + vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + mResult.r[3] = vaddq_f32(vZ, vW); + return mResult; +#elif defined(_XM_AVX2_INTRINSICS_) + __m256 t0 = _mm256_castps128_ps256(M1.r[0]); + t0 = _mm256_insertf128_ps(t0, M1.r[1], 1); + __m256 t1 = _mm256_castps128_ps256(M1.r[2]); + t1 = _mm256_insertf128_ps(t1, M1.r[3], 1); + + __m256 u0 = _mm256_castps128_ps256(M2.r[0]); + u0 = _mm256_insertf128_ps(u0, M2.r[1], 1); + __m256 u1 = _mm256_castps128_ps256(M2.r[2]); + u1 = _mm256_insertf128_ps(u1, M2.r[3], 1); + + __m256 a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 b0 = _mm256_permute2f128_ps(u0, u0, 0x00); + __m256 c0 = _mm256_mul_ps(a0, b0); + __m256 c1 = _mm256_mul_ps(a1, b0); + + a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(1, 1, 1, 1)); + a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(1, 1, 1, 1)); + b0 = _mm256_permute2f128_ps(u0, u0, 0x11); + __m256 c2 = _mm256_fmadd_ps(a0, b0, c0); + __m256 c3 = _mm256_fmadd_ps(a1, b0, c1); + + a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(2, 2, 2, 2)); + a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 b1 = _mm256_permute2f128_ps(u1, u1, 0x00); + __m256 c4 = _mm256_mul_ps(a0, b1); + __m256 c5 = _mm256_mul_ps(a1, b1); + + a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(3, 3, 3, 3)); + a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(3, 3, 3, 3)); + b1 = _mm256_permute2f128_ps(u1, u1, 0x11); + __m256 c6 = _mm256_fmadd_ps(a0, b1, c4); + __m256 c7 = _mm256_fmadd_ps(a1, b1, c5); + + t0 = _mm256_add_ps(c2, c6); + t1 = _mm256_add_ps(c3, c7); + + XMMATRIX mResult; + mResult.r[0] = _mm256_castps256_ps128(t0); + mResult.r[1] = _mm256_extractf128_ps(t0, 1); + mResult.r[2] = _mm256_castps256_ps128(t1); + mResult.r[3] = _mm256_extractf128_ps(t1, 1); + return mResult; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX mResult; + // Splat the component X,Y,Z then W +#if defined(_XM_AVX_INTRINSICS_) + XMVECTOR vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 0); + XMVECTOR vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 1); + XMVECTOR vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 2); + XMVECTOR vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 3); +#else + // Use vW to hold the original row + XMVECTOR vW = M1.r[0]; + XMVECTOR vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + // Perform the operation on the first row + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + // Perform a binary add to reduce cumulative errors + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + mResult.r[0] = vX; + // Repeat for the other 3 rows +#if defined(_XM_AVX_INTRINSICS_) + vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 0); + vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 1); + vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 2); + vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 3); +#else + vW = M1.r[1]; + vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + mResult.r[1] = vX; +#if defined(_XM_AVX_INTRINSICS_) + vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 0); + vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 1); + vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 2); + vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 3); +#else + vW = M1.r[2]; + vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + mResult.r[2] = vX; +#if defined(_XM_AVX_INTRINSICS_) + vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 0); + vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 1); + vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 2); + vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 3); +#else + vW = M1.r[3]; + vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + mResult.r[3] = vX; + return mResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixMultiplyTranspose +( + FXMMATRIX M1, + CXMMATRIX M2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMMATRIX mResult; + // Cache the invariants in registers + float x = M2.m[0][0]; + float y = M2.m[1][0]; + float z = M2.m[2][0]; + float w = M2.m[3][0]; + // Perform the operation on the first row + mResult.m[0][0] = (M1.m[0][0] * x) + (M1.m[0][1] * y) + (M1.m[0][2] * z) + (M1.m[0][3] * w); + mResult.m[0][1] = (M1.m[1][0] * x) + (M1.m[1][1] * y) + (M1.m[1][2] * z) + (M1.m[1][3] * w); + mResult.m[0][2] = (M1.m[2][0] * x) + (M1.m[2][1] * y) + (M1.m[2][2] * z) + (M1.m[2][3] * w); + mResult.m[0][3] = (M1.m[3][0] * x) + (M1.m[3][1] * y) + (M1.m[3][2] * z) + (M1.m[3][3] * w); + // Repeat for all the other rows + x = M2.m[0][1]; + y = M2.m[1][1]; + z = M2.m[2][1]; + w = M2.m[3][1]; + mResult.m[1][0] = (M1.m[0][0] * x) + (M1.m[0][1] * y) + (M1.m[0][2] * z) + (M1.m[0][3] * w); + mResult.m[1][1] = (M1.m[1][0] * x) + (M1.m[1][1] * y) + (M1.m[1][2] * z) + (M1.m[1][3] * w); + mResult.m[1][2] = (M1.m[2][0] * x) + (M1.m[2][1] * y) + (M1.m[2][2] * z) + (M1.m[2][3] * w); + mResult.m[1][3] = (M1.m[3][0] * x) + (M1.m[3][1] * y) + (M1.m[3][2] * z) + (M1.m[3][3] * w); + x = M2.m[0][2]; + y = M2.m[1][2]; + z = M2.m[2][2]; + w = M2.m[3][2]; + mResult.m[2][0] = (M1.m[0][0] * x) + (M1.m[0][1] * y) + (M1.m[0][2] * z) + (M1.m[0][3] * w); + mResult.m[2][1] = (M1.m[1][0] * x) + (M1.m[1][1] * y) + (M1.m[1][2] * z) + (M1.m[1][3] * w); + mResult.m[2][2] = (M1.m[2][0] * x) + (M1.m[2][1] * y) + (M1.m[2][2] * z) + (M1.m[2][3] * w); + mResult.m[2][3] = (M1.m[3][0] * x) + (M1.m[3][1] * y) + (M1.m[3][2] * z) + (M1.m[3][3] * w); + x = M2.m[0][3]; + y = M2.m[1][3]; + z = M2.m[2][3]; + w = M2.m[3][3]; + mResult.m[3][0] = (M1.m[0][0] * x) + (M1.m[0][1] * y) + (M1.m[0][2] * z) + (M1.m[0][3] * w); + mResult.m[3][1] = (M1.m[1][0] * x) + (M1.m[1][1] * y) + (M1.m[1][2] * z) + (M1.m[1][3] * w); + mResult.m[3][2] = (M1.m[2][0] * x) + (M1.m[2][1] * y) + (M1.m[2][2] * z) + (M1.m[2][3] * w); + mResult.m[3][3] = (M1.m[3][0] * x) + (M1.m[3][1] * y) + (M1.m[3][2] * z) + (M1.m[3][3] * w); + return mResult; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(M1.r[0]); + float32x2_t VH = vget_high_f32(M1.r[0]); + // Perform the operation on the first row + float32x4_t vX = vmulq_lane_f32(M2.r[0], VL, 0); + float32x4_t vY = vmulq_lane_f32(M2.r[1], VL, 1); + float32x4_t vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + float32x4_t vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + float32x4_t r0 = vaddq_f32(vZ, vW); + // Repeat for the other 3 rows + VL = vget_low_f32(M1.r[1]); + VH = vget_high_f32(M1.r[1]); + vX = vmulq_lane_f32(M2.r[0], VL, 0); + vY = vmulq_lane_f32(M2.r[1], VL, 1); + vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + float32x4_t r1 = vaddq_f32(vZ, vW); + VL = vget_low_f32(M1.r[2]); + VH = vget_high_f32(M1.r[2]); + vX = vmulq_lane_f32(M2.r[0], VL, 0); + vY = vmulq_lane_f32(M2.r[1], VL, 1); + vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + float32x4_t r2 = vaddq_f32(vZ, vW); + VL = vget_low_f32(M1.r[3]); + VH = vget_high_f32(M1.r[3]); + vX = vmulq_lane_f32(M2.r[0], VL, 0); + vY = vmulq_lane_f32(M2.r[1], VL, 1); + vZ = vmlaq_lane_f32(vX, M2.r[2], VH, 0); + vW = vmlaq_lane_f32(vY, M2.r[3], VH, 1); + float32x4_t r3 = vaddq_f32(vZ, vW); + + // Transpose result + float32x4x2_t P0 = vzipq_f32(r0, r2); + float32x4x2_t P1 = vzipq_f32(r1, r3); + + float32x4x2_t T0 = vzipq_f32(P0.val[0], P1.val[0]); + float32x4x2_t T1 = vzipq_f32(P0.val[1], P1.val[1]); + + XMMATRIX mResult; + mResult.r[0] = T0.val[0]; + mResult.r[1] = T0.val[1]; + mResult.r[2] = T1.val[0]; + mResult.r[3] = T1.val[1]; + return mResult; +#elif defined(_XM_AVX2_INTRINSICS_) + __m256 t0 = _mm256_castps128_ps256(M1.r[0]); + t0 = _mm256_insertf128_ps(t0, M1.r[1], 1); + __m256 t1 = _mm256_castps128_ps256(M1.r[2]); + t1 = _mm256_insertf128_ps(t1, M1.r[3], 1); + + __m256 u0 = _mm256_castps128_ps256(M2.r[0]); + u0 = _mm256_insertf128_ps(u0, M2.r[1], 1); + __m256 u1 = _mm256_castps128_ps256(M2.r[2]); + u1 = _mm256_insertf128_ps(u1, M2.r[3], 1); + + __m256 a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 b0 = _mm256_permute2f128_ps(u0, u0, 0x00); + __m256 c0 = _mm256_mul_ps(a0, b0); + __m256 c1 = _mm256_mul_ps(a1, b0); + + a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(1, 1, 1, 1)); + a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(1, 1, 1, 1)); + b0 = _mm256_permute2f128_ps(u0, u0, 0x11); + __m256 c2 = _mm256_fmadd_ps(a0, b0, c0); + __m256 c3 = _mm256_fmadd_ps(a1, b0, c1); + + a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(2, 2, 2, 2)); + a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 b1 = _mm256_permute2f128_ps(u1, u1, 0x00); + __m256 c4 = _mm256_mul_ps(a0, b1); + __m256 c5 = _mm256_mul_ps(a1, b1); + + a0 = _mm256_shuffle_ps(t0, t0, _MM_SHUFFLE(3, 3, 3, 3)); + a1 = _mm256_shuffle_ps(t1, t1, _MM_SHUFFLE(3, 3, 3, 3)); + b1 = _mm256_permute2f128_ps(u1, u1, 0x11); + __m256 c6 = _mm256_fmadd_ps(a0, b1, c4); + __m256 c7 = _mm256_fmadd_ps(a1, b1, c5); + + t0 = _mm256_add_ps(c2, c6); + t1 = _mm256_add_ps(c3, c7); + + // Transpose result + __m256 vTemp = _mm256_unpacklo_ps(t0, t1); + __m256 vTemp2 = _mm256_unpackhi_ps(t0, t1); + __m256 vTemp3 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x20); + __m256 vTemp4 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x31); + vTemp = _mm256_unpacklo_ps(vTemp3, vTemp4); + vTemp2 = _mm256_unpackhi_ps(vTemp3, vTemp4); + t0 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x20); + t1 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x31); + + XMMATRIX mResult; + mResult.r[0] = _mm256_castps256_ps128(t0); + mResult.r[1] = _mm256_extractf128_ps(t0, 1); + mResult.r[2] = _mm256_castps256_ps128(t1); + mResult.r[3] = _mm256_extractf128_ps(t1, 1); + return mResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the component X,Y,Z then W +#if defined(_XM_AVX_INTRINSICS_) + XMVECTOR vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 0); + XMVECTOR vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 1); + XMVECTOR vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 2); + XMVECTOR vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[0]) + 3); +#else + // Use vW to hold the original row + XMVECTOR vW = M1.r[0]; + XMVECTOR vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + // Perform the operation on the first row + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + // Perform a binary add to reduce cumulative errors + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + XMVECTOR r0 = vX; + // Repeat for the other 3 rows +#if defined(_XM_AVX_INTRINSICS_) + vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 0); + vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 1); + vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 2); + vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[1]) + 3); +#else + vW = M1.r[1]; + vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + XMVECTOR r1 = vX; +#if defined(_XM_AVX_INTRINSICS_) + vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 0); + vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 1); + vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 2); + vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[2]) + 3); +#else + vW = M1.r[2]; + vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + XMVECTOR r2 = vX; +#if defined(_XM_AVX_INTRINSICS_) + vX = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 0); + vY = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 1); + vZ = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 2); + vW = _mm_broadcast_ss(reinterpret_cast(&M1.r[3]) + 3); +#else + vW = M1.r[3]; + vX = XM_PERMUTE_PS(vW, _MM_SHUFFLE(0, 0, 0, 0)); + vY = XM_PERMUTE_PS(vW, _MM_SHUFFLE(1, 1, 1, 1)); + vZ = XM_PERMUTE_PS(vW, _MM_SHUFFLE(2, 2, 2, 2)); + vW = XM_PERMUTE_PS(vW, _MM_SHUFFLE(3, 3, 3, 3)); +#endif + vX = _mm_mul_ps(vX, M2.r[0]); + vY = _mm_mul_ps(vY, M2.r[1]); + vZ = _mm_mul_ps(vZ, M2.r[2]); + vW = _mm_mul_ps(vW, M2.r[3]); + vX = _mm_add_ps(vX, vZ); + vY = _mm_add_ps(vY, vW); + vX = _mm_add_ps(vX, vY); + XMVECTOR r3 = vX; + + // Transpose result + // x.x,x.y,y.x,y.y + XMVECTOR vTemp1 = _mm_shuffle_ps(r0, r1, _MM_SHUFFLE(1, 0, 1, 0)); + // x.z,x.w,y.z,y.w + XMVECTOR vTemp3 = _mm_shuffle_ps(r0, r1, _MM_SHUFFLE(3, 2, 3, 2)); + // z.x,z.y,w.x,w.y + XMVECTOR vTemp2 = _mm_shuffle_ps(r2, r3, _MM_SHUFFLE(1, 0, 1, 0)); + // z.z,z.w,w.z,w.w + XMVECTOR vTemp4 = _mm_shuffle_ps(r2, r3, _MM_SHUFFLE(3, 2, 3, 2)); + + XMMATRIX mResult; + // x.x,y.x,z.x,w.x + mResult.r[0] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + // x.y,y.y,z.y,w.y + mResult.r[1] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + // x.z,y.z,z.z,w.z + mResult.r[2] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + // x.w,y.w,z.w,w.w + mResult.r[3] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(3, 1, 3, 1)); + return mResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixTranspose(FXMMATRIX M) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + // Original matrix: + // + // m00m01m02m03 + // m10m11m12m13 + // m20m21m22m23 + // m30m31m32m33 + + XMMATRIX P; + P.r[0] = XMVectorMergeXY(M.r[0], M.r[2]); // m00m20m01m21 + P.r[1] = XMVectorMergeXY(M.r[1], M.r[3]); // m10m30m11m31 + P.r[2] = XMVectorMergeZW(M.r[0], M.r[2]); // m02m22m03m23 + P.r[3] = XMVectorMergeZW(M.r[1], M.r[3]); // m12m32m13m33 + + XMMATRIX MT; + MT.r[0] = XMVectorMergeXY(P.r[0], P.r[1]); // m00m10m20m30 + MT.r[1] = XMVectorMergeZW(P.r[0], P.r[1]); // m01m11m21m31 + MT.r[2] = XMVectorMergeXY(P.r[2], P.r[3]); // m02m12m22m32 + MT.r[3] = XMVectorMergeZW(P.r[2], P.r[3]); // m03m13m23m33 + return MT; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4x2_t P0 = vzipq_f32(M.r[0], M.r[2]); + float32x4x2_t P1 = vzipq_f32(M.r[1], M.r[3]); + + float32x4x2_t T0 = vzipq_f32(P0.val[0], P1.val[0]); + float32x4x2_t T1 = vzipq_f32(P0.val[1], P1.val[1]); + + XMMATRIX mResult; + mResult.r[0] = T0.val[0]; + mResult.r[1] = T0.val[1]; + mResult.r[2] = T1.val[0]; + mResult.r[3] = T1.val[1]; + return mResult; +#elif defined(_XM_AVX2_INTRINSICS_) + __m256 t0 = _mm256_castps128_ps256(M.r[0]); + t0 = _mm256_insertf128_ps(t0, M.r[1], 1); + __m256 t1 = _mm256_castps128_ps256(M.r[2]); + t1 = _mm256_insertf128_ps(t1, M.r[3], 1); + + __m256 vTemp = _mm256_unpacklo_ps(t0, t1); + __m256 vTemp2 = _mm256_unpackhi_ps(t0, t1); + __m256 vTemp3 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x20); + __m256 vTemp4 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x31); + vTemp = _mm256_unpacklo_ps(vTemp3, vTemp4); + vTemp2 = _mm256_unpackhi_ps(vTemp3, vTemp4); + t0 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x20); + t1 = _mm256_permute2f128_ps(vTemp, vTemp2, 0x31); + + XMMATRIX mResult; + mResult.r[0] = _mm256_castps256_ps128(t0); + mResult.r[1] = _mm256_extractf128_ps(t0, 1); + mResult.r[2] = _mm256_castps256_ps128(t1); + mResult.r[3] = _mm256_extractf128_ps(t1, 1); + return mResult; +#elif defined(_XM_SSE_INTRINSICS_) + // x.x,x.y,y.x,y.y + XMVECTOR vTemp1 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(1, 0, 1, 0)); + // x.z,x.w,y.z,y.w + XMVECTOR vTemp3 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + // z.x,z.y,w.x,w.y + XMVECTOR vTemp2 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(1, 0, 1, 0)); + // z.z,z.w,w.z,w.w + XMVECTOR vTemp4 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + + XMMATRIX mResult; + // x.x,y.x,z.x,w.x + mResult.r[0] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + // x.y,y.y,z.y,w.y + mResult.r[1] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + // x.z,y.z,z.z,w.z + mResult.r[2] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + // x.w,y.w,z.w,w.w + mResult.r[3] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(3, 1, 3, 1)); + return mResult; +#endif +} + +//------------------------------------------------------------------------------ +// Return the inverse and the determinant of a 4x4 matrix +_Use_decl_annotations_ +inline XMMATRIX XM_CALLCONV XMMatrixInverse +( + XMVECTOR* pDeterminant, + FXMMATRIX M +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + XMMATRIX MT = XMMatrixTranspose(M); + + XMVECTOR V0[4], V1[4]; + V0[0] = XMVectorSwizzle(MT.r[2]); + V1[0] = XMVectorSwizzle(MT.r[3]); + V0[1] = XMVectorSwizzle(MT.r[0]); + V1[1] = XMVectorSwizzle(MT.r[1]); + V0[2] = XMVectorPermute(MT.r[2], MT.r[0]); + V1[2] = XMVectorPermute(MT.r[3], MT.r[1]); + + XMVECTOR D0 = XMVectorMultiply(V0[0], V1[0]); + XMVECTOR D1 = XMVectorMultiply(V0[1], V1[1]); + XMVECTOR D2 = XMVectorMultiply(V0[2], V1[2]); + + V0[0] = XMVectorSwizzle(MT.r[2]); + V1[0] = XMVectorSwizzle(MT.r[3]); + V0[1] = XMVectorSwizzle(MT.r[0]); + V1[1] = XMVectorSwizzle(MT.r[1]); + V0[2] = XMVectorPermute(MT.r[2], MT.r[0]); + V1[2] = XMVectorPermute(MT.r[3], MT.r[1]); + + D0 = XMVectorNegativeMultiplySubtract(V0[0], V1[0], D0); + D1 = XMVectorNegativeMultiplySubtract(V0[1], V1[1], D1); + D2 = XMVectorNegativeMultiplySubtract(V0[2], V1[2], D2); + + V0[0] = XMVectorSwizzle(MT.r[1]); + V1[0] = XMVectorPermute(D0, D2); + V0[1] = XMVectorSwizzle(MT.r[0]); + V1[1] = XMVectorPermute(D0, D2); + V0[2] = XMVectorSwizzle(MT.r[3]); + V1[2] = XMVectorPermute(D1, D2); + V0[3] = XMVectorSwizzle(MT.r[2]); + V1[3] = XMVectorPermute(D1, D2); + + XMVECTOR C0 = XMVectorMultiply(V0[0], V1[0]); + XMVECTOR C2 = XMVectorMultiply(V0[1], V1[1]); + XMVECTOR C4 = XMVectorMultiply(V0[2], V1[2]); + XMVECTOR C6 = XMVectorMultiply(V0[3], V1[3]); + + V0[0] = XMVectorSwizzle(MT.r[1]); + V1[0] = XMVectorPermute(D0, D2); + V0[1] = XMVectorSwizzle(MT.r[0]); + V1[1] = XMVectorPermute(D0, D2); + V0[2] = XMVectorSwizzle(MT.r[3]); + V1[2] = XMVectorPermute(D1, D2); + V0[3] = XMVectorSwizzle(MT.r[2]); + V1[3] = XMVectorPermute(D1, D2); + + C0 = XMVectorNegativeMultiplySubtract(V0[0], V1[0], C0); + C2 = XMVectorNegativeMultiplySubtract(V0[1], V1[1], C2); + C4 = XMVectorNegativeMultiplySubtract(V0[2], V1[2], C4); + C6 = XMVectorNegativeMultiplySubtract(V0[3], V1[3], C6); + + V0[0] = XMVectorSwizzle(MT.r[1]); + V1[0] = XMVectorPermute(D0, D2); + V0[1] = XMVectorSwizzle(MT.r[0]); + V1[1] = XMVectorPermute(D0, D2); + V0[2] = XMVectorSwizzle(MT.r[3]); + V1[2] = XMVectorPermute(D1, D2); + V0[3] = XMVectorSwizzle(MT.r[2]); + V1[3] = XMVectorPermute(D1, D2); + + XMVECTOR C1 = XMVectorNegativeMultiplySubtract(V0[0], V1[0], C0); + C0 = XMVectorMultiplyAdd(V0[0], V1[0], C0); + XMVECTOR C3 = XMVectorMultiplyAdd(V0[1], V1[1], C2); + C2 = XMVectorNegativeMultiplySubtract(V0[1], V1[1], C2); + XMVECTOR C5 = XMVectorNegativeMultiplySubtract(V0[2], V1[2], C4); + C4 = XMVectorMultiplyAdd(V0[2], V1[2], C4); + XMVECTOR C7 = XMVectorMultiplyAdd(V0[3], V1[3], C6); + C6 = XMVectorNegativeMultiplySubtract(V0[3], V1[3], C6); + + XMMATRIX R; + R.r[0] = XMVectorSelect(C0, C1, g_XMSelect0101.v); + R.r[1] = XMVectorSelect(C2, C3, g_XMSelect0101.v); + R.r[2] = XMVectorSelect(C4, C5, g_XMSelect0101.v); + R.r[3] = XMVectorSelect(C6, C7, g_XMSelect0101.v); + + XMVECTOR Determinant = XMVector4Dot(R.r[0], MT.r[0]); + + if (pDeterminant != nullptr) + *pDeterminant = Determinant; + + XMVECTOR Reciprocal = XMVectorReciprocal(Determinant); + + XMMATRIX Result; + Result.r[0] = XMVectorMultiply(R.r[0], Reciprocal); + Result.r[1] = XMVectorMultiply(R.r[1], Reciprocal); + Result.r[2] = XMVectorMultiply(R.r[2], Reciprocal); + Result.r[3] = XMVectorMultiply(R.r[3], Reciprocal); + return Result; + +#elif defined(_XM_SSE_INTRINSICS_) + // Transpose matrix + XMVECTOR vTemp1 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(1, 0, 1, 0)); + XMVECTOR vTemp3 = _mm_shuffle_ps(M.r[0], M.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + XMVECTOR vTemp2 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(1, 0, 1, 0)); + XMVECTOR vTemp4 = _mm_shuffle_ps(M.r[2], M.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + + XMMATRIX MT; + MT.r[0] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(2, 0, 2, 0)); + MT.r[1] = _mm_shuffle_ps(vTemp1, vTemp2, _MM_SHUFFLE(3, 1, 3, 1)); + MT.r[2] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(2, 0, 2, 0)); + MT.r[3] = _mm_shuffle_ps(vTemp3, vTemp4, _MM_SHUFFLE(3, 1, 3, 1)); + + XMVECTOR V00 = XM_PERMUTE_PS(MT.r[2], _MM_SHUFFLE(1, 1, 0, 0)); + XMVECTOR V10 = XM_PERMUTE_PS(MT.r[3], _MM_SHUFFLE(3, 2, 3, 2)); + XMVECTOR V01 = XM_PERMUTE_PS(MT.r[0], _MM_SHUFFLE(1, 1, 0, 0)); + XMVECTOR V11 = XM_PERMUTE_PS(MT.r[1], _MM_SHUFFLE(3, 2, 3, 2)); + XMVECTOR V02 = _mm_shuffle_ps(MT.r[2], MT.r[0], _MM_SHUFFLE(2, 0, 2, 0)); + XMVECTOR V12 = _mm_shuffle_ps(MT.r[3], MT.r[1], _MM_SHUFFLE(3, 1, 3, 1)); + + XMVECTOR D0 = _mm_mul_ps(V00, V10); + XMVECTOR D1 = _mm_mul_ps(V01, V11); + XMVECTOR D2 = _mm_mul_ps(V02, V12); + + V00 = XM_PERMUTE_PS(MT.r[2], _MM_SHUFFLE(3, 2, 3, 2)); + V10 = XM_PERMUTE_PS(MT.r[3], _MM_SHUFFLE(1, 1, 0, 0)); + V01 = XM_PERMUTE_PS(MT.r[0], _MM_SHUFFLE(3, 2, 3, 2)); + V11 = XM_PERMUTE_PS(MT.r[1], _MM_SHUFFLE(1, 1, 0, 0)); + V02 = _mm_shuffle_ps(MT.r[2], MT.r[0], _MM_SHUFFLE(3, 1, 3, 1)); + V12 = _mm_shuffle_ps(MT.r[3], MT.r[1], _MM_SHUFFLE(2, 0, 2, 0)); + + D0 = XM_FNMADD_PS(V00, V10, D0); + D1 = XM_FNMADD_PS(V01, V11, D1); + D2 = XM_FNMADD_PS(V02, V12, D2); + // V11 = D0Y,D0W,D2Y,D2Y + V11 = _mm_shuffle_ps(D0, D2, _MM_SHUFFLE(1, 1, 3, 1)); + V00 = XM_PERMUTE_PS(MT.r[1], _MM_SHUFFLE(1, 0, 2, 1)); + V10 = _mm_shuffle_ps(V11, D0, _MM_SHUFFLE(0, 3, 0, 2)); + V01 = XM_PERMUTE_PS(MT.r[0], _MM_SHUFFLE(0, 1, 0, 2)); + V11 = _mm_shuffle_ps(V11, D0, _MM_SHUFFLE(2, 1, 2, 1)); + // V13 = D1Y,D1W,D2W,D2W + XMVECTOR V13 = _mm_shuffle_ps(D1, D2, _MM_SHUFFLE(3, 3, 3, 1)); + V02 = XM_PERMUTE_PS(MT.r[3], _MM_SHUFFLE(1, 0, 2, 1)); + V12 = _mm_shuffle_ps(V13, D1, _MM_SHUFFLE(0, 3, 0, 2)); + XMVECTOR V03 = XM_PERMUTE_PS(MT.r[2], _MM_SHUFFLE(0, 1, 0, 2)); + V13 = _mm_shuffle_ps(V13, D1, _MM_SHUFFLE(2, 1, 2, 1)); + + XMVECTOR C0 = _mm_mul_ps(V00, V10); + XMVECTOR C2 = _mm_mul_ps(V01, V11); + XMVECTOR C4 = _mm_mul_ps(V02, V12); + XMVECTOR C6 = _mm_mul_ps(V03, V13); + + // V11 = D0X,D0Y,D2X,D2X + V11 = _mm_shuffle_ps(D0, D2, _MM_SHUFFLE(0, 0, 1, 0)); + V00 = XM_PERMUTE_PS(MT.r[1], _MM_SHUFFLE(2, 1, 3, 2)); + V10 = _mm_shuffle_ps(D0, V11, _MM_SHUFFLE(2, 1, 0, 3)); + V01 = XM_PERMUTE_PS(MT.r[0], _MM_SHUFFLE(1, 3, 2, 3)); + V11 = _mm_shuffle_ps(D0, V11, _MM_SHUFFLE(0, 2, 1, 2)); + // V13 = D1X,D1Y,D2Z,D2Z + V13 = _mm_shuffle_ps(D1, D2, _MM_SHUFFLE(2, 2, 1, 0)); + V02 = XM_PERMUTE_PS(MT.r[3], _MM_SHUFFLE(2, 1, 3, 2)); + V12 = _mm_shuffle_ps(D1, V13, _MM_SHUFFLE(2, 1, 0, 3)); + V03 = XM_PERMUTE_PS(MT.r[2], _MM_SHUFFLE(1, 3, 2, 3)); + V13 = _mm_shuffle_ps(D1, V13, _MM_SHUFFLE(0, 2, 1, 2)); + + C0 = XM_FNMADD_PS(V00, V10, C0); + C2 = XM_FNMADD_PS(V01, V11, C2); + C4 = XM_FNMADD_PS(V02, V12, C4); + C6 = XM_FNMADD_PS(V03, V13, C6); + + V00 = XM_PERMUTE_PS(MT.r[1], _MM_SHUFFLE(0, 3, 0, 3)); + // V10 = D0Z,D0Z,D2X,D2Y + V10 = _mm_shuffle_ps(D0, D2, _MM_SHUFFLE(1, 0, 2, 2)); + V10 = XM_PERMUTE_PS(V10, _MM_SHUFFLE(0, 2, 3, 0)); + V01 = XM_PERMUTE_PS(MT.r[0], _MM_SHUFFLE(2, 0, 3, 1)); + // V11 = D0X,D0W,D2X,D2Y + V11 = _mm_shuffle_ps(D0, D2, _MM_SHUFFLE(1, 0, 3, 0)); + V11 = XM_PERMUTE_PS(V11, _MM_SHUFFLE(2, 1, 0, 3)); + V02 = XM_PERMUTE_PS(MT.r[3], _MM_SHUFFLE(0, 3, 0, 3)); + // V12 = D1Z,D1Z,D2Z,D2W + V12 = _mm_shuffle_ps(D1, D2, _MM_SHUFFLE(3, 2, 2, 2)); + V12 = XM_PERMUTE_PS(V12, _MM_SHUFFLE(0, 2, 3, 0)); + V03 = XM_PERMUTE_PS(MT.r[2], _MM_SHUFFLE(2, 0, 3, 1)); + // V13 = D1X,D1W,D2Z,D2W + V13 = _mm_shuffle_ps(D1, D2, _MM_SHUFFLE(3, 2, 3, 0)); + V13 = XM_PERMUTE_PS(V13, _MM_SHUFFLE(2, 1, 0, 3)); + + V00 = _mm_mul_ps(V00, V10); + V01 = _mm_mul_ps(V01, V11); + V02 = _mm_mul_ps(V02, V12); + V03 = _mm_mul_ps(V03, V13); + XMVECTOR C1 = _mm_sub_ps(C0, V00); + C0 = _mm_add_ps(C0, V00); + XMVECTOR C3 = _mm_add_ps(C2, V01); + C2 = _mm_sub_ps(C2, V01); + XMVECTOR C5 = _mm_sub_ps(C4, V02); + C4 = _mm_add_ps(C4, V02); + XMVECTOR C7 = _mm_add_ps(C6, V03); + C6 = _mm_sub_ps(C6, V03); + + C0 = _mm_shuffle_ps(C0, C1, _MM_SHUFFLE(3, 1, 2, 0)); + C2 = _mm_shuffle_ps(C2, C3, _MM_SHUFFLE(3, 1, 2, 0)); + C4 = _mm_shuffle_ps(C4, C5, _MM_SHUFFLE(3, 1, 2, 0)); + C6 = _mm_shuffle_ps(C6, C7, _MM_SHUFFLE(3, 1, 2, 0)); + C0 = XM_PERMUTE_PS(C0, _MM_SHUFFLE(3, 1, 2, 0)); + C2 = XM_PERMUTE_PS(C2, _MM_SHUFFLE(3, 1, 2, 0)); + C4 = XM_PERMUTE_PS(C4, _MM_SHUFFLE(3, 1, 2, 0)); + C6 = XM_PERMUTE_PS(C6, _MM_SHUFFLE(3, 1, 2, 0)); + // Get the determinant + XMVECTOR vTemp = XMVector4Dot(C0, MT.r[0]); + if (pDeterminant != nullptr) + *pDeterminant = vTemp; + vTemp = _mm_div_ps(g_XMOne, vTemp); + XMMATRIX mResult; + mResult.r[0] = _mm_mul_ps(C0, vTemp); + mResult.r[1] = _mm_mul_ps(C2, vTemp); + mResult.r[2] = _mm_mul_ps(C4, vTemp); + mResult.r[3] = _mm_mul_ps(C6, vTemp); + return mResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixVectorTensorProduct +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + XMMATRIX mResult; + mResult.r[0] = XMVectorMultiply(XMVectorSwizzle<0, 0, 0, 0>(V1), V2); + mResult.r[1] = XMVectorMultiply(XMVectorSwizzle<1, 1, 1, 1>(V1), V2); + mResult.r[2] = XMVectorMultiply(XMVectorSwizzle<2, 2, 2, 2>(V1), V2); + mResult.r[3] = XMVectorMultiply(XMVectorSwizzle<3, 3, 3, 3>(V1), V2); + return mResult; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMMatrixDeterminant(FXMMATRIX M) noexcept +{ + static const XMVECTORF32 Sign = { { { 1.0f, -1.0f, 1.0f, -1.0f } } }; + + XMVECTOR V0 = XMVectorSwizzle(M.r[2]); + XMVECTOR V1 = XMVectorSwizzle(M.r[3]); + XMVECTOR V2 = XMVectorSwizzle(M.r[2]); + XMVECTOR V3 = XMVectorSwizzle(M.r[3]); + XMVECTOR V4 = XMVectorSwizzle(M.r[2]); + XMVECTOR V5 = XMVectorSwizzle(M.r[3]); + + XMVECTOR P0 = XMVectorMultiply(V0, V1); + XMVECTOR P1 = XMVectorMultiply(V2, V3); + XMVECTOR P2 = XMVectorMultiply(V4, V5); + + V0 = XMVectorSwizzle(M.r[2]); + V1 = XMVectorSwizzle(M.r[3]); + V2 = XMVectorSwizzle(M.r[2]); + V3 = XMVectorSwizzle(M.r[3]); + V4 = XMVectorSwizzle(M.r[2]); + V5 = XMVectorSwizzle(M.r[3]); + + P0 = XMVectorNegativeMultiplySubtract(V0, V1, P0); + P1 = XMVectorNegativeMultiplySubtract(V2, V3, P1); + P2 = XMVectorNegativeMultiplySubtract(V4, V5, P2); + + V0 = XMVectorSwizzle(M.r[1]); + V1 = XMVectorSwizzle(M.r[1]); + V2 = XMVectorSwizzle(M.r[1]); + + XMVECTOR S = XMVectorMultiply(M.r[0], Sign.v); + XMVECTOR R = XMVectorMultiply(V0, P0); + R = XMVectorNegativeMultiplySubtract(V1, P1, R); + R = XMVectorMultiplyAdd(V2, P2, R); + + return XMVector4Dot(S, R); +} + +#define XM3RANKDECOMPOSE(a, b, c, x, y, z) \ + if((x) < (y)) \ + { \ + if((y) < (z)) \ + { \ + (a) = 2; \ + (b) = 1; \ + (c) = 0; \ + } \ + else \ + { \ + (a) = 1; \ + \ + if((x) < (z)) \ + { \ + (b) = 2; \ + (c) = 0; \ + } \ + else \ + { \ + (b) = 0; \ + (c) = 2; \ + } \ + } \ + } \ + else \ + { \ + if((x) < (z)) \ + { \ + (a) = 2; \ + (b) = 0; \ + (c) = 1; \ + } \ + else \ + { \ + (a) = 0; \ + \ + if((y) < (z)) \ + { \ + (b) = 2; \ + (c) = 1; \ + } \ + else \ + { \ + (b) = 1; \ + (c) = 2; \ + } \ + } \ + } + +#define XM3_DECOMP_EPSILON 0.0001f + +_Use_decl_annotations_ +inline bool XM_CALLCONV XMMatrixDecompose +( + XMVECTOR* outScale, + XMVECTOR* outRotQuat, + XMVECTOR* outTrans, + FXMMATRIX M +) noexcept +{ + static const XMVECTOR* pvCanonicalBasis[3] = { + &g_XMIdentityR0.v, + &g_XMIdentityR1.v, + &g_XMIdentityR2.v + }; + + assert(outScale != nullptr); + assert(outRotQuat != nullptr); + assert(outTrans != nullptr); + + // Get the translation + outTrans[0] = M.r[3]; + + XMVECTOR* ppvBasis[3]; + XMMATRIX matTemp; + ppvBasis[0] = &matTemp.r[0]; + ppvBasis[1] = &matTemp.r[1]; + ppvBasis[2] = &matTemp.r[2]; + + matTemp.r[0] = M.r[0]; + matTemp.r[1] = M.r[1]; + matTemp.r[2] = M.r[2]; + matTemp.r[3] = g_XMIdentityR3.v; + + auto pfScales = reinterpret_cast(outScale); + + size_t a, b, c; + XMVectorGetXPtr(&pfScales[0], XMVector3Length(ppvBasis[0][0])); + XMVectorGetXPtr(&pfScales[1], XMVector3Length(ppvBasis[1][0])); + XMVectorGetXPtr(&pfScales[2], XMVector3Length(ppvBasis[2][0])); + pfScales[3] = 0.f; + + XM3RANKDECOMPOSE(a, b, c, pfScales[0], pfScales[1], pfScales[2]) + + if (pfScales[a] < XM3_DECOMP_EPSILON) + { + ppvBasis[a][0] = pvCanonicalBasis[a][0]; + } + ppvBasis[a][0] = XMVector3Normalize(ppvBasis[a][0]); + + if (pfScales[b] < XM3_DECOMP_EPSILON) + { + size_t aa, bb, cc; + float fAbsX, fAbsY, fAbsZ; + + fAbsX = fabsf(XMVectorGetX(ppvBasis[a][0])); + fAbsY = fabsf(XMVectorGetY(ppvBasis[a][0])); + fAbsZ = fabsf(XMVectorGetZ(ppvBasis[a][0])); + + XM3RANKDECOMPOSE(aa, bb, cc, fAbsX, fAbsY, fAbsZ) + + ppvBasis[b][0] = XMVector3Cross(ppvBasis[a][0], pvCanonicalBasis[cc][0]); + } + + ppvBasis[b][0] = XMVector3Normalize(ppvBasis[b][0]); + + if (pfScales[c] < XM3_DECOMP_EPSILON) + { + ppvBasis[c][0] = XMVector3Cross(ppvBasis[a][0], ppvBasis[b][0]); + } + + ppvBasis[c][0] = XMVector3Normalize(ppvBasis[c][0]); + + float fDet = XMVectorGetX(XMMatrixDeterminant(matTemp)); + + // use Kramer's rule to check for handedness of coordinate system + if (fDet < 0.0f) + { + // switch coordinate system by negating the scale and inverting the basis vector on the x-axis + pfScales[a] = -pfScales[a]; + ppvBasis[a][0] = XMVectorNegate(ppvBasis[a][0]); + + fDet = -fDet; + } + + fDet -= 1.0f; + fDet *= fDet; + + if (XM3_DECOMP_EPSILON < fDet) + { + // Non-SRT matrix encountered + return false; + } + + // generate the quaternion from the matrix + outRotQuat[0] = XMQuaternionRotationMatrix(matTemp); + return true; +} + +#undef XM3_DECOMP_EPSILON +#undef XM3RANKDECOMPOSE + +//------------------------------------------------------------------------------ +// Transformation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixIdentity() noexcept +{ + XMMATRIX M; + M.r[0] = g_XMIdentityR0.v; + M.r[1] = g_XMIdentityR1.v; + M.r[2] = g_XMIdentityR2.v; + M.r[3] = g_XMIdentityR3.v; + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixSet +( + float m00, float m01, float m02, float m03, + float m10, float m11, float m12, float m13, + float m20, float m21, float m22, float m23, + float m30, float m31, float m32, float m33 +) noexcept +{ + XMMATRIX M; +#if defined(_XM_NO_INTRINSICS_) + M.m[0][0] = m00; M.m[0][1] = m01; M.m[0][2] = m02; M.m[0][3] = m03; + M.m[1][0] = m10; M.m[1][1] = m11; M.m[1][2] = m12; M.m[1][3] = m13; + M.m[2][0] = m20; M.m[2][1] = m21; M.m[2][2] = m22; M.m[2][3] = m23; + M.m[3][0] = m30; M.m[3][1] = m31; M.m[3][2] = m32; M.m[3][3] = m33; +#else + M.r[0] = XMVectorSet(m00, m01, m02, m03); + M.r[1] = XMVectorSet(m10, m11, m12, m13); + M.r[2] = XMVectorSet(m20, m21, m22, m23); + M.r[3] = XMVectorSet(m30, m31, m32, m33); +#endif + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixTranslation +( + float OffsetX, + float OffsetY, + float OffsetZ +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.m[0][0] = 1.0f; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = 1.0f; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = 1.0f; + M.m[2][3] = 0.0f; + + M.m[3][0] = OffsetX; + M.m[3][1] = OffsetY; + M.m[3][2] = OffsetZ; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_SSE_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX M; + M.r[0] = g_XMIdentityR0.v; + M.r[1] = g_XMIdentityR1.v; + M.r[2] = g_XMIdentityR2.v; + M.r[3] = XMVectorSet(OffsetX, OffsetY, OffsetZ, 1.f); + return M; +#endif +} + + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixTranslationFromVector(FXMVECTOR Offset) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.m[0][0] = 1.0f; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = 1.0f; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = 1.0f; + M.m[2][3] = 0.0f; + + M.m[3][0] = Offset.vector4_f32[0]; + M.m[3][1] = Offset.vector4_f32[1]; + M.m[3][2] = Offset.vector4_f32[2]; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_SSE_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX M; + M.r[0] = g_XMIdentityR0.v; + M.r[1] = g_XMIdentityR1.v; + M.r[2] = g_XMIdentityR2.v; + M.r[3] = XMVectorSelect(g_XMIdentityR3.v, Offset, g_XMSelect1110.v); + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixScaling +( + float ScaleX, + float ScaleY, + float ScaleZ +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.m[0][0] = ScaleX; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = ScaleY; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = ScaleZ; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + const XMVECTOR Zero = vdupq_n_f32(0); + XMMATRIX M; + M.r[0] = vsetq_lane_f32(ScaleX, Zero, 0); + M.r[1] = vsetq_lane_f32(ScaleY, Zero, 1); + M.r[2] = vsetq_lane_f32(ScaleZ, Zero, 2); + M.r[3] = g_XMIdentityR3.v; + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + M.r[0] = _mm_set_ps(0, 0, 0, ScaleX); + M.r[1] = _mm_set_ps(0, 0, ScaleY, 0); + M.r[2] = _mm_set_ps(0, ScaleZ, 0, 0); + M.r[3] = g_XMIdentityR3.v; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixScalingFromVector(FXMVECTOR Scale) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMMATRIX M; + M.m[0][0] = Scale.vector4_f32[0]; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = Scale.vector4_f32[1]; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = Scale.vector4_f32[2]; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX M; + M.r[0] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(Scale), g_XMMaskX)); + M.r[1] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(Scale), g_XMMaskY)); + M.r[2] = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(Scale), g_XMMaskZ)); + M.r[3] = g_XMIdentityR3.v; + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + M.r[0] = _mm_and_ps(Scale, g_XMMaskX); + M.r[1] = _mm_and_ps(Scale, g_XMMaskY); + M.r[2] = _mm_and_ps(Scale, g_XMMaskZ); + M.r[3] = g_XMIdentityR3.v; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationX(float Angle) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + XMMATRIX M; + M.m[0][0] = 1.0f; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = fCosAngle; + M.m[1][2] = fSinAngle; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = -fSinAngle; + M.m[2][2] = fCosAngle; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + const float32x4_t Zero = vdupq_n_f32(0); + + float32x4_t T1 = vsetq_lane_f32(fCosAngle, Zero, 1); + T1 = vsetq_lane_f32(fSinAngle, T1, 2); + + float32x4_t T2 = vsetq_lane_f32(-fSinAngle, Zero, 1); + T2 = vsetq_lane_f32(fCosAngle, T2, 2); + + XMMATRIX M; + M.r[0] = g_XMIdentityR0.v; + M.r[1] = T1; + M.r[2] = T2; + M.r[3] = g_XMIdentityR3.v; + return M; +#elif defined(_XM_SSE_INTRINSICS_) + float SinAngle; + float CosAngle; + XMScalarSinCos(&SinAngle, &CosAngle, Angle); + + XMVECTOR vSin = _mm_set_ss(SinAngle); + XMVECTOR vCos = _mm_set_ss(CosAngle); + // x = 0,y = cos,z = sin, w = 0 + vCos = _mm_shuffle_ps(vCos, vSin, _MM_SHUFFLE(3, 0, 0, 3)); + XMMATRIX M; + M.r[0] = g_XMIdentityR0; + M.r[1] = vCos; + // x = 0,y = sin,z = cos, w = 0 + vCos = XM_PERMUTE_PS(vCos, _MM_SHUFFLE(3, 1, 2, 0)); + // x = 0,y = -sin,z = cos, w = 0 + vCos = _mm_mul_ps(vCos, g_XMNegateY); + M.r[2] = vCos; + M.r[3] = g_XMIdentityR3; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationY(float Angle) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + XMMATRIX M; + M.m[0][0] = fCosAngle; + M.m[0][1] = 0.0f; + M.m[0][2] = -fSinAngle; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = 1.0f; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = fSinAngle; + M.m[2][1] = 0.0f; + M.m[2][2] = fCosAngle; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + const float32x4_t Zero = vdupq_n_f32(0); + + float32x4_t T0 = vsetq_lane_f32(fCosAngle, Zero, 0); + T0 = vsetq_lane_f32(-fSinAngle, T0, 2); + + float32x4_t T2 = vsetq_lane_f32(fSinAngle, Zero, 0); + T2 = vsetq_lane_f32(fCosAngle, T2, 2); + + XMMATRIX M; + M.r[0] = T0; + M.r[1] = g_XMIdentityR1.v; + M.r[2] = T2; + M.r[3] = g_XMIdentityR3.v; + return M; +#elif defined(_XM_SSE_INTRINSICS_) + float SinAngle; + float CosAngle; + XMScalarSinCos(&SinAngle, &CosAngle, Angle); + + XMVECTOR vSin = _mm_set_ss(SinAngle); + XMVECTOR vCos = _mm_set_ss(CosAngle); + // x = sin,y = 0,z = cos, w = 0 + vSin = _mm_shuffle_ps(vSin, vCos, _MM_SHUFFLE(3, 0, 3, 0)); + XMMATRIX M; + M.r[2] = vSin; + M.r[1] = g_XMIdentityR1; + // x = cos,y = 0,z = sin, w = 0 + vSin = XM_PERMUTE_PS(vSin, _MM_SHUFFLE(3, 0, 1, 2)); + // x = cos,y = 0,z = -sin, w = 0 + vSin = _mm_mul_ps(vSin, g_XMNegateZ); + M.r[0] = vSin; + M.r[3] = g_XMIdentityR3; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationZ(float Angle) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + XMMATRIX M; + M.m[0][0] = fCosAngle; + M.m[0][1] = fSinAngle; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = -fSinAngle; + M.m[1][1] = fCosAngle; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = 1.0f; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + const float32x4_t Zero = vdupq_n_f32(0); + + float32x4_t T0 = vsetq_lane_f32(fCosAngle, Zero, 0); + T0 = vsetq_lane_f32(fSinAngle, T0, 1); + + float32x4_t T1 = vsetq_lane_f32(-fSinAngle, Zero, 0); + T1 = vsetq_lane_f32(fCosAngle, T1, 1); + + XMMATRIX M; + M.r[0] = T0; + M.r[1] = T1; + M.r[2] = g_XMIdentityR2.v; + M.r[3] = g_XMIdentityR3.v; + return M; +#elif defined(_XM_SSE_INTRINSICS_) + float SinAngle; + float CosAngle; + XMScalarSinCos(&SinAngle, &CosAngle, Angle); + + XMVECTOR vSin = _mm_set_ss(SinAngle); + XMVECTOR vCos = _mm_set_ss(CosAngle); + // x = cos,y = sin,z = 0, w = 0 + vCos = _mm_unpacklo_ps(vCos, vSin); + XMMATRIX M; + M.r[0] = vCos; + // x = sin,y = cos,z = 0, w = 0 + vCos = XM_PERMUTE_PS(vCos, _MM_SHUFFLE(3, 2, 0, 1)); + // x = cos,y = -sin,z = 0, w = 0 + vCos = _mm_mul_ps(vCos, g_XMNegateX); + M.r[1] = vCos; + M.r[2] = g_XMIdentityR2; + M.r[3] = g_XMIdentityR3; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationRollPitchYaw +( + float Pitch, + float Yaw, + float Roll +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float cp = cosf(Pitch); + float sp = sinf(Pitch); + + float cy = cosf(Yaw); + float sy = sinf(Yaw); + + float cr = cosf(Roll); + float sr = sinf(Roll); + + XMMATRIX M; + M.m[0][0] = cr * cy + sr * sp * sy; + M.m[0][1] = sr * cp; + M.m[0][2] = sr * sp * cy - cr * sy; + M.m[0][3] = 0.0f; + + M.m[1][0] = cr * sp * sy - sr * cy; + M.m[1][1] = cr * cp; + M.m[1][2] = sr * sy + cr * sp * cy; + M.m[1][3] = 0.0f; + + M.m[2][0] = cp * sy; + M.m[2][1] = -sp; + M.m[2][2] = cp * cy; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; +#else + XMVECTOR Angles = XMVectorSet(Pitch, Yaw, Roll, 0.0f); + return XMMatrixRotationRollPitchYawFromVector(Angles); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationRollPitchYawFromVector +( + FXMVECTOR Angles // +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float cp = cosf(Angles.vector4_f32[0]); + float sp = sinf(Angles.vector4_f32[0]); + + float cy = cosf(Angles.vector4_f32[1]); + float sy = sinf(Angles.vector4_f32[1]); + + float cr = cosf(Angles.vector4_f32[2]); + float sr = sinf(Angles.vector4_f32[2]); + + XMMATRIX M; + M.m[0][0] = cr * cy + sr * sp * sy; + M.m[0][1] = sr * cp; + M.m[0][2] = sr * sp * cy - cr * sy; + M.m[0][3] = 0.0f; + + M.m[1][0] = cr * sp * sy - sr * cy; + M.m[1][1] = cr * cp; + M.m[1][2] = sr * sy + cr * sp * cy; + M.m[1][3] = 0.0f; + + M.m[2][0] = cp * sy; + M.m[2][1] = -sp; + M.m[2][2] = cp * cy; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = 0.0f; + M.m[3][3] = 1.0f; + return M; +#else + static const XMVECTORF32 Sign = { { { 1.0f, -1.0f, -1.0f, 1.0f } } }; + + XMVECTOR SinAngles, CosAngles; + XMVectorSinCos(&SinAngles, &CosAngles, Angles); + + XMVECTOR P0 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR Y0 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR P1 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR Y1 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR P2 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR P3 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR Y2 = XMVectorSplatX(SinAngles); + XMVECTOR NS = XMVectorNegate(SinAngles); + + XMVECTOR Q0 = XMVectorMultiply(P0, Y0); + XMVECTOR Q1 = XMVectorMultiply(P1, Sign.v); + Q1 = XMVectorMultiply(Q1, Y1); + XMVECTOR Q2 = XMVectorMultiply(P2, Y2); + Q2 = XMVectorMultiplyAdd(Q2, P3, Q1); + + XMVECTOR V0 = XMVectorPermute(Q0, Q2); + XMVECTOR V1 = XMVectorPermute(Q0, Q2); + XMVECTOR V2 = XMVectorPermute(Q0, NS); + + XMMATRIX M; + M.r[0] = XMVectorSelect(g_XMZero, V0, g_XMSelect1110.v); + M.r[1] = XMVectorSelect(g_XMZero, V1, g_XMSelect1110.v); + M.r[2] = XMVectorSelect(g_XMZero, V2, g_XMSelect1110.v); + M.r[3] = g_XMIdentityR3; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationNormal +( + FXMVECTOR NormalAxis, + float Angle +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + XMVECTOR A = XMVectorSet(fSinAngle, fCosAngle, 1.0f - fCosAngle, 0.0f); + + XMVECTOR C2 = XMVectorSplatZ(A); + XMVECTOR C1 = XMVectorSplatY(A); + XMVECTOR C0 = XMVectorSplatX(A); + + XMVECTOR N0 = XMVectorSwizzle(NormalAxis); + XMVECTOR N1 = XMVectorSwizzle(NormalAxis); + + XMVECTOR V0 = XMVectorMultiply(C2, N0); + V0 = XMVectorMultiply(V0, N1); + + XMVECTOR R0 = XMVectorMultiply(C2, NormalAxis); + R0 = XMVectorMultiplyAdd(R0, NormalAxis, C1); + + XMVECTOR R1 = XMVectorMultiplyAdd(C0, NormalAxis, V0); + XMVECTOR R2 = XMVectorNegativeMultiplySubtract(C0, NormalAxis, V0); + + V0 = XMVectorSelect(A, R0, g_XMSelect1110.v); + XMVECTOR V1 = XMVectorPermute(R1, R2); + XMVECTOR V2 = XMVectorPermute(R1, R2); + + XMMATRIX M; + M.r[0] = XMVectorPermute(V0, V1); + M.r[1] = XMVectorPermute(V0, V1); + M.r[2] = XMVectorPermute(V0, V2); + M.r[3] = g_XMIdentityR3.v; + return M; + +#elif defined(_XM_SSE_INTRINSICS_) + float fSinAngle; + float fCosAngle; + XMScalarSinCos(&fSinAngle, &fCosAngle, Angle); + + XMVECTOR C2 = _mm_set_ps1(1.0f - fCosAngle); + XMVECTOR C1 = _mm_set_ps1(fCosAngle); + XMVECTOR C0 = _mm_set_ps1(fSinAngle); + + XMVECTOR N0 = XM_PERMUTE_PS(NormalAxis, _MM_SHUFFLE(3, 0, 2, 1)); + XMVECTOR N1 = XM_PERMUTE_PS(NormalAxis, _MM_SHUFFLE(3, 1, 0, 2)); + + XMVECTOR V0 = _mm_mul_ps(C2, N0); + V0 = _mm_mul_ps(V0, N1); + + XMVECTOR R0 = _mm_mul_ps(C2, NormalAxis); + R0 = _mm_mul_ps(R0, NormalAxis); + R0 = _mm_add_ps(R0, C1); + + XMVECTOR R1 = _mm_mul_ps(C0, NormalAxis); + R1 = _mm_add_ps(R1, V0); + XMVECTOR R2 = _mm_mul_ps(C0, NormalAxis); + R2 = _mm_sub_ps(V0, R2); + + V0 = _mm_and_ps(R0, g_XMMask3); + XMVECTOR V1 = _mm_shuffle_ps(R1, R2, _MM_SHUFFLE(2, 1, 2, 0)); + V1 = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 3, 2, 1)); + XMVECTOR V2 = _mm_shuffle_ps(R1, R2, _MM_SHUFFLE(0, 0, 1, 1)); + V2 = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 0, 2, 0)); + + R2 = _mm_shuffle_ps(V0, V1, _MM_SHUFFLE(1, 0, 3, 0)); + R2 = XM_PERMUTE_PS(R2, _MM_SHUFFLE(1, 3, 2, 0)); + + XMMATRIX M; + M.r[0] = R2; + + R2 = _mm_shuffle_ps(V0, V1, _MM_SHUFFLE(3, 2, 3, 1)); + R2 = XM_PERMUTE_PS(R2, _MM_SHUFFLE(1, 3, 0, 2)); + M.r[1] = R2; + + V2 = _mm_shuffle_ps(V2, V0, _MM_SHUFFLE(3, 2, 1, 0)); + M.r[2] = V2; + M.r[3] = g_XMIdentityR3.v; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationAxis +( + FXMVECTOR Axis, + float Angle +) noexcept +{ + assert(!XMVector3Equal(Axis, XMVectorZero())); + assert(!XMVector3IsInfinite(Axis)); + + XMVECTOR Normal = XMVector3Normalize(Axis); + return XMMatrixRotationNormal(Normal, Angle); +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixRotationQuaternion(FXMVECTOR Quaternion) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + float qx = Quaternion.vector4_f32[0]; + float qxx = qx * qx; + + float qy = Quaternion.vector4_f32[1]; + float qyy = qy * qy; + + float qz = Quaternion.vector4_f32[2]; + float qzz = qz * qz; + + float qw = Quaternion.vector4_f32[3]; + + XMMATRIX M; + M.m[0][0] = 1.f - 2.f * qyy - 2.f * qzz; + M.m[0][1] = 2.f * qx * qy + 2.f * qz * qw; + M.m[0][2] = 2.f * qx * qz - 2.f * qy * qw; + M.m[0][3] = 0.f; + + M.m[1][0] = 2.f * qx * qy - 2.f * qz * qw; + M.m[1][1] = 1.f - 2.f * qxx - 2.f * qzz; + M.m[1][2] = 2.f * qy * qz + 2.f * qx * qw; + M.m[1][3] = 0.f; + + M.m[2][0] = 2.f * qx * qz + 2.f * qy * qw; + M.m[2][1] = 2.f * qy * qz - 2.f * qx * qw; + M.m[2][2] = 1.f - 2.f * qxx - 2.f * qyy; + M.m[2][3] = 0.f; + + M.m[3][0] = 0.f; + M.m[3][1] = 0.f; + M.m[3][2] = 0.f; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Constant1110 = { { { 1.0f, 1.0f, 1.0f, 0.0f } } }; + + XMVECTOR Q0 = XMVectorAdd(Quaternion, Quaternion); + XMVECTOR Q1 = XMVectorMultiply(Quaternion, Q0); + + XMVECTOR V0 = XMVectorPermute(Q1, Constant1110.v); + XMVECTOR V1 = XMVectorPermute(Q1, Constant1110.v); + XMVECTOR R0 = XMVectorSubtract(Constant1110, V0); + R0 = XMVectorSubtract(R0, V1); + + V0 = XMVectorSwizzle(Quaternion); + V1 = XMVectorSwizzle(Q0); + V0 = XMVectorMultiply(V0, V1); + + V1 = XMVectorSplatW(Quaternion); + XMVECTOR V2 = XMVectorSwizzle(Q0); + V1 = XMVectorMultiply(V1, V2); + + XMVECTOR R1 = XMVectorAdd(V0, V1); + XMVECTOR R2 = XMVectorSubtract(V0, V1); + + V0 = XMVectorPermute(R1, R2); + V1 = XMVectorPermute(R1, R2); + + XMMATRIX M; + M.r[0] = XMVectorPermute(R0, V0); + M.r[1] = XMVectorPermute(R0, V0); + M.r[2] = XMVectorPermute(R0, V1); + M.r[3] = g_XMIdentityR3.v; + return M; + +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Constant1110 = { { { 1.0f, 1.0f, 1.0f, 0.0f } } }; + + XMVECTOR Q0 = _mm_add_ps(Quaternion, Quaternion); + XMVECTOR Q1 = _mm_mul_ps(Quaternion, Q0); + + XMVECTOR V0 = XM_PERMUTE_PS(Q1, _MM_SHUFFLE(3, 0, 0, 1)); + V0 = _mm_and_ps(V0, g_XMMask3); + XMVECTOR V1 = XM_PERMUTE_PS(Q1, _MM_SHUFFLE(3, 1, 2, 2)); + V1 = _mm_and_ps(V1, g_XMMask3); + XMVECTOR R0 = _mm_sub_ps(Constant1110, V0); + R0 = _mm_sub_ps(R0, V1); + + V0 = XM_PERMUTE_PS(Quaternion, _MM_SHUFFLE(3, 1, 0, 0)); + V1 = XM_PERMUTE_PS(Q0, _MM_SHUFFLE(3, 2, 1, 2)); + V0 = _mm_mul_ps(V0, V1); + + V1 = XM_PERMUTE_PS(Quaternion, _MM_SHUFFLE(3, 3, 3, 3)); + XMVECTOR V2 = XM_PERMUTE_PS(Q0, _MM_SHUFFLE(3, 0, 2, 1)); + V1 = _mm_mul_ps(V1, V2); + + XMVECTOR R1 = _mm_add_ps(V0, V1); + XMVECTOR R2 = _mm_sub_ps(V0, V1); + + V0 = _mm_shuffle_ps(R1, R2, _MM_SHUFFLE(1, 0, 2, 1)); + V0 = XM_PERMUTE_PS(V0, _MM_SHUFFLE(1, 3, 2, 0)); + V1 = _mm_shuffle_ps(R1, R2, _MM_SHUFFLE(2, 2, 0, 0)); + V1 = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 0, 2, 0)); + + Q1 = _mm_shuffle_ps(R0, V0, _MM_SHUFFLE(1, 0, 3, 0)); + Q1 = XM_PERMUTE_PS(Q1, _MM_SHUFFLE(1, 3, 2, 0)); + + XMMATRIX M; + M.r[0] = Q1; + + Q1 = _mm_shuffle_ps(R0, V0, _MM_SHUFFLE(3, 2, 3, 1)); + Q1 = XM_PERMUTE_PS(Q1, _MM_SHUFFLE(1, 3, 0, 2)); + M.r[1] = Q1; + + Q1 = _mm_shuffle_ps(V1, R0, _MM_SHUFFLE(3, 2, 1, 0)); + M.r[2] = Q1; + M.r[3] = g_XMIdentityR3; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixTransformation2D +( + FXMVECTOR ScalingOrigin, + float ScalingOrientation, + FXMVECTOR Scaling, + FXMVECTOR RotationOrigin, + float Rotation, + GXMVECTOR Translation +) noexcept +{ + // M = Inverse(MScalingOrigin) * Transpose(MScalingOrientation) * MScaling * MScalingOrientation * + // MScalingOrigin * Inverse(MRotationOrigin) * MRotation * MRotationOrigin * MTranslation; + + XMVECTOR VScalingOrigin = XMVectorSelect(g_XMSelect1100.v, ScalingOrigin, g_XMSelect1100.v); + XMVECTOR NegScalingOrigin = XMVectorNegate(VScalingOrigin); + + XMMATRIX MScalingOriginI = XMMatrixTranslationFromVector(NegScalingOrigin); + XMMATRIX MScalingOrientation = XMMatrixRotationZ(ScalingOrientation); + XMMATRIX MScalingOrientationT = XMMatrixTranspose(MScalingOrientation); + XMVECTOR VScaling = XMVectorSelect(g_XMOne.v, Scaling, g_XMSelect1100.v); + XMMATRIX MScaling = XMMatrixScalingFromVector(VScaling); + XMVECTOR VRotationOrigin = XMVectorSelect(g_XMSelect1100.v, RotationOrigin, g_XMSelect1100.v); + XMMATRIX MRotation = XMMatrixRotationZ(Rotation); + XMVECTOR VTranslation = XMVectorSelect(g_XMSelect1100.v, Translation, g_XMSelect1100.v); + + XMMATRIX M = XMMatrixMultiply(MScalingOriginI, MScalingOrientationT); + M = XMMatrixMultiply(M, MScaling); + M = XMMatrixMultiply(M, MScalingOrientation); + M.r[3] = XMVectorAdd(M.r[3], VScalingOrigin); + M.r[3] = XMVectorSubtract(M.r[3], VRotationOrigin); + M = XMMatrixMultiply(M, MRotation); + M.r[3] = XMVectorAdd(M.r[3], VRotationOrigin); + M.r[3] = XMVectorAdd(M.r[3], VTranslation); + + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixTransformation +( + FXMVECTOR ScalingOrigin, + FXMVECTOR ScalingOrientationQuaternion, + FXMVECTOR Scaling, + GXMVECTOR RotationOrigin, + HXMVECTOR RotationQuaternion, + HXMVECTOR Translation +) noexcept +{ + // M = Inverse(MScalingOrigin) * Transpose(MScalingOrientation) * MScaling * MScalingOrientation * + // MScalingOrigin * Inverse(MRotationOrigin) * MRotation * MRotationOrigin * MTranslation; + + XMVECTOR VScalingOrigin = XMVectorSelect(g_XMSelect1110.v, ScalingOrigin, g_XMSelect1110.v); + XMVECTOR NegScalingOrigin = XMVectorNegate(ScalingOrigin); + + XMMATRIX MScalingOriginI = XMMatrixTranslationFromVector(NegScalingOrigin); + XMMATRIX MScalingOrientation = XMMatrixRotationQuaternion(ScalingOrientationQuaternion); + XMMATRIX MScalingOrientationT = XMMatrixTranspose(MScalingOrientation); + XMMATRIX MScaling = XMMatrixScalingFromVector(Scaling); + XMVECTOR VRotationOrigin = XMVectorSelect(g_XMSelect1110.v, RotationOrigin, g_XMSelect1110.v); + XMMATRIX MRotation = XMMatrixRotationQuaternion(RotationQuaternion); + XMVECTOR VTranslation = XMVectorSelect(g_XMSelect1110.v, Translation, g_XMSelect1110.v); + + XMMATRIX M; + M = XMMatrixMultiply(MScalingOriginI, MScalingOrientationT); + M = XMMatrixMultiply(M, MScaling); + M = XMMatrixMultiply(M, MScalingOrientation); + M.r[3] = XMVectorAdd(M.r[3], VScalingOrigin); + M.r[3] = XMVectorSubtract(M.r[3], VRotationOrigin); + M = XMMatrixMultiply(M, MRotation); + M.r[3] = XMVectorAdd(M.r[3], VRotationOrigin); + M.r[3] = XMVectorAdd(M.r[3], VTranslation); + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixAffineTransformation2D +( + FXMVECTOR Scaling, + FXMVECTOR RotationOrigin, + float Rotation, + FXMVECTOR Translation +) noexcept +{ + // M = MScaling * Inverse(MRotationOrigin) * MRotation * MRotationOrigin * MTranslation; + + XMVECTOR VScaling = XMVectorSelect(g_XMOne.v, Scaling, g_XMSelect1100.v); + XMMATRIX MScaling = XMMatrixScalingFromVector(VScaling); + XMVECTOR VRotationOrigin = XMVectorSelect(g_XMSelect1100.v, RotationOrigin, g_XMSelect1100.v); + XMMATRIX MRotation = XMMatrixRotationZ(Rotation); + XMVECTOR VTranslation = XMVectorSelect(g_XMSelect1100.v, Translation, g_XMSelect1100.v); + + XMMATRIX M; + M = MScaling; + M.r[3] = XMVectorSubtract(M.r[3], VRotationOrigin); + M = XMMatrixMultiply(M, MRotation); + M.r[3] = XMVectorAdd(M.r[3], VRotationOrigin); + M.r[3] = XMVectorAdd(M.r[3], VTranslation); + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixAffineTransformation +( + FXMVECTOR Scaling, + FXMVECTOR RotationOrigin, + FXMVECTOR RotationQuaternion, + GXMVECTOR Translation +) noexcept +{ + // M = MScaling * Inverse(MRotationOrigin) * MRotation * MRotationOrigin * MTranslation; + + XMMATRIX MScaling = XMMatrixScalingFromVector(Scaling); + XMVECTOR VRotationOrigin = XMVectorSelect(g_XMSelect1110.v, RotationOrigin, g_XMSelect1110.v); + XMMATRIX MRotation = XMMatrixRotationQuaternion(RotationQuaternion); + XMVECTOR VTranslation = XMVectorSelect(g_XMSelect1110.v, Translation, g_XMSelect1110.v); + + XMMATRIX M; + M = MScaling; + M.r[3] = XMVectorSubtract(M.r[3], VRotationOrigin); + M = XMMatrixMultiply(M, MRotation); + M.r[3] = XMVectorAdd(M.r[3], VRotationOrigin); + M.r[3] = XMVectorAdd(M.r[3], VTranslation); + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixReflect(FXMVECTOR ReflectionPlane) noexcept +{ + assert(!XMVector3Equal(ReflectionPlane, XMVectorZero())); + assert(!XMPlaneIsInfinite(ReflectionPlane)); + + static const XMVECTORF32 NegativeTwo = { { { -2.0f, -2.0f, -2.0f, 0.0f } } }; + + XMVECTOR P = XMPlaneNormalize(ReflectionPlane); + XMVECTOR S = XMVectorMultiply(P, NegativeTwo); + + XMVECTOR A = XMVectorSplatX(P); + XMVECTOR B = XMVectorSplatY(P); + XMVECTOR C = XMVectorSplatZ(P); + XMVECTOR D = XMVectorSplatW(P); + + XMMATRIX M; + M.r[0] = XMVectorMultiplyAdd(A, S, g_XMIdentityR0.v); + M.r[1] = XMVectorMultiplyAdd(B, S, g_XMIdentityR1.v); + M.r[2] = XMVectorMultiplyAdd(C, S, g_XMIdentityR2.v); + M.r[3] = XMVectorMultiplyAdd(D, S, g_XMIdentityR3.v); + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixShadow +( + FXMVECTOR ShadowPlane, + FXMVECTOR LightPosition +) noexcept +{ + static const XMVECTORU32 Select0001 = { { { XM_SELECT_0, XM_SELECT_0, XM_SELECT_0, XM_SELECT_1 } } }; + + assert(!XMVector3Equal(ShadowPlane, XMVectorZero())); + assert(!XMPlaneIsInfinite(ShadowPlane)); + + XMVECTOR P = XMPlaneNormalize(ShadowPlane); + XMVECTOR Dot = XMPlaneDot(P, LightPosition); + P = XMVectorNegate(P); + XMVECTOR D = XMVectorSplatW(P); + XMVECTOR C = XMVectorSplatZ(P); + XMVECTOR B = XMVectorSplatY(P); + XMVECTOR A = XMVectorSplatX(P); + Dot = XMVectorSelect(Select0001.v, Dot, Select0001.v); + + XMMATRIX M; + M.r[3] = XMVectorMultiplyAdd(D, LightPosition, Dot); + Dot = XMVectorRotateLeft(Dot, 1); + M.r[2] = XMVectorMultiplyAdd(C, LightPosition, Dot); + Dot = XMVectorRotateLeft(Dot, 1); + M.r[1] = XMVectorMultiplyAdd(B, LightPosition, Dot); + Dot = XMVectorRotateLeft(Dot, 1); + M.r[0] = XMVectorMultiplyAdd(A, LightPosition, Dot); + return M; +} + +//------------------------------------------------------------------------------ +// View and projection initialization operations +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixLookAtLH +( + FXMVECTOR EyePosition, + FXMVECTOR FocusPosition, + FXMVECTOR UpDirection +) noexcept +{ + XMVECTOR EyeDirection = XMVectorSubtract(FocusPosition, EyePosition); + return XMMatrixLookToLH(EyePosition, EyeDirection, UpDirection); +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixLookAtRH +( + FXMVECTOR EyePosition, + FXMVECTOR FocusPosition, + FXMVECTOR UpDirection +) noexcept +{ + XMVECTOR NegEyeDirection = XMVectorSubtract(EyePosition, FocusPosition); + return XMMatrixLookToLH(EyePosition, NegEyeDirection, UpDirection); +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixLookToLH +( + FXMVECTOR EyePosition, + FXMVECTOR EyeDirection, + FXMVECTOR UpDirection +) noexcept +{ + assert(!XMVector3Equal(EyeDirection, XMVectorZero())); + assert(!XMVector3IsInfinite(EyeDirection)); + assert(!XMVector3Equal(UpDirection, XMVectorZero())); + assert(!XMVector3IsInfinite(UpDirection)); + + XMVECTOR R2 = XMVector3Normalize(EyeDirection); + + XMVECTOR R0 = XMVector3Cross(UpDirection, R2); + R0 = XMVector3Normalize(R0); + + XMVECTOR R1 = XMVector3Cross(R2, R0); + + XMVECTOR NegEyePosition = XMVectorNegate(EyePosition); + + XMVECTOR D0 = XMVector3Dot(R0, NegEyePosition); + XMVECTOR D1 = XMVector3Dot(R1, NegEyePosition); + XMVECTOR D2 = XMVector3Dot(R2, NegEyePosition); + + XMMATRIX M; + M.r[0] = XMVectorSelect(D0, R0, g_XMSelect1110.v); + M.r[1] = XMVectorSelect(D1, R1, g_XMSelect1110.v); + M.r[2] = XMVectorSelect(D2, R2, g_XMSelect1110.v); + M.r[3] = g_XMIdentityR3.v; + + M = XMMatrixTranspose(M); + + return M; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixLookToRH +( + FXMVECTOR EyePosition, + FXMVECTOR EyeDirection, + FXMVECTOR UpDirection +) noexcept +{ + XMVECTOR NegEyeDirection = XMVectorNegate(EyeDirection); + return XMMatrixLookToLH(EyePosition, NegEyeDirection, UpDirection); +} + +//------------------------------------------------------------------------------ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable:28931, "PREfast noise: Esp:1266") +#endif + +inline XMMATRIX XM_CALLCONV XMMatrixPerspectiveLH +( + float ViewWidth, + float ViewHeight, + float NearZ, + float FarZ +) noexcept +{ + assert(NearZ > 0.f && FarZ > 0.f); + assert(!XMScalarNearEqual(ViewWidth, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(ViewHeight, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float TwoNearZ = NearZ + NearZ; + float fRange = FarZ / (FarZ - NearZ); + + XMMATRIX M; + M.m[0][0] = TwoNearZ / ViewWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = TwoNearZ / ViewHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = 1.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = -fRange * NearZ; + M.m[3][3] = 0.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float TwoNearZ = NearZ + NearZ; + float fRange = FarZ / (FarZ - NearZ); + const float32x4_t Zero = vdupq_n_f32(0); + XMMATRIX M; + M.r[0] = vsetq_lane_f32(TwoNearZ / ViewWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(TwoNearZ / ViewHeight, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, g_XMIdentityR3.v, 2); + M.r[3] = vsetq_lane_f32(-fRange * NearZ, Zero, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float TwoNearZ = NearZ + NearZ; + float fRange = FarZ / (FarZ - NearZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + TwoNearZ / ViewWidth, + TwoNearZ / ViewHeight, + fRange, + -fRange * NearZ + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // TwoNearZ / ViewWidth,0,0,0 + M.r[0] = vTemp; + // 0,TwoNearZ / ViewHeight,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // x=fRange,y=-fRange * NearZ,0,1.0f + vValues = _mm_shuffle_ps(vValues, g_XMIdentityR3, _MM_SHUFFLE(3, 2, 3, 2)); + // 0,0,fRange,1.0f + vTemp = _mm_setzero_ps(); + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(3, 0, 0, 0)); + M.r[2] = vTemp; + // 0,0,-fRange * NearZ,0 + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(2, 1, 0, 0)); + M.r[3] = vTemp; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixPerspectiveRH +( + float ViewWidth, + float ViewHeight, + float NearZ, + float FarZ +) noexcept +{ + assert(NearZ > 0.f && FarZ > 0.f); + assert(!XMScalarNearEqual(ViewWidth, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(ViewHeight, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float TwoNearZ = NearZ + NearZ; + float fRange = FarZ / (NearZ - FarZ); + + XMMATRIX M; + M.m[0][0] = TwoNearZ / ViewWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = TwoNearZ / ViewHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = -1.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = fRange * NearZ; + M.m[3][3] = 0.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float TwoNearZ = NearZ + NearZ; + float fRange = FarZ / (NearZ - FarZ); + const float32x4_t Zero = vdupq_n_f32(0); + + XMMATRIX M; + M.r[0] = vsetq_lane_f32(TwoNearZ / ViewWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(TwoNearZ / ViewHeight, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, g_XMNegIdentityR3.v, 2); + M.r[3] = vsetq_lane_f32(fRange * NearZ, Zero, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float TwoNearZ = NearZ + NearZ; + float fRange = FarZ / (NearZ - FarZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + TwoNearZ / ViewWidth, + TwoNearZ / ViewHeight, + fRange, + fRange * NearZ + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // TwoNearZ / ViewWidth,0,0,0 + M.r[0] = vTemp; + // 0,TwoNearZ / ViewHeight,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // x=fRange,y=-fRange * NearZ,0,-1.0f + vValues = _mm_shuffle_ps(vValues, g_XMNegIdentityR3, _MM_SHUFFLE(3, 2, 3, 2)); + // 0,0,fRange,-1.0f + vTemp = _mm_setzero_ps(); + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(3, 0, 0, 0)); + M.r[2] = vTemp; + // 0,0,-fRange * NearZ,0 + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(2, 1, 0, 0)); + M.r[3] = vTemp; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixPerspectiveFovLH +( + float FovAngleY, + float AspectRatio, + float NearZ, + float FarZ +) noexcept +{ + assert(NearZ > 0.f && FarZ > 0.f); + assert(!XMScalarNearEqual(FovAngleY, 0.0f, 0.00001f * 2.0f)); + assert(!XMScalarNearEqual(AspectRatio, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float SinFov; + float CosFov; + XMScalarSinCos(&SinFov, &CosFov, 0.5f * FovAngleY); + + float Height = CosFov / SinFov; + float Width = Height / AspectRatio; + float fRange = FarZ / (FarZ - NearZ); + + XMMATRIX M; + M.m[0][0] = Width; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = Height; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = 1.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = -fRange * NearZ; + M.m[3][3] = 0.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float SinFov; + float CosFov; + XMScalarSinCos(&SinFov, &CosFov, 0.5f * FovAngleY); + + float fRange = FarZ / (FarZ - NearZ); + float Height = CosFov / SinFov; + float Width = Height / AspectRatio; + const float32x4_t Zero = vdupq_n_f32(0); + + XMMATRIX M; + M.r[0] = vsetq_lane_f32(Width, Zero, 0); + M.r[1] = vsetq_lane_f32(Height, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, g_XMIdentityR3.v, 2); + M.r[3] = vsetq_lane_f32(-fRange * NearZ, Zero, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + float SinFov; + float CosFov; + XMScalarSinCos(&SinFov, &CosFov, 0.5f * FovAngleY); + + float fRange = FarZ / (FarZ - NearZ); + // Note: This is recorded on the stack + float Height = CosFov / SinFov; + XMVECTOR rMem = { + Height / AspectRatio, + Height, + fRange, + -fRange * NearZ + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // Height / AspectRatio,0,0,0 + XMMATRIX M; + M.r[0] = vTemp; + // 0,Height,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // x=fRange,y=-fRange * NearZ,0,1.0f + vTemp = _mm_setzero_ps(); + vValues = _mm_shuffle_ps(vValues, g_XMIdentityR3, _MM_SHUFFLE(3, 2, 3, 2)); + // 0,0,fRange,1.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(3, 0, 0, 0)); + M.r[2] = vTemp; + // 0,0,-fRange * NearZ,0.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(2, 1, 0, 0)); + M.r[3] = vTemp; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixPerspectiveFovRH +( + float FovAngleY, + float AspectRatio, + float NearZ, + float FarZ +) noexcept +{ + assert(NearZ > 0.f && FarZ > 0.f); + assert(!XMScalarNearEqual(FovAngleY, 0.0f, 0.00001f * 2.0f)); + assert(!XMScalarNearEqual(AspectRatio, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float SinFov; + float CosFov; + XMScalarSinCos(&SinFov, &CosFov, 0.5f * FovAngleY); + + float Height = CosFov / SinFov; + float Width = Height / AspectRatio; + float fRange = FarZ / (NearZ - FarZ); + + XMMATRIX M; + M.m[0][0] = Width; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = Height; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = -1.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = fRange * NearZ; + M.m[3][3] = 0.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float SinFov; + float CosFov; + XMScalarSinCos(&SinFov, &CosFov, 0.5f * FovAngleY); + float fRange = FarZ / (NearZ - FarZ); + float Height = CosFov / SinFov; + float Width = Height / AspectRatio; + const float32x4_t Zero = vdupq_n_f32(0); + + XMMATRIX M; + M.r[0] = vsetq_lane_f32(Width, Zero, 0); + M.r[1] = vsetq_lane_f32(Height, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, g_XMNegIdentityR3.v, 2); + M.r[3] = vsetq_lane_f32(fRange * NearZ, Zero, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + float SinFov; + float CosFov; + XMScalarSinCos(&SinFov, &CosFov, 0.5f * FovAngleY); + float fRange = FarZ / (NearZ - FarZ); + // Note: This is recorded on the stack + float Height = CosFov / SinFov; + XMVECTOR rMem = { + Height / AspectRatio, + Height, + fRange, + fRange * NearZ + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // Height / AspectRatio,0,0,0 + XMMATRIX M; + M.r[0] = vTemp; + // 0,Height,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // x=fRange,y=-fRange * NearZ,0,-1.0f + vTemp = _mm_setzero_ps(); + vValues = _mm_shuffle_ps(vValues, g_XMNegIdentityR3, _MM_SHUFFLE(3, 2, 3, 2)); + // 0,0,fRange,-1.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(3, 0, 0, 0)); + M.r[2] = vTemp; + // 0,0,fRange * NearZ,0.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(2, 1, 0, 0)); + M.r[3] = vTemp; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixPerspectiveOffCenterLH +( + float ViewLeft, + float ViewRight, + float ViewBottom, + float ViewTop, + float NearZ, + float FarZ +) noexcept +{ + assert(NearZ > 0.f && FarZ > 0.f); + assert(!XMScalarNearEqual(ViewRight, ViewLeft, 0.00001f)); + assert(!XMScalarNearEqual(ViewTop, ViewBottom, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float TwoNearZ = NearZ + NearZ; + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = FarZ / (FarZ - NearZ); + + XMMATRIX M; + M.m[0][0] = TwoNearZ * ReciprocalWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = TwoNearZ * ReciprocalHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = -(ViewLeft + ViewRight) * ReciprocalWidth; + M.m[2][1] = -(ViewTop + ViewBottom) * ReciprocalHeight; + M.m[2][2] = fRange; + M.m[2][3] = 1.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = -fRange * NearZ; + M.m[3][3] = 0.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float TwoNearZ = NearZ + NearZ; + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = FarZ / (FarZ - NearZ); + const float32x4_t Zero = vdupq_n_f32(0); + + XMMATRIX M; + M.r[0] = vsetq_lane_f32(TwoNearZ * ReciprocalWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(TwoNearZ * ReciprocalHeight, Zero, 1); + M.r[2] = XMVectorSet(-(ViewLeft + ViewRight) * ReciprocalWidth, + -(ViewTop + ViewBottom) * ReciprocalHeight, + fRange, + 1.0f); + M.r[3] = vsetq_lane_f32(-fRange * NearZ, Zero, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float TwoNearZ = NearZ + NearZ; + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = FarZ / (FarZ - NearZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + TwoNearZ * ReciprocalWidth, + TwoNearZ * ReciprocalHeight, + -fRange * NearZ, + 0 + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // TwoNearZ*ReciprocalWidth,0,0,0 + M.r[0] = vTemp; + // 0,TwoNearZ*ReciprocalHeight,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // 0,0,fRange,1.0f + M.r[2] = XMVectorSet(-(ViewLeft + ViewRight) * ReciprocalWidth, + -(ViewTop + ViewBottom) * ReciprocalHeight, + fRange, + 1.0f); + // 0,0,-fRange * NearZ,0.0f + vValues = _mm_and_ps(vValues, g_XMMaskZ); + M.r[3] = vValues; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixPerspectiveOffCenterRH +( + float ViewLeft, + float ViewRight, + float ViewBottom, + float ViewTop, + float NearZ, + float FarZ +) noexcept +{ + assert(NearZ > 0.f && FarZ > 0.f); + assert(!XMScalarNearEqual(ViewRight, ViewLeft, 0.00001f)); + assert(!XMScalarNearEqual(ViewTop, ViewBottom, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float TwoNearZ = NearZ + NearZ; + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = FarZ / (NearZ - FarZ); + + XMMATRIX M; + M.m[0][0] = TwoNearZ * ReciprocalWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = TwoNearZ * ReciprocalHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = (ViewLeft + ViewRight) * ReciprocalWidth; + M.m[2][1] = (ViewTop + ViewBottom) * ReciprocalHeight; + M.m[2][2] = fRange; + M.m[2][3] = -1.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = fRange * NearZ; + M.m[3][3] = 0.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float TwoNearZ = NearZ + NearZ; + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = FarZ / (NearZ - FarZ); + const float32x4_t Zero = vdupq_n_f32(0); + + XMMATRIX M; + M.r[0] = vsetq_lane_f32(TwoNearZ * ReciprocalWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(TwoNearZ * ReciprocalHeight, Zero, 1); + M.r[2] = XMVectorSet((ViewLeft + ViewRight) * ReciprocalWidth, + (ViewTop + ViewBottom) * ReciprocalHeight, + fRange, + -1.0f); + M.r[3] = vsetq_lane_f32(fRange * NearZ, Zero, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float TwoNearZ = NearZ + NearZ; + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = FarZ / (NearZ - FarZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + TwoNearZ * ReciprocalWidth, + TwoNearZ * ReciprocalHeight, + fRange * NearZ, + 0 + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // TwoNearZ*ReciprocalWidth,0,0,0 + M.r[0] = vTemp; + // 0,TwoNearZ*ReciprocalHeight,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // 0,0,fRange,1.0f + M.r[2] = XMVectorSet((ViewLeft + ViewRight) * ReciprocalWidth, + (ViewTop + ViewBottom) * ReciprocalHeight, + fRange, + -1.0f); + // 0,0,-fRange * NearZ,0.0f + vValues = _mm_and_ps(vValues, g_XMMaskZ); + M.r[3] = vValues; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixOrthographicLH +( + float ViewWidth, + float ViewHeight, + float NearZ, + float FarZ +) noexcept +{ + assert(!XMScalarNearEqual(ViewWidth, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(ViewHeight, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float fRange = 1.0f / (FarZ - NearZ); + + XMMATRIX M; + M.m[0][0] = 2.0f / ViewWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = 2.0f / ViewHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = -fRange * NearZ; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fRange = 1.0f / (FarZ - NearZ); + + const float32x4_t Zero = vdupq_n_f32(0); + XMMATRIX M; + M.r[0] = vsetq_lane_f32(2.0f / ViewWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(2.0f / ViewHeight, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, Zero, 2); + M.r[3] = vsetq_lane_f32(-fRange * NearZ, g_XMIdentityR3.v, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float fRange = 1.0f / (FarZ - NearZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + 2.0f / ViewWidth, + 2.0f / ViewHeight, + fRange, + -fRange * NearZ + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // 2.0f / ViewWidth,0,0,0 + M.r[0] = vTemp; + // 0,2.0f / ViewHeight,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // x=fRange,y=-fRange * NearZ,0,1.0f + vTemp = _mm_setzero_ps(); + vValues = _mm_shuffle_ps(vValues, g_XMIdentityR3, _MM_SHUFFLE(3, 2, 3, 2)); + // 0,0,fRange,0.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(2, 0, 0, 0)); + M.r[2] = vTemp; + // 0,0,-fRange * NearZ,1.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(3, 1, 0, 0)); + M.r[3] = vTemp; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixOrthographicRH +( + float ViewWidth, + float ViewHeight, + float NearZ, + float FarZ +) noexcept +{ + assert(!XMScalarNearEqual(ViewWidth, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(ViewHeight, 0.0f, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float fRange = 1.0f / (NearZ - FarZ); + + XMMATRIX M; + M.m[0][0] = 2.0f / ViewWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = 2.0f / ViewHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = 0.0f; + + M.m[3][0] = 0.0f; + M.m[3][1] = 0.0f; + M.m[3][2] = fRange * NearZ; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float fRange = 1.0f / (NearZ - FarZ); + + const float32x4_t Zero = vdupq_n_f32(0); + XMMATRIX M; + M.r[0] = vsetq_lane_f32(2.0f / ViewWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(2.0f / ViewHeight, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, Zero, 2); + M.r[3] = vsetq_lane_f32(fRange * NearZ, g_XMIdentityR3.v, 2); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float fRange = 1.0f / (NearZ - FarZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + 2.0f / ViewWidth, + 2.0f / ViewHeight, + fRange, + fRange * NearZ + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // 2.0f / ViewWidth,0,0,0 + M.r[0] = vTemp; + // 0,2.0f / ViewHeight,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + M.r[1] = vTemp; + // x=fRange,y=fRange * NearZ,0,1.0f + vTemp = _mm_setzero_ps(); + vValues = _mm_shuffle_ps(vValues, g_XMIdentityR3, _MM_SHUFFLE(3, 2, 3, 2)); + // 0,0,fRange,0.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(2, 0, 0, 0)); + M.r[2] = vTemp; + // 0,0,fRange * NearZ,1.0f + vTemp = _mm_shuffle_ps(vTemp, vValues, _MM_SHUFFLE(3, 1, 0, 0)); + M.r[3] = vTemp; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixOrthographicOffCenterLH +( + float ViewLeft, + float ViewRight, + float ViewBottom, + float ViewTop, + float NearZ, + float FarZ +) noexcept +{ + assert(!XMScalarNearEqual(ViewRight, ViewLeft, 0.00001f)); + assert(!XMScalarNearEqual(ViewTop, ViewBottom, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = 1.0f / (FarZ - NearZ); + + XMMATRIX M; + M.m[0][0] = ReciprocalWidth + ReciprocalWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = ReciprocalHeight + ReciprocalHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = 0.0f; + + M.m[3][0] = -(ViewLeft + ViewRight) * ReciprocalWidth; + M.m[3][1] = -(ViewTop + ViewBottom) * ReciprocalHeight; + M.m[3][2] = -fRange * NearZ; + M.m[3][3] = 1.0f; + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = 1.0f / (FarZ - NearZ); + const float32x4_t Zero = vdupq_n_f32(0); + XMMATRIX M; + M.r[0] = vsetq_lane_f32(ReciprocalWidth + ReciprocalWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(ReciprocalHeight + ReciprocalHeight, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, Zero, 2); + M.r[3] = XMVectorSet(-(ViewLeft + ViewRight) * ReciprocalWidth, + -(ViewTop + ViewBottom) * ReciprocalHeight, + -fRange * NearZ, + 1.0f); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float fReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float fReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = 1.0f / (FarZ - NearZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + fReciprocalWidth, + fReciprocalHeight, + fRange, + 1.0f + }; + XMVECTOR rMem2 = { + -(ViewLeft + ViewRight), + -(ViewTop + ViewBottom), + -NearZ, + 1.0f + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // fReciprocalWidth*2,0,0,0 + vTemp = _mm_add_ss(vTemp, vTemp); + M.r[0] = vTemp; + // 0,fReciprocalHeight*2,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + vTemp = _mm_add_ps(vTemp, vTemp); + M.r[1] = vTemp; + // 0,0,fRange,0.0f + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskZ); + M.r[2] = vTemp; + // -(ViewLeft + ViewRight)*fReciprocalWidth,-(ViewTop + ViewBottom)*fReciprocalHeight,fRange*-NearZ,1.0f + vValues = _mm_mul_ps(vValues, rMem2); + M.r[3] = vValues; + return M; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMatrixOrthographicOffCenterRH +( + float ViewLeft, + float ViewRight, + float ViewBottom, + float ViewTop, + float NearZ, + float FarZ +) noexcept +{ + assert(!XMScalarNearEqual(ViewRight, ViewLeft, 0.00001f)); + assert(!XMScalarNearEqual(ViewTop, ViewBottom, 0.00001f)); + assert(!XMScalarNearEqual(FarZ, NearZ, 0.00001f)); + +#if defined(_XM_NO_INTRINSICS_) + + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = 1.0f / (NearZ - FarZ); + + XMMATRIX M; + M.m[0][0] = ReciprocalWidth + ReciprocalWidth; + M.m[0][1] = 0.0f; + M.m[0][2] = 0.0f; + M.m[0][3] = 0.0f; + + M.m[1][0] = 0.0f; + M.m[1][1] = ReciprocalHeight + ReciprocalHeight; + M.m[1][2] = 0.0f; + M.m[1][3] = 0.0f; + + M.m[2][0] = 0.0f; + M.m[2][1] = 0.0f; + M.m[2][2] = fRange; + M.m[2][3] = 0.0f; + + M.r[3] = XMVectorSet(-(ViewLeft + ViewRight) * ReciprocalWidth, + -(ViewTop + ViewBottom) * ReciprocalHeight, + fRange * NearZ, + 1.0f); + return M; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float ReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float ReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = 1.0f / (NearZ - FarZ); + const float32x4_t Zero = vdupq_n_f32(0); + XMMATRIX M; + M.r[0] = vsetq_lane_f32(ReciprocalWidth + ReciprocalWidth, Zero, 0); + M.r[1] = vsetq_lane_f32(ReciprocalHeight + ReciprocalHeight, Zero, 1); + M.r[2] = vsetq_lane_f32(fRange, Zero, 2); + M.r[3] = XMVectorSet(-(ViewLeft + ViewRight) * ReciprocalWidth, + -(ViewTop + ViewBottom) * ReciprocalHeight, + fRange * NearZ, + 1.0f); + return M; +#elif defined(_XM_SSE_INTRINSICS_) + XMMATRIX M; + float fReciprocalWidth = 1.0f / (ViewRight - ViewLeft); + float fReciprocalHeight = 1.0f / (ViewTop - ViewBottom); + float fRange = 1.0f / (NearZ - FarZ); + // Note: This is recorded on the stack + XMVECTOR rMem = { + fReciprocalWidth, + fReciprocalHeight, + fRange, + 1.0f + }; + XMVECTOR rMem2 = { + -(ViewLeft + ViewRight), + -(ViewTop + ViewBottom), + NearZ, + 1.0f + }; + // Copy from memory to SSE register + XMVECTOR vValues = rMem; + XMVECTOR vTemp = _mm_setzero_ps(); + // Copy x only + vTemp = _mm_move_ss(vTemp, vValues); + // fReciprocalWidth*2,0,0,0 + vTemp = _mm_add_ss(vTemp, vTemp); + M.r[0] = vTemp; + // 0,fReciprocalHeight*2,0,0 + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskY); + vTemp = _mm_add_ps(vTemp, vTemp); + M.r[1] = vTemp; + // 0,0,fRange,0.0f + vTemp = vValues; + vTemp = _mm_and_ps(vTemp, g_XMMaskZ); + M.r[2] = vTemp; + // -(ViewLeft + ViewRight)*fReciprocalWidth,-(ViewTop + ViewBottom)*fReciprocalHeight,fRange*-NearZ,1.0f + vValues = _mm_mul_ps(vValues, rMem2); + M.r[3] = vValues; + return M; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +/**************************************************************************** + * + * XMMATRIX operators and methods + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMMATRIX::XMMATRIX +( + float m00, float m01, float m02, float m03, + float m10, float m11, float m12, float m13, + float m20, float m21, float m22, float m23, + float m30, float m31, float m32, float m33 +) noexcept +{ + r[0] = XMVectorSet(m00, m01, m02, m03); + r[1] = XMVectorSet(m10, m11, m12, m13); + r[2] = XMVectorSet(m20, m21, m22, m23); + r[3] = XMVectorSet(m30, m31, m32, m33); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMMATRIX::XMMATRIX(const float* pArray) noexcept +{ + assert(pArray != nullptr); + r[0] = XMLoadFloat4(reinterpret_cast(pArray)); + r[1] = XMLoadFloat4(reinterpret_cast(pArray + 4)); + r[2] = XMLoadFloat4(reinterpret_cast(pArray + 8)); + r[3] = XMLoadFloat4(reinterpret_cast(pArray + 12)); +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XMMATRIX::operator- () const noexcept +{ + XMMATRIX R; + R.r[0] = XMVectorNegate(r[0]); + R.r[1] = XMVectorNegate(r[1]); + R.r[2] = XMVectorNegate(r[2]); + R.r[3] = XMVectorNegate(r[3]); + return R; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX& XM_CALLCONV XMMATRIX::operator+= (FXMMATRIX M) noexcept +{ + r[0] = XMVectorAdd(r[0], M.r[0]); + r[1] = XMVectorAdd(r[1], M.r[1]); + r[2] = XMVectorAdd(r[2], M.r[2]); + r[3] = XMVectorAdd(r[3], M.r[3]); + return *this; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX& XM_CALLCONV XMMATRIX::operator-= (FXMMATRIX M) noexcept +{ + r[0] = XMVectorSubtract(r[0], M.r[0]); + r[1] = XMVectorSubtract(r[1], M.r[1]); + r[2] = XMVectorSubtract(r[2], M.r[2]); + r[3] = XMVectorSubtract(r[3], M.r[3]); + return *this; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX& XM_CALLCONV XMMATRIX::operator*=(FXMMATRIX M) noexcept +{ + *this = XMMatrixMultiply(*this, M); + return *this; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX& XMMATRIX::operator*= (float S) noexcept +{ + r[0] = XMVectorScale(r[0], S); + r[1] = XMVectorScale(r[1], S); + r[2] = XMVectorScale(r[2], S); + r[3] = XMVectorScale(r[3], S); + return *this; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX& XMMATRIX::operator/= (float S) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR vS = XMVectorReplicate(S); + r[0] = XMVectorDivide(r[0], vS); + r[1] = XMVectorDivide(r[1], vS); + r[2] = XMVectorDivide(r[2], vS); + r[3] = XMVectorDivide(r[3], vS); + return *this; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + float32x4_t vS = vdupq_n_f32(S); + r[0] = vdivq_f32(r[0], vS); + r[1] = vdivq_f32(r[1], vS); + r[2] = vdivq_f32(r[2], vS); + r[3] = vdivq_f32(r[3], vS); +#else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x2_t vS = vdup_n_f32(S); + float32x2_t R0 = vrecpe_f32(vS); + float32x2_t S0 = vrecps_f32(R0, vS); + R0 = vmul_f32(S0, R0); + S0 = vrecps_f32(R0, vS); + R0 = vmul_f32(S0, R0); + float32x4_t Reciprocal = vcombine_f32(R0, R0); + r[0] = vmulq_f32(r[0], Reciprocal); + r[1] = vmulq_f32(r[1], Reciprocal); + r[2] = vmulq_f32(r[2], Reciprocal); + r[3] = vmulq_f32(r[3], Reciprocal); +#endif + return *this; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 vS = _mm_set_ps1(S); + r[0] = _mm_div_ps(r[0], vS); + r[1] = _mm_div_ps(r[1], vS); + r[2] = _mm_div_ps(r[2], vS); + r[3] = _mm_div_ps(r[3], vS); + return *this; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMATRIX::operator+ (FXMMATRIX M) const noexcept +{ + XMMATRIX R; + R.r[0] = XMVectorAdd(r[0], M.r[0]); + R.r[1] = XMVectorAdd(r[1], M.r[1]); + R.r[2] = XMVectorAdd(r[2], M.r[2]); + R.r[3] = XMVectorAdd(r[3], M.r[3]); + return R; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMATRIX::operator- (FXMMATRIX M) const noexcept +{ + XMMATRIX R; + R.r[0] = XMVectorSubtract(r[0], M.r[0]); + R.r[1] = XMVectorSubtract(r[1], M.r[1]); + R.r[2] = XMVectorSubtract(r[2], M.r[2]); + R.r[3] = XMVectorSubtract(r[3], M.r[3]); + return R; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV XMMATRIX::operator*(FXMMATRIX M) const noexcept +{ + return XMMatrixMultiply(*this, M); +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XMMATRIX::operator* (float S) const noexcept +{ + XMMATRIX R; + R.r[0] = XMVectorScale(r[0], S); + R.r[1] = XMVectorScale(r[1], S); + R.r[2] = XMVectorScale(r[2], S); + R.r[3] = XMVectorScale(r[3], S); + return R; +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XMMATRIX::operator/ (float S) const noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR vS = XMVectorReplicate(S); + XMMATRIX R; + R.r[0] = XMVectorDivide(r[0], vS); + R.r[1] = XMVectorDivide(r[1], vS); + R.r[2] = XMVectorDivide(r[2], vS); + R.r[3] = XMVectorDivide(r[3], vS); + return R; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + float32x4_t vS = vdupq_n_f32(S); + XMMATRIX R; + R.r[0] = vdivq_f32(r[0], vS); + R.r[1] = vdivq_f32(r[1], vS); + R.r[2] = vdivq_f32(r[2], vS); + R.r[3] = vdivq_f32(r[3], vS); +#else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x2_t vS = vdup_n_f32(S); + float32x2_t R0 = vrecpe_f32(vS); + float32x2_t S0 = vrecps_f32(R0, vS); + R0 = vmul_f32(S0, R0); + S0 = vrecps_f32(R0, vS); + R0 = vmul_f32(S0, R0); + float32x4_t Reciprocal = vcombine_f32(R0, R0); + XMMATRIX R; + R.r[0] = vmulq_f32(r[0], Reciprocal); + R.r[1] = vmulq_f32(r[1], Reciprocal); + R.r[2] = vmulq_f32(r[2], Reciprocal); + R.r[3] = vmulq_f32(r[3], Reciprocal); +#endif + return R; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 vS = _mm_set_ps1(S); + XMMATRIX R; + R.r[0] = _mm_div_ps(r[0], vS); + R.r[1] = _mm_div_ps(r[1], vS); + R.r[2] = _mm_div_ps(r[2], vS); + R.r[3] = _mm_div_ps(r[3], vS); + return R; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMMATRIX XM_CALLCONV operator* +( + float S, + FXMMATRIX M + ) noexcept +{ + XMMATRIX R; + R.r[0] = XMVectorScale(M.r[0], S); + R.r[1] = XMVectorScale(M.r[1], S); + R.r[2] = XMVectorScale(M.r[2], S); + R.r[3] = XMVectorScale(M.r[3], S); + return R; +} + +/**************************************************************************** + * + * XMFLOAT3X3 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMFLOAT3X3::XMFLOAT3X3(const float* pArray) noexcept +{ + assert(pArray != nullptr); + for (size_t Row = 0; Row < 3; Row++) + { + for (size_t Column = 0; Column < 3; Column++) + { + m[Row][Column] = pArray[Row * 3 + Column]; + } + } +} + +/**************************************************************************** + * + * XMFLOAT4X3 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMFLOAT4X3::XMFLOAT4X3(const float* pArray) noexcept +{ + assert(pArray != nullptr); + + m[0][0] = pArray[0]; + m[0][1] = pArray[1]; + m[0][2] = pArray[2]; + + m[1][0] = pArray[3]; + m[1][1] = pArray[4]; + m[1][2] = pArray[5]; + + m[2][0] = pArray[6]; + m[2][1] = pArray[7]; + m[2][2] = pArray[8]; + + m[3][0] = pArray[9]; + m[3][1] = pArray[10]; + m[3][2] = pArray[11]; +} + +/**************************************************************************** +* +* XMFLOAT3X4 operators +* +****************************************************************************/ + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMFLOAT3X4::XMFLOAT3X4(const float* pArray) noexcept +{ + assert(pArray != nullptr); + + m[0][0] = pArray[0]; + m[0][1] = pArray[1]; + m[0][2] = pArray[2]; + m[0][3] = pArray[3]; + + m[1][0] = pArray[4]; + m[1][1] = pArray[5]; + m[1][2] = pArray[6]; + m[1][3] = pArray[7]; + + m[2][0] = pArray[8]; + m[2][1] = pArray[9]; + m[2][2] = pArray[10]; + m[2][3] = pArray[11]; +} + +/**************************************************************************** + * + * XMFLOAT4X4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMFLOAT4X4::XMFLOAT4X4(const float* pArray) noexcept +{ + assert(pArray != nullptr); + + m[0][0] = pArray[0]; + m[0][1] = pArray[1]; + m[0][2] = pArray[2]; + m[0][3] = pArray[3]; + + m[1][0] = pArray[4]; + m[1][1] = pArray[5]; + m[1][2] = pArray[6]; + m[1][3] = pArray[7]; + + m[2][0] = pArray[8]; + m[2][1] = pArray[9]; + m[2][2] = pArray[10]; + m[2][3] = pArray[11]; + + m[3][0] = pArray[12]; + m[3][1] = pArray[13]; + m[3][2] = pArray[14]; + m[3][3] = pArray[15]; +} + diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathMisc.inl b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathMisc.inl new file mode 100644 index 00000000..d7981043 --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathMisc.inl @@ -0,0 +1,2493 @@ +//------------------------------------------------------------------------------------- +// DirectXMathMisc.inl -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +/**************************************************************************** + * + * Quaternion + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMQuaternionEqual +( + FXMVECTOR Q1, + FXMVECTOR Q2 +) noexcept +{ + return XMVector4Equal(Q1, Q2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMQuaternionNotEqual +( + FXMVECTOR Q1, + FXMVECTOR Q2 +) noexcept +{ + return XMVector4NotEqual(Q1, Q2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMQuaternionIsNaN(FXMVECTOR Q) noexcept +{ + return XMVector4IsNaN(Q); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMQuaternionIsInfinite(FXMVECTOR Q) noexcept +{ + return XMVector4IsInfinite(Q); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMQuaternionIsIdentity(FXMVECTOR Q) noexcept +{ + return XMVector4Equal(Q, g_XMIdentityR3.v); +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionDot +( + FXMVECTOR Q1, + FXMVECTOR Q2 +) noexcept +{ + return XMVector4Dot(Q1, Q2); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionMultiply +( + FXMVECTOR Q1, + FXMVECTOR Q2 +) noexcept +{ + // Returns the product Q2*Q1 (which is the concatenation of a rotation Q1 followed by the rotation Q2) + + // [ (Q2.w * Q1.x) + (Q2.x * Q1.w) + (Q2.y * Q1.z) - (Q2.z * Q1.y), + // (Q2.w * Q1.y) - (Q2.x * Q1.z) + (Q2.y * Q1.w) + (Q2.z * Q1.x), + // (Q2.w * Q1.z) + (Q2.x * Q1.y) - (Q2.y * Q1.x) + (Q2.z * Q1.w), + // (Q2.w * Q1.w) - (Q2.x * Q1.x) - (Q2.y * Q1.y) - (Q2.z * Q1.z) ] + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + (Q2.vector4_f32[3] * Q1.vector4_f32[0]) + (Q2.vector4_f32[0] * Q1.vector4_f32[3]) + (Q2.vector4_f32[1] * Q1.vector4_f32[2]) - (Q2.vector4_f32[2] * Q1.vector4_f32[1]), + (Q2.vector4_f32[3] * Q1.vector4_f32[1]) - (Q2.vector4_f32[0] * Q1.vector4_f32[2]) + (Q2.vector4_f32[1] * Q1.vector4_f32[3]) + (Q2.vector4_f32[2] * Q1.vector4_f32[0]), + (Q2.vector4_f32[3] * Q1.vector4_f32[2]) + (Q2.vector4_f32[0] * Q1.vector4_f32[1]) - (Q2.vector4_f32[1] * Q1.vector4_f32[0]) + (Q2.vector4_f32[2] * Q1.vector4_f32[3]), + (Q2.vector4_f32[3] * Q1.vector4_f32[3]) - (Q2.vector4_f32[0] * Q1.vector4_f32[0]) - (Q2.vector4_f32[1] * Q1.vector4_f32[1]) - (Q2.vector4_f32[2] * Q1.vector4_f32[2]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 ControlWZYX = { { { 1.0f, -1.0f, 1.0f, -1.0f } } }; + static const XMVECTORF32 ControlZWXY = { { { 1.0f, 1.0f, -1.0f, -1.0f } } }; + static const XMVECTORF32 ControlYXWZ = { { { -1.0f, 1.0f, 1.0f, -1.0f } } }; + + float32x2_t Q2L = vget_low_f32(Q2); + float32x2_t Q2H = vget_high_f32(Q2); + + float32x4_t Q2X = vdupq_lane_f32(Q2L, 0); + float32x4_t Q2Y = vdupq_lane_f32(Q2L, 1); + float32x4_t Q2Z = vdupq_lane_f32(Q2H, 0); + XMVECTOR vResult = vmulq_lane_f32(Q1, Q2H, 1); + + // Mul by Q1WZYX + float32x4_t vTemp = vrev64q_f32(Q1); + vTemp = vcombine_f32(vget_high_f32(vTemp), vget_low_f32(vTemp)); + Q2X = vmulq_f32(Q2X, vTemp); + vResult = vmlaq_f32(vResult, Q2X, ControlWZYX); + + // Mul by Q1ZWXY + vTemp = vreinterpretq_f32_u32(vrev64q_u32(vreinterpretq_u32_f32(vTemp))); + Q2Y = vmulq_f32(Q2Y, vTemp); + vResult = vmlaq_f32(vResult, Q2Y, ControlZWXY); + + // Mul by Q1YXWZ + vTemp = vreinterpretq_f32_u32(vrev64q_u32(vreinterpretq_u32_f32(vTemp))); + vTemp = vcombine_f32(vget_high_f32(vTemp), vget_low_f32(vTemp)); + Q2Z = vmulq_f32(Q2Z, vTemp); + vResult = vmlaq_f32(vResult, Q2Z, ControlYXWZ); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ControlWZYX = { { { 1.0f, -1.0f, 1.0f, -1.0f } } }; + static const XMVECTORF32 ControlZWXY = { { { 1.0f, 1.0f, -1.0f, -1.0f } } }; + static const XMVECTORF32 ControlYXWZ = { { { -1.0f, 1.0f, 1.0f, -1.0f } } }; + // Copy to SSE registers and use as few as possible for x86 + XMVECTOR Q2X = Q2; + XMVECTOR Q2Y = Q2; + XMVECTOR Q2Z = Q2; + XMVECTOR vResult = Q2; + // Splat with one instruction + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 3, 3, 3)); + Q2X = XM_PERMUTE_PS(Q2X, _MM_SHUFFLE(0, 0, 0, 0)); + Q2Y = XM_PERMUTE_PS(Q2Y, _MM_SHUFFLE(1, 1, 1, 1)); + Q2Z = XM_PERMUTE_PS(Q2Z, _MM_SHUFFLE(2, 2, 2, 2)); + // Retire Q1 and perform Q1*Q2W + vResult = _mm_mul_ps(vResult, Q1); + XMVECTOR Q1Shuffle = Q1; + // Shuffle the copies of Q1 + Q1Shuffle = XM_PERMUTE_PS(Q1Shuffle, _MM_SHUFFLE(0, 1, 2, 3)); + // Mul by Q1WZYX + Q2X = _mm_mul_ps(Q2X, Q1Shuffle); + Q1Shuffle = XM_PERMUTE_PS(Q1Shuffle, _MM_SHUFFLE(2, 3, 0, 1)); + // Flip the signs on y and z + vResult = XM_FMADD_PS(Q2X, ControlWZYX, vResult); + // Mul by Q1ZWXY + Q2Y = _mm_mul_ps(Q2Y, Q1Shuffle); + Q1Shuffle = XM_PERMUTE_PS(Q1Shuffle, _MM_SHUFFLE(0, 1, 2, 3)); + // Flip the signs on z and w + Q2Y = _mm_mul_ps(Q2Y, ControlZWXY); + // Mul by Q1YXWZ + Q2Z = _mm_mul_ps(Q2Z, Q1Shuffle); + // Flip the signs on x and w + Q2Y = XM_FMADD_PS(Q2Z, ControlYXWZ, Q2Y); + vResult = _mm_add_ps(vResult, Q2Y); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionLengthSq(FXMVECTOR Q) noexcept +{ + return XMVector4LengthSq(Q); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionReciprocalLength(FXMVECTOR Q) noexcept +{ + return XMVector4ReciprocalLength(Q); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionLength(FXMVECTOR Q) noexcept +{ + return XMVector4Length(Q); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionNormalizeEst(FXMVECTOR Q) noexcept +{ + return XMVector4NormalizeEst(Q); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionNormalize(FXMVECTOR Q) noexcept +{ + return XMVector4Normalize(Q); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionConjugate(FXMVECTOR Q) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + -Q.vector4_f32[0], + -Q.vector4_f32[1], + -Q.vector4_f32[2], + Q.vector4_f32[3] + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 NegativeOne3 = { { { -1.0f, -1.0f, -1.0f, 1.0f } } }; + return vmulq_f32(Q, NegativeOne3.v); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 NegativeOne3 = { { { -1.0f, -1.0f, -1.0f, 1.0f } } }; + return _mm_mul_ps(Q, NegativeOne3); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionInverse(FXMVECTOR Q) noexcept +{ + XMVECTOR L = XMVector4LengthSq(Q); + XMVECTOR Conjugate = XMQuaternionConjugate(Q); + + XMVECTOR Control = XMVectorLessOrEqual(L, g_XMEpsilon.v); + + XMVECTOR Result = XMVectorDivide(Conjugate, L); + + Result = XMVectorSelect(Result, g_XMZero, Control); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionLn(FXMVECTOR Q) noexcept +{ + static const XMVECTORF32 OneMinusEpsilon = { { { 1.0f - 0.00001f, 1.0f - 0.00001f, 1.0f - 0.00001f, 1.0f - 0.00001f } } }; + + XMVECTOR QW = XMVectorSplatW(Q); + XMVECTOR Q0 = XMVectorSelect(g_XMSelect1110.v, Q, g_XMSelect1110.v); + + XMVECTOR ControlW = XMVectorInBounds(QW, OneMinusEpsilon.v); + + XMVECTOR Theta = XMVectorACos(QW); + XMVECTOR SinTheta = XMVectorSin(Theta); + + XMVECTOR S = XMVectorDivide(Theta, SinTheta); + + XMVECTOR Result = XMVectorMultiply(Q0, S); + Result = XMVectorSelect(Q0, Result, ControlW); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionExp(FXMVECTOR Q) noexcept +{ + XMVECTOR Theta = XMVector3Length(Q); + + XMVECTOR SinTheta, CosTheta; + XMVectorSinCos(&SinTheta, &CosTheta, Theta); + + XMVECTOR S = XMVectorDivide(SinTheta, Theta); + + XMVECTOR Result = XMVectorMultiply(Q, S); + + const XMVECTOR Zero = XMVectorZero(); + XMVECTOR Control = XMVectorNearEqual(Theta, Zero, g_XMEpsilon.v); + Result = XMVectorSelect(Result, Q, Control); + + Result = XMVectorSelect(CosTheta, Result, g_XMSelect1110.v); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionSlerp +( + FXMVECTOR Q0, + FXMVECTOR Q1, + float t +) noexcept +{ + XMVECTOR T = XMVectorReplicate(t); + return XMQuaternionSlerpV(Q0, Q1, T); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionSlerpV +( + FXMVECTOR Q0, + FXMVECTOR Q1, + FXMVECTOR T +) noexcept +{ + assert((XMVectorGetY(T) == XMVectorGetX(T)) && (XMVectorGetZ(T) == XMVectorGetX(T)) && (XMVectorGetW(T) == XMVectorGetX(T))); + + // Result = Q0 * sin((1.0 - t) * Omega) / sin(Omega) + Q1 * sin(t * Omega) / sin(Omega) + +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + const XMVECTORF32 OneMinusEpsilon = { { { 1.0f - 0.00001f, 1.0f - 0.00001f, 1.0f - 0.00001f, 1.0f - 0.00001f } } }; + + XMVECTOR CosOmega = XMQuaternionDot(Q0, Q1); + + const XMVECTOR Zero = XMVectorZero(); + XMVECTOR Control = XMVectorLess(CosOmega, Zero); + XMVECTOR Sign = XMVectorSelect(g_XMOne.v, g_XMNegativeOne.v, Control); + + CosOmega = XMVectorMultiply(CosOmega, Sign); + + Control = XMVectorLess(CosOmega, OneMinusEpsilon); + + XMVECTOR SinOmega = XMVectorNegativeMultiplySubtract(CosOmega, CosOmega, g_XMOne.v); + SinOmega = XMVectorSqrt(SinOmega); + + XMVECTOR Omega = XMVectorATan2(SinOmega, CosOmega); + + XMVECTOR SignMask = XMVectorSplatSignMask(); + XMVECTOR V01 = XMVectorShiftLeft(T, Zero, 2); + SignMask = XMVectorShiftLeft(SignMask, Zero, 3); + V01 = XMVectorXorInt(V01, SignMask); + V01 = XMVectorAdd(g_XMIdentityR0.v, V01); + + XMVECTOR InvSinOmega = XMVectorReciprocal(SinOmega); + + XMVECTOR S0 = XMVectorMultiply(V01, Omega); + S0 = XMVectorSin(S0); + S0 = XMVectorMultiply(S0, InvSinOmega); + + S0 = XMVectorSelect(V01, S0, Control); + + XMVECTOR S1 = XMVectorSplatY(S0); + S0 = XMVectorSplatX(S0); + + S1 = XMVectorMultiply(S1, Sign); + + XMVECTOR Result = XMVectorMultiply(Q0, S0); + Result = XMVectorMultiplyAdd(Q1, S1, Result); + + return Result; + +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 OneMinusEpsilon = { { { 1.0f - 0.00001f, 1.0f - 0.00001f, 1.0f - 0.00001f, 1.0f - 0.00001f } } }; + static const XMVECTORU32 SignMask2 = { { { 0x80000000, 0x00000000, 0x00000000, 0x00000000 } } }; + + XMVECTOR CosOmega = XMQuaternionDot(Q0, Q1); + + const XMVECTOR Zero = XMVectorZero(); + XMVECTOR Control = XMVectorLess(CosOmega, Zero); + XMVECTOR Sign = XMVectorSelect(g_XMOne, g_XMNegativeOne, Control); + + CosOmega = _mm_mul_ps(CosOmega, Sign); + + Control = XMVectorLess(CosOmega, OneMinusEpsilon); + + XMVECTOR SinOmega = _mm_mul_ps(CosOmega, CosOmega); + SinOmega = _mm_sub_ps(g_XMOne, SinOmega); + SinOmega = _mm_sqrt_ps(SinOmega); + + XMVECTOR Omega = XMVectorATan2(SinOmega, CosOmega); + + XMVECTOR V01 = XM_PERMUTE_PS(T, _MM_SHUFFLE(2, 3, 0, 1)); + V01 = _mm_and_ps(V01, g_XMMaskXY); + V01 = _mm_xor_ps(V01, SignMask2); + V01 = _mm_add_ps(g_XMIdentityR0, V01); + + XMVECTOR S0 = _mm_mul_ps(V01, Omega); + S0 = XMVectorSin(S0); + S0 = _mm_div_ps(S0, SinOmega); + + S0 = XMVectorSelect(V01, S0, Control); + + XMVECTOR S1 = XMVectorSplatY(S0); + S0 = XMVectorSplatX(S0); + + S1 = _mm_mul_ps(S1, Sign); + XMVECTOR Result = _mm_mul_ps(Q0, S0); + S1 = _mm_mul_ps(S1, Q1); + Result = _mm_add_ps(Result, S1); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionSquad +( + FXMVECTOR Q0, + FXMVECTOR Q1, + FXMVECTOR Q2, + GXMVECTOR Q3, + float t +) noexcept +{ + XMVECTOR T = XMVectorReplicate(t); + return XMQuaternionSquadV(Q0, Q1, Q2, Q3, T); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionSquadV +( + FXMVECTOR Q0, + FXMVECTOR Q1, + FXMVECTOR Q2, + GXMVECTOR Q3, + HXMVECTOR T +) noexcept +{ + assert((XMVectorGetY(T) == XMVectorGetX(T)) && (XMVectorGetZ(T) == XMVectorGetX(T)) && (XMVectorGetW(T) == XMVectorGetX(T))); + + XMVECTOR TP = T; + const XMVECTOR Two = XMVectorSplatConstant(2, 0); + + XMVECTOR Q03 = XMQuaternionSlerpV(Q0, Q3, T); + XMVECTOR Q12 = XMQuaternionSlerpV(Q1, Q2, T); + + TP = XMVectorNegativeMultiplySubtract(TP, TP, TP); + TP = XMVectorMultiply(TP, Two); + + XMVECTOR Result = XMQuaternionSlerpV(Q03, Q12, TP); + + return Result; +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMQuaternionSquadSetup +( + XMVECTOR* pA, + XMVECTOR* pB, + XMVECTOR* pC, + FXMVECTOR Q0, + FXMVECTOR Q1, + FXMVECTOR Q2, + GXMVECTOR Q3 +) noexcept +{ + assert(pA); + assert(pB); + assert(pC); + + XMVECTOR LS12 = XMQuaternionLengthSq(XMVectorAdd(Q1, Q2)); + XMVECTOR LD12 = XMQuaternionLengthSq(XMVectorSubtract(Q1, Q2)); + XMVECTOR SQ2 = XMVectorNegate(Q2); + + XMVECTOR Control1 = XMVectorLess(LS12, LD12); + SQ2 = XMVectorSelect(Q2, SQ2, Control1); + + XMVECTOR LS01 = XMQuaternionLengthSq(XMVectorAdd(Q0, Q1)); + XMVECTOR LD01 = XMQuaternionLengthSq(XMVectorSubtract(Q0, Q1)); + XMVECTOR SQ0 = XMVectorNegate(Q0); + + XMVECTOR LS23 = XMQuaternionLengthSq(XMVectorAdd(SQ2, Q3)); + XMVECTOR LD23 = XMQuaternionLengthSq(XMVectorSubtract(SQ2, Q3)); + XMVECTOR SQ3 = XMVectorNegate(Q3); + + XMVECTOR Control0 = XMVectorLess(LS01, LD01); + XMVECTOR Control2 = XMVectorLess(LS23, LD23); + + SQ0 = XMVectorSelect(Q0, SQ0, Control0); + SQ3 = XMVectorSelect(Q3, SQ3, Control2); + + XMVECTOR InvQ1 = XMQuaternionInverse(Q1); + XMVECTOR InvQ2 = XMQuaternionInverse(SQ2); + + XMVECTOR LnQ0 = XMQuaternionLn(XMQuaternionMultiply(InvQ1, SQ0)); + XMVECTOR LnQ2 = XMQuaternionLn(XMQuaternionMultiply(InvQ1, SQ2)); + XMVECTOR LnQ1 = XMQuaternionLn(XMQuaternionMultiply(InvQ2, Q1)); + XMVECTOR LnQ3 = XMQuaternionLn(XMQuaternionMultiply(InvQ2, SQ3)); + + const XMVECTOR NegativeOneQuarter = XMVectorSplatConstant(-1, 2); + + XMVECTOR ExpQ02 = XMVectorMultiply(XMVectorAdd(LnQ0, LnQ2), NegativeOneQuarter); + XMVECTOR ExpQ13 = XMVectorMultiply(XMVectorAdd(LnQ1, LnQ3), NegativeOneQuarter); + ExpQ02 = XMQuaternionExp(ExpQ02); + ExpQ13 = XMQuaternionExp(ExpQ13); + + *pA = XMQuaternionMultiply(Q1, ExpQ02); + *pB = XMQuaternionMultiply(SQ2, ExpQ13); + *pC = SQ2; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionBaryCentric +( + FXMVECTOR Q0, + FXMVECTOR Q1, + FXMVECTOR Q2, + float f, + float g +) noexcept +{ + float s = f + g; + + XMVECTOR Result; + if ((s < 0.00001f) && (s > -0.00001f)) + { + Result = Q0; + } + else + { + XMVECTOR Q01 = XMQuaternionSlerp(Q0, Q1, s); + XMVECTOR Q02 = XMQuaternionSlerp(Q0, Q2, s); + + Result = XMQuaternionSlerp(Q01, Q02, g / s); + } + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionBaryCentricV +( + FXMVECTOR Q0, + FXMVECTOR Q1, + FXMVECTOR Q2, + GXMVECTOR F, + HXMVECTOR G +) noexcept +{ + assert((XMVectorGetY(F) == XMVectorGetX(F)) && (XMVectorGetZ(F) == XMVectorGetX(F)) && (XMVectorGetW(F) == XMVectorGetX(F))); + assert((XMVectorGetY(G) == XMVectorGetX(G)) && (XMVectorGetZ(G) == XMVectorGetX(G)) && (XMVectorGetW(G) == XMVectorGetX(G))); + + const XMVECTOR Epsilon = XMVectorSplatConstant(1, 16); + + XMVECTOR S = XMVectorAdd(F, G); + + XMVECTOR Result; + if (XMVector4InBounds(S, Epsilon)) + { + Result = Q0; + } + else + { + XMVECTOR Q01 = XMQuaternionSlerpV(Q0, Q1, S); + XMVECTOR Q02 = XMQuaternionSlerpV(Q0, Q2, S); + XMVECTOR GS = XMVectorReciprocal(S); + GS = XMVectorMultiply(G, GS); + + Result = XMQuaternionSlerpV(Q01, Q02, GS); + } + + return Result; +} + +//------------------------------------------------------------------------------ +// Transformation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionIdentity() noexcept +{ + return g_XMIdentityR3.v; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionRotationRollPitchYaw +( + float Pitch, + float Yaw, + float Roll +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + const float halfpitch = Pitch * 0.5f; + float cp = cosf(halfpitch); + float sp = sinf(halfpitch); + + const float halfyaw = Yaw * 0.5f; + float cy = cosf(halfyaw); + float sy = sinf(halfyaw); + + const float halfroll = Roll * 0.5f; + float cr = cosf(halfroll); + float sr = sinf(halfroll); + + XMVECTORF32 vResult = { { { + cr * sp * cy + sr * cp * sy, + cr * cp * sy - sr * sp * cy, + sr * cp * cy - cr * sp * sy, + cr * cp * cy + sr * sp * sy + } } }; + return vResult; +#else + XMVECTOR Angles = XMVectorSet(Pitch, Yaw, Roll, 0.0f); + return XMQuaternionRotationRollPitchYawFromVector(Angles); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionRotationRollPitchYawFromVector +( + FXMVECTOR Angles // +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + const float halfpitch = Angles.vector4_f32[0] * 0.5f; + float cp = cosf(halfpitch); + float sp = sinf(halfpitch); + + const float halfyaw = Angles.vector4_f32[1] * 0.5f; + float cy = cosf(halfyaw); + float sy = sinf(halfyaw); + + const float halfroll = Angles.vector4_f32[2] * 0.5f; + float cr = cosf(halfroll); + float sr = sinf(halfroll); + + XMVECTORF32 vResult = { { { + cr * sp * cy + sr * cp * sy, + cr * cp * sy - sr * sp * cy, + sr * cp * cy - cr * sp * sy, + cr * cp * cy + sr * sp * sy + } } }; + return vResult; +#else + static const XMVECTORF32 Sign = { { { 1.0f, -1.0f, -1.0f, 1.0f } } }; + + XMVECTOR HalfAngles = XMVectorMultiply(Angles, g_XMOneHalf.v); + + XMVECTOR SinAngles, CosAngles; + XMVectorSinCos(&SinAngles, &CosAngles, HalfAngles); + + XMVECTOR P0 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR Y0 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR R0 = XMVectorPermute(SinAngles, CosAngles); + XMVECTOR P1 = XMVectorPermute(CosAngles, SinAngles); + XMVECTOR Y1 = XMVectorPermute(CosAngles, SinAngles); + XMVECTOR R1 = XMVectorPermute(CosAngles, SinAngles); + + XMVECTOR Q1 = XMVectorMultiply(P1, Sign.v); + XMVECTOR Q0 = XMVectorMultiply(P0, Y0); + Q1 = XMVectorMultiply(Q1, Y1); + Q0 = XMVectorMultiply(Q0, R0); + XMVECTOR Q = XMVectorMultiplyAdd(Q1, R1, Q0); + + return Q; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionRotationNormal +( + FXMVECTOR NormalAxis, + float Angle +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + XMVECTOR N = XMVectorSelect(g_XMOne.v, NormalAxis, g_XMSelect1110.v); + + float SinV, CosV; + XMScalarSinCos(&SinV, &CosV, 0.5f * Angle); + + XMVECTOR Scale = XMVectorSet(SinV, SinV, SinV, CosV); + return XMVectorMultiply(N, Scale); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR N = _mm_and_ps(NormalAxis, g_XMMask3); + N = _mm_or_ps(N, g_XMIdentityR3); + XMVECTOR Scale = _mm_set_ps1(0.5f * Angle); + XMVECTOR vSine; + XMVECTOR vCosine; + XMVectorSinCos(&vSine, &vCosine, Scale); + Scale = _mm_and_ps(vSine, g_XMMask3); + vCosine = _mm_and_ps(vCosine, g_XMMaskW); + Scale = _mm_or_ps(Scale, vCosine); + N = _mm_mul_ps(N, Scale); + return N; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionRotationAxis +( + FXMVECTOR Axis, + float Angle +) noexcept +{ + assert(!XMVector3Equal(Axis, XMVectorZero())); + assert(!XMVector3IsInfinite(Axis)); + + XMVECTOR Normal = XMVector3Normalize(Axis); + XMVECTOR Q = XMQuaternionRotationNormal(Normal, Angle); + return Q; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMQuaternionRotationMatrix(FXMMATRIX M) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 q; + float r22 = M.m[2][2]; + if (r22 <= 0.f) // x^2 + y^2 >= z^2 + w^2 + { + float dif10 = M.m[1][1] - M.m[0][0]; + float omr22 = 1.f - r22; + if (dif10 <= 0.f) // x^2 >= y^2 + { + float fourXSqr = omr22 - dif10; + float inv4x = 0.5f / sqrtf(fourXSqr); + q.f[0] = fourXSqr * inv4x; + q.f[1] = (M.m[0][1] + M.m[1][0]) * inv4x; + q.f[2] = (M.m[0][2] + M.m[2][0]) * inv4x; + q.f[3] = (M.m[1][2] - M.m[2][1]) * inv4x; + } + else // y^2 >= x^2 + { + float fourYSqr = omr22 + dif10; + float inv4y = 0.5f / sqrtf(fourYSqr); + q.f[0] = (M.m[0][1] + M.m[1][0]) * inv4y; + q.f[1] = fourYSqr * inv4y; + q.f[2] = (M.m[1][2] + M.m[2][1]) * inv4y; + q.f[3] = (M.m[2][0] - M.m[0][2]) * inv4y; + } + } + else // z^2 + w^2 >= x^2 + y^2 + { + float sum10 = M.m[1][1] + M.m[0][0]; + float opr22 = 1.f + r22; + if (sum10 <= 0.f) // z^2 >= w^2 + { + float fourZSqr = opr22 - sum10; + float inv4z = 0.5f / sqrtf(fourZSqr); + q.f[0] = (M.m[0][2] + M.m[2][0]) * inv4z; + q.f[1] = (M.m[1][2] + M.m[2][1]) * inv4z; + q.f[2] = fourZSqr * inv4z; + q.f[3] = (M.m[0][1] - M.m[1][0]) * inv4z; + } + else // w^2 >= z^2 + { + float fourWSqr = opr22 + sum10; + float inv4w = 0.5f / sqrtf(fourWSqr); + q.f[0] = (M.m[1][2] - M.m[2][1]) * inv4w; + q.f[1] = (M.m[2][0] - M.m[0][2]) * inv4w; + q.f[2] = (M.m[0][1] - M.m[1][0]) * inv4w; + q.f[3] = fourWSqr * inv4w; + } + } + return q.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 XMPMMP = { { { +1.0f, -1.0f, -1.0f, +1.0f } } }; + static const XMVECTORF32 XMMPMP = { { { -1.0f, +1.0f, -1.0f, +1.0f } } }; + static const XMVECTORF32 XMMMPP = { { { -1.0f, -1.0f, +1.0f, +1.0f } } }; + static const XMVECTORU32 Select0110 = { { { XM_SELECT_0, XM_SELECT_1, XM_SELECT_1, XM_SELECT_0 } } }; + static const XMVECTORU32 Select0010 = { { { XM_SELECT_0, XM_SELECT_0, XM_SELECT_1, XM_SELECT_0 } } }; + + float32x4_t r0 = M.r[0]; + float32x4_t r1 = M.r[1]; + float32x4_t r2 = M.r[2]; + + float32x4_t r00 = vdupq_lane_f32(vget_low_f32(r0), 0); + float32x4_t r11 = vdupq_lane_f32(vget_low_f32(r1), 1); + float32x4_t r22 = vdupq_lane_f32(vget_high_f32(r2), 0); + + // x^2 >= y^2 equivalent to r11 - r00 <= 0 + float32x4_t r11mr00 = vsubq_f32(r11, r00); + uint32x4_t x2gey2 = vcleq_f32(r11mr00, g_XMZero); + + // z^2 >= w^2 equivalent to r11 + r00 <= 0 + float32x4_t r11pr00 = vaddq_f32(r11, r00); + uint32x4_t z2gew2 = vcleq_f32(r11pr00, g_XMZero); + + // x^2 + y^2 >= z^2 + w^2 equivalent to r22 <= 0 + uint32x4_t x2py2gez2pw2 = vcleq_f32(r22, g_XMZero); + + // (4*x^2, 4*y^2, 4*z^2, 4*w^2) + float32x4_t t0 = vmulq_f32(XMPMMP, r00); + float32x4_t x2y2z2w2 = vmlaq_f32(t0, XMMPMP, r11); + x2y2z2w2 = vmlaq_f32(x2y2z2w2, XMMMPP, r22); + x2y2z2w2 = vaddq_f32(x2y2z2w2, g_XMOne); + + // (r01, r02, r12, r11) + t0 = vextq_f32(r0, r0, 1); + float32x4_t t1 = vextq_f32(r1, r1, 1); + t0 = vcombine_f32(vget_low_f32(t0), vrev64_f32(vget_low_f32(t1))); + + // (r10, r20, r21, r10) + t1 = vextq_f32(r2, r2, 3); + float32x4_t r10 = vdupq_lane_f32(vget_low_f32(r1), 0); + t1 = vbslq_f32(Select0110, t1, r10); + + // (4*x*y, 4*x*z, 4*y*z, unused) + float32x4_t xyxzyz = vaddq_f32(t0, t1); + + // (r21, r20, r10, r10) + t0 = vcombine_f32(vrev64_f32(vget_low_f32(r2)), vget_low_f32(r10)); + + // (r12, r02, r01, r12) + float32x4_t t2 = vcombine_f32(vrev64_f32(vget_high_f32(r0)), vrev64_f32(vget_low_f32(r0))); + float32x4_t t3 = vdupq_lane_f32(vget_high_f32(r1), 0); + t1 = vbslq_f32(Select0110, t2, t3); + + // (4*x*w, 4*y*w, 4*z*w, unused) + float32x4_t xwywzw = vsubq_f32(t0, t1); + xwywzw = vmulq_f32(XMMPMP, xwywzw); + + // (4*x*x, 4*x*y, 4*x*z, 4*x*w) + t0 = vextq_f32(xyxzyz, xyxzyz, 3); + t1 = vbslq_f32(Select0110, t0, x2y2z2w2); + t2 = vdupq_lane_f32(vget_low_f32(xwywzw), 0); + float32x4_t tensor0 = vbslq_f32(g_XMSelect1110, t1, t2); + + // (4*y*x, 4*y*y, 4*y*z, 4*y*w) + t0 = vbslq_f32(g_XMSelect1011, xyxzyz, x2y2z2w2); + t1 = vdupq_lane_f32(vget_low_f32(xwywzw), 1); + float32x4_t tensor1 = vbslq_f32(g_XMSelect1110, t0, t1); + + // (4*z*x, 4*z*y, 4*z*z, 4*z*w) + t0 = vextq_f32(xyxzyz, xyxzyz, 1); + t1 = vcombine_f32(vget_low_f32(t0), vrev64_f32(vget_high_f32(xwywzw))); + float32x4_t tensor2 = vbslq_f32(Select0010, x2y2z2w2, t1); + + // (4*w*x, 4*w*y, 4*w*z, 4*w*w) + float32x4_t tensor3 = vbslq_f32(g_XMSelect1110, xwywzw, x2y2z2w2); + + // Select the row of the tensor-product matrix that has the largest + // magnitude. + t0 = vbslq_f32(x2gey2, tensor0, tensor1); + t1 = vbslq_f32(z2gew2, tensor2, tensor3); + t2 = vbslq_f32(x2py2gez2pw2, t0, t1); + + // Normalize the row. No division by zero is possible because the + // quaternion is unit-length (and the row is a nonzero multiple of + // the quaternion). + t0 = XMVector4Length(t2); + return XMVectorDivide(t2, t0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 XMPMMP = { { { +1.0f, -1.0f, -1.0f, +1.0f } } }; + static const XMVECTORF32 XMMPMP = { { { -1.0f, +1.0f, -1.0f, +1.0f } } }; + static const XMVECTORF32 XMMMPP = { { { -1.0f, -1.0f, +1.0f, +1.0f } } }; + + XMVECTOR r0 = M.r[0]; // (r00, r01, r02, 0) + XMVECTOR r1 = M.r[1]; // (r10, r11, r12, 0) + XMVECTOR r2 = M.r[2]; // (r20, r21, r22, 0) + + // (r00, r00, r00, r00) + XMVECTOR r00 = XM_PERMUTE_PS(r0, _MM_SHUFFLE(0, 0, 0, 0)); + // (r11, r11, r11, r11) + XMVECTOR r11 = XM_PERMUTE_PS(r1, _MM_SHUFFLE(1, 1, 1, 1)); + // (r22, r22, r22, r22) + XMVECTOR r22 = XM_PERMUTE_PS(r2, _MM_SHUFFLE(2, 2, 2, 2)); + + // x^2 >= y^2 equivalent to r11 - r00 <= 0 + // (r11 - r00, r11 - r00, r11 - r00, r11 - r00) + XMVECTOR r11mr00 = _mm_sub_ps(r11, r00); + XMVECTOR x2gey2 = _mm_cmple_ps(r11mr00, g_XMZero); + + // z^2 >= w^2 equivalent to r11 + r00 <= 0 + // (r11 + r00, r11 + r00, r11 + r00, r11 + r00) + XMVECTOR r11pr00 = _mm_add_ps(r11, r00); + XMVECTOR z2gew2 = _mm_cmple_ps(r11pr00, g_XMZero); + + // x^2 + y^2 >= z^2 + w^2 equivalent to r22 <= 0 + XMVECTOR x2py2gez2pw2 = _mm_cmple_ps(r22, g_XMZero); + + // (4*x^2, 4*y^2, 4*z^2, 4*w^2) + XMVECTOR t0 = XM_FMADD_PS(XMPMMP, r00, g_XMOne); + XMVECTOR t1 = _mm_mul_ps(XMMPMP, r11); + XMVECTOR t2 = XM_FMADD_PS(XMMMPP, r22, t0); + XMVECTOR x2y2z2w2 = _mm_add_ps(t1, t2); + + // (r01, r02, r12, r11) + t0 = _mm_shuffle_ps(r0, r1, _MM_SHUFFLE(1, 2, 2, 1)); + // (r10, r10, r20, r21) + t1 = _mm_shuffle_ps(r1, r2, _MM_SHUFFLE(1, 0, 0, 0)); + // (r10, r20, r21, r10) + t1 = XM_PERMUTE_PS(t1, _MM_SHUFFLE(1, 3, 2, 0)); + // (4*x*y, 4*x*z, 4*y*z, unused) + XMVECTOR xyxzyz = _mm_add_ps(t0, t1); + + // (r21, r20, r10, r10) + t0 = _mm_shuffle_ps(r2, r1, _MM_SHUFFLE(0, 0, 0, 1)); + // (r12, r12, r02, r01) + t1 = _mm_shuffle_ps(r1, r0, _MM_SHUFFLE(1, 2, 2, 2)); + // (r12, r02, r01, r12) + t1 = XM_PERMUTE_PS(t1, _MM_SHUFFLE(1, 3, 2, 0)); + // (4*x*w, 4*y*w, 4*z*w, unused) + XMVECTOR xwywzw = _mm_sub_ps(t0, t1); + xwywzw = _mm_mul_ps(XMMPMP, xwywzw); + + // (4*x^2, 4*y^2, 4*x*y, unused) + t0 = _mm_shuffle_ps(x2y2z2w2, xyxzyz, _MM_SHUFFLE(0, 0, 1, 0)); + // (4*z^2, 4*w^2, 4*z*w, unused) + t1 = _mm_shuffle_ps(x2y2z2w2, xwywzw, _MM_SHUFFLE(0, 2, 3, 2)); + // (4*x*z, 4*y*z, 4*x*w, 4*y*w) + t2 = _mm_shuffle_ps(xyxzyz, xwywzw, _MM_SHUFFLE(1, 0, 2, 1)); + + // (4*x*x, 4*x*y, 4*x*z, 4*x*w) + XMVECTOR tensor0 = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(2, 0, 2, 0)); + // (4*y*x, 4*y*y, 4*y*z, 4*y*w) + XMVECTOR tensor1 = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(3, 1, 1, 2)); + // (4*z*x, 4*z*y, 4*z*z, 4*z*w) + XMVECTOR tensor2 = _mm_shuffle_ps(t2, t1, _MM_SHUFFLE(2, 0, 1, 0)); + // (4*w*x, 4*w*y, 4*w*z, 4*w*w) + XMVECTOR tensor3 = _mm_shuffle_ps(t2, t1, _MM_SHUFFLE(1, 2, 3, 2)); + + // Select the row of the tensor-product matrix that has the largest + // magnitude. + t0 = _mm_and_ps(x2gey2, tensor0); + t1 = _mm_andnot_ps(x2gey2, tensor1); + t0 = _mm_or_ps(t0, t1); + t1 = _mm_and_ps(z2gew2, tensor2); + t2 = _mm_andnot_ps(z2gew2, tensor3); + t1 = _mm_or_ps(t1, t2); + t0 = _mm_and_ps(x2py2gez2pw2, t0); + t1 = _mm_andnot_ps(x2py2gez2pw2, t1); + t2 = _mm_or_ps(t0, t1); + + // Normalize the row. No division by zero is possible because the + // quaternion is unit-length (and the row is a nonzero multiple of + // the quaternion). + t0 = XMVector4Length(t2); + return _mm_div_ps(t2, t0); +#endif +} + +//------------------------------------------------------------------------------ +// Conversion operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMQuaternionToAxisAngle +( + XMVECTOR* pAxis, + float* pAngle, + FXMVECTOR Q +) noexcept +{ + assert(pAxis); + assert(pAngle); + + *pAxis = Q; + + *pAngle = 2.0f * XMScalarACos(XMVectorGetW(Q)); +} + +/**************************************************************************** + * + * Plane + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMPlaneEqual +( + FXMVECTOR P1, + FXMVECTOR P2 +) noexcept +{ + return XMVector4Equal(P1, P2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMPlaneNearEqual +( + FXMVECTOR P1, + FXMVECTOR P2, + FXMVECTOR Epsilon +) noexcept +{ + XMVECTOR NP1 = XMPlaneNormalize(P1); + XMVECTOR NP2 = XMPlaneNormalize(P2); + return XMVector4NearEqual(NP1, NP2, Epsilon); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMPlaneNotEqual +( + FXMVECTOR P1, + FXMVECTOR P2 +) noexcept +{ + return XMVector4NotEqual(P1, P2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMPlaneIsNaN(FXMVECTOR P) noexcept +{ + return XMVector4IsNaN(P); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMPlaneIsInfinite(FXMVECTOR P) noexcept +{ + return XMVector4IsInfinite(P); +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneDot +( + FXMVECTOR P, + FXMVECTOR V +) noexcept +{ + return XMVector4Dot(P, V); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneDotCoord +( + FXMVECTOR P, + FXMVECTOR V +) noexcept +{ + // Result = P[0] * V[0] + P[1] * V[1] + P[2] * V[2] + P[3] + + XMVECTOR V3 = XMVectorSelect(g_XMOne.v, V, g_XMSelect1110.v); + XMVECTOR Result = XMVector4Dot(P, V3); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneDotNormal +( + FXMVECTOR P, + FXMVECTOR V +) noexcept +{ + return XMVector3Dot(P, V); +} + +//------------------------------------------------------------------------------ +// XMPlaneNormalizeEst uses a reciprocal estimate and +// returns QNaN on zero and infinite vectors. + +inline XMVECTOR XM_CALLCONV XMPlaneNormalizeEst(FXMVECTOR P) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + XMVECTOR Result = XMVector3ReciprocalLengthEst(P); + return XMVectorMultiply(P, Result); + +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(P, P, 0x7f); + XMVECTOR vResult = _mm_rsqrt_ps(vTemp); + return _mm_mul_ps(vResult, P); +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product + XMVECTOR vDot = _mm_mul_ps(P, P); + // x=Dot.y, y=Dot.z + XMVECTOR vTemp = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(2, 1, 2, 1)); + // Result.x = x+y + vDot = _mm_add_ss(vDot, vTemp); + // x=Dot.z + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // Result.x = (x+y)+z + vDot = _mm_add_ss(vDot, vTemp); + // Splat x + vDot = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(0, 0, 0, 0)); + // Get the reciprocal + vDot = _mm_rsqrt_ps(vDot); + // Get the reciprocal + vDot = _mm_mul_ps(vDot, P); + return vDot; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneNormalize(FXMVECTOR P) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float fLengthSq = sqrtf((P.vector4_f32[0] * P.vector4_f32[0]) + (P.vector4_f32[1] * P.vector4_f32[1]) + (P.vector4_f32[2] * P.vector4_f32[2])); + // Prevent divide by zero + if (fLengthSq > 0) + { + fLengthSq = 1.0f / fLengthSq; + } + XMVECTORF32 vResult = { { { + P.vector4_f32[0] * fLengthSq, + P.vector4_f32[1] * fLengthSq, + P.vector4_f32[2] * fLengthSq, + P.vector4_f32[3] * fLengthSq + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR vLength = XMVector3ReciprocalLength(P); + return XMVectorMultiply(P, vLength); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vLengthSq = _mm_dp_ps(P, P, 0x7f); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Reciprocal mul to perform the normalization + vResult = _mm_div_ps(P, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vLengthSq); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y and z only + XMVECTOR vLengthSq = _mm_mul_ps(P, P); + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 1, 2, 1)); + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Reciprocal mul to perform the normalization + vResult = _mm_div_ps(P, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vLengthSq); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneIntersectLine +( + FXMVECTOR P, + FXMVECTOR LinePoint1, + FXMVECTOR LinePoint2 +) noexcept +{ + XMVECTOR V1 = XMVector3Dot(P, LinePoint1); + XMVECTOR V2 = XMVector3Dot(P, LinePoint2); + XMVECTOR D = XMVectorSubtract(V1, V2); + + XMVECTOR VT = XMPlaneDotCoord(P, LinePoint1); + VT = XMVectorDivide(VT, D); + + XMVECTOR Point = XMVectorSubtract(LinePoint2, LinePoint1); + Point = XMVectorMultiplyAdd(Point, VT, LinePoint1); + + const XMVECTOR Zero = XMVectorZero(); + XMVECTOR Control = XMVectorNearEqual(D, Zero, g_XMEpsilon.v); + + return XMVectorSelect(Point, g_XMQNaN.v, Control); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMPlaneIntersectPlane +( + XMVECTOR* pLinePoint1, + XMVECTOR* pLinePoint2, + FXMVECTOR P1, + FXMVECTOR P2 +) noexcept +{ + assert(pLinePoint1); + assert(pLinePoint2); + + XMVECTOR V1 = XMVector3Cross(P2, P1); + + XMVECTOR LengthSq = XMVector3LengthSq(V1); + + XMVECTOR V2 = XMVector3Cross(P2, V1); + + XMVECTOR P1W = XMVectorSplatW(P1); + XMVECTOR Point = XMVectorMultiply(V2, P1W); + + XMVECTOR V3 = XMVector3Cross(V1, P1); + + XMVECTOR P2W = XMVectorSplatW(P2); + Point = XMVectorMultiplyAdd(V3, P2W, Point); + + XMVECTOR LinePoint1 = XMVectorDivide(Point, LengthSq); + + XMVECTOR LinePoint2 = XMVectorAdd(LinePoint1, V1); + + XMVECTOR Control = XMVectorLessOrEqual(LengthSq, g_XMEpsilon.v); + *pLinePoint1 = XMVectorSelect(LinePoint1, g_XMQNaN.v, Control); + *pLinePoint2 = XMVectorSelect(LinePoint2, g_XMQNaN.v, Control); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneTransform +( + FXMVECTOR P, + FXMMATRIX ITM +) noexcept +{ + XMVECTOR W = XMVectorSplatW(P); + XMVECTOR Z = XMVectorSplatZ(P); + XMVECTOR Y = XMVectorSplatY(P); + XMVECTOR X = XMVectorSplatX(P); + + XMVECTOR Result = XMVectorMultiply(W, ITM.r[3]); + Result = XMVectorMultiplyAdd(Z, ITM.r[2], Result); + Result = XMVectorMultiplyAdd(Y, ITM.r[1], Result); + Result = XMVectorMultiplyAdd(X, ITM.r[0], Result); + return Result; +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMFLOAT4* XM_CALLCONV XMPlaneTransformStream +( + XMFLOAT4* pOutputStream, + size_t OutputStride, + const XMFLOAT4* pInputStream, + size_t InputStride, + size_t PlaneCount, + FXMMATRIX ITM +) noexcept +{ + return XMVector4TransformStream(pOutputStream, + OutputStride, + pInputStream, + InputStride, + PlaneCount, + ITM); +} + +//------------------------------------------------------------------------------ +// Conversion operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneFromPointNormal +( + FXMVECTOR Point, + FXMVECTOR Normal +) noexcept +{ + XMVECTOR W = XMVector3Dot(Point, Normal); + W = XMVectorNegate(W); + return XMVectorSelect(W, Normal, g_XMSelect1110.v); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMPlaneFromPoints +( + FXMVECTOR Point1, + FXMVECTOR Point2, + FXMVECTOR Point3 +) noexcept +{ + XMVECTOR V21 = XMVectorSubtract(Point1, Point2); + XMVECTOR V31 = XMVectorSubtract(Point1, Point3); + + XMVECTOR N = XMVector3Cross(V21, V31); + N = XMVector3Normalize(N); + + XMVECTOR D = XMPlaneDotNormal(N, Point1); + D = XMVectorNegate(D); + + XMVECTOR Result = XMVectorSelect(D, N, g_XMSelect1110.v); + + return Result; +} + +/**************************************************************************** + * + * Color + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorEqual +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVector4Equal(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorNotEqual +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVector4NotEqual(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorGreater +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVector4Greater(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorGreaterOrEqual +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVector4GreaterOrEqual(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorLess +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVector4Less(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorLessOrEqual +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVector4LessOrEqual(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorIsNaN(FXMVECTOR C) noexcept +{ + return XMVector4IsNaN(C); +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMColorIsInfinite(FXMVECTOR C) noexcept +{ + return XMVector4IsInfinite(C); +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorNegative(FXMVECTOR vColor) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + 1.0f - vColor.vector4_f32[0], + 1.0f - vColor.vector4_f32[1], + 1.0f - vColor.vector4_f32[2], + vColor.vector4_f32[3] + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vTemp = veorq_u32(vreinterpretq_u32_f32(vColor), g_XMNegate3); + return vaddq_f32(vreinterpretq_f32_u32(vTemp), g_XMOne3); +#elif defined(_XM_SSE_INTRINSICS_) + // Negate only x,y and z. + XMVECTOR vTemp = _mm_xor_ps(vColor, g_XMNegate3); + // Add 1,1,1,0 to -x,-y,-z,w + return _mm_add_ps(vTemp, g_XMOne3); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorModulate +( + FXMVECTOR C1, + FXMVECTOR C2 +) noexcept +{ + return XMVectorMultiply(C1, C2); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorAdjustSaturation +( + FXMVECTOR vColor, + float fSaturation +) noexcept +{ + // Luminance = 0.2125f * C[0] + 0.7154f * C[1] + 0.0721f * C[2]; + // Result = (C - Luminance) * Saturation + Luminance; + + const XMVECTORF32 gvLuminance = { { { 0.2125f, 0.7154f, 0.0721f, 0.0f } } }; +#if defined(_XM_NO_INTRINSICS_) + float fLuminance = (vColor.vector4_f32[0] * gvLuminance.f[0]) + (vColor.vector4_f32[1] * gvLuminance.f[1]) + (vColor.vector4_f32[2] * gvLuminance.f[2]); + XMVECTOR vResult; + vResult.vector4_f32[0] = ((vColor.vector4_f32[0] - fLuminance) * fSaturation) + fLuminance; + vResult.vector4_f32[1] = ((vColor.vector4_f32[1] - fLuminance) * fSaturation) + fLuminance; + vResult.vector4_f32[2] = ((vColor.vector4_f32[2] - fLuminance) * fSaturation) + fLuminance; + vResult.vector4_f32[3] = vColor.vector4_f32[3]; + return vResult; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR vLuminance = XMVector3Dot(vColor, gvLuminance); + XMVECTOR vResult = vsubq_f32(vColor, vLuminance); + vResult = vmlaq_n_f32(vLuminance, vResult, fSaturation); + return vbslq_f32(g_XMSelect1110, vResult, vColor); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vLuminance = XMVector3Dot(vColor, gvLuminance); + // Splat fSaturation + XMVECTOR vSaturation = _mm_set_ps1(fSaturation); + // vResult = ((vColor-vLuminance)*vSaturation)+vLuminance; + XMVECTOR vResult = _mm_sub_ps(vColor, vLuminance); + vResult = XM_FMADD_PS(vResult, vSaturation, vLuminance); + // Retain w from the source color + vLuminance = _mm_shuffle_ps(vResult, vColor, _MM_SHUFFLE(3, 2, 2, 2)); // x = vResult.z,y = vResult.z,z = vColor.z,w=vColor.w + vResult = _mm_shuffle_ps(vResult, vLuminance, _MM_SHUFFLE(3, 0, 1, 0)); // x = vResult.x,y = vResult.y,z = vResult.z,w=vColor.w + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorAdjustContrast +( + FXMVECTOR vColor, + float fContrast +) noexcept +{ + // Result = (vColor - 0.5f) * fContrast + 0.5f; + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + ((vColor.vector4_f32[0] - 0.5f) * fContrast) + 0.5f, + ((vColor.vector4_f32[1] - 0.5f) * fContrast) + 0.5f, + ((vColor.vector4_f32[2] - 0.5f) * fContrast) + 0.5f, + vColor.vector4_f32[3] // Leave W untouched + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR vResult = vsubq_f32(vColor, g_XMOneHalf.v); + vResult = vmlaq_n_f32(g_XMOneHalf.v, vResult, fContrast); + return vbslq_f32(g_XMSelect1110, vResult, vColor); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vScale = _mm_set_ps1(fContrast); // Splat the scale + XMVECTOR vResult = _mm_sub_ps(vColor, g_XMOneHalf); // Subtract 0.5f from the source (Saving source) + vResult = XM_FMADD_PS(vResult, vScale, g_XMOneHalf); +// Retain w from the source color + vScale = _mm_shuffle_ps(vResult, vColor, _MM_SHUFFLE(3, 2, 2, 2)); // x = vResult.z,y = vResult.z,z = vColor.z,w=vColor.w + vResult = _mm_shuffle_ps(vResult, vScale, _MM_SHUFFLE(3, 0, 1, 0)); // x = vResult.x,y = vResult.y,z = vResult.z,w=vColor.w + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToHSL(FXMVECTOR rgb) noexcept +{ + XMVECTOR r = XMVectorSplatX(rgb); + XMVECTOR g = XMVectorSplatY(rgb); + XMVECTOR b = XMVectorSplatZ(rgb); + + XMVECTOR min = XMVectorMin(r, XMVectorMin(g, b)); + XMVECTOR max = XMVectorMax(r, XMVectorMax(g, b)); + + XMVECTOR l = XMVectorMultiply(XMVectorAdd(min, max), g_XMOneHalf); + + XMVECTOR d = XMVectorSubtract(max, min); + + XMVECTOR la = XMVectorSelect(rgb, l, g_XMSelect1110); + + if (XMVector3Less(d, g_XMEpsilon)) + { + // Achromatic, assume H and S of 0 + return XMVectorSelect(la, g_XMZero, g_XMSelect1100); + } + else + { + XMVECTOR s, h; + + XMVECTOR d2 = XMVectorAdd(min, max); + + if (XMVector3Greater(l, g_XMOneHalf)) + { + // d / (2-max-min) + s = XMVectorDivide(d, XMVectorSubtract(g_XMTwo, d2)); + } + else + { + // d / (max+min) + s = XMVectorDivide(d, d2); + } + + if (XMVector3Equal(r, max)) + { + // Red is max + h = XMVectorDivide(XMVectorSubtract(g, b), d); + } + else if (XMVector3Equal(g, max)) + { + // Green is max + h = XMVectorDivide(XMVectorSubtract(b, r), d); + h = XMVectorAdd(h, g_XMTwo); + } + else + { + // Blue is max + h = XMVectorDivide(XMVectorSubtract(r, g), d); + h = XMVectorAdd(h, g_XMFour); + } + + h = XMVectorDivide(h, g_XMSix); + + if (XMVector3Less(h, g_XMZero)) + h = XMVectorAdd(h, g_XMOne); + + XMVECTOR lha = XMVectorSelect(la, h, g_XMSelect1100); + return XMVectorSelect(s, lha, g_XMSelect1011); + } +} + +//------------------------------------------------------------------------------ + +namespace MathInternal +{ + + inline XMVECTOR XM_CALLCONV XMColorHue2Clr(FXMVECTOR p, FXMVECTOR q, FXMVECTOR h) noexcept + { + static const XMVECTORF32 oneSixth = { { { 1.0f / 6.0f, 1.0f / 6.0f, 1.0f / 6.0f, 1.0f / 6.0f } } }; + static const XMVECTORF32 twoThirds = { { { 2.0f / 3.0f, 2.0f / 3.0f, 2.0f / 3.0f, 2.0f / 3.0f } } }; + + XMVECTOR t = h; + + if (XMVector3Less(t, g_XMZero)) + t = XMVectorAdd(t, g_XMOne); + + if (XMVector3Greater(t, g_XMOne)) + t = XMVectorSubtract(t, g_XMOne); + + if (XMVector3Less(t, oneSixth)) + { + // p + (q - p) * 6 * t + XMVECTOR t1 = XMVectorSubtract(q, p); + XMVECTOR t2 = XMVectorMultiply(g_XMSix, t); + return XMVectorMultiplyAdd(t1, t2, p); + } + + if (XMVector3Less(t, g_XMOneHalf)) + return q; + + if (XMVector3Less(t, twoThirds)) + { + // p + (q - p) * 6 * (2/3 - t) + XMVECTOR t1 = XMVectorSubtract(q, p); + XMVECTOR t2 = XMVectorMultiply(g_XMSix, XMVectorSubtract(twoThirds, t)); + return XMVectorMultiplyAdd(t1, t2, p); + } + + return p; + } + +} // namespace MathInternal + +inline XMVECTOR XM_CALLCONV XMColorHSLToRGB(FXMVECTOR hsl) noexcept +{ + static const XMVECTORF32 oneThird = { { { 1.0f / 3.0f, 1.0f / 3.0f, 1.0f / 3.0f, 1.0f / 3.0f } } }; + + XMVECTOR s = XMVectorSplatY(hsl); + XMVECTOR l = XMVectorSplatZ(hsl); + + if (XMVector3NearEqual(s, g_XMZero, g_XMEpsilon)) + { + // Achromatic + return XMVectorSelect(hsl, l, g_XMSelect1110); + } + else + { + XMVECTOR h = XMVectorSplatX(hsl); + + XMVECTOR q; + if (XMVector3Less(l, g_XMOneHalf)) + { + q = XMVectorMultiply(l, XMVectorAdd(g_XMOne, s)); + } + else + { + q = XMVectorSubtract(XMVectorAdd(l, s), XMVectorMultiply(l, s)); + } + + XMVECTOR p = XMVectorSubtract(XMVectorMultiply(g_XMTwo, l), q); + + XMVECTOR r = DirectX::MathInternal::XMColorHue2Clr(p, q, XMVectorAdd(h, oneThird)); + XMVECTOR g = DirectX::MathInternal::XMColorHue2Clr(p, q, h); + XMVECTOR b = DirectX::MathInternal::XMColorHue2Clr(p, q, XMVectorSubtract(h, oneThird)); + + XMVECTOR rg = XMVectorSelect(g, r, g_XMSelect1000); + XMVECTOR ba = XMVectorSelect(hsl, b, g_XMSelect1110); + + return XMVectorSelect(ba, rg, g_XMSelect1100); + } +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToHSV(FXMVECTOR rgb) noexcept +{ + XMVECTOR r = XMVectorSplatX(rgb); + XMVECTOR g = XMVectorSplatY(rgb); + XMVECTOR b = XMVectorSplatZ(rgb); + + XMVECTOR min = XMVectorMin(r, XMVectorMin(g, b)); + XMVECTOR v = XMVectorMax(r, XMVectorMax(g, b)); + + XMVECTOR d = XMVectorSubtract(v, min); + + XMVECTOR s = (XMVector3NearEqual(v, g_XMZero, g_XMEpsilon)) ? g_XMZero : XMVectorDivide(d, v); + + if (XMVector3Less(d, g_XMEpsilon)) + { + // Achromatic, assume H of 0 + XMVECTOR hv = XMVectorSelect(v, g_XMZero, g_XMSelect1000); + XMVECTOR hva = XMVectorSelect(rgb, hv, g_XMSelect1110); + return XMVectorSelect(s, hva, g_XMSelect1011); + } + else + { + XMVECTOR h; + + if (XMVector3Equal(r, v)) + { + // Red is max + h = XMVectorDivide(XMVectorSubtract(g, b), d); + + if (XMVector3Less(g, b)) + h = XMVectorAdd(h, g_XMSix); + } + else if (XMVector3Equal(g, v)) + { + // Green is max + h = XMVectorDivide(XMVectorSubtract(b, r), d); + h = XMVectorAdd(h, g_XMTwo); + } + else + { + // Blue is max + h = XMVectorDivide(XMVectorSubtract(r, g), d); + h = XMVectorAdd(h, g_XMFour); + } + + h = XMVectorDivide(h, g_XMSix); + + XMVECTOR hv = XMVectorSelect(v, h, g_XMSelect1000); + XMVECTOR hva = XMVectorSelect(rgb, hv, g_XMSelect1110); + return XMVectorSelect(s, hva, g_XMSelect1011); + } +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorHSVToRGB(FXMVECTOR hsv) noexcept +{ + XMVECTOR h = XMVectorSplatX(hsv); + XMVECTOR s = XMVectorSplatY(hsv); + XMVECTOR v = XMVectorSplatZ(hsv); + + XMVECTOR h6 = XMVectorMultiply(h, g_XMSix); + + XMVECTOR i = XMVectorFloor(h6); + XMVECTOR f = XMVectorSubtract(h6, i); + + // p = v* (1-s) + XMVECTOR p = XMVectorMultiply(v, XMVectorSubtract(g_XMOne, s)); + + // q = v*(1-f*s) + XMVECTOR q = XMVectorMultiply(v, XMVectorSubtract(g_XMOne, XMVectorMultiply(f, s))); + + // t = v*(1 - (1-f)*s) + XMVECTOR t = XMVectorMultiply(v, XMVectorSubtract(g_XMOne, XMVectorMultiply(XMVectorSubtract(g_XMOne, f), s))); + + auto ii = static_cast(XMVectorGetX(XMVectorMod(i, g_XMSix))); + + XMVECTOR _rgb; + + switch (ii) + { + case 0: // rgb = vtp + { + XMVECTOR vt = XMVectorSelect(t, v, g_XMSelect1000); + _rgb = XMVectorSelect(p, vt, g_XMSelect1100); + } + break; + case 1: // rgb = qvp + { + XMVECTOR qv = XMVectorSelect(v, q, g_XMSelect1000); + _rgb = XMVectorSelect(p, qv, g_XMSelect1100); + } + break; + case 2: // rgb = pvt + { + XMVECTOR pv = XMVectorSelect(v, p, g_XMSelect1000); + _rgb = XMVectorSelect(t, pv, g_XMSelect1100); + } + break; + case 3: // rgb = pqv + { + XMVECTOR pq = XMVectorSelect(q, p, g_XMSelect1000); + _rgb = XMVectorSelect(v, pq, g_XMSelect1100); + } + break; + case 4: // rgb = tpv + { + XMVECTOR tp = XMVectorSelect(p, t, g_XMSelect1000); + _rgb = XMVectorSelect(v, tp, g_XMSelect1100); + } + break; + default: // rgb = vpq + { + XMVECTOR vp = XMVectorSelect(p, v, g_XMSelect1000); + _rgb = XMVectorSelect(q, vp, g_XMSelect1100); + } + break; + } + + return XMVectorSelect(hsv, _rgb, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToYUV(FXMVECTOR rgb) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 0.299f, -0.147f, 0.615f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { 0.587f, -0.289f, -0.515f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 0.114f, 0.436f, -0.100f, 0.0f } } }; + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(rgb, M); + + return XMVectorSelect(rgb, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorYUVToRGB(FXMVECTOR yuv) noexcept +{ + static const XMVECTORF32 Scale1 = { { { 0.0f, -0.395f, 2.032f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 1.140f, -0.581f, 0.0f, 0.0f } } }; + + XMMATRIX M(g_XMOne, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(yuv, M); + + return XMVectorSelect(yuv, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToYUV_HD(FXMVECTOR rgb) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 0.2126f, -0.0997f, 0.6150f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { 0.7152f, -0.3354f, -0.5586f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 0.0722f, 0.4351f, -0.0564f, 0.0f } } }; + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(rgb, M); + + return XMVectorSelect(rgb, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorYUVToRGB_HD(FXMVECTOR yuv) noexcept +{ + static const XMVECTORF32 Scale1 = { { { 0.0f, -0.2153f, 2.1324f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 1.2803f, -0.3806f, 0.0f, 0.0f } } }; + + XMMATRIX M(g_XMOne, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(yuv, M); + + return XMVectorSelect(yuv, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToYUV_UHD(FXMVECTOR rgb) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 0.2627f, -0.1215f, 0.6150f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { 0.6780f, -0.3136f, -0.5655f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 0.0593f, 0.4351f, -0.0495f, 0.0f } } }; + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(rgb, M); + + return XMVectorSelect(rgb, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorYUVToRGB_UHD(FXMVECTOR yuv) noexcept +{ + static const XMVECTORF32 Scale1 = { { { 0.0f, -0.1891f, 2.1620f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 1.1989f, -0.4645f, 0.0f, 0.0f } } }; + + XMMATRIX M(g_XMOne, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(yuv, M); + + return XMVectorSelect(yuv, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToXYZ(FXMVECTOR rgb) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 0.4887180f, 0.1762044f, 0.0000000f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { 0.3106803f, 0.8129847f, 0.0102048f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 0.2006017f, 0.0108109f, 0.9897952f, 0.0f } } }; + static const XMVECTORF32 Scale = { { { 1.f / 0.17697f, 1.f / 0.17697f, 1.f / 0.17697f, 0.0f } } }; + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVectorMultiply(XMVector3Transform(rgb, M), Scale); + + return XMVectorSelect(rgb, clr, g_XMSelect1110); +} + +inline XMVECTOR XM_CALLCONV XMColorXYZToRGB(FXMVECTOR xyz) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 2.3706743f, -0.5138850f, 0.0052982f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { -0.9000405f, 1.4253036f, -0.0146949f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { -0.4706338f, 0.0885814f, 1.0093968f, 0.0f } } }; + static const XMVECTORF32 Scale = { { { 0.17697f, 0.17697f, 0.17697f, 0.0f } } }; + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(XMVectorMultiply(xyz, Scale), M); + + return XMVectorSelect(xyz, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorXYZToSRGB(FXMVECTOR xyz) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 3.2406f, -0.9689f, 0.0557f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { -1.5372f, 1.8758f, -0.2040f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { -0.4986f, 0.0415f, 1.0570f, 0.0f } } }; + static const XMVECTORF32 Cutoff = { { { 0.0031308f, 0.0031308f, 0.0031308f, 0.0f } } }; + static const XMVECTORF32 Exp = { { { 1.0f / 2.4f, 1.0f / 2.4f, 1.0f / 2.4f, 1.0f } } }; + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR lclr = XMVector3Transform(xyz, M); + + XMVECTOR sel = XMVectorGreater(lclr, Cutoff); + + // clr = 12.92 * lclr for lclr <= 0.0031308f + XMVECTOR smallC = XMVectorMultiply(lclr, g_XMsrgbScale); + + // clr = (1+a)*pow(lclr, 1/2.4) - a for lclr > 0.0031308 (where a = 0.055) + XMVECTOR largeC = XMVectorSubtract(XMVectorMultiply(g_XMsrgbA1, XMVectorPow(lclr, Exp)), g_XMsrgbA); + + XMVECTOR clr = XMVectorSelect(smallC, largeC, sel); + + return XMVectorSelect(xyz, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorSRGBToXYZ(FXMVECTOR srgb) noexcept +{ + static const XMVECTORF32 Scale0 = { { { 0.4124f, 0.2126f, 0.0193f, 0.0f } } }; + static const XMVECTORF32 Scale1 = { { { 0.3576f, 0.7152f, 0.1192f, 0.0f } } }; + static const XMVECTORF32 Scale2 = { { { 0.1805f, 0.0722f, 0.9505f, 0.0f } } }; + static const XMVECTORF32 Cutoff = { { { 0.04045f, 0.04045f, 0.04045f, 0.0f } } }; + static const XMVECTORF32 Exp = { { { 2.4f, 2.4f, 2.4f, 1.0f } } }; + + XMVECTOR sel = XMVectorGreater(srgb, Cutoff); + + // lclr = clr / 12.92 + XMVECTOR smallC = XMVectorDivide(srgb, g_XMsrgbScale); + + // lclr = pow( (clr + a) / (1+a), 2.4 ) + XMVECTOR largeC = XMVectorPow(XMVectorDivide(XMVectorAdd(srgb, g_XMsrgbA), g_XMsrgbA1), Exp); + + XMVECTOR lclr = XMVectorSelect(smallC, largeC, sel); + + XMMATRIX M(Scale0, Scale1, Scale2, g_XMZero); + XMVECTOR clr = XMVector3Transform(lclr, M); + + return XMVectorSelect(srgb, clr, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorRGBToSRGB(FXMVECTOR rgb) noexcept +{ + static const XMVECTORF32 Cutoff = { { { 0.0031308f, 0.0031308f, 0.0031308f, 1.f } } }; + static const XMVECTORF32 Linear = { { { 12.92f, 12.92f, 12.92f, 1.f } } }; + static const XMVECTORF32 Scale = { { { 1.055f, 1.055f, 1.055f, 1.f } } }; + static const XMVECTORF32 Bias = { { { 0.055f, 0.055f, 0.055f, 0.f } } }; + static const XMVECTORF32 InvGamma = { { { 1.0f / 2.4f, 1.0f / 2.4f, 1.0f / 2.4f, 1.f } } }; + + XMVECTOR V = XMVectorSaturate(rgb); + XMVECTOR V0 = XMVectorMultiply(V, Linear); + XMVECTOR V1 = XMVectorSubtract(XMVectorMultiply(Scale, XMVectorPow(V, InvGamma)), Bias); + XMVECTOR select = XMVectorLess(V, Cutoff); + V = XMVectorSelect(V1, V0, select); + return XMVectorSelect(rgb, V, g_XMSelect1110); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMColorSRGBToRGB(FXMVECTOR srgb) noexcept +{ + static const XMVECTORF32 Cutoff = { { { 0.04045f, 0.04045f, 0.04045f, 1.f } } }; + static const XMVECTORF32 ILinear = { { { 1.f / 12.92f, 1.f / 12.92f, 1.f / 12.92f, 1.f } } }; + static const XMVECTORF32 Scale = { { { 1.f / 1.055f, 1.f / 1.055f, 1.f / 1.055f, 1.f } } }; + static const XMVECTORF32 Bias = { { { 0.055f, 0.055f, 0.055f, 0.f } } }; + static const XMVECTORF32 Gamma = { { { 2.4f, 2.4f, 2.4f, 1.f } } }; + + XMVECTOR V = XMVectorSaturate(srgb); + XMVECTOR V0 = XMVectorMultiply(V, ILinear); + XMVECTOR V1 = XMVectorPow(XMVectorMultiply(XMVectorAdd(V, Bias), Scale), Gamma); + XMVECTOR select = XMVectorGreater(V, Cutoff); + V = XMVectorSelect(V0, V1, select); + return XMVectorSelect(srgb, V, g_XMSelect1110); +} + +/**************************************************************************** + * + * Miscellaneous + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline bool XMVerifyCPUSupport() noexcept +{ +#if defined(_XM_SSE_INTRINSICS_) && !defined(__powerpc64__) && !defined(_XM_NO_INTRINSICS_) + int CPUInfo[4] = { -1 }; +#if (defined(__clang__) || defined(__GNUC__)) && !defined(_MSC_VER) && !defined(__MINGW32__) + __cpuid(0, CPUInfo[0], CPUInfo[1], CPUInfo[2], CPUInfo[3]); +#else + __cpuid(CPUInfo, 0); +#endif + +#ifdef __AVX2__ + if (CPUInfo[0] < 7) + return false; +#else + if (CPUInfo[0] < 1) + return false; +#endif + +#if (defined(__clang__) || defined(__GNUC__)) && !defined(_MSC_VER) && !defined(__MINGW32__) + __cpuid(1, CPUInfo[0], CPUInfo[1], CPUInfo[2], CPUInfo[3]); +#else + __cpuid(CPUInfo, 1); +#endif + +#if defined(__AVX2__) || defined(_XM_AVX2_INTRINSICS_) + // The compiler can emit FMA3 instructions even without explicit intrinsics use + if ((CPUInfo[2] & 0x38081001) != 0x38081001) + return false; // No F16C/AVX/OSXSAVE/SSE4.1/FMA3/SSE3 support +#elif defined(_XM_FMA3_INTRINSICS_) && defined(_XM_F16C_INTRINSICS_) + if ((CPUInfo[2] & 0x38081001) != 0x38081001) + return false; // No F16C/AVX/OSXSAVE/SSE4.1/FMA3/SSE3 support +#elif defined(_XM_FMA3_INTRINSICS_) + if ((CPUInfo[2] & 0x18081001) != 0x18081001) + return false; // No AVX/OSXSAVE/SSE4.1/FMA3/SSE3 support +#elif defined(_XM_F16C_INTRINSICS_) + if ((CPUInfo[2] & 0x38080001) != 0x38080001) + return false; // No F16C/AVX/OSXSAVE/SSE4.1/SSE3 support +#elif defined(__AVX__) || defined(_XM_AVX_INTRINSICS_) + if ((CPUInfo[2] & 0x18080001) != 0x18080001) + return false; // No AVX/OSXSAVE/SSE4.1/SSE3 support +#elif defined(_XM_SSE4_INTRINSICS_) + if ((CPUInfo[2] & 0x80001) != 0x80001) + return false; // No SSE3/SSE4.1 support +#elif defined(_XM_SSE3_INTRINSICS_) + if (!(CPUInfo[2] & 0x1)) + return false; // No SSE3 support +#endif + + // The x64 processor model requires SSE2 support, but no harm in checking + if ((CPUInfo[3] & 0x6000000) != 0x6000000) + return false; // No SSE2/SSE support + +#if defined(__AVX2__) || defined(_XM_AVX2_INTRINSICS_) +#if (defined(__clang__) || defined(__GNUC__)) && !defined(_MSC_VER) && !defined(__MINGW32__) + __cpuid_count(7, 0, CPUInfo[0], CPUInfo[1], CPUInfo[2], CPUInfo[3]); +#else + __cpuidex(CPUInfo, 7, 0); +#endif + if (!(CPUInfo[1] & 0x20)) + return false; // No AVX2 support +#endif + + return true; +#elif defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + // ARM-NEON support is required for the Windows on ARM platform + return true; +#else + // No intrinsics path always supported + return true; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMFresnelTerm +( + FXMVECTOR CosIncidentAngle, + FXMVECTOR RefractionIndex +) noexcept +{ + assert(!XMVector4IsInfinite(CosIncidentAngle)); + + // Result = 0.5f * (g - c)^2 / (g + c)^2 * ((c * (g + c) - 1)^2 / (c * (g - c) + 1)^2 + 1) where + // c = CosIncidentAngle + // g = sqrt(c^2 + RefractionIndex^2 - 1) + +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + XMVECTOR G = XMVectorMultiplyAdd(RefractionIndex, RefractionIndex, g_XMNegativeOne.v); + G = XMVectorMultiplyAdd(CosIncidentAngle, CosIncidentAngle, G); + G = XMVectorAbs(G); + G = XMVectorSqrt(G); + + XMVECTOR S = XMVectorAdd(G, CosIncidentAngle); + XMVECTOR D = XMVectorSubtract(G, CosIncidentAngle); + + XMVECTOR V0 = XMVectorMultiply(D, D); + XMVECTOR V1 = XMVectorMultiply(S, S); + V1 = XMVectorReciprocal(V1); + V0 = XMVectorMultiply(g_XMOneHalf.v, V0); + V0 = XMVectorMultiply(V0, V1); + + XMVECTOR V2 = XMVectorMultiplyAdd(CosIncidentAngle, S, g_XMNegativeOne.v); + XMVECTOR V3 = XMVectorMultiplyAdd(CosIncidentAngle, D, g_XMOne.v); + V2 = XMVectorMultiply(V2, V2); + V3 = XMVectorMultiply(V3, V3); + V3 = XMVectorReciprocal(V3); + V2 = XMVectorMultiplyAdd(V2, V3, g_XMOne.v); + + XMVECTOR Result = XMVectorMultiply(V0, V2); + + Result = XMVectorSaturate(Result); + + return Result; + +#elif defined(_XM_SSE_INTRINSICS_) + // G = sqrt(abs((RefractionIndex^2-1) + CosIncidentAngle^2)) + XMVECTOR G = _mm_mul_ps(RefractionIndex, RefractionIndex); + XMVECTOR vTemp = _mm_mul_ps(CosIncidentAngle, CosIncidentAngle); + G = _mm_sub_ps(G, g_XMOne); + vTemp = _mm_add_ps(vTemp, G); + // max((0-vTemp),vTemp) == abs(vTemp) + // The abs is needed to deal with refraction and cosine being zero + G = _mm_setzero_ps(); + G = _mm_sub_ps(G, vTemp); + G = _mm_max_ps(G, vTemp); + // Last operation, the sqrt() + G = _mm_sqrt_ps(G); + + // Calc G-C and G+C + XMVECTOR GAddC = _mm_add_ps(G, CosIncidentAngle); + XMVECTOR GSubC = _mm_sub_ps(G, CosIncidentAngle); + // Perform the term (0.5f *(g - c)^2) / (g + c)^2 + XMVECTOR vResult = _mm_mul_ps(GSubC, GSubC); + vTemp = _mm_mul_ps(GAddC, GAddC); + vResult = _mm_mul_ps(vResult, g_XMOneHalf); + vResult = _mm_div_ps(vResult, vTemp); + // Perform the term ((c * (g + c) - 1)^2 / (c * (g - c) + 1)^2 + 1) + GAddC = _mm_mul_ps(GAddC, CosIncidentAngle); + GSubC = _mm_mul_ps(GSubC, CosIncidentAngle); + GAddC = _mm_sub_ps(GAddC, g_XMOne); + GSubC = _mm_add_ps(GSubC, g_XMOne); + GAddC = _mm_mul_ps(GAddC, GAddC); + GSubC = _mm_mul_ps(GSubC, GSubC); + GAddC = _mm_div_ps(GAddC, GSubC); + GAddC = _mm_add_ps(GAddC, g_XMOne); + // Multiply the two term parts + vResult = _mm_mul_ps(vResult, GAddC); + // Clamp to 0.0 - 1.0f + vResult = _mm_max_ps(vResult, g_XMZero); + vResult = _mm_min_ps(vResult, g_XMOne); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XMScalarNearEqual +( + float S1, + float S2, + float Epsilon +) noexcept +{ + float Delta = S1 - S2; + return (fabsf(Delta) <= Epsilon); +} + +//------------------------------------------------------------------------------ +// Modulo the range of the given angle such that -XM_PI <= Angle < XM_PI +inline float XMScalarModAngle(float Angle) noexcept +{ + // Note: The modulo is performed with unsigned math only to work + // around a precision error on numbers that are close to PI + + // Normalize the range from 0.0f to XM_2PI + Angle = Angle + XM_PI; + // Perform the modulo, unsigned + float fTemp = fabsf(Angle); + fTemp = fTemp - (XM_2PI * static_cast(static_cast(fTemp / XM_2PI))); + // Restore the number to the range of -XM_PI to XM_PI-epsilon + fTemp = fTemp - XM_PI; + // If the modulo'd value was negative, restore negation + if (Angle < 0.0f) + { + fTemp = -fTemp; + } + return fTemp; +} + +//------------------------------------------------------------------------------ + +inline float XMScalarSin(float Value) noexcept +{ + // Map Value to y in [-pi,pi], x = 2*pi*quotient + remainder. + float quotient = XM_1DIV2PI * Value; + if (Value >= 0.0f) + { + quotient = static_cast(static_cast(quotient + 0.5f)); + } + else + { + quotient = static_cast(static_cast(quotient - 0.5f)); + } + float y = Value - XM_2PI * quotient; + + // Map y to [-pi/2,pi/2] with sin(y) = sin(Value). + if (y > XM_PIDIV2) + { + y = XM_PI - y; + } + else if (y < -XM_PIDIV2) + { + y = -XM_PI - y; + } + + // 11-degree minimax approximation + float y2 = y * y; + return (((((-2.3889859e-08f * y2 + 2.7525562e-06f) * y2 - 0.00019840874f) * y2 + 0.0083333310f) * y2 - 0.16666667f) * y2 + 1.0f) * y; +} + +//------------------------------------------------------------------------------ + +inline float XMScalarSinEst(float Value) noexcept +{ + // Map Value to y in [-pi,pi], x = 2*pi*quotient + remainder. + float quotient = XM_1DIV2PI * Value; + if (Value >= 0.0f) + { + quotient = static_cast(static_cast(quotient + 0.5f)); + } + else + { + quotient = static_cast(static_cast(quotient - 0.5f)); + } + float y = Value - XM_2PI * quotient; + + // Map y to [-pi/2,pi/2] with sin(y) = sin(Value). + if (y > XM_PIDIV2) + { + y = XM_PI - y; + } + else if (y < -XM_PIDIV2) + { + y = -XM_PI - y; + } + + // 7-degree minimax approximation + float y2 = y * y; + return (((-0.00018524670f * y2 + 0.0083139502f) * y2 - 0.16665852f) * y2 + 1.0f) * y; +} + +//------------------------------------------------------------------------------ + +inline float XMScalarCos(float Value) noexcept +{ + // Map Value to y in [-pi,pi], x = 2*pi*quotient + remainder. + float quotient = XM_1DIV2PI * Value; + if (Value >= 0.0f) + { + quotient = static_cast(static_cast(quotient + 0.5f)); + } + else + { + quotient = static_cast(static_cast(quotient - 0.5f)); + } + float y = Value - XM_2PI * quotient; + + // Map y to [-pi/2,pi/2] with cos(y) = sign*cos(x). + float sign; + if (y > XM_PIDIV2) + { + y = XM_PI - y; + sign = -1.0f; + } + else if (y < -XM_PIDIV2) + { + y = -XM_PI - y; + sign = -1.0f; + } + else + { + sign = +1.0f; + } + + // 10-degree minimax approximation + float y2 = y * y; + float p = ((((-2.6051615e-07f * y2 + 2.4760495e-05f) * y2 - 0.0013888378f) * y2 + 0.041666638f) * y2 - 0.5f) * y2 + 1.0f; + return sign * p; +} + +//------------------------------------------------------------------------------ + +inline float XMScalarCosEst(float Value) noexcept +{ + // Map Value to y in [-pi,pi], x = 2*pi*quotient + remainder. + float quotient = XM_1DIV2PI * Value; + if (Value >= 0.0f) + { + quotient = static_cast(static_cast(quotient + 0.5f)); + } + else + { + quotient = static_cast(static_cast(quotient - 0.5f)); + } + float y = Value - XM_2PI * quotient; + + // Map y to [-pi/2,pi/2] with cos(y) = sign*cos(x). + float sign; + if (y > XM_PIDIV2) + { + y = XM_PI - y; + sign = -1.0f; + } + else if (y < -XM_PIDIV2) + { + y = -XM_PI - y; + sign = -1.0f; + } + else + { + sign = +1.0f; + } + + // 6-degree minimax approximation + float y2 = y * y; + float p = ((-0.0012712436f * y2 + 0.041493919f) * y2 - 0.49992746f) * y2 + 1.0f; + return sign * p; +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline void XMScalarSinCos +( + float* pSin, + float* pCos, + float Value +) noexcept +{ + assert(pSin); + assert(pCos); + + // Map Value to y in [-pi,pi], x = 2*pi*quotient + remainder. + float quotient = XM_1DIV2PI * Value; + if (Value >= 0.0f) + { + quotient = static_cast(static_cast(quotient + 0.5f)); + } + else + { + quotient = static_cast(static_cast(quotient - 0.5f)); + } + float y = Value - XM_2PI * quotient; + + // Map y to [-pi/2,pi/2] with sin(y) = sin(Value). + float sign; + if (y > XM_PIDIV2) + { + y = XM_PI - y; + sign = -1.0f; + } + else if (y < -XM_PIDIV2) + { + y = -XM_PI - y; + sign = -1.0f; + } + else + { + sign = +1.0f; + } + + float y2 = y * y; + + // 11-degree minimax approximation + *pSin = (((((-2.3889859e-08f * y2 + 2.7525562e-06f) * y2 - 0.00019840874f) * y2 + 0.0083333310f) * y2 - 0.16666667f) * y2 + 1.0f) * y; + + // 10-degree minimax approximation + float p = ((((-2.6051615e-07f * y2 + 2.4760495e-05f) * y2 - 0.0013888378f) * y2 + 0.041666638f) * y2 - 0.5f) * y2 + 1.0f; + *pCos = sign * p; +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline void XMScalarSinCosEst +( + float* pSin, + float* pCos, + float Value +) noexcept +{ + assert(pSin); + assert(pCos); + + // Map Value to y in [-pi,pi], x = 2*pi*quotient + remainder. + float quotient = XM_1DIV2PI * Value; + if (Value >= 0.0f) + { + quotient = static_cast(static_cast(quotient + 0.5f)); + } + else + { + quotient = static_cast(static_cast(quotient - 0.5f)); + } + float y = Value - XM_2PI * quotient; + + // Map y to [-pi/2,pi/2] with sin(y) = sin(Value). + float sign; + if (y > XM_PIDIV2) + { + y = XM_PI - y; + sign = -1.0f; + } + else if (y < -XM_PIDIV2) + { + y = -XM_PI - y; + sign = -1.0f; + } + else + { + sign = +1.0f; + } + + float y2 = y * y; + + // 7-degree minimax approximation + *pSin = (((-0.00018524670f * y2 + 0.0083139502f) * y2 - 0.16665852f) * y2 + 1.0f) * y; + + // 6-degree minimax approximation + float p = ((-0.0012712436f * y2 + 0.041493919f) * y2 - 0.49992746f) * y2 + 1.0f; + *pCos = sign * p; +} + +//------------------------------------------------------------------------------ + +inline float XMScalarASin(float Value) noexcept +{ + // Clamp input to [-1,1]. + bool nonnegative = (Value >= 0.0f); + float x = fabsf(Value); + float omx = 1.0f - x; + if (omx < 0.0f) + { + omx = 0.0f; + } + float root = sqrtf(omx); + + // 7-degree minimax approximation + float result = ((((((-0.0012624911f * x + 0.0066700901f) * x - 0.0170881256f) * x + 0.0308918810f) * x - 0.0501743046f) * x + 0.0889789874f) * x - 0.2145988016f) * x + 1.5707963050f; + result *= root; // acos(|x|) + + // acos(x) = pi - acos(-x) when x < 0, asin(x) = pi/2 - acos(x) + return (nonnegative ? XM_PIDIV2 - result : result - XM_PIDIV2); +} + +//------------------------------------------------------------------------------ + +inline float XMScalarASinEst(float Value) noexcept +{ + // Clamp input to [-1,1]. + bool nonnegative = (Value >= 0.0f); + float x = fabsf(Value); + float omx = 1.0f - x; + if (omx < 0.0f) + { + omx = 0.0f; + } + float root = sqrtf(omx); + + // 3-degree minimax approximation + float result = ((-0.0187293f * x + 0.0742610f) * x - 0.2121144f) * x + 1.5707288f; + result *= root; // acos(|x|) + + // acos(x) = pi - acos(-x) when x < 0, asin(x) = pi/2 - acos(x) + return (nonnegative ? XM_PIDIV2 - result : result - XM_PIDIV2); +} + +//------------------------------------------------------------------------------ + +inline float XMScalarACos(float Value) noexcept +{ + // Clamp input to [-1,1]. + bool nonnegative = (Value >= 0.0f); + float x = fabsf(Value); + float omx = 1.0f - x; + if (omx < 0.0f) + { + omx = 0.0f; + } + float root = sqrtf(omx); + + // 7-degree minimax approximation + float result = ((((((-0.0012624911f * x + 0.0066700901f) * x - 0.0170881256f) * x + 0.0308918810f) * x - 0.0501743046f) * x + 0.0889789874f) * x - 0.2145988016f) * x + 1.5707963050f; + result *= root; + + // acos(x) = pi - acos(-x) when x < 0 + return (nonnegative ? result : XM_PI - result); +} + +//------------------------------------------------------------------------------ + +inline float XMScalarACosEst(float Value) noexcept +{ + // Clamp input to [-1,1]. + bool nonnegative = (Value >= 0.0f); + float x = fabsf(Value); + float omx = 1.0f - x; + if (omx < 0.0f) + { + omx = 0.0f; + } + float root = sqrtf(omx); + + // 3-degree minimax approximation + float result = ((-0.0187293f * x + 0.0742610f) * x - 0.2121144f) * x + 1.5707288f; + result *= root; + + // acos(x) = pi - acos(-x) when x < 0 + return (nonnegative ? result : XM_PI - result); +} + diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathVector.inl b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathVector.inl new file mode 100644 index 00000000..20bbab1d --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXMathVector.inl @@ -0,0 +1,14883 @@ +//------------------------------------------------------------------------------------- +// DirectXMathVector.inl -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +#if defined(_XM_NO_INTRINSICS_) +#define XMISNAN(x) isnan(x) +#define XMISINF(x) isinf(x) +#endif + +#if defined(_XM_SSE_INTRINSICS_) + +#define XM3UNPACK3INTO4(l1, l2, l3) \ + XMVECTOR V3 = _mm_shuffle_ps(l2, l3, _MM_SHUFFLE(0, 0, 3, 2));\ + XMVECTOR V2 = _mm_shuffle_ps(l2, l1, _MM_SHUFFLE(3, 3, 1, 0));\ + V2 = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 0, 2));\ + XMVECTOR V4 = _mm_castsi128_ps(_mm_srli_si128(_mm_castps_si128(L3), 32 / 8)) + +#define XM3PACK4INTO3(v2x) \ + v2x = _mm_shuffle_ps(V2, V3, _MM_SHUFFLE(1, 0, 2, 1));\ + V2 = _mm_shuffle_ps(V2, V1, _MM_SHUFFLE(2, 2, 0, 0));\ + V1 = _mm_shuffle_ps(V1, V2, _MM_SHUFFLE(0, 2, 1, 0));\ + V3 = _mm_shuffle_ps(V3, V4, _MM_SHUFFLE(0, 0, 2, 2));\ + V3 = _mm_shuffle_ps(V3, V4, _MM_SHUFFLE(2, 1, 2, 0)) + +#endif + +/**************************************************************************** + * + * General Vector + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Assignment operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ +// Return a vector with all elements equaling zero +inline XMVECTOR XM_CALLCONV XMVectorZero() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { 0.0f, 0.0f, 0.0f, 0.0f } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_n_f32(0); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_setzero_ps(); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with four floating point values +inline XMVECTOR XM_CALLCONV XMVectorSet +( + float x, + float y, + float z, + float w +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { x, y, z, w } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t V0 = vcreate_f32( + static_cast(*reinterpret_cast(&x)) + | (static_cast(*reinterpret_cast(&y)) << 32)); + float32x2_t V1 = vcreate_f32( + static_cast(*reinterpret_cast(&z)) + | (static_cast(*reinterpret_cast(&w)) << 32)); + return vcombine_f32(V0, V1); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_set_ps(w, z, y, x); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with four integer values +inline XMVECTOR XM_CALLCONV XMVectorSetInt +( + uint32_t x, + uint32_t y, + uint32_t z, + uint32_t w +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult = { { { x, y, z, w } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t V0 = vcreate_u32(static_cast(x) | (static_cast(y) << 32)); + uint32x2_t V1 = vcreate_u32(static_cast(z) | (static_cast(w) << 32)); + return vreinterpretq_f32_u32(vcombine_u32(V0, V1)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_set_epi32(static_cast(w), static_cast(z), static_cast(y), static_cast(x)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with a replicated floating point value +inline XMVECTOR XM_CALLCONV XMVectorReplicate(float Value) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = Value; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_n_f32(Value); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_set_ps1(Value); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with a replicated floating point value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorReplicatePtr(const float* pValue) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float Value = pValue[0]; + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = Value; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vld1q_dup_f32(pValue); +#elif defined(_XM_AVX_INTRINSICS_) + return _mm_broadcast_ss(pValue); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_load_ps1(pValue); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with a replicated integer value +inline XMVECTOR XM_CALLCONV XMVectorReplicateInt(uint32_t Value) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = + vResult.u[1] = + vResult.u[2] = + vResult.u[3] = Value; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vdupq_n_u32(Value)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_set1_epi32(static_cast(Value)); + return _mm_castsi128_ps(vTemp); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with a replicated integer value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorReplicateIntPtr(const uint32_t* pValue) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t Value = pValue[0]; + XMVECTORU32 vResult; + vResult.u[0] = + vResult.u[1] = + vResult.u[2] = + vResult.u[3] = Value; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vld1q_dup_u32(pValue)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_load_ps1(reinterpret_cast(pValue)); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with all bits set (true mask) +inline XMVECTOR XM_CALLCONV XMVectorTrueInt() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult = { { { 0xFFFFFFFFU, 0xFFFFFFFFU, 0xFFFFFFFFU, 0xFFFFFFFFU } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_s32(vdupq_n_s32(-1)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_set1_epi32(-1); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +// Initialize a vector with all bits clear (false mask) +inline XMVECTOR XM_CALLCONV XMVectorFalseInt() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { 0.0f, 0.0f, 0.0f, 0.0f } } }; + return vResult; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vdupq_n_u32(0)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_setzero_ps(); +#endif +} + +//------------------------------------------------------------------------------ +// Replicate the x component of the vector +inline XMVECTOR XM_CALLCONV XMVectorSplatX(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = V.vector4_f32[0]; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_lane_f32(vget_low_f32(V), 0); +#elif defined(_XM_AVX2_INTRINSICS_) && defined(_XM_FAVOR_INTEL_) + return _mm_broadcastss_ps(V); +#elif defined(_XM_SSE_INTRINSICS_) + return XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); +#endif +} + +//------------------------------------------------------------------------------ +// Replicate the y component of the vector +inline XMVECTOR XM_CALLCONV XMVectorSplatY(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = V.vector4_f32[1]; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_lane_f32(vget_low_f32(V), 1); +#elif defined(_XM_SSE_INTRINSICS_) + return XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); +#endif +} + +//------------------------------------------------------------------------------ +// Replicate the z component of the vector +inline XMVECTOR XM_CALLCONV XMVectorSplatZ(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = V.vector4_f32[2]; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_lane_f32(vget_high_f32(V), 0); +#elif defined(_XM_SSE_INTRINSICS_) + return XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); +#endif +} + +//------------------------------------------------------------------------------ +// Replicate the w component of the vector +inline XMVECTOR XM_CALLCONV XMVectorSplatW(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = V.vector4_f32[3]; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_lane_f32(vget_high_f32(V), 1); +#elif defined(_XM_SSE_INTRINSICS_) + return XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); +#endif +} + +//------------------------------------------------------------------------------ +// Return a vector of 1.0f,1.0f,1.0f,1.0f +inline XMVECTOR XM_CALLCONV XMVectorSplatOne() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = 1.0f; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vdupq_n_f32(1.0f); +#elif defined(_XM_SSE_INTRINSICS_) + return g_XMOne; +#endif +} + +//------------------------------------------------------------------------------ +// Return a vector of INF,INF,INF,INF +inline XMVECTOR XM_CALLCONV XMVectorSplatInfinity() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = + vResult.u[1] = + vResult.u[2] = + vResult.u[3] = 0x7F800000; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vdupq_n_u32(0x7F800000)); +#elif defined(_XM_SSE_INTRINSICS_) + return g_XMInfinity; +#endif +} + +//------------------------------------------------------------------------------ +// Return a vector of Q_NAN,Q_NAN,Q_NAN,Q_NAN +inline XMVECTOR XM_CALLCONV XMVectorSplatQNaN() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = + vResult.u[1] = + vResult.u[2] = + vResult.u[3] = 0x7FC00000; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vdupq_n_u32(0x7FC00000)); +#elif defined(_XM_SSE_INTRINSICS_) + return g_XMQNaN; +#endif +} + +//------------------------------------------------------------------------------ +// Return a vector of 1.192092896e-7f,1.192092896e-7f,1.192092896e-7f,1.192092896e-7f +inline XMVECTOR XM_CALLCONV XMVectorSplatEpsilon() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = + vResult.u[1] = + vResult.u[2] = + vResult.u[3] = 0x34000000; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vdupq_n_u32(0x34000000)); +#elif defined(_XM_SSE_INTRINSICS_) + return g_XMEpsilon; +#endif +} + +//------------------------------------------------------------------------------ +// Return a vector of -0.0f (0x80000000),-0.0f,-0.0f,-0.0f +inline XMVECTOR XM_CALLCONV XMVectorSplatSignMask() noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 vResult; + vResult.u[0] = + vResult.u[1] = + vResult.u[2] = + vResult.u[3] = 0x80000000U; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vdupq_n_u32(0x80000000U)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_set1_epi32(static_cast(0x80000000)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +// Return a floating point value via an index. This is not a recommended +// function to use due to performance loss. +inline float XM_CALLCONV XMVectorGetByIndex(FXMVECTOR V, size_t i) noexcept +{ + assert(i < 4); + _Analysis_assume_(i < 4); +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_f32[i]; +#else + XMVECTORF32 U; + U.v = V; + return U.f[i]; +#endif +} + +//------------------------------------------------------------------------------ +// Return the X component in an FPU register. +inline float XM_CALLCONV XMVectorGetX(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_f32[0]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_f32(V, 0); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cvtss_f32(V); +#endif +} + +// Return the Y component in an FPU register. +inline float XM_CALLCONV XMVectorGetY(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_f32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_f32(V, 1); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + return _mm_cvtss_f32(vTemp); +#endif +} + +// Return the Z component in an FPU register. +inline float XM_CALLCONV XMVectorGetZ(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_f32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_f32(V, 2); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + return _mm_cvtss_f32(vTemp); +#endif +} + +// Return the W component in an FPU register. +inline float XM_CALLCONV XMVectorGetW(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_f32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_f32(V, 3); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + return _mm_cvtss_f32(vTemp); +#endif +} + +//------------------------------------------------------------------------------ + +// Store a component indexed by i into a 32 bit float location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetByIndexPtr(float* f, FXMVECTOR V, size_t i) noexcept +{ + assert(f != nullptr); + assert(i < 4); + _Analysis_assume_(i < 4); +#if defined(_XM_NO_INTRINSICS_) + *f = V.vector4_f32[i]; +#else + XMVECTORF32 U; + U.v = V; + *f = U.f[i]; +#endif +} + +//------------------------------------------------------------------------------ + +// Store the X component into a 32 bit float location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetXPtr(float* x, FXMVECTOR V) noexcept +{ + assert(x != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *x = V.vector4_f32[0]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_f32(x, V, 0); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_ss(x, V); +#endif +} + +// Store the Y component into a 32 bit float location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetYPtr(float* y, FXMVECTOR V) noexcept +{ + assert(y != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *y = V.vector4_f32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_f32(y, V, 1); +#elif defined(_XM_SSE4_INTRINSICS_) + * (reinterpret_cast(y)) = _mm_extract_ps(V, 1); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + _mm_store_ss(y, vResult); +#endif +} + +// Store the Z component into a 32 bit float location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetZPtr(float* z, FXMVECTOR V) noexcept +{ + assert(z != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *z = V.vector4_f32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_f32(z, V, 2); +#elif defined(_XM_SSE4_INTRINSICS_) + * (reinterpret_cast(z)) = _mm_extract_ps(V, 2); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(z, vResult); +#endif +} + +// Store the W component into a 32 bit float location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetWPtr(float* w, FXMVECTOR V) noexcept +{ + assert(w != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *w = V.vector4_f32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_f32(w, V, 3); +#elif defined(_XM_SSE4_INTRINSICS_) + * (reinterpret_cast(w)) = _mm_extract_ps(V, 3); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + _mm_store_ss(w, vResult); +#endif +} + +//------------------------------------------------------------------------------ + +// Return an integer value via an index. This is not a recommended +// function to use due to performance loss. +inline uint32_t XM_CALLCONV XMVectorGetIntByIndex(FXMVECTOR V, size_t i) noexcept +{ + assert(i < 4); + _Analysis_assume_(i < 4); +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_u32[i]; +#else + XMVECTORU32 U; + U.v = V; + return U.u[i]; +#endif +} + +//------------------------------------------------------------------------------ + +// Return the X component in an integer register. +inline uint32_t XM_CALLCONV XMVectorGetIntX(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_u32[0]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_u32(vreinterpretq_u32_f32(V), 0); +#elif defined(_XM_SSE_INTRINSICS_) + return static_cast(_mm_cvtsi128_si32(_mm_castps_si128(V))); +#endif +} + +// Return the Y component in an integer register. +inline uint32_t XM_CALLCONV XMVectorGetIntY(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_u32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_u32(vreinterpretq_u32_f32(V), 1); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i V1 = _mm_castps_si128(V); + return static_cast(_mm_extract_epi32(V1, 1)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vResulti = _mm_shuffle_epi32(_mm_castps_si128(V), _MM_SHUFFLE(1, 1, 1, 1)); + return static_cast(_mm_cvtsi128_si32(vResulti)); +#endif +} + +// Return the Z component in an integer register. +inline uint32_t XM_CALLCONV XMVectorGetIntZ(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_u32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_u32(vreinterpretq_u32_f32(V), 2); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i V1 = _mm_castps_si128(V); + return static_cast(_mm_extract_epi32(V1, 2)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vResulti = _mm_shuffle_epi32(_mm_castps_si128(V), _MM_SHUFFLE(2, 2, 2, 2)); + return static_cast(_mm_cvtsi128_si32(vResulti)); +#endif +} + +// Return the W component in an integer register. +inline uint32_t XM_CALLCONV XMVectorGetIntW(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return V.vector4_u32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vgetq_lane_u32(vreinterpretq_u32_f32(V), 3); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i V1 = _mm_castps_si128(V); + return static_cast(_mm_extract_epi32(V1, 3)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vResulti = _mm_shuffle_epi32(_mm_castps_si128(V), _MM_SHUFFLE(3, 3, 3, 3)); + return static_cast(_mm_cvtsi128_si32(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ + +// Store a component indexed by i into a 32 bit integer location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetIntByIndexPtr(uint32_t* x, FXMVECTOR V, size_t i) noexcept +{ + assert(x != nullptr); + assert(i < 4); + _Analysis_assume_(i < 4); +#if defined(_XM_NO_INTRINSICS_) + *x = V.vector4_u32[i]; +#else + XMVECTORU32 U; + U.v = V; + *x = U.u[i]; +#endif +} + +//------------------------------------------------------------------------------ + +// Store the X component into a 32 bit integer location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetIntXPtr(uint32_t* x, FXMVECTOR V) noexcept +{ + assert(x != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *x = V.vector4_u32[0]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_u32(x, *reinterpret_cast(&V), 0); +#elif defined(_XM_SSE_INTRINSICS_) + _mm_store_ss(reinterpret_cast(x), V); +#endif +} + +// Store the Y component into a 32 bit integer location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetIntYPtr(uint32_t* y, FXMVECTOR V) noexcept +{ + assert(y != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *y = V.vector4_u32[1]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_u32(y, *reinterpret_cast(&V), 1); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i V1 = _mm_castps_si128(V); + *y = static_cast(_mm_extract_epi32(V1, 1)); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + _mm_store_ss(reinterpret_cast(y), vResult); +#endif +} + +// Store the Z component into a 32 bit integer locaCantion in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetIntZPtr(uint32_t* z, FXMVECTOR V) noexcept +{ + assert(z != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *z = V.vector4_u32[2]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_u32(z, *reinterpret_cast(&V), 2); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i V1 = _mm_castps_si128(V); + *z = static_cast(_mm_extract_epi32(V1, 2)); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + _mm_store_ss(reinterpret_cast(z), vResult); +#endif +} + +// Store the W component into a 32 bit integer location in memory. +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorGetIntWPtr(uint32_t* w, FXMVECTOR V) noexcept +{ + assert(w != nullptr); +#if defined(_XM_NO_INTRINSICS_) + *w = V.vector4_u32[3]; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + vst1q_lane_u32(w, *reinterpret_cast(&V), 3); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i V1 = _mm_castps_si128(V); + *w = static_cast(_mm_extract_epi32(V1, 3)); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + _mm_store_ss(reinterpret_cast(w), vResult); +#endif +} + +//------------------------------------------------------------------------------ + +// Set a single indexed floating point component +inline XMVECTOR XM_CALLCONV XMVectorSetByIndex(FXMVECTOR V, float f, size_t i) noexcept +{ + assert(i < 4); + _Analysis_assume_(i < 4); + XMVECTORF32 U; + U.v = V; + U.f[i] = f; + return U.v; +} + +//------------------------------------------------------------------------------ + +// Sets the X component of a vector to a passed floating point value +inline XMVECTOR XM_CALLCONV XMVectorSetX(FXMVECTOR V, float x) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + x, + V.vector4_f32[1], + V.vector4_f32[2], + V.vector4_f32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vsetq_lane_f32(x, V, 0); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_set_ss(x); + vResult = _mm_move_ss(V, vResult); + return vResult; +#endif +} + +// Sets the Y component of a vector to a passed floating point value +inline XMVECTOR XM_CALLCONV XMVectorSetY(FXMVECTOR V, float y) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + V.vector4_f32[0], + y, + V.vector4_f32[2], + V.vector4_f32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vsetq_lane_f32(y, V, 1); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vResult = _mm_set_ss(y); + vResult = _mm_insert_ps(V, vResult, 0x10); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Swap y and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 2, 0, 1)); + // Convert input to vector + XMVECTOR vTemp = _mm_set_ss(y); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap y and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 2, 0, 1)); + return vResult; +#endif +} +// Sets the Z component of a vector to a passed floating point value +inline XMVECTOR XM_CALLCONV XMVectorSetZ(FXMVECTOR V, float z) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + V.vector4_f32[0], + V.vector4_f32[1], + z, + V.vector4_f32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vsetq_lane_f32(z, V, 2); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vResult = _mm_set_ss(z); + vResult = _mm_insert_ps(V, vResult, 0x20); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Swap z and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 0, 1, 2)); + // Convert input to vector + XMVECTOR vTemp = _mm_set_ss(z); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap z and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 0, 1, 2)); + return vResult; +#endif +} + +// Sets the W component of a vector to a passed floating point value +inline XMVECTOR XM_CALLCONV XMVectorSetW(FXMVECTOR V, float w) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + V.vector4_f32[0], + V.vector4_f32[1], + V.vector4_f32[2], + w + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vsetq_lane_f32(w, V, 3); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vResult = _mm_set_ss(w); + vResult = _mm_insert_ps(V, vResult, 0x30); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Swap w and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 2, 1, 3)); + // Convert input to vector + XMVECTOR vTemp = _mm_set_ss(w); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap w and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 2, 1, 3)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +// Sets a component of a vector to a floating point value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetByIndexPtr(FXMVECTOR V, const float* f, size_t i) noexcept +{ + assert(f != nullptr); + assert(i < 4); + _Analysis_assume_(i < 4); + XMVECTORF32 U; + U.v = V; + U.f[i] = *f; + return U.v; +} + +//------------------------------------------------------------------------------ + +// Sets the X component of a vector to a floating point value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetXPtr(FXMVECTOR V, const float* x) noexcept +{ + assert(x != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + *x, + V.vector4_f32[1], + V.vector4_f32[2], + V.vector4_f32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vld1q_lane_f32(x, V, 0); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_load_ss(x); + vResult = _mm_move_ss(V, vResult); + return vResult; +#endif +} + +// Sets the Y component of a vector to a floating point value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetYPtr(FXMVECTOR V, const float* y) noexcept +{ + assert(y != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + V.vector4_f32[0], + *y, + V.vector4_f32[2], + V.vector4_f32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vld1q_lane_f32(y, V, 1); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap y and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 2, 0, 1)); + // Convert input to vector + XMVECTOR vTemp = _mm_load_ss(y); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap y and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 2, 0, 1)); + return vResult; +#endif +} + +// Sets the Z component of a vector to a floating point value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetZPtr(FXMVECTOR V, const float* z) noexcept +{ + assert(z != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + V.vector4_f32[0], + V.vector4_f32[1], + *z, + V.vector4_f32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vld1q_lane_f32(z, V, 2); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap z and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 0, 1, 2)); + // Convert input to vector + XMVECTOR vTemp = _mm_load_ss(z); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap z and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 0, 1, 2)); + return vResult; +#endif +} + +// Sets the W component of a vector to a floating point value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetWPtr(FXMVECTOR V, const float* w) noexcept +{ + assert(w != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 U = { { { + V.vector4_f32[0], + V.vector4_f32[1], + V.vector4_f32[2], + *w + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vld1q_lane_f32(w, V, 3); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap w and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 2, 1, 3)); + // Convert input to vector + XMVECTOR vTemp = _mm_load_ss(w); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap w and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 2, 1, 3)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +// Sets a component of a vector to an integer passed by value +inline XMVECTOR XM_CALLCONV XMVectorSetIntByIndex(FXMVECTOR V, uint32_t x, size_t i) noexcept +{ + assert(i < 4); + _Analysis_assume_(i < 4); + XMVECTORU32 tmp; + tmp.v = V; + tmp.u[i] = x; + return tmp; +} + +//------------------------------------------------------------------------------ + +// Sets the X component of a vector to an integer passed by value +inline XMVECTOR XM_CALLCONV XMVectorSetIntX(FXMVECTOR V, uint32_t x) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + x, + V.vector4_u32[1], + V.vector4_u32[2], + V.vector4_u32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vsetq_lane_u32(x, vreinterpretq_u32_f32(V), 0)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cvtsi32_si128(static_cast(x)); + XMVECTOR vResult = _mm_move_ss(V, _mm_castsi128_ps(vTemp)); + return vResult; +#endif +} + +// Sets the Y component of a vector to an integer passed by value +inline XMVECTOR XM_CALLCONV XMVectorSetIntY(FXMVECTOR V, uint32_t y) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + V.vector4_u32[0], + y, + V.vector4_u32[2], + V.vector4_u32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vsetq_lane_u32(y, vreinterpretq_u32_f32(V), 1)); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i vResult = _mm_castps_si128(V); + vResult = _mm_insert_epi32(vResult, static_cast(y), 1); + return _mm_castsi128_ps(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap y and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 2, 0, 1)); + // Convert input to vector + __m128i vTemp = _mm_cvtsi32_si128(static_cast(y)); + // Replace the x component + vResult = _mm_move_ss(vResult, _mm_castsi128_ps(vTemp)); + // Swap y and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 2, 0, 1)); + return vResult; +#endif +} + +// Sets the Z component of a vector to an integer passed by value +inline XMVECTOR XM_CALLCONV XMVectorSetIntZ(FXMVECTOR V, uint32_t z) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + V.vector4_u32[0], + V.vector4_u32[1], + z, + V.vector4_u32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vsetq_lane_u32(z, vreinterpretq_u32_f32(V), 2)); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i vResult = _mm_castps_si128(V); + vResult = _mm_insert_epi32(vResult, static_cast(z), 2); + return _mm_castsi128_ps(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap z and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 0, 1, 2)); + // Convert input to vector + __m128i vTemp = _mm_cvtsi32_si128(static_cast(z)); + // Replace the x component + vResult = _mm_move_ss(vResult, _mm_castsi128_ps(vTemp)); + // Swap z and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 0, 1, 2)); + return vResult; +#endif +} + +// Sets the W component of a vector to an integer passed by value +inline XMVECTOR XM_CALLCONV XMVectorSetIntW(FXMVECTOR V, uint32_t w) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + V.vector4_u32[0], + V.vector4_u32[1], + V.vector4_u32[2], + w + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vsetq_lane_u32(w, vreinterpretq_u32_f32(V), 3)); +#elif defined(_XM_SSE4_INTRINSICS_) + __m128i vResult = _mm_castps_si128(V); + vResult = _mm_insert_epi32(vResult, static_cast(w), 3); + return _mm_castsi128_ps(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap w and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 2, 1, 3)); + // Convert input to vector + __m128i vTemp = _mm_cvtsi32_si128(static_cast(w)); + // Replace the x component + vResult = _mm_move_ss(vResult, _mm_castsi128_ps(vTemp)); + // Swap w and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 2, 1, 3)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +// Sets a component of a vector to an integer value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetIntByIndexPtr(FXMVECTOR V, const uint32_t* x, size_t i) noexcept +{ + assert(x != nullptr); + assert(i < 4); + _Analysis_assume_(i < 4); + XMVECTORU32 tmp; + tmp.v = V; + tmp.u[i] = *x; + return tmp; +} + +//------------------------------------------------------------------------------ + +// Sets the X component of a vector to an integer value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetIntXPtr(FXMVECTOR V, const uint32_t* x) noexcept +{ + assert(x != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + *x, + V.vector4_u32[1], + V.vector4_u32[2], + V.vector4_u32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vld1q_lane_u32(x, *reinterpret_cast(&V), 0)); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_load_ss(reinterpret_cast(x)); + XMVECTOR vResult = _mm_move_ss(V, vTemp); + return vResult; +#endif +} + +// Sets the Y component of a vector to an integer value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetIntYPtr(FXMVECTOR V, const uint32_t* y) noexcept +{ + assert(y != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + V.vector4_u32[0], + *y, + V.vector4_u32[2], + V.vector4_u32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vld1q_lane_u32(y, *reinterpret_cast(&V), 1)); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap y and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 2, 0, 1)); + // Convert input to vector + XMVECTOR vTemp = _mm_load_ss(reinterpret_cast(y)); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap y and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 2, 0, 1)); + return vResult; +#endif +} + +// Sets the Z component of a vector to an integer value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetIntZPtr(FXMVECTOR V, const uint32_t* z) noexcept +{ + assert(z != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + V.vector4_u32[0], + V.vector4_u32[1], + *z, + V.vector4_u32[3] + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vld1q_lane_u32(z, *reinterpret_cast(&V), 2)); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap z and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 0, 1, 2)); + // Convert input to vector + XMVECTOR vTemp = _mm_load_ss(reinterpret_cast(z)); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap z and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 0, 1, 2)); + return vResult; +#endif +} + +// Sets the W component of a vector to an integer value passed by pointer +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorSetIntWPtr(FXMVECTOR V, const uint32_t* w) noexcept +{ + assert(w != nullptr); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORU32 U = { { { + V.vector4_u32[0], + V.vector4_u32[1], + V.vector4_u32[2], + *w + } } }; + return U.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vld1q_lane_u32(w, *reinterpret_cast(&V), 3)); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap w and x + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 2, 1, 3)); + // Convert input to vector + XMVECTOR vTemp = _mm_load_ss(reinterpret_cast(w)); + // Replace the x component + vResult = _mm_move_ss(vResult, vTemp); + // Swap w and x again + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 2, 1, 3)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSwizzle +( + FXMVECTOR V, + uint32_t E0, + uint32_t E1, + uint32_t E2, + uint32_t E3 +) noexcept +{ + assert((E0 < 4) && (E1 < 4) && (E2 < 4) && (E3 < 4)); + _Analysis_assume_((E0 < 4) && (E1 < 4) && (E2 < 4) && (E3 < 4)); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + V.vector4_f32[E0], + V.vector4_f32[E1], + V.vector4_f32[E2], + V.vector4_f32[E3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const uint32_t ControlElement[4] = + { + 0x03020100, // XM_SWIZZLE_X + 0x07060504, // XM_SWIZZLE_Y + 0x0B0A0908, // XM_SWIZZLE_Z + 0x0F0E0D0C, // XM_SWIZZLE_W + }; + + uint8x8x2_t tbl; + tbl.val[0] = vreinterpret_u8_f32(vget_low_f32(V)); + tbl.val[1] = vreinterpret_u8_f32(vget_high_f32(V)); + + uint32x2_t idx = vcreate_u32(static_cast(ControlElement[E0]) | (static_cast(ControlElement[E1]) << 32)); + const uint8x8_t rL = vtbl2_u8(tbl, vreinterpret_u8_u32(idx)); + + idx = vcreate_u32(static_cast(ControlElement[E2]) | (static_cast(ControlElement[E3]) << 32)); + const uint8x8_t rH = vtbl2_u8(tbl, vreinterpret_u8_u32(idx)); + + return vcombine_f32(vreinterpret_f32_u8(rL), vreinterpret_f32_u8(rH)); +#elif defined(_XM_AVX_INTRINSICS_) + unsigned int elem[4] = { E0, E1, E2, E3 }; + __m128i vControl = _mm_loadu_si128(reinterpret_cast(&elem[0])); + return _mm_permutevar_ps(V, vControl); +#else +#if defined(__GNUC__) && !defined(__clang__) + // workaround some GCC optimization behavior that breaks this function + XMVECTORU32 T; + T.v = V; + auto aPtr = reinterpret_cast(&T); +#else + auto aPtr = reinterpret_cast(&V); +#endif + + XMVECTORU32 vResult; + vResult.u[0] = aPtr[E0]; + vResult.u[1] = aPtr[E1]; + vResult.u[2] = aPtr[E2]; + vResult.u[3] = aPtr[E3]; + + return vResult.v; +#endif +} + +//------------------------------------------------------------------------------ +inline XMVECTOR XM_CALLCONV XMVectorPermute +( + FXMVECTOR V1, + FXMVECTOR V2, + uint32_t PermuteX, + uint32_t PermuteY, + uint32_t PermuteZ, + uint32_t PermuteW +) noexcept +{ + assert(PermuteX <= 7 && PermuteY <= 7 && PermuteZ <= 7 && PermuteW <= 7); + _Analysis_assume_(PermuteX <= 7 && PermuteY <= 7 && PermuteZ <= 7 && PermuteW <= 7); + +#if defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + static const uint32_t ControlElement[8] = + { + 0x03020100, // XM_PERMUTE_0X + 0x07060504, // XM_PERMUTE_0Y + 0x0B0A0908, // XM_PERMUTE_0Z + 0x0F0E0D0C, // XM_PERMUTE_0W + 0x13121110, // XM_PERMUTE_1X + 0x17161514, // XM_PERMUTE_1Y + 0x1B1A1918, // XM_PERMUTE_1Z + 0x1F1E1D1C, // XM_PERMUTE_1W + }; + + uint8x8x4_t tbl; + tbl.val[0] = vreinterpret_u8_f32(vget_low_f32(V1)); + tbl.val[1] = vreinterpret_u8_f32(vget_high_f32(V1)); + tbl.val[2] = vreinterpret_u8_f32(vget_low_f32(V2)); + tbl.val[3] = vreinterpret_u8_f32(vget_high_f32(V2)); + + uint32x2_t idx = vcreate_u32(static_cast(ControlElement[PermuteX]) | (static_cast(ControlElement[PermuteY]) << 32)); + const uint8x8_t rL = vtbl4_u8(tbl, vreinterpret_u8_u32(idx)); + + idx = vcreate_u32(static_cast(ControlElement[PermuteZ]) | (static_cast(ControlElement[PermuteW]) << 32)); + const uint8x8_t rH = vtbl4_u8(tbl, vreinterpret_u8_u32(idx)); + + return vcombine_f32(vreinterpret_f32_u8(rL), vreinterpret_f32_u8(rH)); +#elif defined(_XM_AVX_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + static const XMVECTORU32 three = { { { 3, 3, 3, 3 } } }; + + XM_ALIGNED_DATA(16) unsigned int elem[4] = { PermuteX, PermuteY, PermuteZ, PermuteW }; + __m128i vControl = _mm_load_si128(reinterpret_cast(&elem[0])); + + __m128i vSelect = _mm_cmpgt_epi32(vControl, three); + vControl = _mm_castps_si128(_mm_and_ps(_mm_castsi128_ps(vControl), three)); + + __m128 shuffled1 = _mm_permutevar_ps(V1, vControl); + __m128 shuffled2 = _mm_permutevar_ps(V2, vControl); + + __m128 masked1 = _mm_andnot_ps(_mm_castsi128_ps(vSelect), shuffled1); + __m128 masked2 = _mm_and_ps(_mm_castsi128_ps(vSelect), shuffled2); + + return _mm_or_ps(masked1, masked2); +#else + + const uint32_t* aPtr[2]; + +#if defined(__GNUC__) && !defined(__clang__) + // workaround some GCC optimization behavior that breaks this function + XMVECTORU32 T1; + T1.v = V1; + XMVECTORU32 T2; + T2.v = V2; + aPtr[0] = reinterpret_cast(&T1); + aPtr[1] = reinterpret_cast(&T2); +#else + aPtr[0] = reinterpret_cast(&V1); + aPtr[1] = reinterpret_cast(&V2); +#endif + + XMVECTORU32 vResult; + const uint32_t i0 = PermuteX & 3; + const uint32_t vi0 = PermuteX >> 2; + vResult.u[0] = aPtr[vi0][i0]; + + const uint32_t i1 = PermuteY & 3; + const uint32_t vi1 = PermuteY >> 2; + vResult.u[1] = aPtr[vi1][i1]; + + const uint32_t i2 = PermuteZ & 3; + const uint32_t vi2 = PermuteZ >> 2; + vResult.u[2] = aPtr[vi2][i2]; + + const uint32_t i3 = PermuteW & 3; + const uint32_t vi3 = PermuteW >> 2; + vResult.u[3] = aPtr[vi3][i3]; + + return vResult.v; +#endif +} + +//------------------------------------------------------------------------------ +// Define a control vector to be used in XMVectorSelect +// operations. The four integers specified in XMVectorSelectControl +// serve as indices to select between components in two vectors. +// The first index controls selection for the first component of +// the vectors involved in a select operation, the second index +// controls selection for the second component etc. A value of +// zero for an index causes the corresponding component from the first +// vector to be selected whereas a one causes the component from the +// second vector to be selected instead. + +inline XMVECTOR XM_CALLCONV XMVectorSelectControl +( + uint32_t VectorIndex0, + uint32_t VectorIndex1, + uint32_t VectorIndex2, + uint32_t VectorIndex3 +) noexcept +{ +#if defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + // x=Index0,y=Index1,z=Index2,w=Index3 + __m128i vTemp = _mm_set_epi32(static_cast(VectorIndex3), static_cast(VectorIndex2), static_cast(VectorIndex1), static_cast(VectorIndex0)); + // Any non-zero entries become 0xFFFFFFFF else 0 + vTemp = _mm_cmpgt_epi32(vTemp, g_XMZero); + return _mm_castsi128_ps(vTemp); +#elif defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + int32x2_t V0 = vcreate_s32(static_cast(VectorIndex0) | (static_cast(VectorIndex1) << 32)); + int32x2_t V1 = vcreate_s32(static_cast(VectorIndex2) | (static_cast(VectorIndex3) << 32)); + int32x4_t vTemp = vcombine_s32(V0, V1); + // Any non-zero entries become 0xFFFFFFFF else 0 + return vreinterpretq_f32_u32(vcgtq_s32(vTemp, g_XMZero)); +#else + XMVECTOR ControlVector; + const uint32_t ControlElement[] = + { + XM_SELECT_0, + XM_SELECT_1 + }; + + assert(VectorIndex0 < 2); + assert(VectorIndex1 < 2); + assert(VectorIndex2 < 2); + assert(VectorIndex3 < 2); + _Analysis_assume_(VectorIndex0 < 2); + _Analysis_assume_(VectorIndex1 < 2); + _Analysis_assume_(VectorIndex2 < 2); + _Analysis_assume_(VectorIndex3 < 2); + + ControlVector.vector4_u32[0] = ControlElement[VectorIndex0]; + ControlVector.vector4_u32[1] = ControlElement[VectorIndex1]; + ControlVector.vector4_u32[2] = ControlElement[VectorIndex2]; + ControlVector.vector4_u32[3] = ControlElement[VectorIndex3]; + + return ControlVector; + +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSelect +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR Control +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + (V1.vector4_u32[0] & ~Control.vector4_u32[0]) | (V2.vector4_u32[0] & Control.vector4_u32[0]), + (V1.vector4_u32[1] & ~Control.vector4_u32[1]) | (V2.vector4_u32[1] & Control.vector4_u32[1]), + (V1.vector4_u32[2] & ~Control.vector4_u32[2]) | (V2.vector4_u32[2] & Control.vector4_u32[2]), + (V1.vector4_u32[3] & ~Control.vector4_u32[3]) | (V2.vector4_u32[3] & Control.vector4_u32[3]), + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vbslq_f32(vreinterpretq_u32_f32(Control), V2, V1); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp1 = _mm_andnot_ps(Control, V1); + XMVECTOR vTemp2 = _mm_and_ps(V2, Control); + return _mm_or_ps(vTemp1, vTemp2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMergeXY +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + V1.vector4_u32[0], + V2.vector4_u32[0], + V1.vector4_u32[1], + V2.vector4_u32[1], + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vzipq_f32(V1, V2).val[0]; +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_unpacklo_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMergeZW +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + V1.vector4_u32[2], + V2.vector4_u32[2], + V1.vector4_u32[3], + V2.vector4_u32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vzipq_f32(V1, V2).val[1]; +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_unpackhi_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorShiftLeft(FXMVECTOR V1, FXMVECTOR V2, uint32_t Elements) noexcept +{ + assert(Elements < 4); + _Analysis_assume_(Elements < 4); + return XMVectorPermute(V1, V2, Elements, ((Elements)+1), ((Elements)+2), ((Elements)+3)); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorRotateLeft(FXMVECTOR V, uint32_t Elements) noexcept +{ + assert(Elements < 4); + _Analysis_assume_(Elements < 4); + return XMVectorSwizzle(V, Elements & 3, (Elements + 1) & 3, (Elements + 2) & 3, (Elements + 3) & 3); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorRotateRight(FXMVECTOR V, uint32_t Elements) noexcept +{ + assert(Elements < 4); + _Analysis_assume_(Elements < 4); + return XMVectorSwizzle(V, (4 - (Elements)) & 3, (5 - (Elements)) & 3, (6 - (Elements)) & 3, (7 - (Elements)) & 3); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorInsert( + FXMVECTOR VD, FXMVECTOR VS, + uint32_t VSLeftRotateElements, + uint32_t Select0, uint32_t Select1, uint32_t Select2, uint32_t Select3) noexcept +{ + XMVECTOR Control = XMVectorSelectControl(Select0 & 1, Select1 & 1, Select2 & 1, Select3 & 1); + return XMVectorSelect(VD, XMVectorRotateLeft(VS, VSLeftRotateElements), Control); +} + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_f32[0] == V2.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[1] == V2.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[2] == V2.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[3] == V2.vector4_f32[3]) ? 0xFFFFFFFF : 0, + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vceqq_f32(V1, V2)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cmpeq_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorEqualR +( + uint32_t* pCR, + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + assert(pCR != nullptr); +#if defined(_XM_NO_INTRINSICS_) + uint32_t ux = (V1.vector4_f32[0] == V2.vector4_f32[0]) ? 0xFFFFFFFFU : 0; + uint32_t uy = (V1.vector4_f32[1] == V2.vector4_f32[1]) ? 0xFFFFFFFFU : 0; + uint32_t uz = (V1.vector4_f32[2] == V2.vector4_f32[2]) ? 0xFFFFFFFFU : 0; + uint32_t uw = (V1.vector4_f32[3] == V2.vector4_f32[3]) ? 0xFFFFFFFFU : 0; + uint32_t CR = 0; + if (ux & uy & uz & uw) + { + // All elements are greater + CR = XM_CRMASK_CR6TRUE; + } + else if (!(ux | uy | uz | uw)) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + + XMVECTORU32 Control = { { { ux, uy, uz, uw } } }; + return Control; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vreinterpret_u8_u32(vget_low_u32(vResult)), vreinterpret_u8_u32(vget_high_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + // All elements are equal + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + // All elements are not equal + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vreinterpretq_f32_u32(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + uint32_t CR = 0; + int iTest = _mm_movemask_ps(vTemp); + if (iTest == 0xf) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +// Treat the components of the vectors as unsigned integers and +// compare individual bits between the two. This is useful for +// comparing control vectors and result vectors returned from +// other comparison operations. + +inline XMVECTOR XM_CALLCONV XMVectorEqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_u32[0] == V2.vector4_u32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_u32[1] == V2.vector4_u32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_u32[2] == V2.vector4_u32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_u32[3] == V2.vector4_u32[3]) ? 0xFFFFFFFF : 0, + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vceqq_s32(vreinterpretq_s32_f32(V1), vreinterpretq_s32_f32(V2))); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorEqualIntR +( + uint32_t* pCR, + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + assert(pCR != nullptr); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Control = XMVectorEqualInt(V1, V2); + + *pCR = 0; + if (XMVector4EqualInt(Control, XMVectorTrueInt())) + { + // All elements are equal + *pCR |= XM_CRMASK_CR6TRUE; + } + else if (XMVector4EqualInt(Control, XMVectorFalseInt())) + { + // All elements are not equal + *pCR |= XM_CRMASK_CR6FALSE; + } + return Control; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + // All elements are equal + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + // All elements are not equal + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vreinterpretq_f32_u32(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + int iTemp = _mm_movemask_ps(_mm_castsi128_ps(V)); + uint32_t CR = 0; + if (iTemp == 0x0F) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTemp) + { + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorNearEqual +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR Epsilon +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + float fDeltax = V1.vector4_f32[0] - V2.vector4_f32[0]; + float fDeltay = V1.vector4_f32[1] - V2.vector4_f32[1]; + float fDeltaz = V1.vector4_f32[2] - V2.vector4_f32[2]; + float fDeltaw = V1.vector4_f32[3] - V2.vector4_f32[3]; + + fDeltax = fabsf(fDeltax); + fDeltay = fabsf(fDeltay); + fDeltaz = fabsf(fDeltaz); + fDeltaw = fabsf(fDeltaw); + + XMVECTORU32 Control = { { { + (fDeltax <= Epsilon.vector4_f32[0]) ? 0xFFFFFFFFU : 0, + (fDeltay <= Epsilon.vector4_f32[1]) ? 0xFFFFFFFFU : 0, + (fDeltaz <= Epsilon.vector4_f32[2]) ? 0xFFFFFFFFU : 0, + (fDeltaw <= Epsilon.vector4_f32[3]) ? 0xFFFFFFFFU : 0, + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vDelta = vsubq_f32(V1, V2); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + return vacleq_f32(vDelta, Epsilon); +#else + return vreinterpretq_f32_u32(vcleq_f32(vabsq_f32(vDelta), Epsilon)); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + // Get the difference + XMVECTOR vDelta = _mm_sub_ps(V1, V2); + // Get the absolute value of the difference + XMVECTOR vTemp = _mm_setzero_ps(); + vTemp = _mm_sub_ps(vTemp, vDelta); + vTemp = _mm_max_ps(vTemp, vDelta); + vTemp = _mm_cmple_ps(vTemp, Epsilon); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorNotEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_f32[0] != V2.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[1] != V2.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[2] != V2.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[3] != V2.vector4_f32[3]) ? 0xFFFFFFFF : 0, + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vmvnq_u32(vceqq_f32(V1, V2))); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cmpneq_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorNotEqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_u32[0] != V2.vector4_u32[0]) ? 0xFFFFFFFFU : 0, + (V1.vector4_u32[1] != V2.vector4_u32[1]) ? 0xFFFFFFFFU : 0, + (V1.vector4_u32[2] != V2.vector4_u32[2]) ? 0xFFFFFFFFU : 0, + (V1.vector4_u32[3] != V2.vector4_u32[3]) ? 0xFFFFFFFFU : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vmvnq_u32( + vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)))); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return _mm_xor_ps(_mm_castsi128_ps(V), g_XMNegOneMask); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorGreater +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_f32[0] > V2.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[1] > V2.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[2] > V2.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[3] > V2.vector4_f32[3]) ? 0xFFFFFFFF : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vcgtq_f32(V1, V2)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cmpgt_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorGreaterR +( + uint32_t* pCR, + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + assert(pCR != nullptr); +#if defined(_XM_NO_INTRINSICS_) + + uint32_t ux = (V1.vector4_f32[0] > V2.vector4_f32[0]) ? 0xFFFFFFFFU : 0; + uint32_t uy = (V1.vector4_f32[1] > V2.vector4_f32[1]) ? 0xFFFFFFFFU : 0; + uint32_t uz = (V1.vector4_f32[2] > V2.vector4_f32[2]) ? 0xFFFFFFFFU : 0; + uint32_t uw = (V1.vector4_f32[3] > V2.vector4_f32[3]) ? 0xFFFFFFFFU : 0; + uint32_t CR = 0; + if (ux & uy & uz & uw) + { + // All elements are greater + CR = XM_CRMASK_CR6TRUE; + } + else if (!(ux | uy | uz | uw)) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + + XMVECTORU32 Control = { { { ux, uy, uz, uw } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgtq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + // All elements are greater + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vreinterpretq_f32_u32(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + uint32_t CR = 0; + int iTest = _mm_movemask_ps(vTemp); + if (iTest == 0xf) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorGreaterOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_f32[0] >= V2.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[1] >= V2.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[2] >= V2.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[3] >= V2.vector4_f32[3]) ? 0xFFFFFFFF : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vcgeq_f32(V1, V2)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cmpge_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorGreaterOrEqualR +( + uint32_t* pCR, + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + assert(pCR != nullptr); +#if defined(_XM_NO_INTRINSICS_) + + uint32_t ux = (V1.vector4_f32[0] >= V2.vector4_f32[0]) ? 0xFFFFFFFFU : 0; + uint32_t uy = (V1.vector4_f32[1] >= V2.vector4_f32[1]) ? 0xFFFFFFFFU : 0; + uint32_t uz = (V1.vector4_f32[2] >= V2.vector4_f32[2]) ? 0xFFFFFFFFU : 0; + uint32_t uw = (V1.vector4_f32[3] >= V2.vector4_f32[3]) ? 0xFFFFFFFFU : 0; + uint32_t CR = 0; + if (ux & uy & uz & uw) + { + // All elements are greater + CR = XM_CRMASK_CR6TRUE; + } + else if (!(ux | uy | uz | uw)) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + + XMVECTORU32 Control = { { { ux, uy, uz, uw } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgeq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + // All elements are greater or equal + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + // All elements are not greater or equal + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vreinterpretq_f32_u32(vResult); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + uint32_t CR = 0; + int iTest = _mm_movemask_ps(vTemp); + if (iTest == 0xf) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + // All elements are not greater + CR = XM_CRMASK_CR6FALSE; + } + *pCR = CR; + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLess +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_f32[0] < V2.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[1] < V2.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[2] < V2.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[3] < V2.vector4_f32[3]) ? 0xFFFFFFFF : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vcltq_f32(V1, V2)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cmplt_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLessOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V1.vector4_f32[0] <= V2.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[1] <= V2.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[2] <= V2.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V1.vector4_f32[3] <= V2.vector4_f32[3]) ? 0xFFFFFFFF : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vcleq_f32(V1, V2)); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_cmple_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorInBounds +( + FXMVECTOR V, + FXMVECTOR Bounds +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + (V.vector4_f32[0] <= Bounds.vector4_f32[0] && V.vector4_f32[0] >= -Bounds.vector4_f32[0]) ? 0xFFFFFFFF : 0, + (V.vector4_f32[1] <= Bounds.vector4_f32[1] && V.vector4_f32[1] >= -Bounds.vector4_f32[1]) ? 0xFFFFFFFF : 0, + (V.vector4_f32[2] <= Bounds.vector4_f32[2] && V.vector4_f32[2] >= -Bounds.vector4_f32[2]) ? 0xFFFFFFFF : 0, + (V.vector4_f32[3] <= Bounds.vector4_f32[3] && V.vector4_f32[3] >= -Bounds.vector4_f32[3]) ? 0xFFFFFFFF : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Test if less than or equal + uint32x4_t vTemp1 = vcleq_f32(V, Bounds); + // Negate the bounds + uint32x4_t vTemp2 = vreinterpretq_u32_f32(vnegq_f32(Bounds)); + // Test if greater or equal (Reversed) + vTemp2 = vcleq_f32(vreinterpretq_f32_u32(vTemp2), V); + // Blend answers + vTemp1 = vandq_u32(vTemp1, vTemp2); + return vreinterpretq_f32_u32(vTemp1); +#elif defined(_XM_SSE_INTRINSICS_) + // Test if less than or equal + XMVECTOR vTemp1 = _mm_cmple_ps(V, Bounds); + // Negate the bounds + XMVECTOR vTemp2 = _mm_mul_ps(Bounds, g_XMNegativeOne); + // Test if greater or equal (Reversed) + vTemp2 = _mm_cmple_ps(vTemp2, V); + // Blend answers + vTemp1 = _mm_and_ps(vTemp1, vTemp2); + return vTemp1; +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMVectorInBoundsR +( + uint32_t* pCR, + FXMVECTOR V, + FXMVECTOR Bounds +) noexcept +{ + assert(pCR != nullptr); +#if defined(_XM_NO_INTRINSICS_) + + uint32_t ux = (V.vector4_f32[0] <= Bounds.vector4_f32[0] && V.vector4_f32[0] >= -Bounds.vector4_f32[0]) ? 0xFFFFFFFFU : 0; + uint32_t uy = (V.vector4_f32[1] <= Bounds.vector4_f32[1] && V.vector4_f32[1] >= -Bounds.vector4_f32[1]) ? 0xFFFFFFFFU : 0; + uint32_t uz = (V.vector4_f32[2] <= Bounds.vector4_f32[2] && V.vector4_f32[2] >= -Bounds.vector4_f32[2]) ? 0xFFFFFFFFU : 0; + uint32_t uw = (V.vector4_f32[3] <= Bounds.vector4_f32[3] && V.vector4_f32[3] >= -Bounds.vector4_f32[3]) ? 0xFFFFFFFFU : 0; + + uint32_t CR = 0; + if (ux & uy & uz & uw) + { + // All elements are in bounds + CR = XM_CRMASK_CR6BOUNDS; + } + *pCR = CR; + + XMVECTORU32 Control = { { { ux, uy, uz, uw } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Test if less than or equal + uint32x4_t vTemp1 = vcleq_f32(V, Bounds); + // Negate the bounds + uint32x4_t vTemp2 = vreinterpretq_u32_f32(vnegq_f32(Bounds)); + // Test if greater or equal (Reversed) + vTemp2 = vcleq_f32(vreinterpretq_f32_u32(vTemp2), V); + // Blend answers + vTemp1 = vandq_u32(vTemp1, vTemp2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vTemp1)), vget_high_u8(vreinterpretq_u8_u32(vTemp1))); + uint16x4x2_t vTemp3 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp3.val[1]), 1); + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + // All elements are in bounds + CR = XM_CRMASK_CR6BOUNDS; + } + *pCR = CR; + return vreinterpretq_f32_u32(vTemp1); +#elif defined(_XM_SSE_INTRINSICS_) + // Test if less than or equal + XMVECTOR vTemp1 = _mm_cmple_ps(V, Bounds); + // Negate the bounds + XMVECTOR vTemp2 = _mm_mul_ps(Bounds, g_XMNegativeOne); + // Test if greater or equal (Reversed) + vTemp2 = _mm_cmple_ps(vTemp2, V); + // Blend answers + vTemp1 = _mm_and_ps(vTemp1, vTemp2); + + uint32_t CR = 0; + if (_mm_movemask_ps(vTemp1) == 0xf) + { + // All elements are in bounds + CR = XM_CRMASK_CR6BOUNDS; + } + *pCR = CR; + return vTemp1; +#endif +} + +//------------------------------------------------------------------------------ + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(push) +#pragma float_control(precise, on) +#endif + +inline XMVECTOR XM_CALLCONV XMVectorIsNaN(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + XMISNAN(V.vector4_f32[0]) ? 0xFFFFFFFFU : 0, + XMISNAN(V.vector4_f32[1]) ? 0xFFFFFFFFU : 0, + XMISNAN(V.vector4_f32[2]) ? 0xFFFFFFFFU : 0, + XMISNAN(V.vector4_f32[3]) ? 0xFFFFFFFFU : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + XMVECTORU32 vResult = { { { + isnan(vgetq_lane_f32(V, 0)) ? 0xFFFFFFFFU : 0, + isnan(vgetq_lane_f32(V, 1)) ? 0xFFFFFFFFU : 0, + isnan(vgetq_lane_f32(V, 2)) ? 0xFFFFFFFFU : 0, + isnan(vgetq_lane_f32(V, 3)) ? 0xFFFFFFFFU : 0 } } }; + return vResult.v; +#else +// Test against itself. NaN is always not equal + uint32x4_t vTempNan = vceqq_f32(V, V); + // Flip results + return vreinterpretq_f32_u32(vmvnq_u32(vTempNan)); +#endif +#elif defined(_XM_SSE_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + XM_ALIGNED_DATA(16) float tmp[4]; + _mm_store_ps(tmp, V); + XMVECTORU32 vResult = { { { + isnan(tmp[0]) ? 0xFFFFFFFFU : 0, + isnan(tmp[1]) ? 0xFFFFFFFFU : 0, + isnan(tmp[2]) ? 0xFFFFFFFFU : 0, + isnan(tmp[3]) ? 0xFFFFFFFFU : 0 } } }; + return vResult.v; +#else +// Test against itself. NaN is always not equal + return _mm_cmpneq_ps(V, V); +#endif +#endif +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(pop) +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorIsInfinite(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Control = { { { + XMISINF(V.vector4_f32[0]) ? 0xFFFFFFFFU : 0, + XMISINF(V.vector4_f32[1]) ? 0xFFFFFFFFU : 0, + XMISINF(V.vector4_f32[2]) ? 0xFFFFFFFFU : 0, + XMISINF(V.vector4_f32[3]) ? 0xFFFFFFFFU : 0 + } } }; + return Control.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Mask off the sign bit + uint32x4_t vTemp = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + // Compare to infinity + vTemp = vceqq_f32(vreinterpretq_f32_u32(vTemp), g_XMInfinity); + // If any are infinity, the signs are true. + return vreinterpretq_f32_u32(vTemp); +#elif defined(_XM_SSE_INTRINSICS_) + // Mask off the sign bit + __m128 vTemp = _mm_and_ps(V, g_XMAbsMask); + // Compare to infinity + vTemp = _mm_cmpeq_ps(vTemp, g_XMInfinity); + // If any are infinity, the signs are true. + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +// Rounding and clamping operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMin +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + (V1.vector4_f32[0] < V2.vector4_f32[0]) ? V1.vector4_f32[0] : V2.vector4_f32[0], + (V1.vector4_f32[1] < V2.vector4_f32[1]) ? V1.vector4_f32[1] : V2.vector4_f32[1], + (V1.vector4_f32[2] < V2.vector4_f32[2]) ? V1.vector4_f32[2] : V2.vector4_f32[2], + (V1.vector4_f32[3] < V2.vector4_f32[3]) ? V1.vector4_f32[3] : V2.vector4_f32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vminq_f32(V1, V2); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_min_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMax +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + (V1.vector4_f32[0] > V2.vector4_f32[0]) ? V1.vector4_f32[0] : V2.vector4_f32[0], + (V1.vector4_f32[1] > V2.vector4_f32[1]) ? V1.vector4_f32[1] : V2.vector4_f32[1], + (V1.vector4_f32[2] > V2.vector4_f32[2]) ? V1.vector4_f32[2] : V2.vector4_f32[2], + (V1.vector4_f32[3] > V2.vector4_f32[3]) ? V1.vector4_f32[3] : V2.vector4_f32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vmaxq_f32(V1, V2); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_max_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +namespace MathInternal +{ + // Round to nearest (even) a.k.a. banker's rounding + inline float round_to_nearest(float x) noexcept + { + float i = floorf(x); + x -= i; + if (x < 0.5f) + return i; + if (x > 0.5f) + return i + 1.f; + + float int_part; + (void)modff(i / 2.f, &int_part); + if ((2.f * int_part) == i) + { + return i; + } + + return i + 1.f; + } +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(push) +#pragma float_control(precise, on) +#endif + +inline XMVECTOR XM_CALLCONV XMVectorRound(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + MathInternal::round_to_nearest(V.vector4_f32[0]), + MathInternal::round_to_nearest(V.vector4_f32[1]), + MathInternal::round_to_nearest(V.vector4_f32[2]), + MathInternal::round_to_nearest(V.vector4_f32[3]) + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vrndnq_f32(V); +#else + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(V), g_XMNegativeZero); + float32x4_t sMagic = vreinterpretq_f32_u32(vorrq_u32(g_XMNoFraction, sign)); + float32x4_t R1 = vaddq_f32(V, sMagic); + R1 = vsubq_f32(R1, sMagic); + float32x4_t R2 = vabsq_f32(V); + uint32x4_t mask = vcleq_f32(R2, g_XMNoFraction); + return vbslq_f32(mask, R1, V); +#endif +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_round_ps(V, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); +#elif defined(_XM_SSE_INTRINSICS_) + __m128 sign = _mm_and_ps(V, g_XMNegativeZero); + __m128 sMagic = _mm_or_ps(g_XMNoFraction, sign); + __m128 R1 = _mm_add_ps(V, sMagic); + R1 = _mm_sub_ps(R1, sMagic); + __m128 R2 = _mm_and_ps(V, g_XMAbsMask); + __m128 mask = _mm_cmple_ps(R2, g_XMNoFraction); + R2 = _mm_andnot_ps(mask, V); + R1 = _mm_and_ps(R1, mask); + XMVECTOR vResult = _mm_xor_ps(R1, R2); + return vResult; +#endif +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(pop) +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorTruncate(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR Result; + uint32_t i; + + // Avoid C4701 + Result.vector4_f32[0] = 0.0f; + + for (i = 0; i < 4; i++) + { + if (XMISNAN(V.vector4_f32[i])) + { + Result.vector4_u32[i] = 0x7FC00000; + } + else if (fabsf(V.vector4_f32[i]) < 8388608.0f) + { + Result.vector4_f32[i] = static_cast(static_cast(V.vector4_f32[i])); + } + else + { + Result.vector4_f32[i] = V.vector4_f32[i]; + } + } + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vrndq_f32(V); +#else + float32x4_t vTest = vabsq_f32(V); + vTest = vreinterpretq_f32_u32(vcltq_f32(vTest, g_XMNoFraction)); + + int32x4_t vInt = vcvtq_s32_f32(V); + float32x4_t vResult = vcvtq_f32_s32(vInt); + + // All numbers less than 8388608 will use the round to int + // All others, use the ORIGINAL value + return vbslq_f32(vreinterpretq_u32_f32(vTest), vResult, V); +#endif +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_round_ps(V, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC); +#elif defined(_XM_SSE_INTRINSICS_) + // To handle NAN, INF and numbers greater than 8388608, use masking + // Get the abs value + __m128i vTest = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + // Test for greater than 8388608 (All floats with NO fractionals, NAN and INF + vTest = _mm_cmplt_epi32(vTest, g_XMNoFraction); + // Convert to int and back to float for rounding with truncation + __m128i vInt = _mm_cvttps_epi32(V); + // Convert back to floats + XMVECTOR vResult = _mm_cvtepi32_ps(vInt); + // All numbers less than 8388608 will use the round to int + vResult = _mm_and_ps(vResult, _mm_castsi128_ps(vTest)); + // All others, use the ORIGINAL value + vTest = _mm_andnot_si128(vTest, _mm_castps_si128(V)); + vResult = _mm_or_ps(vResult, _mm_castsi128_ps(vTest)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorFloor(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + floorf(V.vector4_f32[0]), + floorf(V.vector4_f32[1]), + floorf(V.vector4_f32[2]), + floorf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vrndmq_f32(V); +#else + float32x4_t vTest = vabsq_f32(V); + vTest = vreinterpretq_f32_u32(vcltq_f32(vTest, g_XMNoFraction)); + // Truncate + int32x4_t vInt = vcvtq_s32_f32(V); + float32x4_t vResult = vcvtq_f32_s32(vInt); + uint32x4_t vLargerMask = vcgtq_f32(vResult, V); + // 0 -> 0, 0xffffffff -> -1.0f + float32x4_t vLarger = vcvtq_f32_s32(vreinterpretq_s32_u32(vLargerMask)); + vResult = vaddq_f32(vResult, vLarger); + // All numbers less than 8388608 will use the round to int + // All others, use the ORIGINAL value + return vbslq_f32(vreinterpretq_u32_f32(vTest), vResult, V); +#endif +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_floor_ps(V); +#elif defined(_XM_SSE_INTRINSICS_) + // To handle NAN, INF and numbers greater than 8388608, use masking + __m128i vTest = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + vTest = _mm_cmplt_epi32(vTest, g_XMNoFraction); + // Truncate + __m128i vInt = _mm_cvttps_epi32(V); + XMVECTOR vResult = _mm_cvtepi32_ps(vInt); + __m128 vLarger = _mm_cmpgt_ps(vResult, V); + // 0 -> 0, 0xffffffff -> -1.0f + vLarger = _mm_cvtepi32_ps(_mm_castps_si128(vLarger)); + vResult = _mm_add_ps(vResult, vLarger); + // All numbers less than 8388608 will use the round to int + vResult = _mm_and_ps(vResult, _mm_castsi128_ps(vTest)); + // All others, use the ORIGINAL value + vTest = _mm_andnot_si128(vTest, _mm_castps_si128(V)); + vResult = _mm_or_ps(vResult, _mm_castsi128_ps(vTest)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorCeiling(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + ceilf(V.vector4_f32[0]), + ceilf(V.vector4_f32[1]), + ceilf(V.vector4_f32[2]), + ceilf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vrndpq_f32(V); +#else + float32x4_t vTest = vabsq_f32(V); + vTest = vreinterpretq_f32_u32(vcltq_f32(vTest, g_XMNoFraction)); + // Truncate + int32x4_t vInt = vcvtq_s32_f32(V); + float32x4_t vResult = vcvtq_f32_s32(vInt); + uint32x4_t vSmallerMask = vcltq_f32(vResult, V); + // 0 -> 0, 0xffffffff -> -1.0f + float32x4_t vSmaller = vcvtq_f32_s32(vreinterpretq_s32_u32(vSmallerMask)); + vResult = vsubq_f32(vResult, vSmaller); + // All numbers less than 8388608 will use the round to int + // All others, use the ORIGINAL value + return vbslq_f32(vreinterpretq_u32_f32(vTest), vResult, V); +#endif +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_ceil_ps(V); +#elif defined(_XM_SSE_INTRINSICS_) + // To handle NAN, INF and numbers greater than 8388608, use masking + __m128i vTest = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + vTest = _mm_cmplt_epi32(vTest, g_XMNoFraction); + // Truncate + __m128i vInt = _mm_cvttps_epi32(V); + XMVECTOR vResult = _mm_cvtepi32_ps(vInt); + __m128 vSmaller = _mm_cmplt_ps(vResult, V); + // 0 -> 0, 0xffffffff -> -1.0f + vSmaller = _mm_cvtepi32_ps(_mm_castps_si128(vSmaller)); + vResult = _mm_sub_ps(vResult, vSmaller); + // All numbers less than 8388608 will use the round to int + vResult = _mm_and_ps(vResult, _mm_castsi128_ps(vTest)); + // All others, use the ORIGINAL value + vTest = _mm_andnot_si128(vTest, _mm_castps_si128(V)); + vResult = _mm_or_ps(vResult, _mm_castsi128_ps(vTest)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorClamp +( + FXMVECTOR V, + FXMVECTOR Min, + FXMVECTOR Max +) noexcept +{ + assert(XMVector4LessOrEqual(Min, Max)); + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVectorMax(Min, V); + Result = XMVectorMin(Max, Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmaxq_f32(Min, V); + vResult = vminq_f32(Max, vResult); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult; + vResult = _mm_max_ps(Min, V); + vResult = _mm_min_ps(Max, vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSaturate(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + const XMVECTOR Zero = XMVectorZero(); + + return XMVectorClamp(V, Zero, g_XMOne.v); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Set <0 to 0 + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0)); + // Set>1 to 1 + return vminq_f32(vResult, vdupq_n_f32(1.0f)); +#elif defined(_XM_SSE_INTRINSICS_) + // Set <0 to 0 + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + // Set>1 to 1 + return _mm_min_ps(vResult, g_XMOne); +#endif +} + +//------------------------------------------------------------------------------ +// Bitwise logical operations +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorAndInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + V1.vector4_u32[0] & V2.vector4_u32[0], + V1.vector4_u32[1] & V2.vector4_u32[1], + V1.vector4_u32[2] & V2.vector4_u32[2], + V1.vector4_u32[3] & V2.vector4_u32[3] + } } }; + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2))); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_and_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorAndCInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + V1.vector4_u32[0] & ~V2.vector4_u32[0], + V1.vector4_u32[1] & ~V2.vector4_u32[1], + V1.vector4_u32[2] & ~V2.vector4_u32[2], + V1.vector4_u32[3] & ~V2.vector4_u32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2))); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_andnot_si128(_mm_castps_si128(V2), _mm_castps_si128(V1)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorOrInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + V1.vector4_u32[0] | V2.vector4_u32[0], + V1.vector4_u32[1] | V2.vector4_u32[1], + V1.vector4_u32[2] | V2.vector4_u32[2], + V1.vector4_u32[3] | V2.vector4_u32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2))); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_or_si128(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorNorInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + ~(V1.vector4_u32[0] | V2.vector4_u32[0]), + ~(V1.vector4_u32[1] | V2.vector4_u32[1]), + ~(V1.vector4_u32[2] | V2.vector4_u32[2]), + ~(V1.vector4_u32[3] | V2.vector4_u32[3]) + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t Result = vorrq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + return vreinterpretq_f32_u32(vbicq_u32(g_XMNegOneMask, Result)); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i Result; + Result = _mm_or_si128(_mm_castps_si128(V1), _mm_castps_si128(V2)); + Result = _mm_andnot_si128(Result, g_XMNegOneMask); + return _mm_castsi128_ps(Result); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorXorInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORU32 Result = { { { + V1.vector4_u32[0] ^ V2.vector4_u32[0], + V1.vector4_u32[1] ^ V2.vector4_u32[1], + V1.vector4_u32[2] ^ V2.vector4_u32[2], + V1.vector4_u32[3] ^ V2.vector4_u32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2))); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i V = _mm_xor_si128(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return _mm_castsi128_ps(V); +#endif +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorNegate(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + -V.vector4_f32[0], + -V.vector4_f32[1], + -V.vector4_f32[2], + -V.vector4_f32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vnegq_f32(V); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR Z; + + Z = _mm_setzero_ps(); + + return _mm_sub_ps(Z, V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorAdd +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + V1.vector4_f32[0] + V2.vector4_f32[0], + V1.vector4_f32[1] + V2.vector4_f32[1], + V1.vector4_f32[2] + V2.vector4_f32[2], + V1.vector4_f32[3] + V2.vector4_f32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vaddq_f32(V1, V2); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_add_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSum(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result; + Result.f[0] = + Result.f[1] = + Result.f[2] = + Result.f[3] = V.vector4_f32[0] + V.vector4_f32[1] + V.vector4_f32[2] + V.vector4_f32[3]; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + float32x4_t vTemp = vpaddq_f32(V, V); + return vpaddq_f32(vTemp, vTemp); +#else + float32x2_t v1 = vget_low_f32(V); + float32x2_t v2 = vget_high_f32(V); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + return vcombine_f32(v1, v1); +#endif +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vTemp = _mm_hadd_ps(V, V); + return _mm_hadd_ps(vTemp, vTemp); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 3, 0, 1)); + XMVECTOR vTemp2 = _mm_add_ps(V, vTemp); + vTemp = XM_PERMUTE_PS(vTemp2, _MM_SHUFFLE(1, 0, 3, 2)); + return _mm_add_ps(vTemp, vTemp2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorAddAngles +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + const XMVECTOR Zero = XMVectorZero(); + + // Add the given angles together. If the range of V1 is such + // that -Pi <= V1 < Pi and the range of V2 is such that + // -2Pi <= V2 <= 2Pi, then the range of the resulting angle + // will be -Pi <= Result < Pi. + XMVECTOR Result = XMVectorAdd(V1, V2); + + XMVECTOR Mask = XMVectorLess(Result, g_XMNegativePi.v); + XMVECTOR Offset = XMVectorSelect(Zero, g_XMTwoPi.v, Mask); + + Mask = XMVectorGreaterOrEqual(Result, g_XMPi.v); + Offset = XMVectorSelect(Offset, g_XMNegativeTwoPi.v, Mask); + + Result = XMVectorAdd(Result, Offset); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Adjust the angles + float32x4_t vResult = vaddq_f32(V1, V2); + // Less than Pi? + uint32x4_t vOffset = vcltq_f32(vResult, g_XMNegativePi); + vOffset = vandq_u32(vOffset, g_XMTwoPi); + // Add 2Pi to all entries less than -Pi + vResult = vaddq_f32(vResult, vreinterpretq_f32_u32(vOffset)); + // Greater than or equal to Pi? + vOffset = vcgeq_f32(vResult, g_XMPi); + vOffset = vandq_u32(vOffset, g_XMTwoPi); + // Sub 2Pi to all entries greater than Pi + vResult = vsubq_f32(vResult, vreinterpretq_f32_u32(vOffset)); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Adjust the angles + XMVECTOR vResult = _mm_add_ps(V1, V2); + // Less than Pi? + XMVECTOR vOffset = _mm_cmplt_ps(vResult, g_XMNegativePi); + vOffset = _mm_and_ps(vOffset, g_XMTwoPi); + // Add 2Pi to all entries less than -Pi + vResult = _mm_add_ps(vResult, vOffset); + // Greater than or equal to Pi? + vOffset = _mm_cmpge_ps(vResult, g_XMPi); + vOffset = _mm_and_ps(vOffset, g_XMTwoPi); + // Sub 2Pi to all entries greater than Pi + vResult = _mm_sub_ps(vResult, vOffset); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSubtract +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + V1.vector4_f32[0] - V2.vector4_f32[0], + V1.vector4_f32[1] - V2.vector4_f32[1], + V1.vector4_f32[2] - V2.vector4_f32[2], + V1.vector4_f32[3] - V2.vector4_f32[3] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vsubq_f32(V1, V2); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_sub_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSubtractAngles +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + const XMVECTOR Zero = XMVectorZero(); + + // Subtract the given angles. If the range of V1 is such + // that -Pi <= V1 < Pi and the range of V2 is such that + // -2Pi <= V2 <= 2Pi, then the range of the resulting angle + // will be -Pi <= Result < Pi. + XMVECTOR Result = XMVectorSubtract(V1, V2); + + XMVECTOR Mask = XMVectorLess(Result, g_XMNegativePi.v); + XMVECTOR Offset = XMVectorSelect(Zero, g_XMTwoPi.v, Mask); + + Mask = XMVectorGreaterOrEqual(Result, g_XMPi.v); + Offset = XMVectorSelect(Offset, g_XMNegativeTwoPi.v, Mask); + + Result = XMVectorAdd(Result, Offset); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Adjust the angles + XMVECTOR vResult = vsubq_f32(V1, V2); + // Less than Pi? + uint32x4_t vOffset = vcltq_f32(vResult, g_XMNegativePi); + vOffset = vandq_u32(vOffset, g_XMTwoPi); + // Add 2Pi to all entries less than -Pi + vResult = vaddq_f32(vResult, vreinterpretq_f32_u32(vOffset)); + // Greater than or equal to Pi? + vOffset = vcgeq_f32(vResult, g_XMPi); + vOffset = vandq_u32(vOffset, g_XMTwoPi); + // Sub 2Pi to all entries greater than Pi + vResult = vsubq_f32(vResult, vreinterpretq_f32_u32(vOffset)); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Adjust the angles + XMVECTOR vResult = _mm_sub_ps(V1, V2); + // Less than Pi? + XMVECTOR vOffset = _mm_cmplt_ps(vResult, g_XMNegativePi); + vOffset = _mm_and_ps(vOffset, g_XMTwoPi); + // Add 2Pi to all entries less than -Pi + vResult = _mm_add_ps(vResult, vOffset); + // Greater than or equal to Pi? + vOffset = _mm_cmpge_ps(vResult, g_XMPi); + vOffset = _mm_and_ps(vOffset, g_XMTwoPi); + // Sub 2Pi to all entries greater than Pi + vResult = _mm_sub_ps(vResult, vOffset); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMultiply +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + V1.vector4_f32[0] * V2.vector4_f32[0], + V1.vector4_f32[1] * V2.vector4_f32[1], + V1.vector4_f32[2] * V2.vector4_f32[2], + V1.vector4_f32[3] * V2.vector4_f32[3] + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vmulq_f32(V1, V2); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_mul_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMultiplyAdd +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR V3 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + V1.vector4_f32[0] * V2.vector4_f32[0] + V3.vector4_f32[0], + V1.vector4_f32[1] * V2.vector4_f32[1] + V3.vector4_f32[1], + V1.vector4_f32[2] * V2.vector4_f32[2] + V3.vector4_f32[2], + V1.vector4_f32[3] * V2.vector4_f32[3] + V3.vector4_f32[3] + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vfmaq_f32(V3, V1, V2); +#else + return vmlaq_f32(V3, V1, V2); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + return XM_FMADD_PS(V1, V2, V3); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorDivide +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + V1.vector4_f32[0] / V2.vector4_f32[0], + V1.vector4_f32[1] / V2.vector4_f32[1], + V1.vector4_f32[2] / V2.vector4_f32[2], + V1.vector4_f32[3] / V2.vector4_f32[3] + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vdivq_f32(V1, V2); +#else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x4_t Reciprocal = vrecpeq_f32(V2); + float32x4_t S = vrecpsq_f32(Reciprocal, V2); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, V2); + Reciprocal = vmulq_f32(S, Reciprocal); + return vmulq_f32(V1, Reciprocal); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_div_ps(V1, V2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorNegativeMultiplySubtract +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR V3 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + V3.vector4_f32[0] - (V1.vector4_f32[0] * V2.vector4_f32[0]), + V3.vector4_f32[1] - (V1.vector4_f32[1] * V2.vector4_f32[1]), + V3.vector4_f32[2] - (V1.vector4_f32[2] * V2.vector4_f32[2]), + V3.vector4_f32[3] - (V1.vector4_f32[3] * V2.vector4_f32[3]) + } } }; + return Result; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + return vfmsq_f32(V3, V1, V2); +#else + return vmlsq_f32(V3, V1, V2); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + return XM_FNMADD_PS(V1, V2, V3); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorScale +( + FXMVECTOR V, + float ScaleFactor +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + V.vector4_f32[0] * ScaleFactor, + V.vector4_f32[1] * ScaleFactor, + V.vector4_f32[2] * ScaleFactor, + V.vector4_f32[3] * ScaleFactor + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vmulq_n_f32(V, ScaleFactor); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_set_ps1(ScaleFactor); + return _mm_mul_ps(vResult, V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorReciprocalEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + 1.f / V.vector4_f32[0], + 1.f / V.vector4_f32[1], + 1.f / V.vector4_f32[2], + 1.f / V.vector4_f32[3] + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vrecpeq_f32(V); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_rcp_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorReciprocal(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + 1.f / V.vector4_f32[0], + 1.f / V.vector4_f32[1], + 1.f / V.vector4_f32[2], + 1.f / V.vector4_f32[3] + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + float32x4_t one = vdupq_n_f32(1.0f); + return vdivq_f32(one, V); +#else + // 2 iterations of Newton-Raphson refinement + float32x4_t Reciprocal = vrecpeq_f32(V); + float32x4_t S = vrecpsq_f32(Reciprocal, V); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, V); + return vmulq_f32(S, Reciprocal); +#endif +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_div_ps(g_XMOne, V); +#endif +} + +//------------------------------------------------------------------------------ +// Return an estimated square root +inline XMVECTOR XM_CALLCONV XMVectorSqrtEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + sqrtf(V.vector4_f32[0]), + sqrtf(V.vector4_f32[1]), + sqrtf(V.vector4_f32[2]), + sqrtf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // 1 iteration of Newton-Raphson refinment of sqrt + float32x4_t S0 = vrsqrteq_f32(V); + float32x4_t P0 = vmulq_f32(V, S0); + float32x4_t R0 = vrsqrtsq_f32(P0, S0); + float32x4_t S1 = vmulq_f32(S0, R0); + + XMVECTOR VEqualsInfinity = XMVectorEqualInt(V, g_XMInfinity.v); + XMVECTOR VEqualsZero = XMVectorEqual(V, vdupq_n_f32(0)); + XMVECTOR Result = vmulq_f32(V, S1); + XMVECTOR Select = XMVectorEqualInt(VEqualsInfinity, VEqualsZero); + return XMVectorSelect(V, Result, Select); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_sqrt_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSqrt(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + sqrtf(V.vector4_f32[0]), + sqrtf(V.vector4_f32[1]), + sqrtf(V.vector4_f32[2]), + sqrtf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // 3 iterations of Newton-Raphson refinment of sqrt + float32x4_t S0 = vrsqrteq_f32(V); + float32x4_t P0 = vmulq_f32(V, S0); + float32x4_t R0 = vrsqrtsq_f32(P0, S0); + float32x4_t S1 = vmulq_f32(S0, R0); + float32x4_t P1 = vmulq_f32(V, S1); + float32x4_t R1 = vrsqrtsq_f32(P1, S1); + float32x4_t S2 = vmulq_f32(S1, R1); + float32x4_t P2 = vmulq_f32(V, S2); + float32x4_t R2 = vrsqrtsq_f32(P2, S2); + float32x4_t S3 = vmulq_f32(S2, R2); + + XMVECTOR VEqualsInfinity = XMVectorEqualInt(V, g_XMInfinity.v); + XMVECTOR VEqualsZero = XMVectorEqual(V, vdupq_n_f32(0)); + XMVECTOR Result = vmulq_f32(V, S3); + XMVECTOR Select = XMVectorEqualInt(VEqualsInfinity, VEqualsZero); + return XMVectorSelect(V, Result, Select); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_sqrt_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorReciprocalSqrtEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + 1.f / sqrtf(V.vector4_f32[0]), + 1.f / sqrtf(V.vector4_f32[1]), + 1.f / sqrtf(V.vector4_f32[2]), + 1.f / sqrtf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vrsqrteq_f32(V); +#elif defined(_XM_SSE_INTRINSICS_) + return _mm_rsqrt_ps(V); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorReciprocalSqrt(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + 1.f / sqrtf(V.vector4_f32[0]), + 1.f / sqrtf(V.vector4_f32[1]), + 1.f / sqrtf(V.vector4_f32[2]), + 1.f / sqrtf(V.vector4_f32[3]) + } } }; + return Result; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x4_t S0 = vrsqrteq_f32(V); + + float32x4_t P0 = vmulq_f32(V, S0); + float32x4_t R0 = vrsqrtsq_f32(P0, S0); + + float32x4_t S1 = vmulq_f32(S0, R0); + float32x4_t P1 = vmulq_f32(V, S1); + float32x4_t R1 = vrsqrtsq_f32(P1, S1); + + return vmulq_f32(S1, R1); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_sqrt_ps(V); + vResult = _mm_div_ps(g_XMOne, vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorExp2(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + exp2f(V.vector4_f32[0]), + exp2f(V.vector4_f32[1]), + exp2f(V.vector4_f32[2]), + exp2f(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t itrunc = vcvtq_s32_f32(V); + float32x4_t ftrunc = vcvtq_f32_s32(itrunc); + float32x4_t y = vsubq_f32(V, ftrunc); + + float32x4_t poly = vmlaq_f32(g_XMExpEst6, g_XMExpEst7, y); + poly = vmlaq_f32(g_XMExpEst5, poly, y); + poly = vmlaq_f32(g_XMExpEst4, poly, y); + poly = vmlaq_f32(g_XMExpEst3, poly, y); + poly = vmlaq_f32(g_XMExpEst2, poly, y); + poly = vmlaq_f32(g_XMExpEst1, poly, y); + poly = vmlaq_f32(g_XMOne, poly, y); + + int32x4_t biased = vaddq_s32(itrunc, g_XMExponentBias); + biased = vshlq_n_s32(biased, 23); + float32x4_t result0 = XMVectorDivide(vreinterpretq_f32_s32(biased), poly); + + biased = vaddq_s32(itrunc, g_XM253); + biased = vshlq_n_s32(biased, 23); + float32x4_t result1 = XMVectorDivide(vreinterpretq_f32_s32(biased), poly); + result1 = vmulq_f32(g_XMMinNormal.v, result1); + + // Use selection to handle the cases + // if (V is NaN) -> QNaN; + // else if (V sign bit set) + // if (V > -150) + // if (V.exponent < -126) -> result1 + // else -> result0 + // else -> +0 + // else + // if (V < 128) -> result0 + // else -> +inf + + uint32x4_t comp = vcltq_s32(vreinterpretq_s32_f32(V), g_XMBin128); + float32x4_t result2 = vbslq_f32(comp, result0, g_XMInfinity); + + comp = vcltq_s32(itrunc, g_XMSubnormalExponent); + float32x4_t result3 = vbslq_f32(comp, result1, result0); + + comp = vcltq_s32(vreinterpretq_s32_f32(V), g_XMBinNeg150); + float32x4_t result4 = vbslq_f32(comp, result3, g_XMZero); + + int32x4_t sign = vandq_s32(vreinterpretq_s32_f32(V), g_XMNegativeZero); + comp = vceqq_s32(sign, g_XMNegativeZero); + float32x4_t result5 = vbslq_f32(comp, result4, result2); + + int32x4_t t0 = vandq_s32(vreinterpretq_s32_f32(V), g_XMQNaNTest); + int32x4_t t1 = vandq_s32(vreinterpretq_s32_f32(V), g_XMInfinity); + t0 = vreinterpretq_s32_u32(vceqq_s32(t0, g_XMZero)); + t1 = vreinterpretq_s32_u32(vceqq_s32(t1, g_XMInfinity)); + int32x4_t isNaN = vbicq_s32(t1, t0); + + float32x4_t vResult = vbslq_f32(vreinterpretq_u32_s32(isNaN), g_XMQNaN, result5); + return vResult; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_exp2_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i itrunc = _mm_cvttps_epi32(V); + __m128 ftrunc = _mm_cvtepi32_ps(itrunc); + __m128 y = _mm_sub_ps(V, ftrunc); + + __m128 poly = XM_FMADD_PS(g_XMExpEst7, y, g_XMExpEst6); + poly = XM_FMADD_PS(poly, y, g_XMExpEst5); + poly = XM_FMADD_PS(poly, y, g_XMExpEst4); + poly = XM_FMADD_PS(poly, y, g_XMExpEst3); + poly = XM_FMADD_PS(poly, y, g_XMExpEst2); + poly = XM_FMADD_PS(poly, y, g_XMExpEst1); + poly = XM_FMADD_PS(poly, y, g_XMOne); + + __m128i biased = _mm_add_epi32(itrunc, g_XMExponentBias); + biased = _mm_slli_epi32(biased, 23); + __m128 result0 = _mm_div_ps(_mm_castsi128_ps(biased), poly); + + biased = _mm_add_epi32(itrunc, g_XM253); + biased = _mm_slli_epi32(biased, 23); + __m128 result1 = _mm_div_ps(_mm_castsi128_ps(biased), poly); + result1 = _mm_mul_ps(g_XMMinNormal.v, result1); + + // Use selection to handle the cases + // if (V is NaN) -> QNaN; + // else if (V sign bit set) + // if (V > -150) + // if (V.exponent < -126) -> result1 + // else -> result0 + // else -> +0 + // else + // if (V < 128) -> result0 + // else -> +inf + + __m128i comp = _mm_cmplt_epi32(_mm_castps_si128(V), g_XMBin128); + __m128i select0 = _mm_and_si128(comp, _mm_castps_si128(result0)); + __m128i select1 = _mm_andnot_si128(comp, g_XMInfinity); + __m128i result2 = _mm_or_si128(select0, select1); + + comp = _mm_cmplt_epi32(itrunc, g_XMSubnormalExponent); + select1 = _mm_and_si128(comp, _mm_castps_si128(result1)); + select0 = _mm_andnot_si128(comp, _mm_castps_si128(result0)); + __m128i result3 = _mm_or_si128(select0, select1); + + comp = _mm_cmplt_epi32(_mm_castps_si128(V), g_XMBinNeg150); + select0 = _mm_and_si128(comp, result3); + select1 = _mm_andnot_si128(comp, g_XMZero); + __m128i result4 = _mm_or_si128(select0, select1); + + __m128i sign = _mm_and_si128(_mm_castps_si128(V), g_XMNegativeZero); + comp = _mm_cmpeq_epi32(sign, g_XMNegativeZero); + select0 = _mm_and_si128(comp, result4); + select1 = _mm_andnot_si128(comp, result2); + __m128i result5 = _mm_or_si128(select0, select1); + + __m128i t0 = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i t1 = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + t0 = _mm_cmpeq_epi32(t0, g_XMZero); + t1 = _mm_cmpeq_epi32(t1, g_XMInfinity); + __m128i isNaN = _mm_andnot_si128(t0, t1); + + select0 = _mm_and_si128(isNaN, g_XMQNaN); + select1 = _mm_andnot_si128(isNaN, result5); + __m128i vResult = _mm_or_si128(select0, select1); + + return _mm_castsi128_ps(vResult); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorExp10(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + powf(10.0f, V.vector4_f32[0]), + powf(10.0f, V.vector4_f32[1]), + powf(10.0f, V.vector4_f32[2]), + powf(10.0f, V.vector4_f32[3]) + } } }; + return Result.v; + +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_exp10_ps(V); + return Result; +#else + // exp10(V) = exp2(vin*log2(10)) + XMVECTOR Vten = XMVectorMultiply(g_XMLg10, V); + return XMVectorExp2(Vten); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorExpE(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + expf(V.vector4_f32[0]), + expf(V.vector4_f32[1]), + expf(V.vector4_f32[2]), + expf(V.vector4_f32[3]) + } } }; + return Result.v; + +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_exp_ps(V); + return Result; +#else + // expE(V) = exp2(vin*log2(e)) + XMVECTOR Ve = XMVectorMultiply(g_XMLgE, V); + return XMVectorExp2(Ve); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorExp(FXMVECTOR V) noexcept +{ + return XMVectorExp2(V); +} + +//------------------------------------------------------------------------------ + +#if defined(_XM_SSE_INTRINSICS_) + +namespace MathInternal +{ + inline __m128i multi_sll_epi32(__m128i value, __m128i count) noexcept + { + __m128i v = _mm_shuffle_epi32(value, _MM_SHUFFLE(0, 0, 0, 0)); + __m128i c = _mm_shuffle_epi32(count, _MM_SHUFFLE(0, 0, 0, 0)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r0 = _mm_sll_epi32(v, c); + + v = _mm_shuffle_epi32(value, _MM_SHUFFLE(1, 1, 1, 1)); + c = _mm_shuffle_epi32(count, _MM_SHUFFLE(1, 1, 1, 1)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r1 = _mm_sll_epi32(v, c); + + v = _mm_shuffle_epi32(value, _MM_SHUFFLE(2, 2, 2, 2)); + c = _mm_shuffle_epi32(count, _MM_SHUFFLE(2, 2, 2, 2)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r2 = _mm_sll_epi32(v, c); + + v = _mm_shuffle_epi32(value, _MM_SHUFFLE(3, 3, 3, 3)); + c = _mm_shuffle_epi32(count, _MM_SHUFFLE(3, 3, 3, 3)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r3 = _mm_sll_epi32(v, c); + + // (r0,r0,r1,r1) + __m128 r01 = _mm_shuffle_ps(_mm_castsi128_ps(r0), _mm_castsi128_ps(r1), _MM_SHUFFLE(0, 0, 0, 0)); + // (r2,r2,r3,r3) + __m128 r23 = _mm_shuffle_ps(_mm_castsi128_ps(r2), _mm_castsi128_ps(r3), _MM_SHUFFLE(0, 0, 0, 0)); + // (r0,r1,r2,r3) + __m128 result = _mm_shuffle_ps(r01, r23, _MM_SHUFFLE(2, 0, 2, 0)); + return _mm_castps_si128(result); + } + + inline __m128i multi_srl_epi32(__m128i value, __m128i count) noexcept + { + __m128i v = _mm_shuffle_epi32(value, _MM_SHUFFLE(0, 0, 0, 0)); + __m128i c = _mm_shuffle_epi32(count, _MM_SHUFFLE(0, 0, 0, 0)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r0 = _mm_srl_epi32(v, c); + + v = _mm_shuffle_epi32(value, _MM_SHUFFLE(1, 1, 1, 1)); + c = _mm_shuffle_epi32(count, _MM_SHUFFLE(1, 1, 1, 1)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r1 = _mm_srl_epi32(v, c); + + v = _mm_shuffle_epi32(value, _MM_SHUFFLE(2, 2, 2, 2)); + c = _mm_shuffle_epi32(count, _MM_SHUFFLE(2, 2, 2, 2)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r2 = _mm_srl_epi32(v, c); + + v = _mm_shuffle_epi32(value, _MM_SHUFFLE(3, 3, 3, 3)); + c = _mm_shuffle_epi32(count, _MM_SHUFFLE(3, 3, 3, 3)); + c = _mm_and_si128(c, g_XMMaskX); + __m128i r3 = _mm_srl_epi32(v, c); + + // (r0,r0,r1,r1) + __m128 r01 = _mm_shuffle_ps(_mm_castsi128_ps(r0), _mm_castsi128_ps(r1), _MM_SHUFFLE(0, 0, 0, 0)); + // (r2,r2,r3,r3) + __m128 r23 = _mm_shuffle_ps(_mm_castsi128_ps(r2), _mm_castsi128_ps(r3), _MM_SHUFFLE(0, 0, 0, 0)); + // (r0,r1,r2,r3) + __m128 result = _mm_shuffle_ps(r01, r23, _MM_SHUFFLE(2, 0, 2, 0)); + return _mm_castps_si128(result); + } + + inline __m128i GetLeadingBit(const __m128i value) noexcept + { + static const XMVECTORI32 g_XM0000FFFF = { { { 0x0000FFFF, 0x0000FFFF, 0x0000FFFF, 0x0000FFFF } } }; + static const XMVECTORI32 g_XM000000FF = { { { 0x000000FF, 0x000000FF, 0x000000FF, 0x000000FF } } }; + static const XMVECTORI32 g_XM0000000F = { { { 0x0000000F, 0x0000000F, 0x0000000F, 0x0000000F } } }; + static const XMVECTORI32 g_XM00000003 = { { { 0x00000003, 0x00000003, 0x00000003, 0x00000003 } } }; + + __m128i v = value, r, c, b, s; + + c = _mm_cmpgt_epi32(v, g_XM0000FFFF); // c = (v > 0xFFFF) + b = _mm_srli_epi32(c, 31); // b = (c ? 1 : 0) + r = _mm_slli_epi32(b, 4); // r = (b << 4) + v = multi_srl_epi32(v, r); // v = (v >> r) + + c = _mm_cmpgt_epi32(v, g_XM000000FF); // c = (v > 0xFF) + b = _mm_srli_epi32(c, 31); // b = (c ? 1 : 0) + s = _mm_slli_epi32(b, 3); // s = (b << 3) + v = multi_srl_epi32(v, s); // v = (v >> s) + r = _mm_or_si128(r, s); // r = (r | s) + + c = _mm_cmpgt_epi32(v, g_XM0000000F); // c = (v > 0xF) + b = _mm_srli_epi32(c, 31); // b = (c ? 1 : 0) + s = _mm_slli_epi32(b, 2); // s = (b << 2) + v = multi_srl_epi32(v, s); // v = (v >> s) + r = _mm_or_si128(r, s); // r = (r | s) + + c = _mm_cmpgt_epi32(v, g_XM00000003); // c = (v > 0x3) + b = _mm_srli_epi32(c, 31); // b = (c ? 1 : 0) + s = _mm_slli_epi32(b, 1); // s = (b << 1) + v = multi_srl_epi32(v, s); // v = (v >> s) + r = _mm_or_si128(r, s); // r = (r | s) + + s = _mm_srli_epi32(v, 1); + r = _mm_or_si128(r, s); + return r; + } +} // namespace MathInternal + +#endif // _XM_SSE_INTRINSICS_ + +#if defined(_XM_ARM_NEON_INTRINSICS_) + +namespace MathInternal +{ + inline int32x4_t GetLeadingBit(const int32x4_t value) noexcept + { + static const XMVECTORI32 g_XM0000FFFF = { { { 0x0000FFFF, 0x0000FFFF, 0x0000FFFF, 0x0000FFFF } } }; + static const XMVECTORI32 g_XM000000FF = { { { 0x000000FF, 0x000000FF, 0x000000FF, 0x000000FF } } }; + static const XMVECTORI32 g_XM0000000F = { { { 0x0000000F, 0x0000000F, 0x0000000F, 0x0000000F } } }; + static const XMVECTORI32 g_XM00000003 = { { { 0x00000003, 0x00000003, 0x00000003, 0x00000003 } } }; + + uint32x4_t c = vcgtq_s32(value, g_XM0000FFFF); // c = (v > 0xFFFF) + int32x4_t b = vshrq_n_s32(vreinterpretq_s32_u32(c), 31); // b = (c ? 1 : 0) + int32x4_t r = vshlq_n_s32(b, 4); // r = (b << 4) + r = vnegq_s32(r); + int32x4_t v = vshlq_s32(value, r); // v = (v >> r) + + c = vcgtq_s32(v, g_XM000000FF); // c = (v > 0xFF) + b = vshrq_n_s32(vreinterpretq_s32_u32(c), 31); // b = (c ? 1 : 0) + int32x4_t s = vshlq_n_s32(b, 3); // s = (b << 3) + s = vnegq_s32(s); + v = vshlq_s32(v, s); // v = (v >> s) + r = vorrq_s32(r, s); // r = (r | s) + + c = vcgtq_s32(v, g_XM0000000F); // c = (v > 0xF) + b = vshrq_n_s32(vreinterpretq_s32_u32(c), 31); // b = (c ? 1 : 0) + s = vshlq_n_s32(b, 2); // s = (b << 2) + s = vnegq_s32(s); + v = vshlq_s32(v, s); // v = (v >> s) + r = vorrq_s32(r, s); // r = (r | s) + + c = vcgtq_s32(v, g_XM00000003); // c = (v > 0x3) + b = vshrq_n_s32(vreinterpretq_s32_u32(c), 31); // b = (c ? 1 : 0) + s = vshlq_n_s32(b, 1); // s = (b << 1) + s = vnegq_s32(s); + v = vshlq_s32(v, s); // v = (v >> s) + r = vorrq_s32(r, s); // r = (r | s) + + s = vshrq_n_s32(v, 1); + r = vorrq_s32(r, s); + return r; + } + +} // namespace MathInternal + +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLog2(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + log2f(V.vector4_f32[0]), + log2f(V.vector4_f32[1]), + log2f(V.vector4_f32[2]), + log2f(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t rawBiased = vandq_s32(vreinterpretq_s32_f32(V), g_XMInfinity); + int32x4_t trailing = vandq_s32(vreinterpretq_s32_f32(V), g_XMQNaNTest); + uint32x4_t isExponentZero = vceqq_s32(vreinterpretq_s32_f32(g_XMZero), rawBiased); + + // Compute exponent and significand for normals. + int32x4_t biased = vshrq_n_s32(rawBiased, 23); + int32x4_t exponentNor = vsubq_s32(biased, g_XMExponentBias); + int32x4_t trailingNor = trailing; + + // Compute exponent and significand for subnormals. + int32x4_t leading = MathInternal::GetLeadingBit(trailing); + int32x4_t shift = vsubq_s32(g_XMNumTrailing, leading); + int32x4_t exponentSub = vsubq_s32(g_XMSubnormalExponent, shift); + int32x4_t trailingSub = vshlq_s32(trailing, shift); + trailingSub = vandq_s32(trailingSub, g_XMQNaNTest); + int32x4_t e = vbslq_s32(isExponentZero, exponentSub, exponentNor); + int32x4_t t = vbslq_s32(isExponentZero, trailingSub, trailingNor); + + // Compute the approximation. + int32x4_t tmp = vorrq_s32(vreinterpretq_s32_f32(g_XMOne), t); + float32x4_t y = vsubq_f32(vreinterpretq_f32_s32(tmp), g_XMOne); + + float32x4_t log2 = vmlaq_f32(g_XMLogEst6, g_XMLogEst7, y); + log2 = vmlaq_f32(g_XMLogEst5, log2, y); + log2 = vmlaq_f32(g_XMLogEst4, log2, y); + log2 = vmlaq_f32(g_XMLogEst3, log2, y); + log2 = vmlaq_f32(g_XMLogEst2, log2, y); + log2 = vmlaq_f32(g_XMLogEst1, log2, y); + log2 = vmlaq_f32(g_XMLogEst0, log2, y); + log2 = vmlaq_f32(vcvtq_f32_s32(e), log2, y); + + // if (x is NaN) -> QNaN + // else if (V is positive) + // if (V is infinite) -> +inf + // else -> log2(V) + // else + // if (V is zero) -> -inf + // else -> -QNaN + + uint32x4_t isInfinite = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + isInfinite = vceqq_u32(isInfinite, g_XMInfinity); + + uint32x4_t isGreaterZero = vcgtq_f32(V, g_XMZero); + uint32x4_t isNotFinite = vcgtq_f32(V, g_XMInfinity); + uint32x4_t isPositive = vbicq_u32(isGreaterZero, isNotFinite); + + uint32x4_t isZero = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + isZero = vceqq_u32(isZero, g_XMZero); + + uint32x4_t t0 = vandq_u32(vreinterpretq_u32_f32(V), g_XMQNaNTest); + uint32x4_t t1 = vandq_u32(vreinterpretq_u32_f32(V), g_XMInfinity); + t0 = vceqq_u32(t0, g_XMZero); + t1 = vceqq_u32(t1, g_XMInfinity); + uint32x4_t isNaN = vbicq_u32(t1, t0); + + float32x4_t result = vbslq_f32(isInfinite, g_XMInfinity, log2); + float32x4_t tmp2 = vbslq_f32(isZero, g_XMNegInfinity, g_XMNegQNaN); + result = vbslq_f32(isPositive, result, tmp2); + result = vbslq_f32(isNaN, g_XMQNaN, result); + return result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_log2_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i rawBiased = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + __m128i trailing = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i isExponentZero = _mm_cmpeq_epi32(g_XMZero, rawBiased); + + // Compute exponent and significand for normals. + __m128i biased = _mm_srli_epi32(rawBiased, 23); + __m128i exponentNor = _mm_sub_epi32(biased, g_XMExponentBias); + __m128i trailingNor = trailing; + + // Compute exponent and significand for subnormals. + __m128i leading = MathInternal::GetLeadingBit(trailing); + __m128i shift = _mm_sub_epi32(g_XMNumTrailing, leading); + __m128i exponentSub = _mm_sub_epi32(g_XMSubnormalExponent, shift); + __m128i trailingSub = MathInternal::multi_sll_epi32(trailing, shift); + trailingSub = _mm_and_si128(trailingSub, g_XMQNaNTest); + + __m128i select0 = _mm_and_si128(isExponentZero, exponentSub); + __m128i select1 = _mm_andnot_si128(isExponentZero, exponentNor); + __m128i e = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isExponentZero, trailingSub); + select1 = _mm_andnot_si128(isExponentZero, trailingNor); + __m128i t = _mm_or_si128(select0, select1); + + // Compute the approximation. + __m128i tmp = _mm_or_si128(g_XMOne, t); + __m128 y = _mm_sub_ps(_mm_castsi128_ps(tmp), g_XMOne); + + __m128 log2 = XM_FMADD_PS(g_XMLogEst7, y, g_XMLogEst6); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst5); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst4); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst3); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst2); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst1); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst0); + log2 = XM_FMADD_PS(log2, y, _mm_cvtepi32_ps(e)); + + // if (x is NaN) -> QNaN + // else if (V is positive) + // if (V is infinite) -> +inf + // else -> log2(V) + // else + // if (V is zero) -> -inf + // else -> -QNaN + + __m128i isInfinite = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + isInfinite = _mm_cmpeq_epi32(isInfinite, g_XMInfinity); + + __m128i isGreaterZero = _mm_cmpgt_epi32(_mm_castps_si128(V), g_XMZero); + __m128i isNotFinite = _mm_cmpgt_epi32(_mm_castps_si128(V), g_XMInfinity); + __m128i isPositive = _mm_andnot_si128(isNotFinite, isGreaterZero); + + __m128i isZero = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + isZero = _mm_cmpeq_epi32(isZero, g_XMZero); + + __m128i t0 = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i t1 = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + t0 = _mm_cmpeq_epi32(t0, g_XMZero); + t1 = _mm_cmpeq_epi32(t1, g_XMInfinity); + __m128i isNaN = _mm_andnot_si128(t0, t1); + + select0 = _mm_and_si128(isInfinite, g_XMInfinity); + select1 = _mm_andnot_si128(isInfinite, _mm_castps_si128(log2)); + __m128i result = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isZero, g_XMNegInfinity); + select1 = _mm_andnot_si128(isZero, g_XMNegQNaN); + tmp = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isPositive, result); + select1 = _mm_andnot_si128(isPositive, tmp); + result = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isNaN, g_XMQNaN); + select1 = _mm_andnot_si128(isNaN, result); + result = _mm_or_si128(select0, select1); + + return _mm_castsi128_ps(result); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLog10(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + log10f(V.vector4_f32[0]), + log10f(V.vector4_f32[1]), + log10f(V.vector4_f32[2]), + log10f(V.vector4_f32[3]) + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t rawBiased = vandq_s32(vreinterpretq_s32_f32(V), g_XMInfinity); + int32x4_t trailing = vandq_s32(vreinterpretq_s32_f32(V), g_XMQNaNTest); + uint32x4_t isExponentZero = vceqq_s32(g_XMZero, rawBiased); + + // Compute exponent and significand for normals. + int32x4_t biased = vshrq_n_s32(rawBiased, 23); + int32x4_t exponentNor = vsubq_s32(biased, g_XMExponentBias); + int32x4_t trailingNor = trailing; + + // Compute exponent and significand for subnormals. + int32x4_t leading = MathInternal::GetLeadingBit(trailing); + int32x4_t shift = vsubq_s32(g_XMNumTrailing, leading); + int32x4_t exponentSub = vsubq_s32(g_XMSubnormalExponent, shift); + int32x4_t trailingSub = vshlq_s32(trailing, shift); + trailingSub = vandq_s32(trailingSub, g_XMQNaNTest); + int32x4_t e = vbslq_s32(isExponentZero, exponentSub, exponentNor); + int32x4_t t = vbslq_s32(isExponentZero, trailingSub, trailingNor); + + // Compute the approximation. + int32x4_t tmp = vorrq_s32(g_XMOne, t); + float32x4_t y = vsubq_f32(vreinterpretq_f32_s32(tmp), g_XMOne); + + float32x4_t log2 = vmlaq_f32(g_XMLogEst6, g_XMLogEst7, y); + log2 = vmlaq_f32(g_XMLogEst5, log2, y); + log2 = vmlaq_f32(g_XMLogEst4, log2, y); + log2 = vmlaq_f32(g_XMLogEst3, log2, y); + log2 = vmlaq_f32(g_XMLogEst2, log2, y); + log2 = vmlaq_f32(g_XMLogEst1, log2, y); + log2 = vmlaq_f32(g_XMLogEst0, log2, y); + log2 = vmlaq_f32(vcvtq_f32_s32(e), log2, y); + + log2 = vmulq_f32(g_XMInvLg10, log2); + + // if (x is NaN) -> QNaN + // else if (V is positive) + // if (V is infinite) -> +inf + // else -> log2(V) + // else + // if (V is zero) -> -inf + // else -> -QNaN + + uint32x4_t isInfinite = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + isInfinite = vceqq_u32(isInfinite, g_XMInfinity); + + uint32x4_t isGreaterZero = vcgtq_s32(vreinterpretq_s32_f32(V), g_XMZero); + uint32x4_t isNotFinite = vcgtq_s32(vreinterpretq_s32_f32(V), g_XMInfinity); + uint32x4_t isPositive = vbicq_u32(isGreaterZero, isNotFinite); + + uint32x4_t isZero = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + isZero = vceqq_u32(isZero, g_XMZero); + + uint32x4_t t0 = vandq_u32(vreinterpretq_u32_f32(V), g_XMQNaNTest); + uint32x4_t t1 = vandq_u32(vreinterpretq_u32_f32(V), g_XMInfinity); + t0 = vceqq_u32(t0, g_XMZero); + t1 = vceqq_u32(t1, g_XMInfinity); + uint32x4_t isNaN = vbicq_u32(t1, t0); + + float32x4_t result = vbslq_f32(isInfinite, g_XMInfinity, log2); + float32x4_t tmp2 = vbslq_f32(isZero, g_XMNegInfinity, g_XMNegQNaN); + result = vbslq_f32(isPositive, result, tmp2); + result = vbslq_f32(isNaN, g_XMQNaN, result); + return result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_log10_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i rawBiased = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + __m128i trailing = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i isExponentZero = _mm_cmpeq_epi32(g_XMZero, rawBiased); + + // Compute exponent and significand for normals. + __m128i biased = _mm_srli_epi32(rawBiased, 23); + __m128i exponentNor = _mm_sub_epi32(biased, g_XMExponentBias); + __m128i trailingNor = trailing; + + // Compute exponent and significand for subnormals. + __m128i leading = MathInternal::GetLeadingBit(trailing); + __m128i shift = _mm_sub_epi32(g_XMNumTrailing, leading); + __m128i exponentSub = _mm_sub_epi32(g_XMSubnormalExponent, shift); + __m128i trailingSub = MathInternal::multi_sll_epi32(trailing, shift); + trailingSub = _mm_and_si128(trailingSub, g_XMQNaNTest); + + __m128i select0 = _mm_and_si128(isExponentZero, exponentSub); + __m128i select1 = _mm_andnot_si128(isExponentZero, exponentNor); + __m128i e = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isExponentZero, trailingSub); + select1 = _mm_andnot_si128(isExponentZero, trailingNor); + __m128i t = _mm_or_si128(select0, select1); + + // Compute the approximation. + __m128i tmp = _mm_or_si128(g_XMOne, t); + __m128 y = _mm_sub_ps(_mm_castsi128_ps(tmp), g_XMOne); + + __m128 log2 = XM_FMADD_PS(g_XMLogEst7, y, g_XMLogEst6); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst5); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst4); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst3); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst2); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst1); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst0); + log2 = XM_FMADD_PS(log2, y, _mm_cvtepi32_ps(e)); + + log2 = _mm_mul_ps(g_XMInvLg10, log2); + + // if (x is NaN) -> QNaN + // else if (V is positive) + // if (V is infinite) -> +inf + // else -> log2(V) + // else + // if (V is zero) -> -inf + // else -> -QNaN + + __m128i isInfinite = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + isInfinite = _mm_cmpeq_epi32(isInfinite, g_XMInfinity); + + __m128i isGreaterZero = _mm_cmpgt_epi32(_mm_castps_si128(V), g_XMZero); + __m128i isNotFinite = _mm_cmpgt_epi32(_mm_castps_si128(V), g_XMInfinity); + __m128i isPositive = _mm_andnot_si128(isNotFinite, isGreaterZero); + + __m128i isZero = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + isZero = _mm_cmpeq_epi32(isZero, g_XMZero); + + __m128i t0 = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i t1 = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + t0 = _mm_cmpeq_epi32(t0, g_XMZero); + t1 = _mm_cmpeq_epi32(t1, g_XMInfinity); + __m128i isNaN = _mm_andnot_si128(t0, t1); + + select0 = _mm_and_si128(isInfinite, g_XMInfinity); + select1 = _mm_andnot_si128(isInfinite, _mm_castps_si128(log2)); + __m128i result = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isZero, g_XMNegInfinity); + select1 = _mm_andnot_si128(isZero, g_XMNegQNaN); + tmp = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isPositive, result); + select1 = _mm_andnot_si128(isPositive, tmp); + result = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isNaN, g_XMQNaN); + select1 = _mm_andnot_si128(isNaN, result); + result = _mm_or_si128(select0, select1); + + return _mm_castsi128_ps(result); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLogE(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + logf(V.vector4_f32[0]), + logf(V.vector4_f32[1]), + logf(V.vector4_f32[2]), + logf(V.vector4_f32[3]) + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int32x4_t rawBiased = vandq_s32(vreinterpretq_s32_f32(V), g_XMInfinity); + int32x4_t trailing = vandq_s32(vreinterpretq_s32_f32(V), g_XMQNaNTest); + uint32x4_t isExponentZero = vceqq_s32(g_XMZero, rawBiased); + + // Compute exponent and significand for normals. + int32x4_t biased = vshrq_n_s32(rawBiased, 23); + int32x4_t exponentNor = vsubq_s32(biased, g_XMExponentBias); + int32x4_t trailingNor = trailing; + + // Compute exponent and significand for subnormals. + int32x4_t leading = MathInternal::GetLeadingBit(trailing); + int32x4_t shift = vsubq_s32(g_XMNumTrailing, leading); + int32x4_t exponentSub = vsubq_s32(g_XMSubnormalExponent, shift); + int32x4_t trailingSub = vshlq_s32(trailing, shift); + trailingSub = vandq_s32(trailingSub, g_XMQNaNTest); + int32x4_t e = vbslq_s32(isExponentZero, exponentSub, exponentNor); + int32x4_t t = vbslq_s32(isExponentZero, trailingSub, trailingNor); + + // Compute the approximation. + int32x4_t tmp = vorrq_s32(g_XMOne, t); + float32x4_t y = vsubq_f32(vreinterpretq_f32_s32(tmp), g_XMOne); + + float32x4_t log2 = vmlaq_f32(g_XMLogEst6, g_XMLogEst7, y); + log2 = vmlaq_f32(g_XMLogEst5, log2, y); + log2 = vmlaq_f32(g_XMLogEst4, log2, y); + log2 = vmlaq_f32(g_XMLogEst3, log2, y); + log2 = vmlaq_f32(g_XMLogEst2, log2, y); + log2 = vmlaq_f32(g_XMLogEst1, log2, y); + log2 = vmlaq_f32(g_XMLogEst0, log2, y); + log2 = vmlaq_f32(vcvtq_f32_s32(e), log2, y); + + log2 = vmulq_f32(g_XMInvLgE, log2); + + // if (x is NaN) -> QNaN + // else if (V is positive) + // if (V is infinite) -> +inf + // else -> log2(V) + // else + // if (V is zero) -> -inf + // else -> -QNaN + + uint32x4_t isInfinite = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + isInfinite = vceqq_u32(isInfinite, g_XMInfinity); + + uint32x4_t isGreaterZero = vcgtq_s32(vreinterpretq_s32_f32(V), g_XMZero); + uint32x4_t isNotFinite = vcgtq_s32(vreinterpretq_s32_f32(V), g_XMInfinity); + uint32x4_t isPositive = vbicq_u32(isGreaterZero, isNotFinite); + + uint32x4_t isZero = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + isZero = vceqq_u32(isZero, g_XMZero); + + uint32x4_t t0 = vandq_u32(vreinterpretq_u32_f32(V), g_XMQNaNTest); + uint32x4_t t1 = vandq_u32(vreinterpretq_u32_f32(V), g_XMInfinity); + t0 = vceqq_u32(t0, g_XMZero); + t1 = vceqq_u32(t1, g_XMInfinity); + uint32x4_t isNaN = vbicq_u32(t1, t0); + + float32x4_t result = vbslq_f32(isInfinite, g_XMInfinity, log2); + float32x4_t tmp2 = vbslq_f32(isZero, g_XMNegInfinity, g_XMNegQNaN); + result = vbslq_f32(isPositive, result, tmp2); + result = vbslq_f32(isNaN, g_XMQNaN, result); + return result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_log_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i rawBiased = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + __m128i trailing = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i isExponentZero = _mm_cmpeq_epi32(g_XMZero, rawBiased); + + // Compute exponent and significand for normals. + __m128i biased = _mm_srli_epi32(rawBiased, 23); + __m128i exponentNor = _mm_sub_epi32(biased, g_XMExponentBias); + __m128i trailingNor = trailing; + + // Compute exponent and significand for subnormals. + __m128i leading = MathInternal::GetLeadingBit(trailing); + __m128i shift = _mm_sub_epi32(g_XMNumTrailing, leading); + __m128i exponentSub = _mm_sub_epi32(g_XMSubnormalExponent, shift); + __m128i trailingSub = MathInternal::multi_sll_epi32(trailing, shift); + trailingSub = _mm_and_si128(trailingSub, g_XMQNaNTest); + + __m128i select0 = _mm_and_si128(isExponentZero, exponentSub); + __m128i select1 = _mm_andnot_si128(isExponentZero, exponentNor); + __m128i e = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isExponentZero, trailingSub); + select1 = _mm_andnot_si128(isExponentZero, trailingNor); + __m128i t = _mm_or_si128(select0, select1); + + // Compute the approximation. + __m128i tmp = _mm_or_si128(g_XMOne, t); + __m128 y = _mm_sub_ps(_mm_castsi128_ps(tmp), g_XMOne); + + __m128 log2 = XM_FMADD_PS(g_XMLogEst7, y, g_XMLogEst6); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst5); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst4); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst3); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst2); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst1); + log2 = XM_FMADD_PS(log2, y, g_XMLogEst0); + log2 = XM_FMADD_PS(log2, y, _mm_cvtepi32_ps(e)); + + log2 = _mm_mul_ps(g_XMInvLgE, log2); + + // if (x is NaN) -> QNaN + // else if (V is positive) + // if (V is infinite) -> +inf + // else -> log2(V) + // else + // if (V is zero) -> -inf + // else -> -QNaN + + __m128i isInfinite = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + isInfinite = _mm_cmpeq_epi32(isInfinite, g_XMInfinity); + + __m128i isGreaterZero = _mm_cmpgt_epi32(_mm_castps_si128(V), g_XMZero); + __m128i isNotFinite = _mm_cmpgt_epi32(_mm_castps_si128(V), g_XMInfinity); + __m128i isPositive = _mm_andnot_si128(isNotFinite, isGreaterZero); + + __m128i isZero = _mm_and_si128(_mm_castps_si128(V), g_XMAbsMask); + isZero = _mm_cmpeq_epi32(isZero, g_XMZero); + + __m128i t0 = _mm_and_si128(_mm_castps_si128(V), g_XMQNaNTest); + __m128i t1 = _mm_and_si128(_mm_castps_si128(V), g_XMInfinity); + t0 = _mm_cmpeq_epi32(t0, g_XMZero); + t1 = _mm_cmpeq_epi32(t1, g_XMInfinity); + __m128i isNaN = _mm_andnot_si128(t0, t1); + + select0 = _mm_and_si128(isInfinite, g_XMInfinity); + select1 = _mm_andnot_si128(isInfinite, _mm_castps_si128(log2)); + __m128i result = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isZero, g_XMNegInfinity); + select1 = _mm_andnot_si128(isZero, g_XMNegQNaN); + tmp = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isPositive, result); + select1 = _mm_andnot_si128(isPositive, tmp); + result = _mm_or_si128(select0, select1); + + select0 = _mm_and_si128(isNaN, g_XMQNaN); + select1 = _mm_andnot_si128(isNaN, result); + result = _mm_or_si128(select0, select1); + + return _mm_castsi128_ps(result); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLog(FXMVECTOR V) noexcept +{ + return XMVectorLog2(V); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorPow +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + powf(V1.vector4_f32[0], V2.vector4_f32[0]), + powf(V1.vector4_f32[1], V2.vector4_f32[1]), + powf(V1.vector4_f32[2], V2.vector4_f32[2]), + powf(V1.vector4_f32[3], V2.vector4_f32[3]) + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTORF32 vResult = { { { + powf(vgetq_lane_f32(V1, 0), vgetq_lane_f32(V2, 0)), + powf(vgetq_lane_f32(V1, 1), vgetq_lane_f32(V2, 1)), + powf(vgetq_lane_f32(V1, 2), vgetq_lane_f32(V2, 2)), + powf(vgetq_lane_f32(V1, 3), vgetq_lane_f32(V2, 3)) + } } }; + return vResult.v; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_pow_ps(V1, V2); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + XM_ALIGNED_DATA(16) float a[4]; + XM_ALIGNED_DATA(16) float b[4]; + _mm_store_ps(a, V1); + _mm_store_ps(b, V2); + XMVECTOR vResult = _mm_setr_ps( + powf(a[0], b[0]), + powf(a[1], b[1]), + powf(a[2], b[2]), + powf(a[3], b[3])); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorAbs(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + fabsf(V.vector4_f32[0]), + fabsf(V.vector4_f32[1]), + fabsf(V.vector4_f32[2]), + fabsf(V.vector4_f32[3]) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + return vabsq_f32(V); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_setzero_ps(); + vResult = _mm_sub_ps(vResult, V); + vResult = _mm_max_ps(vResult, V); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorMod +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + // V1 % V2 = V1 - V2 * truncate(V1 / V2) + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Quotient = XMVectorDivide(V1, V2); + Quotient = XMVectorTruncate(Quotient); + XMVECTOR Result = XMVectorNegativeMultiplySubtract(V2, Quotient, V1); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR vResult = XMVectorDivide(V1, V2); + vResult = XMVectorTruncate(vResult); + return vmlsq_f32(V1, vResult, V2); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_div_ps(V1, V2); + vResult = XMVectorTruncate(vResult); + return XM_FNMADD_PS(vResult, V2, V1); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorModAngles(FXMVECTOR Angles) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR V; + XMVECTOR Result; + + // Modulo the range of the given angles such that -XM_PI <= Angles < XM_PI + V = XMVectorMultiply(Angles, g_XMReciprocalTwoPi.v); + V = XMVectorRound(V); + Result = XMVectorNegativeMultiplySubtract(g_XMTwoPi.v, V, Angles); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Modulo the range of the given angles such that -XM_PI <= Angles < XM_PI + XMVECTOR vResult = vmulq_f32(Angles, g_XMReciprocalTwoPi); + // Use the inline function due to complexity for rounding + vResult = XMVectorRound(vResult); + return vmlsq_f32(Angles, vResult, g_XMTwoPi); +#elif defined(_XM_SSE_INTRINSICS_) + // Modulo the range of the given angles such that -XM_PI <= Angles < XM_PI + XMVECTOR vResult = _mm_mul_ps(Angles, g_XMReciprocalTwoPi); + // Use the inline function due to complexity for rounding + vResult = XMVectorRound(vResult); + return XM_FNMADD_PS(vResult, g_XMTwoPi, Angles); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSin(FXMVECTOR V) noexcept +{ + // 11-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + sinf(V.vector4_f32[0]), + sinf(V.vector4_f32[1]), + sinf(V.vector4_f32[2]), + sinf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with sin(y) = sin(x). + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(x), g_XMNegativeZero); + uint32x4_t c = vorrq_u32(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + float32x4_t absx = vabsq_f32(x); + float32x4_t rflx = vsubq_f32(vreinterpretq_f32_u32(c), x); + uint32x4_t comp = vcleq_f32(absx, g_XMHalfPi); + x = vbslq_f32(comp, x, rflx); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation + const XMVECTOR SC1 = g_XMSinCoefficients1; + const XMVECTOR SC0 = g_XMSinCoefficients0; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(SC0), 1); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_low_f32(SC1), 0); + + vConstants = vdupq_lane_f32(vget_high_f32(SC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(SC0), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(SC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + Result = vmulq_f32(Result, x); + return Result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_sin_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with sin(y) = sin(x). + __m128 sign = _mm_and_ps(x, g_XMNegativeZero); + __m128 c = _mm_or_ps(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + __m128 absx = _mm_andnot_ps(sign, x); // |x| + __m128 rflx = _mm_sub_ps(c, x); + __m128 comp = _mm_cmple_ps(absx, g_XMHalfPi); + __m128 select0 = _mm_and_ps(comp, x); + __m128 select1 = _mm_andnot_ps(comp, rflx); + x = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation + const XMVECTOR SC1 = g_XMSinCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(SC1, _MM_SHUFFLE(0, 0, 0, 0)); + const XMVECTOR SC0 = g_XMSinCoefficients0; + __m128 vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(2, 2, 2, 2)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, x); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorCos(FXMVECTOR V) noexcept +{ + // 10-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + cosf(V.vector4_f32[0]), + cosf(V.vector4_f32[1]), + cosf(V.vector4_f32[2]), + cosf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Map V to x in [-pi,pi]. + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with cos(y) = sign*cos(x). + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(x), g_XMNegativeZero); + uint32x4_t c = vorrq_u32(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + float32x4_t absx = vabsq_f32(x); + float32x4_t rflx = vsubq_f32(vreinterpretq_f32_u32(c), x); + uint32x4_t comp = vcleq_f32(absx, g_XMHalfPi); + x = vbslq_f32(comp, x, rflx); + float32x4_t fsign = vbslq_f32(comp, g_XMOne, g_XMNegativeOne); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation + const XMVECTOR CC1 = g_XMCosCoefficients1; + const XMVECTOR CC0 = g_XMCosCoefficients0; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(CC0), 1); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_low_f32(CC1), 0); + + vConstants = vdupq_lane_f32(vget_high_f32(CC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(CC0), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(CC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + Result = vmulq_f32(Result, fsign); + return Result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_cos_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + // Map V to x in [-pi,pi]. + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with cos(y) = sign*cos(x). + XMVECTOR sign = _mm_and_ps(x, g_XMNegativeZero); + __m128 c = _mm_or_ps(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + __m128 absx = _mm_andnot_ps(sign, x); // |x| + __m128 rflx = _mm_sub_ps(c, x); + __m128 comp = _mm_cmple_ps(absx, g_XMHalfPi); + __m128 select0 = _mm_and_ps(comp, x); + __m128 select1 = _mm_andnot_ps(comp, rflx); + x = _mm_or_ps(select0, select1); + select0 = _mm_and_ps(comp, g_XMOne); + select1 = _mm_andnot_ps(comp, g_XMNegativeOne); + sign = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation + const XMVECTOR CC1 = g_XMCosCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(CC1, _MM_SHUFFLE(0, 0, 0, 0)); + const XMVECTOR CC0 = g_XMCosCoefficients0; + __m128 vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(2, 2, 2, 2)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, sign); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorSinCos +( + XMVECTOR* pSin, + XMVECTOR* pCos, + FXMVECTOR V +) noexcept +{ + assert(pSin != nullptr); + assert(pCos != nullptr); + + // 11/10-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Sin = { { { + sinf(V.vector4_f32[0]), + sinf(V.vector4_f32[1]), + sinf(V.vector4_f32[2]), + sinf(V.vector4_f32[3]) + } } }; + + XMVECTORF32 Cos = { { { + cosf(V.vector4_f32[0]), + cosf(V.vector4_f32[1]), + cosf(V.vector4_f32[2]), + cosf(V.vector4_f32[3]) + } } }; + + *pSin = Sin.v; + *pCos = Cos.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with cos(y) = sign*cos(x). + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(x), g_XMNegativeZero); + uint32x4_t c = vorrq_u32(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + float32x4_t absx = vabsq_f32(x); + float32x4_t rflx = vsubq_f32(vreinterpretq_f32_u32(c), x); + uint32x4_t comp = vcleq_f32(absx, g_XMHalfPi); + x = vbslq_f32(comp, x, rflx); + float32x4_t fsign = vbslq_f32(comp, g_XMOne, g_XMNegativeOne); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation for sine + const XMVECTOR SC1 = g_XMSinCoefficients1; + const XMVECTOR SC0 = g_XMSinCoefficients0; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(SC0), 1); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_low_f32(SC1), 0); + + vConstants = vdupq_lane_f32(vget_high_f32(SC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(SC0), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(SC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + *pSin = vmulq_f32(Result, x); + + // Compute polynomial approximation for cosine + const XMVECTOR CC1 = g_XMCosCoefficients1; + const XMVECTOR CC0 = g_XMCosCoefficients0; + vConstants = vdupq_lane_f32(vget_high_f32(CC0), 1); + Result = vmlaq_lane_f32(vConstants, x2, vget_low_f32(CC1), 0); + + vConstants = vdupq_lane_f32(vget_high_f32(CC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(CC0), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(CC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + *pCos = vmulq_f32(Result, fsign); +#elif defined(_XM_SVML_INTRINSICS_) + *pSin = _mm_sincos_ps(pCos, V); +#elif defined(_XM_SSE_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with sin(y) = sin(x), cos(y) = sign*cos(x). + XMVECTOR sign = _mm_and_ps(x, g_XMNegativeZero); + __m128 c = _mm_or_ps(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + __m128 absx = _mm_andnot_ps(sign, x); // |x| + __m128 rflx = _mm_sub_ps(c, x); + __m128 comp = _mm_cmple_ps(absx, g_XMHalfPi); + __m128 select0 = _mm_and_ps(comp, x); + __m128 select1 = _mm_andnot_ps(comp, rflx); + x = _mm_or_ps(select0, select1); + select0 = _mm_and_ps(comp, g_XMOne); + select1 = _mm_andnot_ps(comp, g_XMNegativeOne); + sign = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation of sine + const XMVECTOR SC1 = g_XMSinCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(SC1, _MM_SHUFFLE(0, 0, 0, 0)); + const XMVECTOR SC0 = g_XMSinCoefficients0; + __m128 vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(2, 2, 2, 2)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SC0, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, x); + *pSin = Result; + + // Compute polynomial approximation of cosine + const XMVECTOR CC1 = g_XMCosCoefficients1; + vConstantsB = XM_PERMUTE_PS(CC1, _MM_SHUFFLE(0, 0, 0, 0)); + const XMVECTOR CC0 = g_XMCosCoefficients0; + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(3, 3, 3, 3)); + Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(2, 2, 2, 2)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CC0, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, sign); + *pCos = Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorTan(FXMVECTOR V) noexcept +{ + // Cody and Waite algorithm to compute tangent. + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + tanf(V.vector4_f32[0]), + tanf(V.vector4_f32[1]), + tanf(V.vector4_f32[2]), + tanf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_tan_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + static const XMVECTORF32 TanCoefficients0 = { { { 1.0f, -4.667168334e-1f, 2.566383229e-2f, -3.118153191e-4f } } }; + static const XMVECTORF32 TanCoefficients1 = { { { 4.981943399e-7f, -1.333835001e-1f, 3.424887824e-3f, -1.786170734e-5f } } }; + static const XMVECTORF32 TanConstants = { { { 1.570796371f, 6.077100628e-11f, 0.000244140625f, 0.63661977228f /*2 / Pi*/ } } }; + static const XMVECTORU32 Mask = { { { 0x1, 0x1, 0x1, 0x1 } } }; + + XMVECTOR TwoDivPi = XMVectorSplatW(TanConstants.v); + + XMVECTOR Zero = XMVectorZero(); + + XMVECTOR C0 = XMVectorSplatX(TanConstants.v); + XMVECTOR C1 = XMVectorSplatY(TanConstants.v); + XMVECTOR Epsilon = XMVectorSplatZ(TanConstants.v); + + XMVECTOR VA = XMVectorMultiply(V, TwoDivPi); + + VA = XMVectorRound(VA); + + XMVECTOR VC = XMVectorNegativeMultiplySubtract(VA, C0, V); + + XMVECTOR VB = XMVectorAbs(VA); + + VC = XMVectorNegativeMultiplySubtract(VA, C1, VC); + +#if defined(_XM_ARM_NEON_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + VB = vreinterpretq_f32_u32(vcvtq_u32_f32(VB)); +#elif defined(_XM_SSE_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + reinterpret_cast<__m128i*>(&VB)[0] = _mm_cvttps_epi32(VB); +#else + for (size_t i = 0; i < 4; i++) + { + VB.vector4_u32[i] = static_cast(VB.vector4_f32[i]); + } +#endif + + XMVECTOR VC2 = XMVectorMultiply(VC, VC); + + XMVECTOR T7 = XMVectorSplatW(TanCoefficients1.v); + XMVECTOR T6 = XMVectorSplatZ(TanCoefficients1.v); + XMVECTOR T4 = XMVectorSplatX(TanCoefficients1.v); + XMVECTOR T3 = XMVectorSplatW(TanCoefficients0.v); + XMVECTOR T5 = XMVectorSplatY(TanCoefficients1.v); + XMVECTOR T2 = XMVectorSplatZ(TanCoefficients0.v); + XMVECTOR T1 = XMVectorSplatY(TanCoefficients0.v); + XMVECTOR T0 = XMVectorSplatX(TanCoefficients0.v); + + XMVECTOR VBIsEven = XMVectorAndInt(VB, Mask.v); + VBIsEven = XMVectorEqualInt(VBIsEven, Zero); + + XMVECTOR N = XMVectorMultiplyAdd(VC2, T7, T6); + XMVECTOR D = XMVectorMultiplyAdd(VC2, T4, T3); + N = XMVectorMultiplyAdd(VC2, N, T5); + D = XMVectorMultiplyAdd(VC2, D, T2); + N = XMVectorMultiply(VC2, N); + D = XMVectorMultiplyAdd(VC2, D, T1); + N = XMVectorMultiplyAdd(VC, N, VC); + XMVECTOR VCNearZero = XMVectorInBounds(VC, Epsilon); + D = XMVectorMultiplyAdd(VC2, D, T0); + + N = XMVectorSelect(N, VC, VCNearZero); + D = XMVectorSelect(D, g_XMOne.v, VCNearZero); + + XMVECTOR R0 = XMVectorNegate(N); + XMVECTOR R1 = XMVectorDivide(N, D); + R0 = XMVectorDivide(D, R0); + + XMVECTOR VIsZero = XMVectorEqual(V, Zero); + + XMVECTOR Result = XMVectorSelect(R0, R1, VBIsEven); + + Result = XMVectorSelect(Result, Zero, VIsZero); + + return Result; + +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSinH(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + sinhf(V.vector4_f32[0]), + sinhf(V.vector4_f32[1]), + sinhf(V.vector4_f32[2]), + sinhf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.442695040888963f, 1.442695040888963f, 1.442695040888963f, 1.442695040888963f } } }; // 1.0f / ln(2.0f) + + XMVECTOR V1 = vmlaq_f32(g_XMNegativeOne.v, V, Scale.v); + XMVECTOR V2 = vmlsq_f32(g_XMNegativeOne.v, V, Scale.v); + XMVECTOR E1 = XMVectorExp(V1); + XMVECTOR E2 = XMVectorExp(V2); + + return vsubq_f32(E1, E2); +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_sinh_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.442695040888963f, 1.442695040888963f, 1.442695040888963f, 1.442695040888963f } } }; // 1.0f / ln(2.0f) + + XMVECTOR V1 = XM_FMADD_PS(V, Scale, g_XMNegativeOne); + XMVECTOR V2 = XM_FNMADD_PS(V, Scale, g_XMNegativeOne); + XMVECTOR E1 = XMVectorExp(V1); + XMVECTOR E2 = XMVectorExp(V2); + + return _mm_sub_ps(E1, E2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorCosH(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + coshf(V.vector4_f32[0]), + coshf(V.vector4_f32[1]), + coshf(V.vector4_f32[2]), + coshf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.442695040888963f, 1.442695040888963f, 1.442695040888963f, 1.442695040888963f } } }; // 1.0f / ln(2.0f) + + XMVECTOR V1 = vmlaq_f32(g_XMNegativeOne.v, V, Scale.v); + XMVECTOR V2 = vmlsq_f32(g_XMNegativeOne.v, V, Scale.v); + XMVECTOR E1 = XMVectorExp(V1); + XMVECTOR E2 = XMVectorExp(V2); + return vaddq_f32(E1, E2); +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_cosh_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.442695040888963f, 1.442695040888963f, 1.442695040888963f, 1.442695040888963f } } }; // 1.0f / ln(2.0f) + + XMVECTOR V1 = XM_FMADD_PS(V, Scale.v, g_XMNegativeOne.v); + XMVECTOR V2 = XM_FNMADD_PS(V, Scale.v, g_XMNegativeOne.v); + XMVECTOR E1 = XMVectorExp(V1); + XMVECTOR E2 = XMVectorExp(V2); + return _mm_add_ps(E1, E2); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorTanH(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + tanhf(V.vector4_f32[0]), + tanhf(V.vector4_f32[1]), + tanhf(V.vector4_f32[2]), + tanhf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 2.8853900817779268f, 2.8853900817779268f, 2.8853900817779268f, 2.8853900817779268f } } }; // 2.0f / ln(2.0f) + + XMVECTOR E = vmulq_f32(V, Scale.v); + E = XMVectorExp(E); + E = vmlaq_f32(g_XMOneHalf.v, E, g_XMOneHalf.v); + E = XMVectorReciprocal(E); + return vsubq_f32(g_XMOne.v, E); +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_tanh_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 2.8853900817779268f, 2.8853900817779268f, 2.8853900817779268f, 2.8853900817779268f } } }; // 2.0f / ln(2.0f) + + XMVECTOR E = _mm_mul_ps(V, Scale.v); + E = XMVectorExp(E); + E = XM_FMADD_PS(E, g_XMOneHalf.v, g_XMOneHalf.v); + E = _mm_div_ps(g_XMOne.v, E); + return _mm_sub_ps(g_XMOne.v, E); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorASin(FXMVECTOR V) noexcept +{ + // 7-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + asinf(V.vector4_f32[0]), + asinf(V.vector4_f32[1]), + asinf(V.vector4_f32[2]), + asinf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t nonnegative = vcgeq_f32(V, g_XMZero); + float32x4_t x = vabsq_f32(V); + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + float32x4_t oneMValue = vsubq_f32(g_XMOne, x); + float32x4_t clampOneMValue = vmaxq_f32(g_XMZero, oneMValue); + float32x4_t root = XMVectorSqrt(clampOneMValue); + + // Compute polynomial approximation + const XMVECTOR AC1 = g_XMArcCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(AC1), 0); + XMVECTOR t0 = vmlaq_lane_f32(vConstants, x, vget_high_f32(AC1), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(AC1), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AC1), 0); + t0 = vmlaq_f32(vConstants, t0, x); + + const XMVECTOR AC0 = g_XMArcCoefficients0; + vConstants = vdupq_lane_f32(vget_high_f32(AC0), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_high_f32(AC0), 0); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AC0), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AC0), 0); + t0 = vmlaq_f32(vConstants, t0, x); + t0 = vmulq_f32(t0, root); + + float32x4_t t1 = vsubq_f32(g_XMPi, t0); + t0 = vbslq_f32(nonnegative, t0, t1); + t0 = vsubq_f32(g_XMHalfPi, t0); + return t0; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_asin_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 nonnegative = _mm_cmpge_ps(V, g_XMZero); + __m128 mvalue = _mm_sub_ps(g_XMZero, V); + __m128 x = _mm_max_ps(V, mvalue); // |V| + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + __m128 oneMValue = _mm_sub_ps(g_XMOne, x); + __m128 clampOneMValue = _mm_max_ps(g_XMZero, oneMValue); + __m128 root = _mm_sqrt_ps(clampOneMValue); // sqrt(1-|V|) + + // Compute polynomial approximation + const XMVECTOR AC1 = g_XMArcCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 t0 = XM_FMADD_PS(vConstantsB, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(1, 1, 1, 1)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(0, 0, 0, 0)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + const XMVECTOR AC0 = g_XMArcCoefficients0; + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(3, 3, 3, 3)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(2, 2, 2, 2)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(1, 1, 1, 1)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(0, 0, 0, 0)); + t0 = XM_FMADD_PS(t0, x, vConstants); + t0 = _mm_mul_ps(t0, root); + + __m128 t1 = _mm_sub_ps(g_XMPi, t0); + t0 = _mm_and_ps(nonnegative, t0); + t1 = _mm_andnot_ps(nonnegative, t1); + t0 = _mm_or_ps(t0, t1); + t0 = _mm_sub_ps(g_XMHalfPi, t0); + return t0; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorACos(FXMVECTOR V) noexcept +{ + // 7-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + acosf(V.vector4_f32[0]), + acosf(V.vector4_f32[1]), + acosf(V.vector4_f32[2]), + acosf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t nonnegative = vcgeq_f32(V, g_XMZero); + float32x4_t x = vabsq_f32(V); + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + float32x4_t oneMValue = vsubq_f32(g_XMOne, x); + float32x4_t clampOneMValue = vmaxq_f32(g_XMZero, oneMValue); + float32x4_t root = XMVectorSqrt(clampOneMValue); + + // Compute polynomial approximation + const XMVECTOR AC1 = g_XMArcCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(AC1), 0); + XMVECTOR t0 = vmlaq_lane_f32(vConstants, x, vget_high_f32(AC1), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(AC1), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AC1), 0); + t0 = vmlaq_f32(vConstants, t0, x); + + const XMVECTOR AC0 = g_XMArcCoefficients0; + vConstants = vdupq_lane_f32(vget_high_f32(AC0), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_high_f32(AC0), 0); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AC0), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AC0), 0); + t0 = vmlaq_f32(vConstants, t0, x); + t0 = vmulq_f32(t0, root); + + float32x4_t t1 = vsubq_f32(g_XMPi, t0); + t0 = vbslq_f32(nonnegative, t0, t1); + return t0; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_acos_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 nonnegative = _mm_cmpge_ps(V, g_XMZero); + __m128 mvalue = _mm_sub_ps(g_XMZero, V); + __m128 x = _mm_max_ps(V, mvalue); // |V| + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + __m128 oneMValue = _mm_sub_ps(g_XMOne, x); + __m128 clampOneMValue = _mm_max_ps(g_XMZero, oneMValue); + __m128 root = _mm_sqrt_ps(clampOneMValue); // sqrt(1-|V|) + + // Compute polynomial approximation + const XMVECTOR AC1 = g_XMArcCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 t0 = XM_FMADD_PS(vConstantsB, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(1, 1, 1, 1)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC1, _MM_SHUFFLE(0, 0, 0, 0)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + const XMVECTOR AC0 = g_XMArcCoefficients0; + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(3, 3, 3, 3)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(2, 2, 2, 2)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(1, 1, 1, 1)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AC0, _MM_SHUFFLE(0, 0, 0, 0)); + t0 = XM_FMADD_PS(t0, x, vConstants); + t0 = _mm_mul_ps(t0, root); + + __m128 t1 = _mm_sub_ps(g_XMPi, t0); + t0 = _mm_and_ps(nonnegative, t0); + t1 = _mm_andnot_ps(nonnegative, t1); + t0 = _mm_or_ps(t0, t1); + return t0; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorATan(FXMVECTOR V) noexcept +{ + // 17-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + atanf(V.vector4_f32[0]), + atanf(V.vector4_f32[1]), + atanf(V.vector4_f32[2]), + atanf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t absV = vabsq_f32(V); + float32x4_t invV = XMVectorReciprocal(V); + uint32x4_t comp = vcgtq_f32(V, g_XMOne); + float32x4_t sign = vbslq_f32(comp, g_XMOne, g_XMNegativeOne); + comp = vcleq_f32(absV, g_XMOne); + sign = vbslq_f32(comp, g_XMZero, sign); + float32x4_t x = vbslq_f32(comp, V, invV); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation + const XMVECTOR TC1 = g_XMATanCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(TC1), 0); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_high_f32(TC1), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(TC1), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(TC1), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + const XMVECTOR TC0 = g_XMATanCoefficients0; + vConstants = vdupq_lane_f32(vget_high_f32(TC0), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_high_f32(TC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(TC0), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(TC0), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + Result = vmulq_f32(Result, x); + + float32x4_t result1 = vmulq_f32(sign, g_XMHalfPi); + result1 = vsubq_f32(result1, Result); + + comp = vceqq_f32(sign, g_XMZero); + Result = vbslq_f32(comp, Result, result1); + return Result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_atan_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 absV = XMVectorAbs(V); + __m128 invV = _mm_div_ps(g_XMOne, V); + __m128 comp = _mm_cmpgt_ps(V, g_XMOne); + __m128 select0 = _mm_and_ps(comp, g_XMOne); + __m128 select1 = _mm_andnot_ps(comp, g_XMNegativeOne); + __m128 sign = _mm_or_ps(select0, select1); + comp = _mm_cmple_ps(absV, g_XMOne); + select0 = _mm_and_ps(comp, g_XMZero); + select1 = _mm_andnot_ps(comp, sign); + sign = _mm_or_ps(select0, select1); + select0 = _mm_and_ps(comp, V); + select1 = _mm_andnot_ps(comp, invV); + __m128 x = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation + const XMVECTOR TC1 = g_XMATanCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(TC1, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(TC1, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(TC1, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(TC1, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + const XMVECTOR TC0 = g_XMATanCoefficients0; + vConstants = XM_PERMUTE_PS(TC0, _MM_SHUFFLE(3, 3, 3, 3)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(TC0, _MM_SHUFFLE(2, 2, 2, 2)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(TC0, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(TC0, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + Result = XM_FMADD_PS(Result, x2, g_XMOne); + + Result = _mm_mul_ps(Result, x); + __m128 result1 = _mm_mul_ps(sign, g_XMHalfPi); + result1 = _mm_sub_ps(result1, Result); + + comp = _mm_cmpeq_ps(sign, g_XMZero); + select0 = _mm_and_ps(comp, Result); + select1 = _mm_andnot_ps(comp, result1); + Result = _mm_or_ps(select0, select1); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorATan2 +( + FXMVECTOR Y, + FXMVECTOR X +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + atan2f(Y.vector4_f32[0], X.vector4_f32[0]), + atan2f(Y.vector4_f32[1], X.vector4_f32[1]), + atan2f(Y.vector4_f32[2], X.vector4_f32[2]), + atan2f(Y.vector4_f32[3], X.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_atan2_ps(Y, X); + return Result; +#else + + // Return the inverse tangent of Y / X in the range of -Pi to Pi with the following exceptions: + + // Y == 0 and X is Negative -> Pi with the sign of Y + // y == 0 and x is positive -> 0 with the sign of y + // Y != 0 and X == 0 -> Pi / 2 with the sign of Y + // Y != 0 and X is Negative -> atan(y/x) + (PI with the sign of Y) + // X == -Infinity and Finite Y -> Pi with the sign of Y + // X == +Infinity and Finite Y -> 0 with the sign of Y + // Y == Infinity and X is Finite -> Pi / 2 with the sign of Y + // Y == Infinity and X == -Infinity -> 3Pi / 4 with the sign of Y + // Y == Infinity and X == +Infinity -> Pi / 4 with the sign of Y + + static const XMVECTORF32 ATan2Constants = { { { XM_PI, XM_PIDIV2, XM_PIDIV4, XM_PI * 3.0f / 4.0f } } }; + + XMVECTOR Zero = XMVectorZero(); + XMVECTOR ATanResultValid = XMVectorTrueInt(); + + XMVECTOR Pi = XMVectorSplatX(ATan2Constants); + XMVECTOR PiOverTwo = XMVectorSplatY(ATan2Constants); + XMVECTOR PiOverFour = XMVectorSplatZ(ATan2Constants); + XMVECTOR ThreePiOverFour = XMVectorSplatW(ATan2Constants); + + XMVECTOR YEqualsZero = XMVectorEqual(Y, Zero); + XMVECTOR XEqualsZero = XMVectorEqual(X, Zero); + XMVECTOR XIsPositive = XMVectorAndInt(X, g_XMNegativeZero.v); + XIsPositive = XMVectorEqualInt(XIsPositive, Zero); + XMVECTOR YEqualsInfinity = XMVectorIsInfinite(Y); + XMVECTOR XEqualsInfinity = XMVectorIsInfinite(X); + + XMVECTOR YSign = XMVectorAndInt(Y, g_XMNegativeZero.v); + Pi = XMVectorOrInt(Pi, YSign); + PiOverTwo = XMVectorOrInt(PiOverTwo, YSign); + PiOverFour = XMVectorOrInt(PiOverFour, YSign); + ThreePiOverFour = XMVectorOrInt(ThreePiOverFour, YSign); + + XMVECTOR R1 = XMVectorSelect(Pi, YSign, XIsPositive); + XMVECTOR R2 = XMVectorSelect(ATanResultValid, PiOverTwo, XEqualsZero); + XMVECTOR R3 = XMVectorSelect(R2, R1, YEqualsZero); + XMVECTOR R4 = XMVectorSelect(ThreePiOverFour, PiOverFour, XIsPositive); + XMVECTOR R5 = XMVectorSelect(PiOverTwo, R4, XEqualsInfinity); + XMVECTOR Result = XMVectorSelect(R3, R5, YEqualsInfinity); + ATanResultValid = XMVectorEqualInt(Result, ATanResultValid); + + XMVECTOR V = XMVectorDivide(Y, X); + + XMVECTOR R0 = XMVectorATan(V); + + R1 = XMVectorSelect(Pi, g_XMNegativeZero, XIsPositive); + R2 = XMVectorAdd(R0, R1); + + return XMVectorSelect(Result, R2, ATanResultValid); + +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorSinEst(FXMVECTOR V) noexcept +{ + // 7-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + sinf(V.vector4_f32[0]), + sinf(V.vector4_f32[1]), + sinf(V.vector4_f32[2]), + sinf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with sin(y) = sin(x). + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(x), g_XMNegativeZero); + uint32x4_t c = vorrq_u32(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + float32x4_t absx = vabsq_f32(x); + float32x4_t rflx = vsubq_f32(vreinterpretq_f32_u32(c), x); + uint32x4_t comp = vcleq_f32(absx, g_XMHalfPi); + x = vbslq_f32(comp, x, rflx); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation + const XMVECTOR SEC = g_XMSinCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(SEC), 0); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_high_f32(SEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(SEC), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + Result = vmulq_f32(Result, x); + return Result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_sin_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with sin(y) = sin(x). + __m128 sign = _mm_and_ps(x, g_XMNegativeZero); + __m128 c = _mm_or_ps(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + __m128 absx = _mm_andnot_ps(sign, x); // |x| + __m128 rflx = _mm_sub_ps(c, x); + __m128 comp = _mm_cmple_ps(absx, g_XMHalfPi); + __m128 select0 = _mm_and_ps(comp, x); + __m128 select1 = _mm_andnot_ps(comp, rflx); + x = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation + const XMVECTOR SEC = g_XMSinCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(SEC, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(SEC, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SEC, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, x); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorCosEst(FXMVECTOR V) noexcept +{ + // 6-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + cosf(V.vector4_f32[0]), + cosf(V.vector4_f32[1]), + cosf(V.vector4_f32[2]), + cosf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Map V to x in [-pi,pi]. + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with cos(y) = sign*cos(x). + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(x), g_XMNegativeZero); + uint32x4_t c = vorrq_u32(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + float32x4_t absx = vabsq_f32(x); + float32x4_t rflx = vsubq_f32(vreinterpretq_f32_u32(c), x); + uint32x4_t comp = vcleq_f32(absx, g_XMHalfPi); + x = vbslq_f32(comp, x, rflx); + float32x4_t fsign = vbslq_f32(comp, g_XMOne, g_XMNegativeOne); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation + const XMVECTOR CEC = g_XMCosCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(CEC), 0); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_high_f32(CEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(CEC), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + Result = vmulq_f32(Result, fsign); + return Result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_cos_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + // Map V to x in [-pi,pi]. + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with cos(y) = sign*cos(x). + XMVECTOR sign = _mm_and_ps(x, g_XMNegativeZero); + __m128 c = _mm_or_ps(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + __m128 absx = _mm_andnot_ps(sign, x); // |x| + __m128 rflx = _mm_sub_ps(c, x); + __m128 comp = _mm_cmple_ps(absx, g_XMHalfPi); + __m128 select0 = _mm_and_ps(comp, x); + __m128 select1 = _mm_andnot_ps(comp, rflx); + x = _mm_or_ps(select0, select1); + select0 = _mm_and_ps(comp, g_XMOne); + select1 = _mm_andnot_ps(comp, g_XMNegativeOne); + sign = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation + const XMVECTOR CEC = g_XMCosCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(CEC, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(CEC, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CEC, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, sign); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline void XM_CALLCONV XMVectorSinCosEst +( + XMVECTOR* pSin, + XMVECTOR* pCos, + FXMVECTOR V +) noexcept +{ + assert(pSin != nullptr); + assert(pCos != nullptr); + + // 7/6-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Sin = { { { + sinf(V.vector4_f32[0]), + sinf(V.vector4_f32[1]), + sinf(V.vector4_f32[2]), + sinf(V.vector4_f32[3]) + } } }; + + XMVECTORF32 Cos = { { { + cosf(V.vector4_f32[0]), + cosf(V.vector4_f32[1]), + cosf(V.vector4_f32[2]), + cosf(V.vector4_f32[3]) + } } }; + + *pSin = Sin.v; + *pCos = Cos.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with cos(y) = sign*cos(x). + uint32x4_t sign = vandq_u32(vreinterpretq_u32_f32(x), g_XMNegativeZero); + uint32x4_t c = vorrq_u32(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + float32x4_t absx = vabsq_f32(x); + float32x4_t rflx = vsubq_f32(vreinterpretq_f32_u32(c), x); + uint32x4_t comp = vcleq_f32(absx, g_XMHalfPi); + x = vbslq_f32(comp, x, rflx); + float32x4_t fsign = vbslq_f32(comp, g_XMOne, g_XMNegativeOne); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation for sine + const XMVECTOR SEC = g_XMSinCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(SEC), 0); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_high_f32(SEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(SEC), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + *pSin = vmulq_f32(Result, x); + + // Compute polynomial approximation + const XMVECTOR CEC = g_XMCosCoefficients1; + vConstants = vdupq_lane_f32(vget_high_f32(CEC), 0); + Result = vmlaq_lane_f32(vConstants, x2, vget_high_f32(CEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(CEC), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + Result = vmlaq_f32(g_XMOne, Result, x2); + *pCos = vmulq_f32(Result, fsign); +#elif defined(_XM_SSE_INTRINSICS_) + // Force the value within the bounds of pi + XMVECTOR x = XMVectorModAngles(V); + + // Map in [-pi/2,pi/2] with sin(y) = sin(x), cos(y) = sign*cos(x). + XMVECTOR sign = _mm_and_ps(x, g_XMNegativeZero); + __m128 c = _mm_or_ps(g_XMPi, sign); // pi when x >= 0, -pi when x < 0 + __m128 absx = _mm_andnot_ps(sign, x); // |x| + __m128 rflx = _mm_sub_ps(c, x); + __m128 comp = _mm_cmple_ps(absx, g_XMHalfPi); + __m128 select0 = _mm_and_ps(comp, x); + __m128 select1 = _mm_andnot_ps(comp, rflx); + x = _mm_or_ps(select0, select1); + select0 = _mm_and_ps(comp, g_XMOne); + select1 = _mm_andnot_ps(comp, g_XMNegativeOne); + sign = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation for sine + const XMVECTOR SEC = g_XMSinCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(SEC, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(SEC, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(SEC, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, x); + *pSin = Result; + + // Compute polynomial approximation for cosine + const XMVECTOR CEC = g_XMCosCoefficients1; + vConstantsB = XM_PERMUTE_PS(CEC, _MM_SHUFFLE(3, 3, 3, 3)); + vConstants = XM_PERMUTE_PS(CEC, _MM_SHUFFLE(2, 2, 2, 2)); + Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(CEC, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + Result = XM_FMADD_PS(Result, x2, g_XMOne); + Result = _mm_mul_ps(Result, sign); + *pCos = Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorTanEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + tanf(V.vector4_f32[0]), + tanf(V.vector4_f32[1]), + tanf(V.vector4_f32[2]), + tanf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_tan_ps(V); + return Result; +#else + + XMVECTOR OneOverPi = XMVectorSplatW(g_XMTanEstCoefficients.v); + + XMVECTOR V1 = XMVectorMultiply(V, OneOverPi); + V1 = XMVectorRound(V1); + + V1 = XMVectorNegativeMultiplySubtract(g_XMPi.v, V1, V); + + XMVECTOR T0 = XMVectorSplatX(g_XMTanEstCoefficients.v); + XMVECTOR T1 = XMVectorSplatY(g_XMTanEstCoefficients.v); + XMVECTOR T2 = XMVectorSplatZ(g_XMTanEstCoefficients.v); + + XMVECTOR V2T2 = XMVectorNegativeMultiplySubtract(V1, V1, T2); + XMVECTOR V2 = XMVectorMultiply(V1, V1); + XMVECTOR V1T0 = XMVectorMultiply(V1, T0); + XMVECTOR V1T1 = XMVectorMultiply(V1, T1); + + XMVECTOR D = XMVectorReciprocalEst(V2T2); + XMVECTOR N = XMVectorMultiplyAdd(V2, V1T1, V1T0); + + return XMVectorMultiply(N, D); + +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorASinEst(FXMVECTOR V) noexcept +{ + // 3-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result; + Result.f[0] = asinf(V.vector4_f32[0]); + Result.f[1] = asinf(V.vector4_f32[1]); + Result.f[2] = asinf(V.vector4_f32[2]); + Result.f[3] = asinf(V.vector4_f32[3]); + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t nonnegative = vcgeq_f32(V, g_XMZero); + float32x4_t x = vabsq_f32(V); + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + float32x4_t oneMValue = vsubq_f32(g_XMOne, x); + float32x4_t clampOneMValue = vmaxq_f32(g_XMZero, oneMValue); + float32x4_t root = XMVectorSqrt(clampOneMValue); + + // Compute polynomial approximation + const XMVECTOR AEC = g_XMArcEstCoefficients; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(AEC), 0); + XMVECTOR t0 = vmlaq_lane_f32(vConstants, x, vget_high_f32(AEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(AEC), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AEC), 0); + t0 = vmlaq_f32(vConstants, t0, x); + t0 = vmulq_f32(t0, root); + + float32x4_t t1 = vsubq_f32(g_XMPi, t0); + t0 = vbslq_f32(nonnegative, t0, t1); + t0 = vsubq_f32(g_XMHalfPi, t0); + return t0; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_asin_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 nonnegative = _mm_cmpge_ps(V, g_XMZero); + __m128 mvalue = _mm_sub_ps(g_XMZero, V); + __m128 x = _mm_max_ps(V, mvalue); // |V| + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + __m128 oneMValue = _mm_sub_ps(g_XMOne, x); + __m128 clampOneMValue = _mm_max_ps(g_XMZero, oneMValue); + __m128 root = _mm_sqrt_ps(clampOneMValue); // sqrt(1-|V|) + + // Compute polynomial approximation + const XMVECTOR AEC = g_XMArcEstCoefficients; + __m128 vConstantsB = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 t0 = XM_FMADD_PS(vConstantsB, x, vConstants); + + vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(1, 1, 1, 1)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(0, 0, 0, 0)); + t0 = XM_FMADD_PS(t0, x, vConstants); + t0 = _mm_mul_ps(t0, root); + + __m128 t1 = _mm_sub_ps(g_XMPi, t0); + t0 = _mm_and_ps(nonnegative, t0); + t1 = _mm_andnot_ps(nonnegative, t1); + t0 = _mm_or_ps(t0, t1); + t0 = _mm_sub_ps(g_XMHalfPi, t0); + return t0; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorACosEst(FXMVECTOR V) noexcept +{ + // 3-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + acosf(V.vector4_f32[0]), + acosf(V.vector4_f32[1]), + acosf(V.vector4_f32[2]), + acosf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t nonnegative = vcgeq_f32(V, g_XMZero); + float32x4_t x = vabsq_f32(V); + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + float32x4_t oneMValue = vsubq_f32(g_XMOne, x); + float32x4_t clampOneMValue = vmaxq_f32(g_XMZero, oneMValue); + float32x4_t root = XMVectorSqrt(clampOneMValue); + + // Compute polynomial approximation + const XMVECTOR AEC = g_XMArcEstCoefficients; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(AEC), 0); + XMVECTOR t0 = vmlaq_lane_f32(vConstants, x, vget_high_f32(AEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(AEC), 1); + t0 = vmlaq_f32(vConstants, t0, x); + + vConstants = vdupq_lane_f32(vget_low_f32(AEC), 0); + t0 = vmlaq_f32(vConstants, t0, x); + t0 = vmulq_f32(t0, root); + + float32x4_t t1 = vsubq_f32(g_XMPi, t0); + t0 = vbslq_f32(nonnegative, t0, t1); + return t0; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_acos_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 nonnegative = _mm_cmpge_ps(V, g_XMZero); + __m128 mvalue = _mm_sub_ps(g_XMZero, V); + __m128 x = _mm_max_ps(V, mvalue); // |V| + + // Compute (1-|V|), clamp to zero to avoid sqrt of negative number. + __m128 oneMValue = _mm_sub_ps(g_XMOne, x); + __m128 clampOneMValue = _mm_max_ps(g_XMZero, oneMValue); + __m128 root = _mm_sqrt_ps(clampOneMValue); // sqrt(1-|V|) + + // Compute polynomial approximation + const XMVECTOR AEC = g_XMArcEstCoefficients; + __m128 vConstantsB = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 t0 = XM_FMADD_PS(vConstantsB, x, vConstants); + + vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(1, 1, 1, 1)); + t0 = XM_FMADD_PS(t0, x, vConstants); + + vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(0, 0, 0, 0)); + t0 = XM_FMADD_PS(t0, x, vConstants); + t0 = _mm_mul_ps(t0, root); + + __m128 t1 = _mm_sub_ps(g_XMPi, t0); + t0 = _mm_and_ps(nonnegative, t0); + t1 = _mm_andnot_ps(nonnegative, t1); + t0 = _mm_or_ps(t0, t1); + return t0; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorATanEst(FXMVECTOR V) noexcept +{ + // 9-degree minimax approximation + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + atanf(V.vector4_f32[0]), + atanf(V.vector4_f32[1]), + atanf(V.vector4_f32[2]), + atanf(V.vector4_f32[3]) + } } }; + return Result.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t absV = vabsq_f32(V); + float32x4_t invV = XMVectorReciprocalEst(V); + uint32x4_t comp = vcgtq_f32(V, g_XMOne); + float32x4_t sign = vbslq_f32(comp, g_XMOne, g_XMNegativeOne); + comp = vcleq_f32(absV, g_XMOne); + sign = vbslq_f32(comp, g_XMZero, sign); + float32x4_t x = vbslq_f32(comp, V, invV); + + float32x4_t x2 = vmulq_f32(x, x); + + // Compute polynomial approximation + const XMVECTOR AEC = g_XMATanEstCoefficients1; + XMVECTOR vConstants = vdupq_lane_f32(vget_high_f32(AEC), 0); + XMVECTOR Result = vmlaq_lane_f32(vConstants, x2, vget_high_f32(AEC), 1); + + vConstants = vdupq_lane_f32(vget_low_f32(AEC), 1); + Result = vmlaq_f32(vConstants, Result, x2); + + vConstants = vdupq_lane_f32(vget_low_f32(AEC), 0); + Result = vmlaq_f32(vConstants, Result, x2); + + // ATanEstCoefficients0 is already splatted + Result = vmlaq_f32(g_XMATanEstCoefficients0, Result, x2); + Result = vmulq_f32(Result, x); + + float32x4_t result1 = vmulq_f32(sign, g_XMHalfPi); + result1 = vsubq_f32(result1, Result); + + comp = vceqq_f32(sign, g_XMZero); + Result = vbslq_f32(comp, Result, result1); + return Result; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_atan_ps(V); + return Result; +#elif defined(_XM_SSE_INTRINSICS_) + __m128 absV = XMVectorAbs(V); + __m128 invV = _mm_div_ps(g_XMOne, V); + __m128 comp = _mm_cmpgt_ps(V, g_XMOne); + __m128 select0 = _mm_and_ps(comp, g_XMOne); + __m128 select1 = _mm_andnot_ps(comp, g_XMNegativeOne); + __m128 sign = _mm_or_ps(select0, select1); + comp = _mm_cmple_ps(absV, g_XMOne); + select0 = _mm_and_ps(comp, g_XMZero); + select1 = _mm_andnot_ps(comp, sign); + sign = _mm_or_ps(select0, select1); + select0 = _mm_and_ps(comp, V); + select1 = _mm_andnot_ps(comp, invV); + __m128 x = _mm_or_ps(select0, select1); + + __m128 x2 = _mm_mul_ps(x, x); + + // Compute polynomial approximation + const XMVECTOR AEC = g_XMATanEstCoefficients1; + __m128 vConstantsB = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(3, 3, 3, 3)); + __m128 vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 Result = XM_FMADD_PS(vConstantsB, x2, vConstants); + + vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(1, 1, 1, 1)); + Result = XM_FMADD_PS(Result, x2, vConstants); + + vConstants = XM_PERMUTE_PS(AEC, _MM_SHUFFLE(0, 0, 0, 0)); + Result = XM_FMADD_PS(Result, x2, vConstants); + // ATanEstCoefficients0 is already splatted + Result = XM_FMADD_PS(Result, x2, g_XMATanEstCoefficients0); + Result = _mm_mul_ps(Result, x); + __m128 result1 = _mm_mul_ps(sign, g_XMHalfPi); + result1 = _mm_sub_ps(result1, Result); + + comp = _mm_cmpeq_ps(sign, g_XMZero); + select0 = _mm_and_ps(comp, Result); + select1 = _mm_andnot_ps(comp, result1); + Result = _mm_or_ps(select0, select1); + return Result; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorATan2Est +( + FXMVECTOR Y, + FXMVECTOR X +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 Result = { { { + atan2f(Y.vector4_f32[0], X.vector4_f32[0]), + atan2f(Y.vector4_f32[1], X.vector4_f32[1]), + atan2f(Y.vector4_f32[2], X.vector4_f32[2]), + atan2f(Y.vector4_f32[3], X.vector4_f32[3]), + } } }; + return Result.v; +#elif defined(_XM_SVML_INTRINSICS_) + XMVECTOR Result = _mm_atan2_ps(Y, X); + return Result; +#else + + static const XMVECTORF32 ATan2Constants = { { { XM_PI, XM_PIDIV2, XM_PIDIV4, 2.3561944905f /* Pi*3/4 */ } } }; + + const XMVECTOR Zero = XMVectorZero(); + XMVECTOR ATanResultValid = XMVectorTrueInt(); + + XMVECTOR Pi = XMVectorSplatX(ATan2Constants); + XMVECTOR PiOverTwo = XMVectorSplatY(ATan2Constants); + XMVECTOR PiOverFour = XMVectorSplatZ(ATan2Constants); + XMVECTOR ThreePiOverFour = XMVectorSplatW(ATan2Constants); + + XMVECTOR YEqualsZero = XMVectorEqual(Y, Zero); + XMVECTOR XEqualsZero = XMVectorEqual(X, Zero); + XMVECTOR XIsPositive = XMVectorAndInt(X, g_XMNegativeZero.v); + XIsPositive = XMVectorEqualInt(XIsPositive, Zero); + XMVECTOR YEqualsInfinity = XMVectorIsInfinite(Y); + XMVECTOR XEqualsInfinity = XMVectorIsInfinite(X); + + XMVECTOR YSign = XMVectorAndInt(Y, g_XMNegativeZero.v); + Pi = XMVectorOrInt(Pi, YSign); + PiOverTwo = XMVectorOrInt(PiOverTwo, YSign); + PiOverFour = XMVectorOrInt(PiOverFour, YSign); + ThreePiOverFour = XMVectorOrInt(ThreePiOverFour, YSign); + + XMVECTOR R1 = XMVectorSelect(Pi, YSign, XIsPositive); + XMVECTOR R2 = XMVectorSelect(ATanResultValid, PiOverTwo, XEqualsZero); + XMVECTOR R3 = XMVectorSelect(R2, R1, YEqualsZero); + XMVECTOR R4 = XMVectorSelect(ThreePiOverFour, PiOverFour, XIsPositive); + XMVECTOR R5 = XMVectorSelect(PiOverTwo, R4, XEqualsInfinity); + XMVECTOR Result = XMVectorSelect(R3, R5, YEqualsInfinity); + ATanResultValid = XMVectorEqualInt(Result, ATanResultValid); + + XMVECTOR Reciprocal = XMVectorReciprocalEst(X); + XMVECTOR V = XMVectorMultiply(Y, Reciprocal); + XMVECTOR R0 = XMVectorATanEst(V); + + R1 = XMVectorSelect(Pi, g_XMNegativeZero, XIsPositive); + R2 = XMVectorAdd(R0, R1); + + Result = XMVectorSelect(Result, R2, ATanResultValid); + + return Result; + +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLerp +( + FXMVECTOR V0, + FXMVECTOR V1, + float t +) noexcept +{ + // V0 + t * (V1 - V0) + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Scale = XMVectorReplicate(t); + XMVECTOR Length = XMVectorSubtract(V1, V0); + return XMVectorMultiplyAdd(Length, Scale, V0); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR L = vsubq_f32(V1, V0); + return vmlaq_n_f32(V0, L, t); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR L = _mm_sub_ps(V1, V0); + XMVECTOR S = _mm_set_ps1(t); + return XM_FMADD_PS(L, S, V0); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorLerpV +( + FXMVECTOR V0, + FXMVECTOR V1, + FXMVECTOR T +) noexcept +{ + // V0 + T * (V1 - V0) + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Length = XMVectorSubtract(V1, V0); + return XMVectorMultiplyAdd(Length, T, V0); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR L = vsubq_f32(V1, V0); + return vmlaq_f32(V0, L, T); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR Length = _mm_sub_ps(V1, V0); + return XM_FMADD_PS(Length, T, V0); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorHermite +( + FXMVECTOR Position0, + FXMVECTOR Tangent0, + FXMVECTOR Position1, + GXMVECTOR Tangent1, + float t +) noexcept +{ + // Result = (2 * t^3 - 3 * t^2 + 1) * Position0 + + // (t^3 - 2 * t^2 + t) * Tangent0 + + // (-2 * t^3 + 3 * t^2) * Position1 + + // (t^3 - t^2) * Tangent1 + +#if defined(_XM_NO_INTRINSICS_) + + float t2 = t * t; + float t3 = t * t2; + + XMVECTOR P0 = XMVectorReplicate(2.0f * t3 - 3.0f * t2 + 1.0f); + XMVECTOR T0 = XMVectorReplicate(t3 - 2.0f * t2 + t); + XMVECTOR P1 = XMVectorReplicate(-2.0f * t3 + 3.0f * t2); + XMVECTOR T1 = XMVectorReplicate(t3 - t2); + + XMVECTOR Result = XMVectorMultiply(P0, Position0); + Result = XMVectorMultiplyAdd(T0, Tangent0, Result); + Result = XMVectorMultiplyAdd(P1, Position1, Result); + Result = XMVectorMultiplyAdd(T1, Tangent1, Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float t2 = t * t; + float t3 = t * t2; + + float p0 = 2.0f * t3 - 3.0f * t2 + 1.0f; + float t0 = t3 - 2.0f * t2 + t; + float p1 = -2.0f * t3 + 3.0f * t2; + float t1 = t3 - t2; + + XMVECTOR vResult = vmulq_n_f32(Position0, p0); + vResult = vmlaq_n_f32(vResult, Tangent0, t0); + vResult = vmlaq_n_f32(vResult, Position1, p1); + vResult = vmlaq_n_f32(vResult, Tangent1, t1); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + float t2 = t * t; + float t3 = t * t2; + + XMVECTOR P0 = _mm_set_ps1(2.0f * t3 - 3.0f * t2 + 1.0f); + XMVECTOR T0 = _mm_set_ps1(t3 - 2.0f * t2 + t); + XMVECTOR P1 = _mm_set_ps1(-2.0f * t3 + 3.0f * t2); + XMVECTOR T1 = _mm_set_ps1(t3 - t2); + + XMVECTOR vResult = _mm_mul_ps(P0, Position0); + vResult = XM_FMADD_PS(Tangent0, T0, vResult); + vResult = XM_FMADD_PS(Position1, P1, vResult); + vResult = XM_FMADD_PS(Tangent1, T1, vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorHermiteV +( + FXMVECTOR Position0, + FXMVECTOR Tangent0, + FXMVECTOR Position1, + GXMVECTOR Tangent1, + HXMVECTOR T +) noexcept +{ + // Result = (2 * t^3 - 3 * t^2 + 1) * Position0 + + // (t^3 - 2 * t^2 + t) * Tangent0 + + // (-2 * t^3 + 3 * t^2) * Position1 + + // (t^3 - t^2) * Tangent1 + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR T2 = XMVectorMultiply(T, T); + XMVECTOR T3 = XMVectorMultiply(T, T2); + + XMVECTOR P0 = XMVectorReplicate(2.0f * T3.vector4_f32[0] - 3.0f * T2.vector4_f32[0] + 1.0f); + XMVECTOR T0 = XMVectorReplicate(T3.vector4_f32[1] - 2.0f * T2.vector4_f32[1] + T.vector4_f32[1]); + XMVECTOR P1 = XMVectorReplicate(-2.0f * T3.vector4_f32[2] + 3.0f * T2.vector4_f32[2]); + XMVECTOR T1 = XMVectorReplicate(T3.vector4_f32[3] - T2.vector4_f32[3]); + + XMVECTOR Result = XMVectorMultiply(P0, Position0); + Result = XMVectorMultiplyAdd(T0, Tangent0, Result); + Result = XMVectorMultiplyAdd(P1, Position1, Result); + Result = XMVectorMultiplyAdd(T1, Tangent1, Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 CatMulT2 = { { { -3.0f, -2.0f, 3.0f, -1.0f } } }; + static const XMVECTORF32 CatMulT3 = { { { 2.0f, 1.0f, -2.0f, 1.0f } } }; + + XMVECTOR T2 = vmulq_f32(T, T); + XMVECTOR T3 = vmulq_f32(T, T2); + // Mul by the constants against t^2 + T2 = vmulq_f32(T2, CatMulT2); + // Mul by the constants against t^3 + T3 = vmlaq_f32(T2, T3, CatMulT3); + // T3 now has the pre-result. + // I need to add t.y only + T2 = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(T), g_XMMaskY)); + T3 = vaddq_f32(T3, T2); + // Add 1.0f to x + T3 = vaddq_f32(T3, g_XMIdentityR0); + // Now, I have the constants created + // Mul the x constant to Position0 + XMVECTOR vResult = vmulq_lane_f32(Position0, vget_low_f32(T3), 0); // T3[0] + // Mul the y constant to Tangent0 + vResult = vmlaq_lane_f32(vResult, Tangent0, vget_low_f32(T3), 1); // T3[1] + // Mul the z constant to Position1 + vResult = vmlaq_lane_f32(vResult, Position1, vget_high_f32(T3), 0); // T3[2] + // Mul the w constant to Tangent1 + vResult = vmlaq_lane_f32(vResult, Tangent1, vget_high_f32(T3), 1); // T3[3] + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 CatMulT2 = { { { -3.0f, -2.0f, 3.0f, -1.0f } } }; + static const XMVECTORF32 CatMulT3 = { { { 2.0f, 1.0f, -2.0f, 1.0f } } }; + + XMVECTOR T2 = _mm_mul_ps(T, T); + XMVECTOR T3 = _mm_mul_ps(T, T2); + // Mul by the constants against t^2 + T2 = _mm_mul_ps(T2, CatMulT2); + // Mul by the constants against t^3 + T3 = XM_FMADD_PS(T3, CatMulT3, T2); + // T3 now has the pre-result. + // I need to add t.y only + T2 = _mm_and_ps(T, g_XMMaskY); + T3 = _mm_add_ps(T3, T2); + // Add 1.0f to x + T3 = _mm_add_ps(T3, g_XMIdentityR0); + // Now, I have the constants created + // Mul the x constant to Position0 + XMVECTOR vResult = XM_PERMUTE_PS(T3, _MM_SHUFFLE(0, 0, 0, 0)); + vResult = _mm_mul_ps(vResult, Position0); + // Mul the y constant to Tangent0 + T2 = XM_PERMUTE_PS(T3, _MM_SHUFFLE(1, 1, 1, 1)); + vResult = XM_FMADD_PS(T2, Tangent0, vResult); + // Mul the z constant to Position1 + T2 = XM_PERMUTE_PS(T3, _MM_SHUFFLE(2, 2, 2, 2)); + vResult = XM_FMADD_PS(T2, Position1, vResult); + // Mul the w constant to Tangent1 + T3 = XM_PERMUTE_PS(T3, _MM_SHUFFLE(3, 3, 3, 3)); + vResult = XM_FMADD_PS(T3, Tangent1, vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorCatmullRom +( + FXMVECTOR Position0, + FXMVECTOR Position1, + FXMVECTOR Position2, + GXMVECTOR Position3, + float t +) noexcept +{ + // Result = ((-t^3 + 2 * t^2 - t) * Position0 + + // (3 * t^3 - 5 * t^2 + 2) * Position1 + + // (-3 * t^3 + 4 * t^2 + t) * Position2 + + // (t^3 - t^2) * Position3) * 0.5 + +#if defined(_XM_NO_INTRINSICS_) + + float t2 = t * t; + float t3 = t * t2; + + XMVECTOR P0 = XMVectorReplicate((-t3 + 2.0f * t2 - t) * 0.5f); + XMVECTOR P1 = XMVectorReplicate((3.0f * t3 - 5.0f * t2 + 2.0f) * 0.5f); + XMVECTOR P2 = XMVectorReplicate((-3.0f * t3 + 4.0f * t2 + t) * 0.5f); + XMVECTOR P3 = XMVectorReplicate((t3 - t2) * 0.5f); + + XMVECTOR Result = XMVectorMultiply(P0, Position0); + Result = XMVectorMultiplyAdd(P1, Position1, Result); + Result = XMVectorMultiplyAdd(P2, Position2, Result); + Result = XMVectorMultiplyAdd(P3, Position3, Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float t2 = t * t; + float t3 = t * t2; + + float p0 = (-t3 + 2.0f * t2 - t) * 0.5f; + float p1 = (3.0f * t3 - 5.0f * t2 + 2.0f) * 0.5f; + float p2 = (-3.0f * t3 + 4.0f * t2 + t) * 0.5f; + float p3 = (t3 - t2) * 0.5f; + + XMVECTOR P1 = vmulq_n_f32(Position1, p1); + XMVECTOR P0 = vmlaq_n_f32(P1, Position0, p0); + XMVECTOR P3 = vmulq_n_f32(Position3, p3); + XMVECTOR P2 = vmlaq_n_f32(P3, Position2, p2); + P0 = vaddq_f32(P0, P2); + return P0; +#elif defined(_XM_SSE_INTRINSICS_) + float t2 = t * t; + float t3 = t * t2; + + XMVECTOR P0 = _mm_set_ps1((-t3 + 2.0f * t2 - t) * 0.5f); + XMVECTOR P1 = _mm_set_ps1((3.0f * t3 - 5.0f * t2 + 2.0f) * 0.5f); + XMVECTOR P2 = _mm_set_ps1((-3.0f * t3 + 4.0f * t2 + t) * 0.5f); + XMVECTOR P3 = _mm_set_ps1((t3 - t2) * 0.5f); + + P1 = _mm_mul_ps(Position1, P1); + P0 = XM_FMADD_PS(Position0, P0, P1); + P3 = _mm_mul_ps(Position3, P3); + P2 = XM_FMADD_PS(Position2, P2, P3); + P0 = _mm_add_ps(P0, P2); + return P0; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorCatmullRomV +( + FXMVECTOR Position0, + FXMVECTOR Position1, + FXMVECTOR Position2, + GXMVECTOR Position3, + HXMVECTOR T +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float fx = T.vector4_f32[0]; + float fy = T.vector4_f32[1]; + float fz = T.vector4_f32[2]; + float fw = T.vector4_f32[3]; + XMVECTORF32 vResult = { { { + 0.5f * ((-fx * fx * fx + 2 * fx * fx - fx) * Position0.vector4_f32[0] + + (3 * fx * fx * fx - 5 * fx * fx + 2) * Position1.vector4_f32[0] + + (-3 * fx * fx * fx + 4 * fx * fx + fx) * Position2.vector4_f32[0] + + (fx * fx * fx - fx * fx) * Position3.vector4_f32[0]), + + 0.5f * ((-fy * fy * fy + 2 * fy * fy - fy) * Position0.vector4_f32[1] + + (3 * fy * fy * fy - 5 * fy * fy + 2) * Position1.vector4_f32[1] + + (-3 * fy * fy * fy + 4 * fy * fy + fy) * Position2.vector4_f32[1] + + (fy * fy * fy - fy * fy) * Position3.vector4_f32[1]), + + 0.5f * ((-fz * fz * fz + 2 * fz * fz - fz) * Position0.vector4_f32[2] + + (3 * fz * fz * fz - 5 * fz * fz + 2) * Position1.vector4_f32[2] + + (-3 * fz * fz * fz + 4 * fz * fz + fz) * Position2.vector4_f32[2] + + (fz * fz * fz - fz * fz) * Position3.vector4_f32[2]), + + 0.5f * ((-fw * fw * fw + 2 * fw * fw - fw) * Position0.vector4_f32[3] + + (3 * fw * fw * fw - 5 * fw * fw + 2) * Position1.vector4_f32[3] + + (-3 * fw * fw * fw + 4 * fw * fw + fw) * Position2.vector4_f32[3] + + (fw * fw * fw - fw * fw) * Position3.vector4_f32[3]) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Catmul2 = { { { 2.0f, 2.0f, 2.0f, 2.0f } } }; + static const XMVECTORF32 Catmul3 = { { { 3.0f, 3.0f, 3.0f, 3.0f } } }; + static const XMVECTORF32 Catmul4 = { { { 4.0f, 4.0f, 4.0f, 4.0f } } }; + static const XMVECTORF32 Catmul5 = { { { 5.0f, 5.0f, 5.0f, 5.0f } } }; + // Cache T^2 and T^3 + XMVECTOR T2 = vmulq_f32(T, T); + XMVECTOR T3 = vmulq_f32(T, T2); + // Perform the Position0 term + XMVECTOR vResult = vaddq_f32(T2, T2); + vResult = vsubq_f32(vResult, T); + vResult = vsubq_f32(vResult, T3); + vResult = vmulq_f32(vResult, Position0); + // Perform the Position1 term and add + XMVECTOR vTemp = vmulq_f32(T3, Catmul3); + vTemp = vmlsq_f32(vTemp, T2, Catmul5); + vTemp = vaddq_f32(vTemp, Catmul2); + vResult = vmlaq_f32(vResult, vTemp, Position1); + // Perform the Position2 term and add + vTemp = vmulq_f32(T2, Catmul4); + vTemp = vmlsq_f32(vTemp, T3, Catmul3); + vTemp = vaddq_f32(vTemp, T); + vResult = vmlaq_f32(vResult, vTemp, Position2); + // Position3 is the last term + T3 = vsubq_f32(T3, T2); + vResult = vmlaq_f32(vResult, T3, Position3); + // Multiply by 0.5f and exit + vResult = vmulq_f32(vResult, g_XMOneHalf); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Catmul2 = { { { 2.0f, 2.0f, 2.0f, 2.0f } } }; + static const XMVECTORF32 Catmul3 = { { { 3.0f, 3.0f, 3.0f, 3.0f } } }; + static const XMVECTORF32 Catmul4 = { { { 4.0f, 4.0f, 4.0f, 4.0f } } }; + static const XMVECTORF32 Catmul5 = { { { 5.0f, 5.0f, 5.0f, 5.0f } } }; + // Cache T^2 and T^3 + XMVECTOR T2 = _mm_mul_ps(T, T); + XMVECTOR T3 = _mm_mul_ps(T, T2); + // Perform the Position0 term + XMVECTOR vResult = _mm_add_ps(T2, T2); + vResult = _mm_sub_ps(vResult, T); + vResult = _mm_sub_ps(vResult, T3); + vResult = _mm_mul_ps(vResult, Position0); + // Perform the Position1 term and add + XMVECTOR vTemp = _mm_mul_ps(T3, Catmul3); + vTemp = XM_FNMADD_PS(T2, Catmul5, vTemp); + vTemp = _mm_add_ps(vTemp, Catmul2); + vResult = XM_FMADD_PS(vTemp, Position1, vResult); + // Perform the Position2 term and add + vTemp = _mm_mul_ps(T2, Catmul4); + vTemp = XM_FNMADD_PS(T3, Catmul3, vTemp); + vTemp = _mm_add_ps(vTemp, T); + vResult = XM_FMADD_PS(vTemp, Position2, vResult); + // Position3 is the last term + T3 = _mm_sub_ps(T3, T2); + vResult = XM_FMADD_PS(T3, Position3, vResult); + // Multiply by 0.5f and exit + vResult = _mm_mul_ps(vResult, g_XMOneHalf); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorBaryCentric +( + FXMVECTOR Position0, + FXMVECTOR Position1, + FXMVECTOR Position2, + float f, + float g +) noexcept +{ + // Result = Position0 + f * (Position1 - Position0) + g * (Position2 - Position0) + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR P10 = XMVectorSubtract(Position1, Position0); + XMVECTOR ScaleF = XMVectorReplicate(f); + + XMVECTOR P20 = XMVectorSubtract(Position2, Position0); + XMVECTOR ScaleG = XMVectorReplicate(g); + + XMVECTOR Result = XMVectorMultiplyAdd(P10, ScaleF, Position0); + Result = XMVectorMultiplyAdd(P20, ScaleG, Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR R1 = vsubq_f32(Position1, Position0); + XMVECTOR R2 = vsubq_f32(Position2, Position0); + R1 = vmlaq_n_f32(Position0, R1, f); + return vmlaq_n_f32(R1, R2, g); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR R1 = _mm_sub_ps(Position1, Position0); + XMVECTOR R2 = _mm_sub_ps(Position2, Position0); + XMVECTOR SF = _mm_set_ps1(f); + R1 = XM_FMADD_PS(R1, SF, Position0); + XMVECTOR SG = _mm_set_ps1(g); + return XM_FMADD_PS(R2, SG, R1); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVectorBaryCentricV +( + FXMVECTOR Position0, + FXMVECTOR Position1, + FXMVECTOR Position2, + GXMVECTOR F, + HXMVECTOR G +) noexcept +{ + // Result = Position0 + f * (Position1 - Position0) + g * (Position2 - Position0) + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR P10 = XMVectorSubtract(Position1, Position0); + XMVECTOR P20 = XMVectorSubtract(Position2, Position0); + + XMVECTOR Result = XMVectorMultiplyAdd(P10, F, Position0); + Result = XMVectorMultiplyAdd(P20, G, Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR R1 = vsubq_f32(Position1, Position0); + XMVECTOR R2 = vsubq_f32(Position2, Position0); + R1 = vmlaq_f32(Position0, R1, F); + return vmlaq_f32(R1, R2, G); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR R1 = _mm_sub_ps(Position1, Position0); + XMVECTOR R2 = _mm_sub_ps(Position2, Position0); + R1 = XM_FMADD_PS(R1, F, Position0); + return XM_FMADD_PS(R2, G, R1); +#endif +} + +/**************************************************************************** + * + * 2D Vector + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2Equal +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] == V2.vector4_f32[0]) && (V1.vector4_f32[1] == V2.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vceq_f32(vget_low_f32(V1), vget_low_f32(V2)); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + // z and w are don't care + return (((_mm_movemask_ps(vTemp) & 3) == 3) != 0); +#endif +} + + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector2EqualR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + uint32_t CR = 0; + if ((V1.vector4_f32[0] == V2.vector4_f32[0]) && + (V1.vector4_f32[1] == V2.vector4_f32[1])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] != V2.vector4_f32[0]) && + (V1.vector4_f32[1] != V2.vector4_f32[1])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vceq_f32(vget_low_f32(V1), vget_low_f32(V2)); + uint64_t r = vget_lane_u64(vreinterpret_u64_u32(vTemp), 0); + uint32_t CR = 0; + if (r == 0xFFFFFFFFFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + // z and w are don't care + int iTest = _mm_movemask_ps(vTemp) & 3; + uint32_t CR = 0; + if (iTest == 3) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2EqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_u32[0] == V2.vector4_u32[0]) && (V1.vector4_u32[1] == V2.vector4_u32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vceq_u32(vget_low_u32(vreinterpretq_u32_f32(V1)), vget_low_u32(vreinterpretq_u32_f32(V2))); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return (((_mm_movemask_ps(_mm_castsi128_ps(vTemp)) & 3) == 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector2EqualIntR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + uint32_t CR = 0; + if ((V1.vector4_u32[0] == V2.vector4_u32[0]) && + (V1.vector4_u32[1] == V2.vector4_u32[1])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_u32[0] != V2.vector4_u32[0]) && + (V1.vector4_u32[1] != V2.vector4_u32[1])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vceq_u32(vget_low_u32(vreinterpretq_u32_f32(V1)), vget_low_u32(vreinterpretq_u32_f32(V2))); + uint64_t r = vget_lane_u64(vreinterpret_u64_u32(vTemp), 0); + uint32_t CR = 0; + if (r == 0xFFFFFFFFFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + int iTest = _mm_movemask_ps(_mm_castsi128_ps(vTemp)) & 3; + uint32_t CR = 0; + if (iTest == 3) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2NearEqual +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR Epsilon +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float dx = fabsf(V1.vector4_f32[0] - V2.vector4_f32[0]); + float dy = fabsf(V1.vector4_f32[1] - V2.vector4_f32[1]); + return ((dx <= Epsilon.vector4_f32[0]) && + (dy <= Epsilon.vector4_f32[1])); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t vDelta = vsub_f32(vget_low_f32(V1), vget_low_f32(V2)); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + uint32x2_t vTemp = vacle_f32(vDelta, vget_low_u32(Epsilon)); +#else + uint32x2_t vTemp = vcle_f32(vabs_f32(vDelta), vget_low_f32(Epsilon)); +#endif + uint64_t r = vget_lane_u64(vreinterpret_u64_u32(vTemp), 0); + return (r == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + // Get the difference + XMVECTOR vDelta = _mm_sub_ps(V1, V2); + // Get the absolute value of the difference + XMVECTOR vTemp = _mm_setzero_ps(); + vTemp = _mm_sub_ps(vTemp, vDelta); + vTemp = _mm_max_ps(vTemp, vDelta); + vTemp = _mm_cmple_ps(vTemp, Epsilon); + // z and w are don't care + return (((_mm_movemask_ps(vTemp) & 3) == 0x3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2NotEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] != V2.vector4_f32[0]) || (V1.vector4_f32[1] != V2.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vceq_f32(vget_low_f32(V1), vget_low_f32(V2)); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) != 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + // z and w are don't care + return (((_mm_movemask_ps(vTemp) & 3) != 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2NotEqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_u32[0] != V2.vector4_u32[0]) || (V1.vector4_u32[1] != V2.vector4_u32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vceq_u32(vget_low_u32(vreinterpretq_u32_f32(V1)), vget_low_u32(vreinterpretq_u32_f32(V2))); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) != 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return (((_mm_movemask_ps(_mm_castsi128_ps(vTemp)) & 3) != 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2Greater +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] > V2.vector4_f32[0]) && (V1.vector4_f32[1] > V2.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vcgt_f32(vget_low_f32(V1), vget_low_f32(V2)); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + // z and w are don't care + return (((_mm_movemask_ps(vTemp) & 3) == 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector2GreaterR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + uint32_t CR = 0; + if ((V1.vector4_f32[0] > V2.vector4_f32[0]) && + (V1.vector4_f32[1] > V2.vector4_f32[1])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] <= V2.vector4_f32[0]) && + (V1.vector4_f32[1] <= V2.vector4_f32[1])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vcgt_f32(vget_low_f32(V1), vget_low_f32(V2)); + uint64_t r = vget_lane_u64(vreinterpret_u64_u32(vTemp), 0); + uint32_t CR = 0; + if (r == 0xFFFFFFFFFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + int iTest = _mm_movemask_ps(vTemp) & 3; + uint32_t CR = 0; + if (iTest == 3) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2GreaterOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] >= V2.vector4_f32[0]) && (V1.vector4_f32[1] >= V2.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vcge_f32(vget_low_f32(V1), vget_low_f32(V2)); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 3) == 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector2GreaterOrEqualR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + uint32_t CR = 0; + if ((V1.vector4_f32[0] >= V2.vector4_f32[0]) && + (V1.vector4_f32[1] >= V2.vector4_f32[1])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] < V2.vector4_f32[0]) && + (V1.vector4_f32[1] < V2.vector4_f32[1])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vcge_f32(vget_low_f32(V1), vget_low_f32(V2)); + uint64_t r = vget_lane_u64(vreinterpret_u64_u32(vTemp), 0); + uint32_t CR = 0; + if (r == 0xFFFFFFFFFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + int iTest = _mm_movemask_ps(vTemp) & 3; + uint32_t CR = 0; + if (iTest == 3) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2Less +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] < V2.vector4_f32[0]) && (V1.vector4_f32[1] < V2.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vclt_f32(vget_low_f32(V1), vget_low_f32(V2)); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmplt_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 3) == 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2LessOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] <= V2.vector4_f32[0]) && (V1.vector4_f32[1] <= V2.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vTemp = vcle_f32(vget_low_f32(V1), vget_low_f32(V2)); + return (vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmple_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 3) == 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2InBounds +( + FXMVECTOR V, + FXMVECTOR Bounds +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V.vector4_f32[0] <= Bounds.vector4_f32[0] && V.vector4_f32[0] >= -Bounds.vector4_f32[0]) && + (V.vector4_f32[1] <= Bounds.vector4_f32[1] && V.vector4_f32[1] >= -Bounds.vector4_f32[1])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + float32x2_t B = vget_low_f32(Bounds); + // Test if less than or equal + uint32x2_t ivTemp1 = vcle_f32(VL, B); + // Negate the bounds + float32x2_t vTemp2 = vneg_f32(B); + // Test if greater or equal (Reversed) + uint32x2_t ivTemp2 = vcle_f32(vTemp2, VL); + // Blend answers + ivTemp1 = vand_u32(ivTemp1, ivTemp2); + // x and y in bounds? + return (vget_lane_u64(vreinterpret_u64_u32(ivTemp1), 0) == 0xFFFFFFFFFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + // Test if less than or equal + XMVECTOR vTemp1 = _mm_cmple_ps(V, Bounds); + // Negate the bounds + XMVECTOR vTemp2 = _mm_mul_ps(Bounds, g_XMNegativeOne); + // Test if greater or equal (Reversed) + vTemp2 = _mm_cmple_ps(vTemp2, V); + // Blend answers + vTemp1 = _mm_and_ps(vTemp1, vTemp2); + // x and y in bounds? (z and w are don't care) + return (((_mm_movemask_ps(vTemp1) & 0x3) == 0x3) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(push) +#pragma float_control(precise, on) +#endif + +inline bool XM_CALLCONV XMVector2IsNaN(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (XMISNAN(V.vector4_f32[0]) || + XMISNAN(V.vector4_f32[1])); +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + return isnan(vgetq_lane_f32(V, 0)) || isnan(vgetq_lane_f32(V, 1)); +#else + float32x2_t VL = vget_low_f32(V); + // Test against itself. NaN is always not equal + uint32x2_t vTempNan = vceq_f32(VL, VL); + // If x or y are NaN, the mask is zero + return (vget_lane_u64(vreinterpret_u64_u32(vTempNan), 0) != 0xFFFFFFFFFFFFFFFFU); +#endif +#elif defined(_XM_SSE_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + XM_ALIGNED_DATA(16) float tmp[4]; + _mm_store_ps(tmp, V); + return isnan(tmp[0]) || isnan(tmp[1]); +#else +// Test against itself. NaN is always not equal + XMVECTOR vTempNan = _mm_cmpneq_ps(V, V); + // If x or y are NaN, the mask is non-zero + return ((_mm_movemask_ps(vTempNan) & 3) != 0); +#endif +#endif +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(pop) +#endif + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector2IsInfinite(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + return (XMISINF(V.vector4_f32[0]) || + XMISINF(V.vector4_f32[1])); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Mask off the sign bit + uint32x2_t vTemp = vand_u32(vget_low_u32(vreinterpretq_u32_f32(V)), vget_low_u32(g_XMAbsMask)); + // Compare to infinity + vTemp = vceq_f32(vreinterpret_f32_u32(vTemp), vget_low_f32(g_XMInfinity)); + // If any are infinity, the signs are true. + return vget_lane_u64(vreinterpret_u64_u32(vTemp), 0) != 0; +#elif defined(_XM_SSE_INTRINSICS_) + // Mask off the sign bit + __m128 vTemp = _mm_and_ps(V, g_XMAbsMask); + // Compare to infinity + vTemp = _mm_cmpeq_ps(vTemp, g_XMInfinity); + // If x or z are infinity, the signs are true. + return ((_mm_movemask_ps(vTemp) & 3) != 0); +#endif +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Dot +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result; + Result.f[0] = + Result.f[1] = + Result.f[2] = + Result.f[3] = V1.vector4_f32[0] * V2.vector4_f32[0] + V1.vector4_f32[1] * V2.vector4_f32[1]; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Perform the dot product on x and y + float32x2_t vTemp = vmul_f32(vget_low_f32(V1), vget_low_f32(V2)); + vTemp = vpadd_f32(vTemp, vTemp); + return vcombine_f32(vTemp, vTemp); +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_dp_ps(V1, V2, 0x3f); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vDot = _mm_mul_ps(V1, V2); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_moveldup_ps(vDot); + return vDot; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y + XMVECTOR vLengthSq = _mm_mul_ps(V1, V2); + // vTemp has y splatted + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + // x+y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Cross +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + // [ V1.x*V2.y - V1.y*V2.x, V1.x*V2.y - V1.y*V2.x ] + +#if defined(_XM_NO_INTRINSICS_) + float fCross = (V1.vector4_f32[0] * V2.vector4_f32[1]) - (V1.vector4_f32[1] * V2.vector4_f32[0]); + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = fCross; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Negate = { { { 1.f, -1.f, 0, 0 } } }; + + float32x2_t vTemp = vmul_f32(vget_low_f32(V1), vrev64_f32(vget_low_f32(V2))); + vTemp = vmul_f32(vTemp, vget_low_f32(Negate)); + vTemp = vpadd_f32(vTemp, vTemp); + return vcombine_f32(vTemp, vTemp); +#elif defined(_XM_SSE_INTRINSICS_) + // Swap x and y + XMVECTOR vResult = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 1, 0, 1)); + // Perform the muls + vResult = _mm_mul_ps(vResult, V1); + // Splat y + XMVECTOR vTemp = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(1, 1, 1, 1)); + // Sub the values + vResult = _mm_sub_ss(vResult, vTemp); + // Splat the cross product + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 0, 0, 0)); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2LengthSq(FXMVECTOR V) noexcept +{ + return XMVector2Dot(V, V); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2ReciprocalLengthEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector2LengthSq(V); + Result = XMVectorReciprocalSqrtEst(Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + // Dot2 + float32x2_t vTemp = vmul_f32(VL, VL); + vTemp = vpadd_f32(vTemp, vTemp); + // Reciprocal sqrt (estimate) + vTemp = vrsqrte_f32(vTemp); + return vcombine_f32(vTemp, vTemp); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x3f); + return _mm_rsqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + XMVECTOR vTemp = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_rsqrt_ss(vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has y splatted + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + // x+y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = _mm_rsqrt_ss(vLengthSq); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2ReciprocalLength(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector2LengthSq(V); + Result = XMVectorReciprocalSqrt(Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + // Dot2 + float32x2_t vTemp = vmul_f32(VL, VL); + vTemp = vpadd_f32(vTemp, vTemp); + // Reciprocal sqrt + float32x2_t S0 = vrsqrte_f32(vTemp); + float32x2_t P0 = vmul_f32(vTemp, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(vTemp, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t Result = vmul_f32(S1, R1); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x3f); + XMVECTOR vLengthSq = _mm_sqrt_ps(vTemp); + return _mm_div_ps(g_XMOne, vLengthSq); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + XMVECTOR vTemp = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ss(vTemp); + vLengthSq = _mm_div_ss(g_XMOne, vLengthSq); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has y splatted + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + // x+y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = _mm_sqrt_ss(vLengthSq); + vLengthSq = _mm_div_ss(g_XMOne, vLengthSq); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2LengthEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector2LengthSq(V); + Result = XMVectorSqrtEst(Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + // Dot2 + float32x2_t vTemp = vmul_f32(VL, VL); + vTemp = vpadd_f32(vTemp, vTemp); + const float32x2_t zero = vdup_n_f32(0); + uint32x2_t VEqualsZero = vceq_f32(vTemp, zero); + // Sqrt (estimate) + float32x2_t Result = vrsqrte_f32(vTemp); + Result = vmul_f32(vTemp, Result); + Result = vbsl_f32(VEqualsZero, zero, Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x3f); + return _mm_sqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + XMVECTOR vTemp = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ss(vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has y splatted + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + // x+y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = _mm_sqrt_ss(vLengthSq); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Length(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector2LengthSq(V); + Result = XMVectorSqrt(Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + // Dot2 + float32x2_t vTemp = vmul_f32(VL, VL); + vTemp = vpadd_f32(vTemp, vTemp); + const float32x2_t zero = vdup_n_f32(0); + uint32x2_t VEqualsZero = vceq_f32(vTemp, zero); + // Sqrt + float32x2_t S0 = vrsqrte_f32(vTemp); + float32x2_t P0 = vmul_f32(vTemp, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(vTemp, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t Result = vmul_f32(S1, R1); + Result = vmul_f32(vTemp, Result); + Result = vbsl_f32(VEqualsZero, zero, Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x3f); + return _mm_sqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + XMVECTOR vTemp = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ss(vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has y splatted + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + // x+y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ +// XMVector2NormalizeEst uses a reciprocal estimate and +// returns QNaN on zero and infinite vectors. + +inline XMVECTOR XM_CALLCONV XMVector2NormalizeEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector2ReciprocalLength(V); + Result = XMVectorMultiply(V, Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + // Dot2 + float32x2_t vTemp = vmul_f32(VL, VL); + vTemp = vpadd_f32(vTemp, vTemp); + // Reciprocal sqrt (estimate) + vTemp = vrsqrte_f32(vTemp); + // Normalize + float32x2_t Result = vmul_f32(VL, vTemp); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x3f); + XMVECTOR vResult = _mm_rsqrt_ps(vTemp); + return _mm_mul_ps(vResult, V); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_rsqrt_ss(vLengthSq); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + vLengthSq = _mm_mul_ps(vLengthSq, V); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has y splatted + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + // x+y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = _mm_rsqrt_ss(vLengthSq); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + vLengthSq = _mm_mul_ps(vLengthSq, V); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Normalize(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR vResult = XMVector2Length(V); + float fLength = vResult.vector4_f32[0]; + + // Prevent divide by zero + if (fLength > 0) + { + fLength = 1.0f / fLength; + } + + vResult.vector4_f32[0] = V.vector4_f32[0] * fLength; + vResult.vector4_f32[1] = V.vector4_f32[1] * fLength; + vResult.vector4_f32[2] = V.vector4_f32[2] * fLength; + vResult.vector4_f32[3] = V.vector4_f32[3] * fLength; + return vResult; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + // Dot2 + float32x2_t vTemp = vmul_f32(VL, VL); + vTemp = vpadd_f32(vTemp, vTemp); + uint32x2_t VEqualsZero = vceq_f32(vTemp, vdup_n_f32(0)); + uint32x2_t VEqualsInf = vceq_f32(vTemp, vget_low_f32(g_XMInfinity)); + // Reciprocal sqrt (2 iterations of Newton-Raphson) + float32x2_t S0 = vrsqrte_f32(vTemp); + float32x2_t P0 = vmul_f32(vTemp, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(vTemp, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + vTemp = vmul_f32(S1, R1); + // Normalize + float32x2_t Result = vmul_f32(VL, vTemp); + Result = vbsl_f32(VEqualsZero, vdup_n_f32(0), Result); + Result = vbsl_f32(VEqualsInf, vget_low_f32(g_XMQNaN), Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vLengthSq = _mm_dp_ps(V, V, 0x3f); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Reciprocal mul to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#elif defined(_XM_SSE3_INTRINSICS_) + // Perform the dot product on x and y only + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_moveldup_ps(vLengthSq); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Reciprocal mul to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x and y only + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 1, 1, 1)); + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Reciprocal mul to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2ClampLength +( + FXMVECTOR V, + float LengthMin, + float LengthMax +) noexcept +{ + XMVECTOR ClampMax = XMVectorReplicate(LengthMax); + XMVECTOR ClampMin = XMVectorReplicate(LengthMin); + return XMVector2ClampLengthV(V, ClampMin, ClampMax); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2ClampLengthV +( + FXMVECTOR V, + FXMVECTOR LengthMin, + FXMVECTOR LengthMax +) noexcept +{ + assert((XMVectorGetY(LengthMin) == XMVectorGetX(LengthMin))); + assert((XMVectorGetY(LengthMax) == XMVectorGetX(LengthMax))); + assert(XMVector2GreaterOrEqual(LengthMin, g_XMZero)); + assert(XMVector2GreaterOrEqual(LengthMax, g_XMZero)); + assert(XMVector2GreaterOrEqual(LengthMax, LengthMin)); + + XMVECTOR LengthSq = XMVector2LengthSq(V); + + const XMVECTOR Zero = XMVectorZero(); + + XMVECTOR RcpLength = XMVectorReciprocalSqrt(LengthSq); + + XMVECTOR InfiniteLength = XMVectorEqualInt(LengthSq, g_XMInfinity.v); + XMVECTOR ZeroLength = XMVectorEqual(LengthSq, Zero); + + XMVECTOR Length = XMVectorMultiply(LengthSq, RcpLength); + + XMVECTOR Normal = XMVectorMultiply(V, RcpLength); + + XMVECTOR Select = XMVectorEqualInt(InfiniteLength, ZeroLength); + Length = XMVectorSelect(LengthSq, Length, Select); + Normal = XMVectorSelect(LengthSq, Normal, Select); + + XMVECTOR ControlMax = XMVectorGreater(Length, LengthMax); + XMVECTOR ControlMin = XMVectorLess(Length, LengthMin); + + XMVECTOR ClampLength = XMVectorSelect(Length, LengthMax, ControlMax); + ClampLength = XMVectorSelect(ClampLength, LengthMin, ControlMin); + + XMVECTOR Result = XMVectorMultiply(Normal, ClampLength); + + // Preserve the original vector (with no precision loss) if the length falls within the given range + XMVECTOR Control = XMVectorEqualInt(ControlMax, ControlMin); + Result = XMVectorSelect(Result, V, Control); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Reflect +( + FXMVECTOR Incident, + FXMVECTOR Normal +) noexcept +{ + // Result = Incident - (2 * dot(Incident, Normal)) * Normal + + XMVECTOR Result; + Result = XMVector2Dot(Incident, Normal); + Result = XMVectorAdd(Result, Result); + Result = XMVectorNegativeMultiplySubtract(Result, Normal, Incident); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Refract +( + FXMVECTOR Incident, + FXMVECTOR Normal, + float RefractionIndex +) noexcept +{ + XMVECTOR Index = XMVectorReplicate(RefractionIndex); + return XMVector2RefractV(Incident, Normal, Index); +} + +//------------------------------------------------------------------------------ + +// Return the refraction of a 2D vector +inline XMVECTOR XM_CALLCONV XMVector2RefractV +( + FXMVECTOR Incident, + FXMVECTOR Normal, + FXMVECTOR RefractionIndex +) noexcept +{ + // Result = RefractionIndex * Incident - Normal * (RefractionIndex * dot(Incident, Normal) + + // sqrt(1 - RefractionIndex * RefractionIndex * (1 - dot(Incident, Normal) * dot(Incident, Normal)))) + +#if defined(_XM_NO_INTRINSICS_) + + float IDotN = (Incident.vector4_f32[0] * Normal.vector4_f32[0]) + (Incident.vector4_f32[1] * Normal.vector4_f32[1]); + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + float RY = 1.0f - (IDotN * IDotN); + float RX = 1.0f - (RY * RefractionIndex.vector4_f32[0] * RefractionIndex.vector4_f32[0]); + RY = 1.0f - (RY * RefractionIndex.vector4_f32[1] * RefractionIndex.vector4_f32[1]); + if (RX >= 0.0f) + { + RX = (RefractionIndex.vector4_f32[0] * Incident.vector4_f32[0]) - (Normal.vector4_f32[0] * ((RefractionIndex.vector4_f32[0] * IDotN) + sqrtf(RX))); + } + else + { + RX = 0.0f; + } + if (RY >= 0.0f) + { + RY = (RefractionIndex.vector4_f32[1] * Incident.vector4_f32[1]) - (Normal.vector4_f32[1] * ((RefractionIndex.vector4_f32[1] * IDotN) + sqrtf(RY))); + } + else + { + RY = 0.0f; + } + + XMVECTOR vResult; + vResult.vector4_f32[0] = RX; + vResult.vector4_f32[1] = RY; + vResult.vector4_f32[2] = 0.0f; + vResult.vector4_f32[3] = 0.0f; + return vResult; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t IL = vget_low_f32(Incident); + float32x2_t NL = vget_low_f32(Normal); + float32x2_t RIL = vget_low_f32(RefractionIndex); + // Get the 2D Dot product of Incident-Normal + float32x2_t vTemp = vmul_f32(IL, NL); + float32x2_t IDotN = vpadd_f32(vTemp, vTemp); + // vTemp = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + vTemp = vmls_f32(vget_low_f32(g_XMOne), IDotN, IDotN); + vTemp = vmul_f32(vTemp, RIL); + vTemp = vmls_f32(vget_low_f32(g_XMOne), vTemp, RIL); + // If any terms are <=0, sqrt() will fail, punt to zero + uint32x2_t vMask = vcgt_f32(vTemp, vget_low_f32(g_XMZero)); + // Sqrt(vTemp) + float32x2_t S0 = vrsqrte_f32(vTemp); + float32x2_t P0 = vmul_f32(vTemp, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(vTemp, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t S2 = vmul_f32(S1, R1); + vTemp = vmul_f32(vTemp, S2); + // R = RefractionIndex * IDotN + sqrt(R) + vTemp = vmla_f32(vTemp, RIL, IDotN); + // Result = RefractionIndex * Incident - Normal * R + float32x2_t vResult = vmul_f32(RIL, IL); + vResult = vmls_f32(vResult, vTemp, NL); + vResult = vreinterpret_f32_u32(vand_u32(vreinterpret_u32_f32(vResult), vMask)); + return vcombine_f32(vResult, vResult); +#elif defined(_XM_SSE_INTRINSICS_) + // Result = RefractionIndex * Incident - Normal * (RefractionIndex * dot(Incident, Normal) + + // sqrt(1 - RefractionIndex * RefractionIndex * (1 - dot(Incident, Normal) * dot(Incident, Normal)))) + // Get the 2D Dot product of Incident-Normal + XMVECTOR IDotN = XMVector2Dot(Incident, Normal); + // vTemp = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + XMVECTOR vTemp = XM_FNMADD_PS(IDotN, IDotN, g_XMOne); + vTemp = _mm_mul_ps(vTemp, RefractionIndex); + vTemp = XM_FNMADD_PS(vTemp, RefractionIndex, g_XMOne); + // If any terms are <=0, sqrt() will fail, punt to zero + XMVECTOR vMask = _mm_cmpgt_ps(vTemp, g_XMZero); + // R = RefractionIndex * IDotN + sqrt(R) + vTemp = _mm_sqrt_ps(vTemp); + vTemp = XM_FMADD_PS(RefractionIndex, IDotN, vTemp); + // Result = RefractionIndex * Incident - Normal * R + XMVECTOR vResult = _mm_mul_ps(RefractionIndex, Incident); + vResult = XM_FNMADD_PS(vTemp, Normal, vResult); + vResult = _mm_and_ps(vResult, vMask); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Orthogonal(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + -V.vector4_f32[1], + V.vector4_f32[0], + 0.f, + 0.f + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Negate = { { { -1.f, 1.f, 0, 0 } } }; + const float32x2_t zero = vdup_n_f32(0); + + float32x2_t VL = vget_low_f32(V); + float32x2_t Result = vmul_f32(vrev64_f32(VL), vget_low_f32(Negate)); + return vcombine_f32(Result, zero); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 2, 0, 1)); + vResult = _mm_mul_ps(vResult, g_XMNegateX); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2AngleBetweenNormalsEst +( + FXMVECTOR N1, + FXMVECTOR N2 +) noexcept +{ + XMVECTOR Result = XMVector2Dot(N1, N2); + Result = XMVectorClamp(Result, g_XMNegativeOne.v, g_XMOne.v); + Result = XMVectorACosEst(Result); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2AngleBetweenNormals +( + FXMVECTOR N1, + FXMVECTOR N2 +) noexcept +{ + XMVECTOR Result = XMVector2Dot(N1, N2); + Result = XMVectorClamp(Result, g_XMNegativeOne, g_XMOne); + Result = XMVectorACos(Result); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2AngleBetweenVectors +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + XMVECTOR L1 = XMVector2ReciprocalLength(V1); + XMVECTOR L2 = XMVector2ReciprocalLength(V2); + + XMVECTOR Dot = XMVector2Dot(V1, V2); + + L1 = XMVectorMultiply(L1, L2); + + XMVECTOR CosAngle = XMVectorMultiply(Dot, L1); + CosAngle = XMVectorClamp(CosAngle, g_XMNegativeOne.v, g_XMOne.v); + + return XMVectorACos(CosAngle); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2LinePointDistance +( + FXMVECTOR LinePoint1, + FXMVECTOR LinePoint2, + FXMVECTOR Point +) noexcept +{ + // Given a vector PointVector from LinePoint1 to Point and a vector + // LineVector from LinePoint1 to LinePoint2, the scaled distance + // PointProjectionScale from LinePoint1 to the perpendicular projection + // of PointVector onto the line is defined as: + // + // PointProjectionScale = dot(PointVector, LineVector) / LengthSq(LineVector) + + XMVECTOR PointVector = XMVectorSubtract(Point, LinePoint1); + XMVECTOR LineVector = XMVectorSubtract(LinePoint2, LinePoint1); + + XMVECTOR LengthSq = XMVector2LengthSq(LineVector); + + XMVECTOR PointProjectionScale = XMVector2Dot(PointVector, LineVector); + PointProjectionScale = XMVectorDivide(PointProjectionScale, LengthSq); + + XMVECTOR DistanceVector = XMVectorMultiply(LineVector, PointProjectionScale); + DistanceVector = XMVectorSubtract(PointVector, DistanceVector); + + return XMVector2Length(DistanceVector); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2IntersectLine +( + FXMVECTOR Line1Point1, + FXMVECTOR Line1Point2, + FXMVECTOR Line2Point1, + GXMVECTOR Line2Point2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) || defined(_XM_ARM_NEON_INTRINSICS_) + + XMVECTOR V1 = XMVectorSubtract(Line1Point2, Line1Point1); + XMVECTOR V2 = XMVectorSubtract(Line2Point2, Line2Point1); + XMVECTOR V3 = XMVectorSubtract(Line1Point1, Line2Point1); + + XMVECTOR C1 = XMVector2Cross(V1, V2); + XMVECTOR C2 = XMVector2Cross(V2, V3); + + XMVECTOR Result; + const XMVECTOR Zero = XMVectorZero(); + if (XMVector2NearEqual(C1, Zero, g_XMEpsilon.v)) + { + if (XMVector2NearEqual(C2, Zero, g_XMEpsilon.v)) + { + // Coincident + Result = g_XMInfinity.v; + } + else + { + // Parallel + Result = g_XMQNaN.v; + } + } + else + { + // Intersection point = Line1Point1 + V1 * (C2 / C1) + XMVECTOR Scale = XMVectorReciprocal(C1); + Scale = XMVectorMultiply(C2, Scale); + Result = XMVectorMultiplyAdd(V1, Scale, Line1Point1); + } + + return Result; + +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR V1 = _mm_sub_ps(Line1Point2, Line1Point1); + XMVECTOR V2 = _mm_sub_ps(Line2Point2, Line2Point1); + XMVECTOR V3 = _mm_sub_ps(Line1Point1, Line2Point1); + // Generate the cross products + XMVECTOR C1 = XMVector2Cross(V1, V2); + XMVECTOR C2 = XMVector2Cross(V2, V3); + // If C1 is not close to epsilon, use the calculated value + XMVECTOR vResultMask = _mm_setzero_ps(); + vResultMask = _mm_sub_ps(vResultMask, C1); + vResultMask = _mm_max_ps(vResultMask, C1); + // 0xFFFFFFFF if the calculated value is to be used + vResultMask = _mm_cmpgt_ps(vResultMask, g_XMEpsilon); + // If C1 is close to epsilon, which fail type is it? INFINITY or NAN? + XMVECTOR vFailMask = _mm_setzero_ps(); + vFailMask = _mm_sub_ps(vFailMask, C2); + vFailMask = _mm_max_ps(vFailMask, C2); + vFailMask = _mm_cmple_ps(vFailMask, g_XMEpsilon); + XMVECTOR vFail = _mm_and_ps(vFailMask, g_XMInfinity); + vFailMask = _mm_andnot_ps(vFailMask, g_XMQNaN); + // vFail is NAN or INF + vFail = _mm_or_ps(vFail, vFailMask); + // Intersection point = Line1Point1 + V1 * (C2 / C1) + XMVECTOR vResult = _mm_div_ps(C2, C1); + vResult = XM_FMADD_PS(vResult, V1, Line1Point1); + // Use result, or failure value + vResult = _mm_and_ps(vResult, vResultMask); + vResultMask = _mm_andnot_ps(vResultMask, vFail); + vResult = _mm_or_ps(vResult, vResultMask); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2Transform +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Y, M.r[1], M.r[3]); + Result = XMVectorMultiplyAdd(X, M.r[0], Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + float32x4_t Result = vmlaq_lane_f32(M.r[3], M.r[1], VL, 1); // Y + return vmlaq_lane_f32(Result, M.r[0], VL, 0); // X +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); // Y + vResult = XM_FMADD_PS(vResult, M.r[1], M.r[3]); + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); // X + vResult = XM_FMADD_PS(vTemp, M.r[0], vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMFLOAT4* XM_CALLCONV XMVector2TransformStream +( + XMFLOAT4* pOutputStream, + size_t OutputStride, + const XMFLOAT2* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT2)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT2)); + + assert(OutputStride >= sizeof(XMFLOAT4)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT4)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat2(reinterpret_cast(pInputVector)); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Y, row1, row3); + Result = XMVectorMultiplyAdd(X, row0, Result); + + #ifdef _PREFAST_ + #pragma prefast(push) + #pragma prefast(disable : 26015, "PREfast noise: Esp:1307" ) + #endif + + XMStoreFloat4(reinterpret_cast(pOutputVector), Result); + + #ifdef _PREFAST_ + #pragma prefast(pop) + #endif + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT2)) && (OutputStride == sizeof(XMFLOAT4))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x2_t V = vld2q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + float32x2_t r3 = vget_low_f32(row3); + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Ax+M + XMVECTOR vResult1 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Bx+N + + XM_PREFETCH(pInputVector); + + r3 = vget_high_f32(row3); + r = vget_high_f32(row0); + XMVECTOR vResult2 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Cx+O + XMVECTOR vResult3 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Dx+P + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(row1); + vResult2 = vmlaq_lane_f32(vResult2, V.val[1], r, 0); // Cx+Gy+O + vResult3 = vmlaq_lane_f32(vResult3, V.val[1], r, 1); // Dx+Hy+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + float32x4x4_t R; + R.val[0] = vResult0; + R.val[1] = vResult1; + R.val[2] = vResult2; + R.val[3] = vResult3; + + vst4q_f32(reinterpret_cast(pOutputVector), R); + pOutputVector += sizeof(XMFLOAT4) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + float32x2_t V = vld1_f32(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vResult = vmlaq_lane_f32(row3, row0, V, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, V, 1); // Y + + vst1q_f32(reinterpret_cast(pOutputVector), vResult); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_AVX2_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + __m256 row0 = _mm256_broadcast_ps(&M.r[0]); + __m256 row1 = _mm256_broadcast_ps(&M.r[1]); + __m256 row3 = _mm256_broadcast_ps(&M.r[3]); + + if (InputStride == sizeof(XMFLOAT2)) + { + if (OutputStride == sizeof(XMFLOAT4)) + { + if (!(reinterpret_cast(pOutputStream) & 0x1F)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempB = _mm256_fmadd_ps(Y1, row1, row3); + __m256 vTempB2 = _mm256_fmadd_ps(Y2, row1, row3); + __m256 vTempA = _mm256_mul_ps(X1, row0); + __m256 vTempA2 = _mm256_mul_ps(X2, row0); + vTempA = _mm256_add_ps(vTempA, vTempB); + vTempA2 = _mm256_add_ps(vTempA2, vTempB2); + + X1 = _mm256_insertf128_ps(vTempA, _mm256_castps256_ps128(vTempA2), 1); + XM256_STREAM_PS(reinterpret_cast(pOutputVector), X1); + pOutputVector += sizeof(XMFLOAT4) * 2; + + X2 = _mm256_insertf128_ps(vTempA2, _mm256_extractf128_ps(vTempA, 1), 0); + XM256_STREAM_PS(reinterpret_cast(pOutputVector), X2); + pOutputVector += sizeof(XMFLOAT4) * 2; + + i += 4; + } + } + else + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempB = _mm256_fmadd_ps(Y1, row1, row3); + __m256 vTempB2 = _mm256_fmadd_ps(Y2, row1, row3); + __m256 vTempA = _mm256_mul_ps(X1, row0); + __m256 vTempA2 = _mm256_mul_ps(X2, row0); + vTempA = _mm256_add_ps(vTempA, vTempB); + vTempA2 = _mm256_add_ps(vTempA2, vTempB2); + + X1 = _mm256_insertf128_ps(vTempA, _mm256_castps256_ps128(vTempA2), 1); + _mm256_storeu_ps(reinterpret_cast(pOutputVector), X1); + pOutputVector += sizeof(XMFLOAT4) * 2; + + X2 = _mm256_insertf128_ps(vTempA2, _mm256_extractf128_ps(vTempA, 1), 0); + _mm256_storeu_ps(reinterpret_cast(pOutputVector), X2); + pOutputVector += sizeof(XMFLOAT4) * 2; + + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempB = _mm256_fmadd_ps(Y1, row1, row3); + __m256 vTempB2 = _mm256_fmadd_ps(Y2, row1, row3); + __m256 vTempA = _mm256_mul_ps(X1, row0); + __m256 vTempA2 = _mm256_mul_ps(X2, row0); + vTempA = _mm256_add_ps(vTempA, vTempB); + vTempA2 = _mm256_add_ps(vTempA2, vTempB2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), _mm256_castps256_ps128(vTempA)); + pOutputVector += OutputStride; + + _mm_storeu_ps(reinterpret_cast(pOutputVector), _mm256_castps256_ps128(vTempA2)); + pOutputVector += OutputStride; + + _mm_storeu_ps(reinterpret_cast(pOutputVector), _mm256_extractf128_ps(vTempA, 1)); + pOutputVector += OutputStride; + + _mm_storeu_ps(reinterpret_cast(pOutputVector), _mm256_extractf128_ps(vTempA2, 1)); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + if (i < VectorCount) + { + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + for (; i < VectorCount; i++) + { + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(xy, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(xy, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t two = VectorCount >> 1; + if (two > 0) + { + if (InputStride == sizeof(XMFLOAT2)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF) && !(OutputStride & 0xF)) + { + // Packed input, aligned output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = XM_FMADD_PS(Y, row1, row3); + vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 2; + } + } + else + { + // Packed input, unaligned output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = XM_FMADD_PS(Y, row1, row3); + vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 2; + } + } + } + } + + if (!(reinterpret_cast(pInputVector) & 0xF) && !(InputStride & 0xF)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF) && !(OutputStride & 0xF)) + { + // Aligned input, aligned output + for (; i < VectorCount; i++) + { + XMVECTOR V = _mm_castsi128_ps(_mm_loadl_epi64(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + } + else + { + // Aligned input, unaligned output + for (; i < VectorCount; i++) + { + XMVECTOR V = _mm_castsi128_ps(_mm_loadl_epi64(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + } + } + else + { + // Unaligned input + for (; i < VectorCount; i++) + { + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(xy, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(xy, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2TransformCoord +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Y, M.r[1], M.r[3]); + Result = XMVectorMultiplyAdd(X, M.r[0], Result); + + XMVECTOR W = XMVectorSplatW(Result); + return XMVectorDivide(Result, W); +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMFLOAT2* XM_CALLCONV XMVector2TransformCoordStream +( + XMFLOAT2* pOutputStream, + size_t OutputStride, + const XMFLOAT2* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT2)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT2)); + + assert(OutputStride >= sizeof(XMFLOAT2)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT2)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat2(reinterpret_cast(pInputVector)); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Y, row1, row3); + Result = XMVectorMultiplyAdd(X, row0, Result); + + XMVECTOR W = XMVectorSplatW(Result); + + Result = XMVectorDivide(Result, W); + + #ifdef _PREFAST_ + #pragma prefast(push) + #pragma prefast(disable : 26015, "PREfast noise: Esp:1307" ) + #endif + + XMStoreFloat2(reinterpret_cast(pOutputVector), Result); + + #ifdef _PREFAST_ + #pragma prefast(pop) + #endif + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT2)) && (OutputStride == sizeof(XMFLOAT2))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x2_t V = vld2q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + float32x2_t r3 = vget_low_f32(row3); + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Ax+M + XMVECTOR vResult1 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Bx+N + + XM_PREFETCH(pInputVector); + + r3 = vget_high_f32(row3); + r = vget_high_f32(row0); + XMVECTOR W = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Dx+P + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(row1); + W = vmlaq_lane_f32(W, V.val[1], r, 1); // Dx+Hy+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + V.val[0] = vdivq_f32(vResult0, W); + V.val[1] = vdivq_f32(vResult1, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + V.val[0] = vmulq_f32(vResult0, Reciprocal); + V.val[1] = vmulq_f32(vResult1, Reciprocal); + #endif + + vst2q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT2) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + float32x2_t V = vld1_f32(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vResult = vmlaq_lane_f32(row3, row0, V, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, V, 1); // Y + + V = vget_high_f32(vResult); + float32x2_t W = vdup_lane_f32(V, 1); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + V = vget_low_f32(vResult); + V = vdiv_f32(V, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal for W + float32x2_t Reciprocal = vrecpe_f32(W); + float32x2_t S = vrecps_f32(Reciprocal, W); + Reciprocal = vmul_f32(S, Reciprocal); + S = vrecps_f32(Reciprocal, W); + Reciprocal = vmul_f32(S, Reciprocal); + + V = vget_low_f32(vResult); + V = vmul_f32(V, Reciprocal); + #endif + + vst1_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_AVX2_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + __m256 row0 = _mm256_broadcast_ps(&M.r[0]); + __m256 row1 = _mm256_broadcast_ps(&M.r[1]); + __m256 row3 = _mm256_broadcast_ps(&M.r[3]); + + if (InputStride == sizeof(XMFLOAT2)) + { + if (OutputStride == sizeof(XMFLOAT2)) + { + if (!(reinterpret_cast(pOutputStream) & 0x1F)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempB = _mm256_fmadd_ps(Y1, row1, row3); + __m256 vTempB2 = _mm256_fmadd_ps(Y2, row1, row3); + __m256 vTempA = _mm256_mul_ps(X1, row0); + __m256 vTempA2 = _mm256_mul_ps(X2, row0); + vTempA = _mm256_add_ps(vTempA, vTempB); + vTempA2 = _mm256_add_ps(vTempA2, vTempB2); + + __m256 W = _mm256_shuffle_ps(vTempA, vTempA, _MM_SHUFFLE(3, 3, 3, 3)); + vTempA = _mm256_div_ps(vTempA, W); + + W = _mm256_shuffle_ps(vTempA2, vTempA2, _MM_SHUFFLE(3, 3, 3, 3)); + vTempA2 = _mm256_div_ps(vTempA2, W); + + X1 = _mm256_shuffle_ps(vTempA, vTempA2, 0x44); + XM256_STREAM_PS(reinterpret_cast(pOutputVector), X1); + pOutputVector += sizeof(XMFLOAT2) * 4; + + i += 4; + } + } + else + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempB = _mm256_fmadd_ps(Y1, row1, row3); + __m256 vTempB2 = _mm256_fmadd_ps(Y2, row1, row3); + __m256 vTempA = _mm256_mul_ps(X1, row0); + __m256 vTempA2 = _mm256_mul_ps(X2, row0); + vTempA = _mm256_add_ps(vTempA, vTempB); + vTempA2 = _mm256_add_ps(vTempA2, vTempB2); + + __m256 W = _mm256_shuffle_ps(vTempA, vTempA, _MM_SHUFFLE(3, 3, 3, 3)); + vTempA = _mm256_div_ps(vTempA, W); + + W = _mm256_shuffle_ps(vTempA2, vTempA2, _MM_SHUFFLE(3, 3, 3, 3)); + vTempA2 = _mm256_div_ps(vTempA2, W); + + X1 = _mm256_shuffle_ps(vTempA, vTempA2, 0x44); + _mm256_storeu_ps(reinterpret_cast(pOutputVector), X1); + pOutputVector += sizeof(XMFLOAT2) * 4; + + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempB = _mm256_fmadd_ps(Y1, row1, row3); + __m256 vTempB2 = _mm256_fmadd_ps(Y2, row1, row3); + __m256 vTempA = _mm256_mul_ps(X1, row0); + __m256 vTempA2 = _mm256_mul_ps(X2, row0); + vTempA = _mm256_add_ps(vTempA, vTempB); + vTempA2 = _mm256_add_ps(vTempA2, vTempB2); + + __m256 W = _mm256_shuffle_ps(vTempA, vTempA, _MM_SHUFFLE(3, 3, 3, 3)); + vTempA = _mm256_div_ps(vTempA, W); + + W = _mm256_shuffle_ps(vTempA2, vTempA2, _MM_SHUFFLE(3, 3, 3, 3)); + vTempA2 = _mm256_div_ps(vTempA2, W); + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_castps256_ps128(vTempA))); + pOutputVector += OutputStride; + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_castps256_ps128(vTempA2))); + pOutputVector += OutputStride; + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_extractf128_ps(vTempA, 1))); + pOutputVector += OutputStride; + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_extractf128_ps(vTempA2, 1))); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + if (i < VectorCount) + { + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + for (; i < VectorCount; i++) + { + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(xy, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(xy, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t two = VectorCount >> 1; + if (two > 0) + { + if (InputStride == sizeof(XMFLOAT2)) + { + if (OutputStride == sizeof(XMFLOAT2)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + // Result 1 + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + XMVECTOR V1 = _mm_div_ps(vTemp, W); + + // Result 2 + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = XM_FMADD_PS(Y, row1, row3); + vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + XMVECTOR V2 = _mm_div_ps(vTemp, W); + + vTemp = _mm_movelh_ps(V1, V2); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += sizeof(XMFLOAT2) * 2; + + i += 2; + } + } + else + { + // Packed input, unaligned & packed output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + // Result 1 + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + XMVECTOR V1 = _mm_div_ps(vTemp, W); + + // Result 2 + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = XM_FMADD_PS(Y, row1, row3); + vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + XMVECTOR V2 = _mm_div_ps(vTemp, W); + + vTemp = _mm_movelh_ps(V1, V2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += sizeof(XMFLOAT2) * 2; + + i += 2; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + // Result 1 + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + + // Result 2 + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = XM_FMADD_PS(Y, row1, row3); + vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + + i += 2; + } + } + } + } + + if (!(reinterpret_cast(pInputVector) & 0xF) && !(InputStride & 0xF)) + { + // Aligned input + for (; i < VectorCount; i++) + { + XMVECTOR V = _mm_castsi128_ps(_mm_loadl_epi64(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + } + } + else + { + // Unaligned input + for (; i < VectorCount; i++) + { + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(xy, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(xy, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Y, row1, row3); + XMVECTOR vTemp2 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector2TransformNormal +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiply(Y, M.r[1]); + Result = XMVectorMultiplyAdd(X, M.r[0], Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + float32x4_t Result = vmulq_lane_f32(M.r[1], VL, 1); // Y + return vmlaq_lane_f32(Result, M.r[0], VL, 0); // X +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); // Y + vResult = _mm_mul_ps(vResult, M.r[1]); + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); // X + vResult = XM_FMADD_PS(vTemp, M.r[0], vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMFLOAT2* XM_CALLCONV XMVector2TransformNormalStream +( + XMFLOAT2* pOutputStream, + size_t OutputStride, + const XMFLOAT2* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT2)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT2)); + + assert(OutputStride >= sizeof(XMFLOAT2)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT2)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat2(reinterpret_cast(pInputVector)); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiply(Y, row1); + Result = XMVectorMultiplyAdd(X, row0, Result); + + #ifdef _PREFAST_ + #pragma prefast(push) + #pragma prefast(disable : 26015, "PREfast noise: Esp:1307" ) + #endif + + XMStoreFloat2(reinterpret_cast(pOutputVector), Result); + + #ifdef _PREFAST_ + #pragma prefast(pop) + #endif + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT2)) && (OutputStride == sizeof(XMFLOAT2))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x2_t V = vld2q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmulq_lane_f32(V.val[0], r, 0); // Ax + XMVECTOR vResult1 = vmulq_lane_f32(V.val[0], r, 1); // Bx + + XM_PREFETCH(pInputVector); + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + V.val[0] = vResult0; + V.val[1] = vResult1; + + vst2q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT2) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + float32x2_t V = vld1_f32(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vResult = vmulq_lane_f32(row0, V, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, V, 1); // Y + + V = vget_low_f32(vResult); + vst1_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_AVX2_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + __m256 row0 = _mm256_broadcast_ps(&M.r[0]); + __m256 row1 = _mm256_broadcast_ps(&M.r[1]); + + if (InputStride == sizeof(XMFLOAT2)) + { + if (OutputStride == sizeof(XMFLOAT2)) + { + if (!(reinterpret_cast(pOutputStream) & 0x1F)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempA = _mm256_mul_ps(Y1, row1); + __m256 vTempB = _mm256_mul_ps(Y2, row1); + vTempA = _mm256_fmadd_ps(X1, row0, vTempA); + vTempB = _mm256_fmadd_ps(X2, row0, vTempB); + + X1 = _mm256_shuffle_ps(vTempA, vTempB, 0x44); + XM256_STREAM_PS(reinterpret_cast(pOutputVector), X1); + pOutputVector += sizeof(XMFLOAT2) * 4; + + i += 4; + } + } + else + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempA = _mm256_mul_ps(Y1, row1); + __m256 vTempB = _mm256_mul_ps(Y2, row1); + vTempA = _mm256_fmadd_ps(X1, row0, vTempA); + vTempB = _mm256_fmadd_ps(X2, row0, vTempB); + + X1 = _mm256_shuffle_ps(vTempA, vTempB, 0x44); + _mm256_storeu_ps(reinterpret_cast(pOutputVector), X1); + pOutputVector += sizeof(XMFLOAT2) * 4; + + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 4; + + __m256 Y2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + __m256 X2 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 Y1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 X1 = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256 vTempA = _mm256_mul_ps(Y1, row1); + __m256 vTempB = _mm256_mul_ps(Y2, row1); + vTempA = _mm256_fmadd_ps(X1, row0, vTempA); + vTempB = _mm256_fmadd_ps(X2, row0, vTempB); + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_castps256_ps128(vTempA))); + pOutputVector += OutputStride; + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_castps256_ps128(vTempB))); + pOutputVector += OutputStride; + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_extractf128_ps(vTempA, 1))); + pOutputVector += OutputStride; + + _mm_store_sd(reinterpret_cast(pOutputVector), + _mm_castps_pd(_mm256_extractf128_ps(vTempB, 1))); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + if (i < VectorCount) + { + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + + for (; i < VectorCount; i++) + { + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(xy, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(xy, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Y, row1); + vTemp = XM_FMADD_PS(X, row0, vTemp); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + + size_t i = 0; + size_t two = VectorCount >> 1; + if (two > 0) + { + if (InputStride == sizeof(XMFLOAT2)) + { + if (OutputStride == sizeof(XMFLOAT2)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + // Result 1 + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Y, row1); + XMVECTOR V1 = XM_FMADD_PS(X, row0, vTemp); + + // Result 2 + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = _mm_mul_ps(Y, row1); + XMVECTOR V2 = XM_FMADD_PS(X, row0, vTemp); + + vTemp = _mm_movelh_ps(V1, V2); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += sizeof(XMFLOAT2) * 2; + + i += 2; + } + } + else + { + // Packed input, unaligned & packed output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + // Result 1 + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Y, row1); + XMVECTOR V1 = XM_FMADD_PS(X, row0, vTemp); + + // Result 2 + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = _mm_mul_ps(Y, row1); + XMVECTOR V2 = XM_FMADD_PS(X, row0, vTemp); + + vTemp = _mm_movelh_ps(V1, V2); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += sizeof(XMFLOAT2) * 2; + + i += 2; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < two; ++j) + { + XMVECTOR V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT2) * 2; + + // Result 1 + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Y, row1); + vTemp = XM_FMADD_PS(X, row0, vTemp); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + + // Result 2 + Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + X = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + + vTemp = _mm_mul_ps(Y, row1); + vTemp = XM_FMADD_PS(X, row0, vTemp); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + + i += 2; + } + } + } + } + + if (!(reinterpret_cast(pInputVector) & 0xF) && !(InputStride & 0xF)) + { + // Aligned input + for (; i < VectorCount; i++) + { + XMVECTOR V = _mm_castsi128_ps(_mm_loadl_epi64(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Y, row1); + vTemp = XM_FMADD_PS(X, row0, vTemp); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + } + } + else + { + // Unaligned input + for (; i < VectorCount; i++) + { + __m128 xy = _mm_castpd_ps(_mm_load_sd(reinterpret_cast(pInputVector))); + pInputVector += InputStride; + + XMVECTOR Y = XM_PERMUTE_PS(xy, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(xy, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Y, row1); + vTemp = XM_FMADD_PS(X, row0, vTemp); + + _mm_store_sd(reinterpret_cast(pOutputVector), _mm_castps_pd(vTemp)); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +/**************************************************************************** + * + * 3D Vector + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3Equal +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] == V2.vector4_f32[0]) && (V1.vector4_f32[1] == V2.vector4_f32[1]) && (V1.vector4_f32[2] == V2.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 7) == 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector3EqualR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t CR = 0; + if ((V1.vector4_f32[0] == V2.vector4_f32[0]) && + (V1.vector4_f32[1] == V2.vector4_f32[1]) && + (V1.vector4_f32[2] == V2.vector4_f32[2])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] != V2.vector4_f32[0]) && + (V1.vector4_f32[1] != V2.vector4_f32[1]) && + (V1.vector4_f32[2] != V2.vector4_f32[2])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU; + + uint32_t CR = 0; + if (r == 0xFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + int iTest = _mm_movemask_ps(vTemp) & 7; + uint32_t CR = 0; + if (iTest == 7) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3EqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_u32[0] == V2.vector4_u32[0]) && (V1.vector4_u32[1] == V2.vector4_u32[1]) && (V1.vector4_u32[2] == V2.vector4_u32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return (((_mm_movemask_ps(_mm_castsi128_ps(vTemp)) & 7) == 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector3EqualIntR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t CR = 0; + if ((V1.vector4_u32[0] == V2.vector4_u32[0]) && + (V1.vector4_u32[1] == V2.vector4_u32[1]) && + (V1.vector4_u32[2] == V2.vector4_u32[2])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_u32[0] != V2.vector4_u32[0]) && + (V1.vector4_u32[1] != V2.vector4_u32[1]) && + (V1.vector4_u32[2] != V2.vector4_u32[2])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU; + + uint32_t CR = 0; + if (r == 0xFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + int iTemp = _mm_movemask_ps(_mm_castsi128_ps(vTemp)) & 7; + uint32_t CR = 0; + if (iTemp == 7) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTemp) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3NearEqual +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR Epsilon +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float dx, dy, dz; + + dx = fabsf(V1.vector4_f32[0] - V2.vector4_f32[0]); + dy = fabsf(V1.vector4_f32[1] - V2.vector4_f32[1]); + dz = fabsf(V1.vector4_f32[2] - V2.vector4_f32[2]); + return (((dx <= Epsilon.vector4_f32[0]) && + (dy <= Epsilon.vector4_f32[1]) && + (dz <= Epsilon.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vDelta = vsubq_f32(V1, V2); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + uint32x4_t vResult = vacleq_f32(vDelta, Epsilon); +#else + uint32x4_t vResult = vcleq_f32(vabsq_f32(vDelta), Epsilon); +#endif + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + // Get the difference + XMVECTOR vDelta = _mm_sub_ps(V1, V2); + // Get the absolute value of the difference + XMVECTOR vTemp = _mm_setzero_ps(); + vTemp = _mm_sub_ps(vTemp, vDelta); + vTemp = _mm_max_ps(vTemp, vDelta); + vTemp = _mm_cmple_ps(vTemp, Epsilon); + // w is don't care + return (((_mm_movemask_ps(vTemp) & 7) == 0x7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3NotEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] != V2.vector4_f32[0]) || (V1.vector4_f32[1] != V2.vector4_f32[1]) || (V1.vector4_f32[2] != V2.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) != 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 7) != 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3NotEqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_u32[0] != V2.vector4_u32[0]) || (V1.vector4_u32[1] != V2.vector4_u32[1]) || (V1.vector4_u32[2] != V2.vector4_u32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) != 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return (((_mm_movemask_ps(_mm_castsi128_ps(vTemp)) & 7) != 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3Greater +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] > V2.vector4_f32[0]) && (V1.vector4_f32[1] > V2.vector4_f32[1]) && (V1.vector4_f32[2] > V2.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgtq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 7) == 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector3GreaterR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t CR = 0; + if ((V1.vector4_f32[0] > V2.vector4_f32[0]) && + (V1.vector4_f32[1] > V2.vector4_f32[1]) && + (V1.vector4_f32[2] > V2.vector4_f32[2])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] <= V2.vector4_f32[0]) && + (V1.vector4_f32[1] <= V2.vector4_f32[1]) && + (V1.vector4_f32[2] <= V2.vector4_f32[2])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgtq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU; + + uint32_t CR = 0; + if (r == 0xFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + uint32_t CR = 0; + int iTest = _mm_movemask_ps(vTemp) & 7; + if (iTest == 7) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3GreaterOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] >= V2.vector4_f32[0]) && (V1.vector4_f32[1] >= V2.vector4_f32[1]) && (V1.vector4_f32[2] >= V2.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgeq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 7) == 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector3GreaterOrEqualR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + uint32_t CR = 0; + if ((V1.vector4_f32[0] >= V2.vector4_f32[0]) && + (V1.vector4_f32[1] >= V2.vector4_f32[1]) && + (V1.vector4_f32[2] >= V2.vector4_f32[2])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] < V2.vector4_f32[0]) && + (V1.vector4_f32[1] < V2.vector4_f32[1]) && + (V1.vector4_f32[2] < V2.vector4_f32[2])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgeq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU; + + uint32_t CR = 0; + if (r == 0xFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + uint32_t CR = 0; + int iTest = _mm_movemask_ps(vTemp) & 7; + if (iTest == 7) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3Less +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] < V2.vector4_f32[0]) && (V1.vector4_f32[1] < V2.vector4_f32[1]) && (V1.vector4_f32[2] < V2.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcltq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmplt_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 7) == 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3LessOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] <= V2.vector4_f32[0]) && (V1.vector4_f32[1] <= V2.vector4_f32[1]) && (V1.vector4_f32[2] <= V2.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcleq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmple_ps(V1, V2); + return (((_mm_movemask_ps(vTemp) & 7) == 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3InBounds +( + FXMVECTOR V, + FXMVECTOR Bounds +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V.vector4_f32[0] <= Bounds.vector4_f32[0] && V.vector4_f32[0] >= -Bounds.vector4_f32[0]) && + (V.vector4_f32[1] <= Bounds.vector4_f32[1] && V.vector4_f32[1] >= -Bounds.vector4_f32[1]) && + (V.vector4_f32[2] <= Bounds.vector4_f32[2] && V.vector4_f32[2] >= -Bounds.vector4_f32[2])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Test if less than or equal + uint32x4_t ivTemp1 = vcleq_f32(V, Bounds); + // Negate the bounds + float32x4_t vTemp2 = vnegq_f32(Bounds); + // Test if greater or equal (Reversed) + uint32x4_t ivTemp2 = vcleq_f32(vTemp2, V); + // Blend answers + ivTemp1 = vandq_u32(ivTemp1, ivTemp2); + // in bounds? + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(ivTemp1)), vget_high_u8(vreinterpretq_u8_u32(ivTemp1))); + uint16x4x2_t vTemp3 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp3.val[1]), 1) & 0xFFFFFFU) == 0xFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + // Test if less than or equal + XMVECTOR vTemp1 = _mm_cmple_ps(V, Bounds); + // Negate the bounds + XMVECTOR vTemp2 = _mm_mul_ps(Bounds, g_XMNegativeOne); + // Test if greater or equal (Reversed) + vTemp2 = _mm_cmple_ps(vTemp2, V); + // Blend answers + vTemp1 = _mm_and_ps(vTemp1, vTemp2); + // x,y and z in bounds? (w is don't care) + return (((_mm_movemask_ps(vTemp1) & 0x7) == 0x7) != 0); +#else + return XMComparisonAllInBounds(XMVector3InBoundsR(V, Bounds)); +#endif +} + +//------------------------------------------------------------------------------ + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(push) +#pragma float_control(precise, on) +#endif + +inline bool XM_CALLCONV XMVector3IsNaN(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + return (XMISNAN(V.vector4_f32[0]) || + XMISNAN(V.vector4_f32[1]) || + XMISNAN(V.vector4_f32[2])); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + return isnan(vgetq_lane_f32(V, 0)) || isnan(vgetq_lane_f32(V, 1)) || isnan(vgetq_lane_f32(V, 2)); +#else +// Test against itself. NaN is always not equal + uint32x4_t vTempNan = vceqq_f32(V, V); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vTempNan)), vget_high_u8(vreinterpretq_u8_u32(vTempNan))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + // If x or y or z are NaN, the mask is zero + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) != 0xFFFFFFU); +#endif +#elif defined(_XM_SSE_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + XM_ALIGNED_DATA(16) float tmp[4]; + _mm_store_ps(tmp, V); + return isnan(tmp[0]) || isnan(tmp[1]) || isnan(tmp[2]); +#else +// Test against itself. NaN is always not equal + XMVECTOR vTempNan = _mm_cmpneq_ps(V, V); + // If x or y or z are NaN, the mask is non-zero + return ((_mm_movemask_ps(vTempNan) & 7) != 0); +#endif +#endif +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(pop) +#endif + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector3IsInfinite(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (XMISINF(V.vector4_f32[0]) || + XMISINF(V.vector4_f32[1]) || + XMISINF(V.vector4_f32[2])); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Mask off the sign bit + uint32x4_t vTempInf = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + // Compare to infinity + vTempInf = vceqq_f32(vreinterpretq_f32_u32(vTempInf), g_XMInfinity); + // If any are infinity, the signs are true. + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vTempInf)), vget_high_u8(vreinterpretq_u8_u32(vTempInf))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return ((vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) & 0xFFFFFFU) != 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Mask off the sign bit + __m128 vTemp = _mm_and_ps(V, g_XMAbsMask); + // Compare to infinity + vTemp = _mm_cmpeq_ps(vTemp, g_XMInfinity); + // If x,y or z are infinity, the signs are true. + return ((_mm_movemask_ps(vTemp) & 7) != 0); +#endif +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Dot +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float fValue = V1.vector4_f32[0] * V2.vector4_f32[0] + V1.vector4_f32[1] * V2.vector4_f32[1] + V1.vector4_f32[2] * V2.vector4_f32[2]; + XMVECTORF32 vResult; + vResult.f[0] = + vResult.f[1] = + vResult.f[2] = + vResult.f[3] = fValue; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vTemp = vmulq_f32(V1, V2); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + return vcombine_f32(v1, v1); +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_dp_ps(V1, V2, 0x7f); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vTemp = _mm_mul_ps(V1, V2); + vTemp = _mm_and_ps(vTemp, g_XMMask3); + vTemp = _mm_hadd_ps(vTemp, vTemp); + return _mm_hadd_ps(vTemp, vTemp); +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product + XMVECTOR vDot = _mm_mul_ps(V1, V2); + // x=Dot.vector4_f32[1], y=Dot.vector4_f32[2] + XMVECTOR vTemp = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(2, 1, 2, 1)); + // Result.vector4_f32[0] = x+y + vDot = _mm_add_ss(vDot, vTemp); + // x=Dot.vector4_f32[2] + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // Result.vector4_f32[0] = (x+y)+z + vDot = _mm_add_ss(vDot, vTemp); + // Splat x + return XM_PERMUTE_PS(vDot, _MM_SHUFFLE(0, 0, 0, 0)); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Cross +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + // [ V1.y*V2.z - V1.z*V2.y, V1.z*V2.x - V1.x*V2.z, V1.x*V2.y - V1.y*V2.x ] + +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + (V1.vector4_f32[1] * V2.vector4_f32[2]) - (V1.vector4_f32[2] * V2.vector4_f32[1]), + (V1.vector4_f32[2] * V2.vector4_f32[0]) - (V1.vector4_f32[0] * V2.vector4_f32[2]), + (V1.vector4_f32[0] * V2.vector4_f32[1]) - (V1.vector4_f32[1] * V2.vector4_f32[0]), + 0.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t v1xy = vget_low_f32(V1); + float32x2_t v2xy = vget_low_f32(V2); + + float32x2_t v1yx = vrev64_f32(v1xy); + float32x2_t v2yx = vrev64_f32(v2xy); + + float32x2_t v1zz = vdup_lane_f32(vget_high_f32(V1), 0); + float32x2_t v2zz = vdup_lane_f32(vget_high_f32(V2), 0); + + XMVECTOR vResult = vmulq_f32(vcombine_f32(v1yx, v1xy), vcombine_f32(v2zz, v2yx)); + vResult = vmlsq_f32(vResult, vcombine_f32(v1zz, v1yx), vcombine_f32(v2yx, v2xy)); + vResult = vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(vResult), g_XMFlipY)); + return vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(vResult), g_XMMask3)); +#elif defined(_XM_SSE_INTRINSICS_) + // y1,z1,x1,w1 + XMVECTOR vTemp1 = XM_PERMUTE_PS(V1, _MM_SHUFFLE(3, 0, 2, 1)); + // z2,x2,y2,w2 + XMVECTOR vTemp2 = XM_PERMUTE_PS(V2, _MM_SHUFFLE(3, 1, 0, 2)); + // Perform the left operation + XMVECTOR vResult = _mm_mul_ps(vTemp1, vTemp2); + // z1,x1,y1,w1 + vTemp1 = XM_PERMUTE_PS(vTemp1, _MM_SHUFFLE(3, 0, 2, 1)); + // y2,z2,x2,w2 + vTemp2 = XM_PERMUTE_PS(vTemp2, _MM_SHUFFLE(3, 1, 0, 2)); + // Perform the right operation + vResult = XM_FNMADD_PS(vTemp1, vTemp2, vResult); + // Set w to zero + return _mm_and_ps(vResult, g_XMMask3); +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3LengthSq(FXMVECTOR V) noexcept +{ + return XMVector3Dot(V, V); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3ReciprocalLengthEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector3LengthSq(V); + Result = XMVectorReciprocalSqrtEst(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot3 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + // Reciprocal sqrt (estimate) + v2 = vrsqrte_f32(v1); + return vcombine_f32(v2, v2); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x7f); + return _mm_rsqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_and_ps(vLengthSq, g_XMMask3); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_rsqrt_ps(vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y and z + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and y + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 2, 1, 2)); + // x+z, y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + // y,y,y,y + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // x+z+y,??,??,?? + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + // Splat the length squared + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + // Get the reciprocal + vLengthSq = _mm_rsqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3ReciprocalLength(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector3LengthSq(V); + Result = XMVectorReciprocalSqrt(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot3 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + // Reciprocal sqrt + float32x2_t S0 = vrsqrte_f32(v1); + float32x2_t P0 = vmul_f32(v1, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(v1, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t Result = vmul_f32(S1, R1); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x7f); + XMVECTOR vLengthSq = _mm_sqrt_ps(vTemp); + return _mm_div_ps(g_XMOne, vLengthSq); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vDot = _mm_mul_ps(V, V); + vDot = _mm_and_ps(vDot, g_XMMask3); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_sqrt_ps(vDot); + vDot = _mm_div_ps(g_XMOne, vDot); + return vDot; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product + XMVECTOR vDot = _mm_mul_ps(V, V); + // x=Dot.y, y=Dot.z + XMVECTOR vTemp = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(2, 1, 2, 1)); + // Result.x = x+y + vDot = _mm_add_ss(vDot, vTemp); + // x=Dot.z + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // Result.x = (x+y)+z + vDot = _mm_add_ss(vDot, vTemp); + // Splat x + vDot = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(0, 0, 0, 0)); + // Get the reciprocal + vDot = _mm_sqrt_ps(vDot); + // Get the reciprocal + vDot = _mm_div_ps(g_XMOne, vDot); + return vDot; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3LengthEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector3LengthSq(V); + Result = XMVectorSqrtEst(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot3 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + const float32x2_t zero = vdup_n_f32(0); + uint32x2_t VEqualsZero = vceq_f32(v1, zero); + // Sqrt (estimate) + float32x2_t Result = vrsqrte_f32(v1); + Result = vmul_f32(v1, Result); + Result = vbsl_f32(VEqualsZero, zero, Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x7f); + return _mm_sqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_and_ps(vLengthSq, g_XMMask3); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y and z + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and y + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 2, 1, 2)); + // x+z, y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + // y,y,y,y + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // x+z+y,??,??,?? + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + // Splat the length squared + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + // Get the length + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Length(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector3LengthSq(V); + Result = XMVectorSqrt(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot3 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + const float32x2_t zero = vdup_n_f32(0); + uint32x2_t VEqualsZero = vceq_f32(v1, zero); + // Sqrt + float32x2_t S0 = vrsqrte_f32(v1); + float32x2_t P0 = vmul_f32(v1, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(v1, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t Result = vmul_f32(S1, R1); + Result = vmul_f32(v1, Result); + Result = vbsl_f32(VEqualsZero, zero, Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x7f); + return _mm_sqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_and_ps(vLengthSq, g_XMMask3); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y and z + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and y + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 2, 1, 2)); + // x+z, y + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + // y,y,y,y + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // x+z+y,??,??,?? + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + // Splat the length squared + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + // Get the length + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ +// XMVector3NormalizeEst uses a reciprocal estimate and +// returns QNaN on zero and infinite vectors. + +inline XMVECTOR XM_CALLCONV XMVector3NormalizeEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector3ReciprocalLength(V); + Result = XMVectorMultiply(V, Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot3 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + // Reciprocal sqrt (estimate) + v2 = vrsqrte_f32(v1); + // Normalize + return vmulq_f32(V, vcombine_f32(v2, v2)); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0x7f); + XMVECTOR vResult = _mm_rsqrt_ps(vTemp); + return _mm_mul_ps(vResult, V); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vDot = _mm_mul_ps(V, V); + vDot = _mm_and_ps(vDot, g_XMMask3); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_rsqrt_ps(vDot); + vDot = _mm_mul_ps(vDot, V); + return vDot; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product + XMVECTOR vDot = _mm_mul_ps(V, V); + // x=Dot.y, y=Dot.z + XMVECTOR vTemp = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(2, 1, 2, 1)); + // Result.x = x+y + vDot = _mm_add_ss(vDot, vTemp); + // x=Dot.z + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + // Result.x = (x+y)+z + vDot = _mm_add_ss(vDot, vTemp); + // Splat x + vDot = XM_PERMUTE_PS(vDot, _MM_SHUFFLE(0, 0, 0, 0)); + // Get the reciprocal + vDot = _mm_rsqrt_ps(vDot); + // Perform the normalization + vDot = _mm_mul_ps(vDot, V); + return vDot; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Normalize(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float fLength; + XMVECTOR vResult; + + vResult = XMVector3Length(V); + fLength = vResult.vector4_f32[0]; + + // Prevent divide by zero + if (fLength > 0) + { + fLength = 1.0f / fLength; + } + + vResult.vector4_f32[0] = V.vector4_f32[0] * fLength; + vResult.vector4_f32[1] = V.vector4_f32[1] * fLength; + vResult.vector4_f32[2] = V.vector4_f32[2] * fLength; + vResult.vector4_f32[3] = V.vector4_f32[3] * fLength; + return vResult; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot3 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vpadd_f32(v1, v1); + v2 = vdup_lane_f32(v2, 0); + v1 = vadd_f32(v1, v2); + uint32x2_t VEqualsZero = vceq_f32(v1, vdup_n_f32(0)); + uint32x2_t VEqualsInf = vceq_f32(v1, vget_low_f32(g_XMInfinity)); + // Reciprocal sqrt (2 iterations of Newton-Raphson) + float32x2_t S0 = vrsqrte_f32(v1); + float32x2_t P0 = vmul_f32(v1, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(v1, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + v2 = vmul_f32(S1, R1); + // Normalize + XMVECTOR vResult = vmulq_f32(V, vcombine_f32(v2, v2)); + vResult = vbslq_f32(vcombine_u32(VEqualsZero, VEqualsZero), vdupq_n_f32(0), vResult); + return vbslq_f32(vcombine_u32(VEqualsInf, VEqualsInf), g_XMQNaN, vResult); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vLengthSq = _mm_dp_ps(V, V, 0x7f); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Divide to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#elif defined(_XM_SSE3_INTRINSICS_) + // Perform the dot product on x,y and z only + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_and_ps(vLengthSq, g_XMMask3); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Divide to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y and z only + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 1, 2, 1)); + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(1, 1, 1, 1)); + vLengthSq = _mm_add_ss(vLengthSq, vTemp); + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(0, 0, 0, 0)); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Divide to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3ClampLength +( + FXMVECTOR V, + float LengthMin, + float LengthMax +) noexcept +{ + XMVECTOR ClampMax = XMVectorReplicate(LengthMax); + XMVECTOR ClampMin = XMVectorReplicate(LengthMin); + + return XMVector3ClampLengthV(V, ClampMin, ClampMax); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3ClampLengthV +( + FXMVECTOR V, + FXMVECTOR LengthMin, + FXMVECTOR LengthMax +) noexcept +{ + assert((XMVectorGetY(LengthMin) == XMVectorGetX(LengthMin)) && (XMVectorGetZ(LengthMin) == XMVectorGetX(LengthMin))); + assert((XMVectorGetY(LengthMax) == XMVectorGetX(LengthMax)) && (XMVectorGetZ(LengthMax) == XMVectorGetX(LengthMax))); + assert(XMVector3GreaterOrEqual(LengthMin, XMVectorZero())); + assert(XMVector3GreaterOrEqual(LengthMax, XMVectorZero())); + assert(XMVector3GreaterOrEqual(LengthMax, LengthMin)); + + XMVECTOR LengthSq = XMVector3LengthSq(V); + + const XMVECTOR Zero = XMVectorZero(); + + XMVECTOR RcpLength = XMVectorReciprocalSqrt(LengthSq); + + XMVECTOR InfiniteLength = XMVectorEqualInt(LengthSq, g_XMInfinity.v); + XMVECTOR ZeroLength = XMVectorEqual(LengthSq, Zero); + + XMVECTOR Normal = XMVectorMultiply(V, RcpLength); + + XMVECTOR Length = XMVectorMultiply(LengthSq, RcpLength); + + XMVECTOR Select = XMVectorEqualInt(InfiniteLength, ZeroLength); + Length = XMVectorSelect(LengthSq, Length, Select); + Normal = XMVectorSelect(LengthSq, Normal, Select); + + XMVECTOR ControlMax = XMVectorGreater(Length, LengthMax); + XMVECTOR ControlMin = XMVectorLess(Length, LengthMin); + + XMVECTOR ClampLength = XMVectorSelect(Length, LengthMax, ControlMax); + ClampLength = XMVectorSelect(ClampLength, LengthMin, ControlMin); + + XMVECTOR Result = XMVectorMultiply(Normal, ClampLength); + + // Preserve the original vector (with no precision loss) if the length falls within the given range + XMVECTOR Control = XMVectorEqualInt(ControlMax, ControlMin); + Result = XMVectorSelect(Result, V, Control); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Reflect +( + FXMVECTOR Incident, + FXMVECTOR Normal +) noexcept +{ + // Result = Incident - (2 * dot(Incident, Normal)) * Normal + + XMVECTOR Result = XMVector3Dot(Incident, Normal); + Result = XMVectorAdd(Result, Result); + Result = XMVectorNegativeMultiplySubtract(Result, Normal, Incident); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Refract +( + FXMVECTOR Incident, + FXMVECTOR Normal, + float RefractionIndex +) noexcept +{ + XMVECTOR Index = XMVectorReplicate(RefractionIndex); + return XMVector3RefractV(Incident, Normal, Index); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3RefractV +( + FXMVECTOR Incident, + FXMVECTOR Normal, + FXMVECTOR RefractionIndex +) noexcept +{ + // Result = RefractionIndex * Incident - Normal * (RefractionIndex * dot(Incident, Normal) + + // sqrt(1 - RefractionIndex * RefractionIndex * (1 - dot(Incident, Normal) * dot(Incident, Normal)))) + +#if defined(_XM_NO_INTRINSICS_) + + const XMVECTOR Zero = XMVectorZero(); + + XMVECTOR IDotN = XMVector3Dot(Incident, Normal); + + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + XMVECTOR R = XMVectorNegativeMultiplySubtract(IDotN, IDotN, g_XMOne.v); + R = XMVectorMultiply(R, RefractionIndex); + R = XMVectorNegativeMultiplySubtract(R, RefractionIndex, g_XMOne.v); + + if (XMVector4LessOrEqual(R, Zero)) + { + // Total internal reflection + return Zero; + } + else + { + // R = RefractionIndex * IDotN + sqrt(R) + R = XMVectorSqrt(R); + R = XMVectorMultiplyAdd(RefractionIndex, IDotN, R); + + // Result = RefractionIndex * Incident - Normal * R + XMVECTOR Result = XMVectorMultiply(RefractionIndex, Incident); + Result = XMVectorNegativeMultiplySubtract(Normal, R, Result); + + return Result; + } + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR IDotN = XMVector3Dot(Incident, Normal); + + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + float32x4_t R = vmlsq_f32(g_XMOne, IDotN, IDotN); + R = vmulq_f32(R, RefractionIndex); + R = vmlsq_f32(g_XMOne, R, RefractionIndex); + + uint32x4_t isrzero = vcleq_f32(R, g_XMZero); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(isrzero)), vget_high_u8(vreinterpretq_u8_u32(isrzero))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + + float32x4_t vResult; + if (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU) + { + // Total internal reflection + vResult = g_XMZero; + } + else + { + // Sqrt(R) + float32x4_t S0 = vrsqrteq_f32(R); + float32x4_t P0 = vmulq_f32(R, S0); + float32x4_t R0 = vrsqrtsq_f32(P0, S0); + float32x4_t S1 = vmulq_f32(S0, R0); + float32x4_t P1 = vmulq_f32(R, S1); + float32x4_t R1 = vrsqrtsq_f32(P1, S1); + float32x4_t S2 = vmulq_f32(S1, R1); + R = vmulq_f32(R, S2); + // R = RefractionIndex * IDotN + sqrt(R) + R = vmlaq_f32(R, RefractionIndex, IDotN); + // Result = RefractionIndex * Incident - Normal * R + vResult = vmulq_f32(RefractionIndex, Incident); + vResult = vmlsq_f32(vResult, R, Normal); + } + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Result = RefractionIndex * Incident - Normal * (RefractionIndex * dot(Incident, Normal) + + // sqrt(1 - RefractionIndex * RefractionIndex * (1 - dot(Incident, Normal) * dot(Incident, Normal)))) + XMVECTOR IDotN = XMVector3Dot(Incident, Normal); + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + XMVECTOR R = XM_FNMADD_PS(IDotN, IDotN, g_XMOne); + XMVECTOR R2 = _mm_mul_ps(RefractionIndex, RefractionIndex); + R = XM_FNMADD_PS(R, R2, g_XMOne); + + XMVECTOR vResult = _mm_cmple_ps(R, g_XMZero); + if (_mm_movemask_ps(vResult) == 0x0f) + { + // Total internal reflection + vResult = g_XMZero; + } + else + { + // R = RefractionIndex * IDotN + sqrt(R) + R = _mm_sqrt_ps(R); + R = XM_FMADD_PS(RefractionIndex, IDotN, R); + // Result = RefractionIndex * Incident - Normal * R + vResult = _mm_mul_ps(RefractionIndex, Incident); + vResult = XM_FNMADD_PS(R, Normal, vResult); + } + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Orthogonal(FXMVECTOR V) noexcept +{ + XMVECTOR Zero = XMVectorZero(); + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR YZYY = XMVectorSwizzle(V); + + XMVECTOR NegativeV = XMVectorSubtract(Zero, V); + + XMVECTOR ZIsNegative = XMVectorLess(Z, Zero); + XMVECTOR YZYYIsNegative = XMVectorLess(YZYY, Zero); + + XMVECTOR S = XMVectorAdd(YZYY, Z); + XMVECTOR D = XMVectorSubtract(YZYY, Z); + + XMVECTOR Select = XMVectorEqualInt(ZIsNegative, YZYYIsNegative); + + XMVECTOR R0 = XMVectorPermute(NegativeV, S); + XMVECTOR R1 = XMVectorPermute(V, D); + + return XMVectorSelect(R1, R0, Select); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3AngleBetweenNormalsEst +( + FXMVECTOR N1, + FXMVECTOR N2 +) noexcept +{ + XMVECTOR Result = XMVector3Dot(N1, N2); + Result = XMVectorClamp(Result, g_XMNegativeOne.v, g_XMOne.v); + Result = XMVectorACosEst(Result); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3AngleBetweenNormals +( + FXMVECTOR N1, + FXMVECTOR N2 +) noexcept +{ + XMVECTOR Result = XMVector3Dot(N1, N2); + Result = XMVectorClamp(Result, g_XMNegativeOne.v, g_XMOne.v); + Result = XMVectorACos(Result); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3AngleBetweenVectors +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + XMVECTOR L1 = XMVector3ReciprocalLength(V1); + XMVECTOR L2 = XMVector3ReciprocalLength(V2); + + XMVECTOR Dot = XMVector3Dot(V1, V2); + + L1 = XMVectorMultiply(L1, L2); + + XMVECTOR CosAngle = XMVectorMultiply(Dot, L1); + CosAngle = XMVectorClamp(CosAngle, g_XMNegativeOne.v, g_XMOne.v); + + return XMVectorACos(CosAngle); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3LinePointDistance +( + FXMVECTOR LinePoint1, + FXMVECTOR LinePoint2, + FXMVECTOR Point +) noexcept +{ + // Given a vector PointVector from LinePoint1 to Point and a vector + // LineVector from LinePoint1 to LinePoint2, the scaled distance + // PointProjectionScale from LinePoint1 to the perpendicular projection + // of PointVector onto the line is defined as: + // + // PointProjectionScale = dot(PointVector, LineVector) / LengthSq(LineVector) + + XMVECTOR PointVector = XMVectorSubtract(Point, LinePoint1); + XMVECTOR LineVector = XMVectorSubtract(LinePoint2, LinePoint1); + + XMVECTOR LengthSq = XMVector3LengthSq(LineVector); + + XMVECTOR PointProjectionScale = XMVector3Dot(PointVector, LineVector); + PointProjectionScale = XMVectorDivide(PointProjectionScale, LengthSq); + + XMVECTOR DistanceVector = XMVectorMultiply(LineVector, PointProjectionScale); + DistanceVector = XMVectorSubtract(PointVector, DistanceVector); + + return XMVector3Length(DistanceVector); +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline void XM_CALLCONV XMVector3ComponentsFromNormal +( + XMVECTOR* pParallel, + XMVECTOR* pPerpendicular, + FXMVECTOR V, + FXMVECTOR Normal +) noexcept +{ + assert(pParallel != nullptr); + assert(pPerpendicular != nullptr); + + XMVECTOR Scale = XMVector3Dot(V, Normal); + + XMVECTOR Parallel = XMVectorMultiply(Normal, Scale); + + *pParallel = Parallel; + *pPerpendicular = XMVectorSubtract(V, Parallel); +} + +//------------------------------------------------------------------------------ +// Transform a vector using a rotation expressed as a unit quaternion + +inline XMVECTOR XM_CALLCONV XMVector3Rotate +( + FXMVECTOR V, + FXMVECTOR RotationQuaternion +) noexcept +{ + XMVECTOR A = XMVectorSelect(g_XMSelect1110.v, V, g_XMSelect1110.v); + XMVECTOR Q = XMQuaternionConjugate(RotationQuaternion); + XMVECTOR Result = XMQuaternionMultiply(Q, A); + return XMQuaternionMultiply(Result, RotationQuaternion); +} + +//------------------------------------------------------------------------------ +// Transform a vector using the inverse of a rotation expressed as a unit quaternion + +inline XMVECTOR XM_CALLCONV XMVector3InverseRotate +( + FXMVECTOR V, + FXMVECTOR RotationQuaternion +) noexcept +{ + XMVECTOR A = XMVectorSelect(g_XMSelect1110.v, V, g_XMSelect1110.v); + XMVECTOR Result = XMQuaternionMultiply(RotationQuaternion, A); + XMVECTOR Q = XMQuaternionConjugate(RotationQuaternion); + return XMQuaternionMultiply(Result, Q); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Transform +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Z, M.r[2], M.r[3]); + Result = XMVectorMultiplyAdd(Y, M.r[1], Result); + Result = XMVectorMultiplyAdd(X, M.r[0], Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + XMVECTOR vResult = vmlaq_lane_f32(M.r[3], M.r[0], VL, 0); // X + vResult = vmlaq_lane_f32(vResult, M.r[1], VL, 1); // Y + return vmlaq_lane_f32(vResult, M.r[2], vget_high_f32(V), 0); // Z +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); // Z + vResult = XM_FMADD_PS(vResult, M.r[2], M.r[3]); + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); // Y + vResult = XM_FMADD_PS(vTemp, M.r[1], vResult); + vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); // X + vResult = XM_FMADD_PS(vTemp, M.r[0], vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 26015 26019, "PREfast noise: Esp:1307" ) +#endif + +_Use_decl_annotations_ +inline XMFLOAT4* XM_CALLCONV XMVector3TransformStream +( + XMFLOAT4* pOutputStream, + size_t OutputStride, + const XMFLOAT3* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT3)); + + assert(OutputStride >= sizeof(XMFLOAT4)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT4)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Z, row2, row3); + Result = XMVectorMultiplyAdd(Y, row1, Result); + Result = XMVectorMultiplyAdd(X, row0, Result); + + XMStoreFloat4(reinterpret_cast(pOutputVector), Result); + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT3)) && (OutputStride == sizeof(XMFLOAT4))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x3_t V = vld3q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT3) * 4; + + float32x2_t r3 = vget_low_f32(row3); + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Ax+M + XMVECTOR vResult1 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Bx+N + + XM_PREFETCH(pInputVector); + + r3 = vget_high_f32(row3); + r = vget_high_f32(row0); + XMVECTOR vResult2 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Cx+O + XMVECTOR vResult3 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Dx+P + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(row1); + vResult2 = vmlaq_lane_f32(vResult2, V.val[1], r, 0); // Cx+Gy+O + vResult3 = vmlaq_lane_f32(vResult3, V.val[1], r, 1); // Dx+Hy+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + r = vget_low_f32(row2); + vResult0 = vmlaq_lane_f32(vResult0, V.val[2], r, 0); // Ax+Ey+Iz+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[2], r, 1); // Bx+Fy+Jz+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 4)); + + r = vget_high_f32(row2); + vResult2 = vmlaq_lane_f32(vResult2, V.val[2], r, 0); // Cx+Gy+Kz+O + vResult3 = vmlaq_lane_f32(vResult3, V.val[2], r, 1); // Dx+Hy+Lz+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 5)); + + float32x4x4_t R; + R.val[0] = vResult0; + R.val[1] = vResult1; + R.val[2] = vResult2; + R.val[3] = vResult3; + + vst4q_f32(reinterpret_cast(pOutputVector), R); + pOutputVector += sizeof(XMFLOAT4) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + float32x2_t VL = vld1_f32(reinterpret_cast(pInputVector)); + float32x2_t zero = vdup_n_f32(0); + float32x2_t VH = vld1_lane_f32(reinterpret_cast(pInputVector) + 2, zero, 0); + pInputVector += InputStride; + + XMVECTOR vResult = vmlaq_lane_f32(row3, row0, VL, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, VL, 1); // Y + vResult = vmlaq_lane_f32(vResult, row2, VH, 0); // Z + + vst1q_f32(reinterpret_cast(pOutputVector), vResult); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(XMFLOAT3)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF) && !(OutputStride & 0xF)) + { + // Packed input, aligned output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 4; + } + } + else + { + // Packed input, unaligned output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + if (!(reinterpret_cast(pOutputStream) & 0xF) && !(OutputStride & 0xF)) + { + // Aligned output + for (; i < VectorCount; ++i) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR Z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + } + else + { + // Unaligned output + for (; i < VectorCount; ++i) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR Z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3TransformCoord +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Z, M.r[2], M.r[3]); + Result = XMVectorMultiplyAdd(Y, M.r[1], Result); + Result = XMVectorMultiplyAdd(X, M.r[0], Result); + + XMVECTOR W = XMVectorSplatW(Result); + return XMVectorDivide(Result, W); +} + +//------------------------------------------------------------------------------ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 26015 26019, "PREfast noise: Esp:1307" ) +#endif + +_Use_decl_annotations_ +inline XMFLOAT3* XM_CALLCONV XMVector3TransformCoordStream +( + XMFLOAT3* pOutputStream, + size_t OutputStride, + const XMFLOAT3* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT3)); + + assert(OutputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT3)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiplyAdd(Z, row2, row3); + Result = XMVectorMultiplyAdd(Y, row1, Result); + Result = XMVectorMultiplyAdd(X, row0, Result); + + XMVECTOR W = XMVectorSplatW(Result); + + Result = XMVectorDivide(Result, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), Result); + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT3)) && (OutputStride == sizeof(XMFLOAT3))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x3_t V = vld3q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT3) * 4; + + float32x2_t r3 = vget_low_f32(row3); + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Ax+M + XMVECTOR vResult1 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Bx+N + + XM_PREFETCH(pInputVector); + + r3 = vget_high_f32(row3); + r = vget_high_f32(row0); + XMVECTOR vResult2 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Cx+O + XMVECTOR W = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Dx+P + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(row1); + vResult2 = vmlaq_lane_f32(vResult2, V.val[1], r, 0); // Cx+Gy+O + W = vmlaq_lane_f32(W, V.val[1], r, 1); // Dx+Hy+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + r = vget_low_f32(row2); + vResult0 = vmlaq_lane_f32(vResult0, V.val[2], r, 0); // Ax+Ey+Iz+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[2], r, 1); // Bx+Fy+Jz+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 4)); + + r = vget_high_f32(row2); + vResult2 = vmlaq_lane_f32(vResult2, V.val[2], r, 0); // Cx+Gy+Kz+O + W = vmlaq_lane_f32(W, V.val[2], r, 1); // Dx+Hy+Lz+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 5)); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + V.val[0] = vdivq_f32(vResult0, W); + V.val[1] = vdivq_f32(vResult1, W); + V.val[2] = vdivq_f32(vResult2, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + V.val[0] = vmulq_f32(vResult0, Reciprocal); + V.val[1] = vmulq_f32(vResult1, Reciprocal); + V.val[2] = vmulq_f32(vResult2, Reciprocal); + #endif + + vst3q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT3) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + float32x2_t VL = vld1_f32(reinterpret_cast(pInputVector)); + float32x2_t zero = vdup_n_f32(0); + float32x2_t VH = vld1_lane_f32(reinterpret_cast(pInputVector) + 2, zero, 0); + pInputVector += InputStride; + + XMVECTOR vResult = vmlaq_lane_f32(row3, row0, VL, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, VL, 1); // Y + vResult = vmlaq_lane_f32(vResult, row2, VH, 0); // Z + + VH = vget_high_f32(vResult); + XMVECTOR W = vdupq_lane_f32(VH, 1); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + vResult = vdivq_f32(vResult, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal for W + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + vResult = vmulq_f32(vResult, Reciprocal); + #endif + + VL = vget_low_f32(vResult); + vst1_f32(reinterpret_cast(pOutputVector), VL); + vst1q_lane_f32(reinterpret_cast(pOutputVector) + 2, vResult, 2); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(XMFLOAT3)) + { + if (OutputStride == sizeof(XMFLOAT3)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V1 = _mm_div_ps(vTemp, W); + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V2 = _mm_div_ps(vTemp, W); + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V3 = _mm_div_ps(vTemp, W); + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V4 = _mm_div_ps(vTemp, W); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector), V1); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 16), vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + else + { + // Packed input, unaligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V1 = _mm_div_ps(vTemp, W); + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V2 = _mm_div_ps(vTemp, W); + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V3 = _mm_div_ps(vTemp, W); + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + V4 = _mm_div_ps(vTemp, W); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector), V1); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 16), vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, row2, row3); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + for (; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR Z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, row2, row3); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + + vTemp = _mm_div_ps(vTemp, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3TransformNormal +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiply(Z, M.r[2]); + Result = XMVectorMultiplyAdd(Y, M.r[1], Result); + Result = XMVectorMultiplyAdd(X, M.r[0], Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + XMVECTOR vResult = vmulq_lane_f32(M.r[0], VL, 0); // X + vResult = vmlaq_lane_f32(vResult, M.r[1], VL, 1); // Y + return vmlaq_lane_f32(vResult, M.r[2], vget_high_f32(V), 0); // Z +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); // Z + vResult = _mm_mul_ps(vResult, M.r[2]); + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); // Y + vResult = XM_FMADD_PS(vTemp, M.r[1], vResult); + vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); // X + vResult = XM_FMADD_PS(vTemp, M.r[0], vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 26015 26019, "PREfast noise: Esp:1307" ) +#endif + +_Use_decl_annotations_ +inline XMFLOAT3* XM_CALLCONV XMVector3TransformNormalStream +( + XMFLOAT3* pOutputStream, + size_t OutputStride, + const XMFLOAT3* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT3)); + + assert(OutputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT3)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiply(Z, row2); + Result = XMVectorMultiplyAdd(Y, row1, Result); + Result = XMVectorMultiplyAdd(X, row0, Result); + + XMStoreFloat3(reinterpret_cast(pOutputVector), Result); + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT3)) && (OutputStride == sizeof(XMFLOAT3))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x3_t V = vld3q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT3) * 4; + + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmulq_lane_f32(V.val[0], r, 0); // Ax + XMVECTOR vResult1 = vmulq_lane_f32(V.val[0], r, 1); // Bx + + XM_PREFETCH(pInputVector); + + r = vget_high_f32(row0); + XMVECTOR vResult2 = vmulq_lane_f32(V.val[0], r, 0); // Cx + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(row1); + vResult2 = vmlaq_lane_f32(vResult2, V.val[1], r, 0); // Cx+Gy + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + r = vget_low_f32(row2); + vResult0 = vmlaq_lane_f32(vResult0, V.val[2], r, 0); // Ax+Ey+Iz + vResult1 = vmlaq_lane_f32(vResult1, V.val[2], r, 1); // Bx+Fy+Jz + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 4)); + + r = vget_high_f32(row2); + vResult2 = vmlaq_lane_f32(vResult2, V.val[2], r, 0); // Cx+Gy+Kz + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 5)); + + V.val[0] = vResult0; + V.val[1] = vResult1; + V.val[2] = vResult2; + + vst3q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT3) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + float32x2_t VL = vld1_f32(reinterpret_cast(pInputVector)); + float32x2_t zero = vdup_n_f32(0); + float32x2_t VH = vld1_lane_f32(reinterpret_cast(pInputVector) + 2, zero, 0); + pInputVector += InputStride; + + XMVECTOR vResult = vmulq_lane_f32(row0, VL, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, VL, 1); // Y + vResult = vmlaq_lane_f32(vResult, row2, VH, 0); // Z + + VL = vget_low_f32(vResult); + vst1_f32(reinterpret_cast(pOutputVector), VL); + vst1q_lane_f32(reinterpret_cast(pOutputVector) + 2, vResult, 2); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(XMFLOAT3)) + { + if (OutputStride == sizeof(XMFLOAT3)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Z, row2); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V1 = _mm_add_ps(vTemp, vTemp3); + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V2 = _mm_add_ps(vTemp, vTemp3); + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V3 = _mm_add_ps(vTemp, vTemp3); + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V4 = _mm_add_ps(vTemp, vTemp3); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector), V1); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 16), vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + else + { + // Packed input, unaligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Z, row2); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V1 = _mm_add_ps(vTemp, vTemp3); + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V2 = _mm_add_ps(vTemp, vTemp3); + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V3 = _mm_add_ps(vTemp, vTemp3); + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + V4 = _mm_add_ps(vTemp, vTemp3); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector), V1); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 16), vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Z, row2); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = _mm_mul_ps(Z, row2); + vTemp2 = _mm_mul_ps(Y, row1); + vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + for (; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR Z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = _mm_mul_ps(Z, row2); + XMVECTOR vTemp2 = _mm_mul_ps(Y, row1); + XMVECTOR vTemp3 = _mm_mul_ps(X, row0); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Project +( + FXMVECTOR V, + float ViewportX, + float ViewportY, + float ViewportWidth, + float ViewportHeight, + float ViewportMinZ, + float ViewportMaxZ, + FXMMATRIX Projection, + CXMMATRIX View, + CXMMATRIX World +) noexcept +{ + const float HalfViewportWidth = ViewportWidth * 0.5f; + const float HalfViewportHeight = ViewportHeight * 0.5f; + + XMVECTOR Scale = XMVectorSet(HalfViewportWidth, -HalfViewportHeight, ViewportMaxZ - ViewportMinZ, 0.0f); + XMVECTOR Offset = XMVectorSet(ViewportX + HalfViewportWidth, ViewportY + HalfViewportHeight, ViewportMinZ, 0.0f); + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + + XMVECTOR Result = XMVector3TransformCoord(V, Transform); + + Result = XMVectorMultiplyAdd(Result, Scale, Offset); + + return Result; +} + +//------------------------------------------------------------------------------ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 26015 26019, "PREfast noise: Esp:1307" ) +#endif + +_Use_decl_annotations_ +inline XMFLOAT3* XM_CALLCONV XMVector3ProjectStream +( + XMFLOAT3* pOutputStream, + size_t OutputStride, + const XMFLOAT3* pInputStream, + size_t InputStride, + size_t VectorCount, + float ViewportX, + float ViewportY, + float ViewportWidth, + float ViewportHeight, + float ViewportMinZ, + float ViewportMaxZ, + FXMMATRIX Projection, + CXMMATRIX View, + CXMMATRIX World +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT3)); + + assert(OutputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT3)); + +#if defined(_XM_NO_INTRINSICS_) + + const float HalfViewportWidth = ViewportWidth * 0.5f; + const float HalfViewportHeight = ViewportHeight * 0.5f; + + XMVECTOR Scale = XMVectorSet(HalfViewportWidth, -HalfViewportHeight, ViewportMaxZ - ViewportMinZ, 1.0f); + XMVECTOR Offset = XMVectorSet(ViewportX + HalfViewportWidth, ViewportY + HalfViewportHeight, ViewportMinZ, 0.0f); + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + + XMVECTOR Result = XMVector3TransformCoord(V, Transform); + Result = XMVectorMultiplyAdd(Result, Scale, Offset); + + XMStoreFloat3(reinterpret_cast(pOutputVector), Result); + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + const float HalfViewportWidth = ViewportWidth * 0.5f; + const float HalfViewportHeight = ViewportHeight * 0.5f; + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT3)) && (OutputStride == sizeof(XMFLOAT3))) + { + XMVECTOR ScaleX = vdupq_n_f32(HalfViewportWidth); + XMVECTOR ScaleY = vdupq_n_f32(-HalfViewportHeight); + XMVECTOR ScaleZ = vdupq_n_f32(ViewportMaxZ - ViewportMinZ); + + XMVECTOR OffsetX = vdupq_n_f32(ViewportX + HalfViewportWidth); + XMVECTOR OffsetY = vdupq_n_f32(ViewportY + HalfViewportHeight); + XMVECTOR OffsetZ = vdupq_n_f32(ViewportMinZ); + + for (size_t j = 0; j < four; ++j) + { + float32x4x3_t V = vld3q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT3) * 4; + + float32x2_t r3 = vget_low_f32(Transform.r[3]); + float32x2_t r = vget_low_f32(Transform.r[0]); + XMVECTOR vResult0 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Ax+M + XMVECTOR vResult1 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Bx+N + + XM_PREFETCH(pInputVector); + + r3 = vget_high_f32(Transform.r[3]); + r = vget_high_f32(Transform.r[0]); + XMVECTOR vResult2 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), V.val[0], r, 0); // Cx+O + XMVECTOR W = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), V.val[0], r, 1); // Dx+P + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(Transform.r[1]); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(Transform.r[1]); + vResult2 = vmlaq_lane_f32(vResult2, V.val[1], r, 0); // Cx+Gy+O + W = vmlaq_lane_f32(W, V.val[1], r, 1); // Dx+Hy+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + r = vget_low_f32(Transform.r[2]); + vResult0 = vmlaq_lane_f32(vResult0, V.val[2], r, 0); // Ax+Ey+Iz+M + vResult1 = vmlaq_lane_f32(vResult1, V.val[2], r, 1); // Bx+Fy+Jz+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 4)); + + r = vget_high_f32(Transform.r[2]); + vResult2 = vmlaq_lane_f32(vResult2, V.val[2], r, 0); // Cx+Gy+Kz+O + W = vmlaq_lane_f32(W, V.val[2], r, 1); // Dx+Hy+Lz+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 5)); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + vResult0 = vdivq_f32(vResult0, W); + vResult1 = vdivq_f32(vResult1, W); + vResult2 = vdivq_f32(vResult2, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + vResult0 = vmulq_f32(vResult0, Reciprocal); + vResult1 = vmulq_f32(vResult1, Reciprocal); + vResult2 = vmulq_f32(vResult2, Reciprocal); + #endif + + V.val[0] = vmlaq_f32(OffsetX, vResult0, ScaleX); + V.val[1] = vmlaq_f32(OffsetY, vResult1, ScaleY); + V.val[2] = vmlaq_f32(OffsetZ, vResult2, ScaleZ); + + vst3q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT3) * 4; + + i += 4; + } + } + } + + if (i < VectorCount) + { + XMVECTOR Scale = XMVectorSet(HalfViewportWidth, -HalfViewportHeight, ViewportMaxZ - ViewportMinZ, 1.0f); + XMVECTOR Offset = XMVectorSet(ViewportX + HalfViewportWidth, ViewportY + HalfViewportHeight, ViewportMinZ, 0.0f); + + for (; i < VectorCount; i++) + { + float32x2_t VL = vld1_f32(reinterpret_cast(pInputVector)); + float32x2_t zero = vdup_n_f32(0); + float32x2_t VH = vld1_lane_f32(reinterpret_cast(pInputVector) + 2, zero, 0); + pInputVector += InputStride; + + XMVECTOR vResult = vmlaq_lane_f32(Transform.r[3], Transform.r[0], VL, 0); // X + vResult = vmlaq_lane_f32(vResult, Transform.r[1], VL, 1); // Y + vResult = vmlaq_lane_f32(vResult, Transform.r[2], VH, 0); // Z + + VH = vget_high_f32(vResult); + XMVECTOR W = vdupq_lane_f32(VH, 1); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + vResult = vdivq_f32(vResult, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal for W + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + vResult = vmulq_f32(vResult, Reciprocal); + #endif + + vResult = vmlaq_f32(Offset, vResult, Scale); + + VL = vget_low_f32(vResult); + vst1_f32(reinterpret_cast(pOutputVector), VL); + vst1q_lane_f32(reinterpret_cast(pOutputVector) + 2, vResult, 2); + pOutputVector += OutputStride; + } + } + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + const float HalfViewportWidth = ViewportWidth * 0.5f; + const float HalfViewportHeight = ViewportHeight * 0.5f; + + XMVECTOR Scale = XMVectorSet(HalfViewportWidth, -HalfViewportHeight, ViewportMaxZ - ViewportMinZ, 1.0f); + XMVECTOR Offset = XMVectorSet(ViewportX + HalfViewportWidth, ViewportY + HalfViewportHeight, ViewportMinZ, 0.0f); + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(XMFLOAT3)) + { + if (OutputStride == sizeof(XMFLOAT3)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V1 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V2 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V3 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V4 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector), V1); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 16), vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + else + { + // Packed input, unaligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V1 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V2 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V3 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + V4 = XM_FMADD_PS(vTemp, Scale, Offset); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector), V1); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 16), vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + vTemp = XM_FMADD_PS(vTemp, Scale, Offset); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 2 + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + vTemp = XM_FMADD_PS(vTemp, Scale, Offset); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 3 + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + vTemp = XM_FMADD_PS(vTemp, Scale, Offset); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 4 + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + vTemp = XM_FMADD_PS(vTemp, Scale, Offset); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + for (; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR Z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + vTemp = XM_FMADD_PS(vTemp, Scale, Offset); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector3Unproject +( + FXMVECTOR V, + float ViewportX, + float ViewportY, + float ViewportWidth, + float ViewportHeight, + float ViewportMinZ, + float ViewportMaxZ, + FXMMATRIX Projection, + CXMMATRIX View, + CXMMATRIX World +) noexcept +{ + static const XMVECTORF32 D = { { { -1.0f, 1.0f, 0.0f, 0.0f } } }; + + XMVECTOR Scale = XMVectorSet(ViewportWidth * 0.5f, -ViewportHeight * 0.5f, ViewportMaxZ - ViewportMinZ, 1.0f); + Scale = XMVectorReciprocal(Scale); + + XMVECTOR Offset = XMVectorSet(-ViewportX, -ViewportY, -ViewportMinZ, 0.0f); + Offset = XMVectorMultiplyAdd(Scale, Offset, D.v); + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + Transform = XMMatrixInverse(nullptr, Transform); + + XMVECTOR Result = XMVectorMultiplyAdd(V, Scale, Offset); + + return XMVector3TransformCoord(Result, Transform); +} + +//------------------------------------------------------------------------------ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 26015 26019, "PREfast noise: Esp:1307" ) +#endif + +_Use_decl_annotations_ +inline XMFLOAT3* XM_CALLCONV XMVector3UnprojectStream +( + XMFLOAT3* pOutputStream, + size_t OutputStride, + const XMFLOAT3* pInputStream, + size_t InputStride, + size_t VectorCount, + float ViewportX, + float ViewportY, + float ViewportWidth, + float ViewportHeight, + float ViewportMinZ, + float ViewportMaxZ, + FXMMATRIX Projection, + CXMMATRIX View, + CXMMATRIX World +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT3)); + + assert(OutputStride >= sizeof(XMFLOAT3)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT3)); + +#if defined(_XM_NO_INTRINSICS_) + + static const XMVECTORF32 D = { { { -1.0f, 1.0f, 0.0f, 0.0f } } }; + + XMVECTOR Scale = XMVectorSet(ViewportWidth * 0.5f, -ViewportHeight * 0.5f, ViewportMaxZ - ViewportMinZ, 1.0f); + Scale = XMVectorReciprocal(Scale); + + XMVECTOR Offset = XMVectorSet(-ViewportX, -ViewportY, -ViewportMinZ, 0.0f); + Offset = XMVectorMultiplyAdd(Scale, Offset, D.v); + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + Transform = XMMatrixInverse(nullptr, Transform); + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + + XMVECTOR Result = XMVectorMultiplyAdd(V, Scale, Offset); + + Result = XMVector3TransformCoord(Result, Transform); + + XMStoreFloat3(reinterpret_cast(pOutputVector), Result); + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + Transform = XMMatrixInverse(nullptr, Transform); + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + float sx = 1.f / (ViewportWidth * 0.5f); + float sy = 1.f / (-ViewportHeight * 0.5f); + float sz = 1.f / (ViewportMaxZ - ViewportMinZ); + + float ox = (-ViewportX * sx) - 1.f; + float oy = (-ViewportY * sy) + 1.f; + float oz = (-ViewportMinZ * sz); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT3)) && (OutputStride == sizeof(XMFLOAT3))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x3_t V = vld3q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT3) * 4; + + XMVECTOR ScaleX = vdupq_n_f32(sx); + XMVECTOR OffsetX = vdupq_n_f32(ox); + XMVECTOR VX = vmlaq_f32(OffsetX, ScaleX, V.val[0]); + + float32x2_t r3 = vget_low_f32(Transform.r[3]); + float32x2_t r = vget_low_f32(Transform.r[0]); + XMVECTOR vResult0 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), VX, r, 0); // Ax+M + XMVECTOR vResult1 = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), VX, r, 1); // Bx+N + + XM_PREFETCH(pInputVector); + + r3 = vget_high_f32(Transform.r[3]); + r = vget_high_f32(Transform.r[0]); + XMVECTOR vResult2 = vmlaq_lane_f32(vdupq_lane_f32(r3, 0), VX, r, 0); // Cx+O + XMVECTOR W = vmlaq_lane_f32(vdupq_lane_f32(r3, 1), VX, r, 1); // Dx+P + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + XMVECTOR ScaleY = vdupq_n_f32(sy); + XMVECTOR OffsetY = vdupq_n_f32(oy); + XMVECTOR VY = vmlaq_f32(OffsetY, ScaleY, V.val[1]); + + r = vget_low_f32(Transform.r[1]); + vResult0 = vmlaq_lane_f32(vResult0, VY, r, 0); // Ax+Ey+M + vResult1 = vmlaq_lane_f32(vResult1, VY, r, 1); // Bx+Fy+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(Transform.r[1]); + vResult2 = vmlaq_lane_f32(vResult2, VY, r, 0); // Cx+Gy+O + W = vmlaq_lane_f32(W, VY, r, 1); // Dx+Hy+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + XMVECTOR ScaleZ = vdupq_n_f32(sz); + XMVECTOR OffsetZ = vdupq_n_f32(oz); + XMVECTOR VZ = vmlaq_f32(OffsetZ, ScaleZ, V.val[2]); + + r = vget_low_f32(Transform.r[2]); + vResult0 = vmlaq_lane_f32(vResult0, VZ, r, 0); // Ax+Ey+Iz+M + vResult1 = vmlaq_lane_f32(vResult1, VZ, r, 1); // Bx+Fy+Jz+N + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 4)); + + r = vget_high_f32(Transform.r[2]); + vResult2 = vmlaq_lane_f32(vResult2, VZ, r, 0); // Cx+Gy+Kz+O + W = vmlaq_lane_f32(W, VZ, r, 1); // Dx+Hy+Lz+P + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 5)); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + V.val[0] = vdivq_f32(vResult0, W); + V.val[1] = vdivq_f32(vResult1, W); + V.val[2] = vdivq_f32(vResult2, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + V.val[0] = vmulq_f32(vResult0, Reciprocal); + V.val[1] = vmulq_f32(vResult1, Reciprocal); + V.val[2] = vmulq_f32(vResult2, Reciprocal); + #endif + + vst3q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT3) * 4; + + i += 4; + } + } + } + + if (i < VectorCount) + { + float32x2_t ScaleL = vcreate_f32( + static_cast(*reinterpret_cast(&sx)) + | (static_cast(*reinterpret_cast(&sy)) << 32)); + float32x2_t ScaleH = vcreate_f32(static_cast(*reinterpret_cast(&sz))); + + float32x2_t OffsetL = vcreate_f32( + static_cast(*reinterpret_cast(&ox)) + | (static_cast(*reinterpret_cast(&oy)) << 32)); + float32x2_t OffsetH = vcreate_f32(static_cast(*reinterpret_cast(&oz))); + + for (; i < VectorCount; i++) + { + float32x2_t VL = vld1_f32(reinterpret_cast(pInputVector)); + float32x2_t zero = vdup_n_f32(0); + float32x2_t VH = vld1_lane_f32(reinterpret_cast(pInputVector) + 2, zero, 0); + pInputVector += InputStride; + + VL = vmla_f32(OffsetL, VL, ScaleL); + VH = vmla_f32(OffsetH, VH, ScaleH); + + XMVECTOR vResult = vmlaq_lane_f32(Transform.r[3], Transform.r[0], VL, 0); // X + vResult = vmlaq_lane_f32(vResult, Transform.r[1], VL, 1); // Y + vResult = vmlaq_lane_f32(vResult, Transform.r[2], VH, 0); // Z + + VH = vget_high_f32(vResult); + XMVECTOR W = vdupq_lane_f32(VH, 1); + + #if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + vResult = vdivq_f32(vResult, W); + #else + // 2 iterations of Newton-Raphson refinement of reciprocal for W + float32x4_t Reciprocal = vrecpeq_f32(W); + float32x4_t S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + S = vrecpsq_f32(Reciprocal, W); + Reciprocal = vmulq_f32(S, Reciprocal); + + vResult = vmulq_f32(vResult, Reciprocal); + #endif + + VL = vget_low_f32(vResult); + vst1_f32(reinterpret_cast(pOutputVector), VL); + vst1q_lane_f32(reinterpret_cast(pOutputVector) + 2, vResult, 2); + pOutputVector += OutputStride; + } + } + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 D = { { { -1.0f, 1.0f, 0.0f, 0.0f } } }; + + XMVECTOR Scale = XMVectorSet(ViewportWidth * 0.5f, -ViewportHeight * 0.5f, ViewportMaxZ - ViewportMinZ, 1.0f); + Scale = XMVectorReciprocal(Scale); + + XMVECTOR Offset = XMVectorSet(-ViewportX, -ViewportY, -ViewportMinZ, 0.0f); + Offset = _mm_mul_ps(Scale, Offset); + Offset = _mm_add_ps(Offset, D); + + XMMATRIX Transform = XMMatrixMultiply(World, View); + Transform = XMMatrixMultiply(Transform, Projection); + Transform = XMMatrixInverse(nullptr, Transform); + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(XMFLOAT3)) + { + if (OutputStride == sizeof(XMFLOAT3)) + { + if (!(reinterpret_cast(pOutputStream) & 0xF)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + V1 = XM_FMADD_PS(V1, Scale, Offset); + + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V1 = _mm_div_ps(vTemp, W); + + // Result 2 + V2 = XM_FMADD_PS(V2, Scale, Offset); + + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V2 = _mm_div_ps(vTemp, W); + + // Result 3 + V3 = XM_FMADD_PS(V3, Scale, Offset); + + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V3 = _mm_div_ps(vTemp, W); + + // Result 4 + V4 = XM_FMADD_PS(V4, Scale, Offset); + + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V4 = _mm_div_ps(vTemp, W); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector), V1); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 16), vTemp); + XM_STREAM_PS(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + else + { + // Packed input, unaligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + V1 = XM_FMADD_PS(V1, Scale, Offset); + + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V1 = _mm_div_ps(vTemp, W); + + // Result 2 + V2 = XM_FMADD_PS(V2, Scale, Offset); + + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V2 = _mm_div_ps(vTemp, W); + + // Result 3 + V3 = XM_FMADD_PS(V3, Scale, Offset); + + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V3 = _mm_div_ps(vTemp, W); + + // Result 4 + V4 = XM_FMADD_PS(V4, Scale, Offset); + + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + V4 = _mm_div_ps(vTemp, W); + + // Pack and store the vectors + XM3PACK4INTO3(vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector), V1); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 16), vTemp); + _mm_storeu_ps(reinterpret_cast(pOutputVector + 32), V3); + pOutputVector += sizeof(XMFLOAT3) * 4; + i += 4; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < four; ++j) + { + __m128 V1 = _mm_loadu_ps(reinterpret_cast(pInputVector)); + __m128 L2 = _mm_loadu_ps(reinterpret_cast(pInputVector + 16)); + __m128 L3 = _mm_loadu_ps(reinterpret_cast(pInputVector + 32)); + pInputVector += sizeof(XMFLOAT3) * 4; + + // Unpack the 4 vectors (.w components are junk) + XM3UNPACK3INTO4(V1, L2, L3); + + // Result 1 + V1 = XM_FMADD_PS(V1, Scale, Offset); + + XMVECTOR Z = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 2 + V2 = XM_FMADD_PS(V2, Scale, Offset); + + Z = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V2, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 3 + V3 = XM_FMADD_PS(V3, Scale, Offset); + + Z = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + // Result 4 + V4 = XM_FMADD_PS(V4, Scale, Offset); + + Z = XM_PERMUTE_PS(V4, _MM_SHUFFLE(2, 2, 2, 2)); + Y = XM_PERMUTE_PS(V4, _MM_SHUFFLE(1, 1, 1, 1)); + X = XM_PERMUTE_PS(V4, _MM_SHUFFLE(0, 0, 0, 0)); + + vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + + i += 4; + } + } + } + } + + for (; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + V = _mm_mul_ps(V, Scale); + V = _mm_add_ps(V, Offset); + + XMVECTOR Z = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR Y = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR X = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + + XMVECTOR vTemp = XM_FMADD_PS(Z, Transform.r[2], Transform.r[3]); + XMVECTOR vTemp2 = _mm_mul_ps(Y, Transform.r[1]); + XMVECTOR vTemp3 = _mm_mul_ps(X, Transform.r[0]); + vTemp = _mm_add_ps(vTemp, vTemp2); + vTemp = _mm_add_ps(vTemp, vTemp3); + + XMVECTOR W = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 3, 3, 3)); + vTemp = _mm_div_ps(vTemp, W); + + XMStoreFloat3(reinterpret_cast(pOutputVector), vTemp); + pOutputVector += OutputStride; + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +/**************************************************************************** + * + * 4D Vector + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ +// Comparison operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4Equal +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] == V2.vector4_f32[0]) && (V1.vector4_f32[1] == V2.vector4_f32[1]) && (V1.vector4_f32[2] == V2.vector4_f32[2]) && (V1.vector4_f32[3] == V2.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + return ((_mm_movemask_ps(vTemp) == 0x0f) != 0); +#else + return XMComparisonAllTrue(XMVector4EqualR(V1, V2)); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector4EqualR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + uint32_t CR = 0; + + if ((V1.vector4_f32[0] == V2.vector4_f32[0]) && + (V1.vector4_f32[1] == V2.vector4_f32[1]) && + (V1.vector4_f32[2] == V2.vector4_f32[2]) && + (V1.vector4_f32[3] == V2.vector4_f32[3])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] != V2.vector4_f32[0]) && + (V1.vector4_f32[1] != V2.vector4_f32[1]) && + (V1.vector4_f32[2] != V2.vector4_f32[2]) && + (V1.vector4_f32[3] != V2.vector4_f32[3])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpeq_ps(V1, V2); + int iTest = _mm_movemask_ps(vTemp); + uint32_t CR = 0; + if (iTest == 0xf) // All equal? + { + CR = XM_CRMASK_CR6TRUE; + } + else if (iTest == 0) // All not equal? + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4EqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_u32[0] == V2.vector4_u32[0]) && (V1.vector4_u32[1] == V2.vector4_u32[1]) && (V1.vector4_u32[2] == V2.vector4_u32[2]) && (V1.vector4_u32[3] == V2.vector4_u32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return ((_mm_movemask_ps(_mm_castsi128_ps(vTemp)) == 0xf) != 0); +#else + return XMComparisonAllTrue(XMVector4EqualIntR(V1, V2)); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector4EqualIntR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t CR = 0; + if (V1.vector4_u32[0] == V2.vector4_u32[0] && + V1.vector4_u32[1] == V2.vector4_u32[1] && + V1.vector4_u32[2] == V2.vector4_u32[2] && + V1.vector4_u32[3] == V2.vector4_u32[3]) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (V1.vector4_u32[0] != V2.vector4_u32[0] && + V1.vector4_u32[1] != V2.vector4_u32[1] && + V1.vector4_u32[2] != V2.vector4_u32[2] && + V1.vector4_u32[3] != V2.vector4_u32[3]) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + int iTest = _mm_movemask_ps(_mm_castsi128_ps(vTemp)); + uint32_t CR = 0; + if (iTest == 0xf) // All equal? + { + CR = XM_CRMASK_CR6TRUE; + } + else if (iTest == 0) // All not equal? + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +inline bool XM_CALLCONV XMVector4NearEqual +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR Epsilon +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float dx, dy, dz, dw; + + dx = fabsf(V1.vector4_f32[0] - V2.vector4_f32[0]); + dy = fabsf(V1.vector4_f32[1] - V2.vector4_f32[1]); + dz = fabsf(V1.vector4_f32[2] - V2.vector4_f32[2]); + dw = fabsf(V1.vector4_f32[3] - V2.vector4_f32[3]); + return (((dx <= Epsilon.vector4_f32[0]) && + (dy <= Epsilon.vector4_f32[1]) && + (dz <= Epsilon.vector4_f32[2]) && + (dw <= Epsilon.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vDelta = vsubq_f32(V1, V2); +#if defined(_MSC_VER) && !defined(__clang__) && !defined(_ARM64_DISTINCT_NEON_TYPES) + uint32x4_t vResult = vacleq_f32(vDelta, Epsilon); +#else + uint32x4_t vResult = vcleq_f32(vabsq_f32(vDelta), Epsilon); +#endif + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + // Get the difference + XMVECTOR vDelta = _mm_sub_ps(V1, V2); + // Get the absolute value of the difference + XMVECTOR vTemp = _mm_setzero_ps(); + vTemp = _mm_sub_ps(vTemp, vDelta); + vTemp = _mm_max_ps(vTemp, vDelta); + vTemp = _mm_cmple_ps(vTemp, Epsilon); + return ((_mm_movemask_ps(vTemp) == 0xf) != 0); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4NotEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] != V2.vector4_f32[0]) || (V1.vector4_f32[1] != V2.vector4_f32[1]) || (V1.vector4_f32[2] != V2.vector4_f32[2]) || (V1.vector4_f32[3] != V2.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) != 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpneq_ps(V1, V2); + return ((_mm_movemask_ps(vTemp)) != 0); +#else + return XMComparisonAnyFalse(XMVector4EqualR(V1, V2)); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4NotEqualInt +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_u32[0] != V2.vector4_u32[0]) || (V1.vector4_u32[1] != V2.vector4_u32[1]) || (V1.vector4_u32[2] != V2.vector4_u32[2]) || (V1.vector4_u32[3] != V2.vector4_u32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vceqq_u32(vreinterpretq_u32_f32(V1), vreinterpretq_u32_f32(V2)); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) != 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + __m128i vTemp = _mm_cmpeq_epi32(_mm_castps_si128(V1), _mm_castps_si128(V2)); + return ((_mm_movemask_ps(_mm_castsi128_ps(vTemp)) != 0xF) != 0); +#else + return XMComparisonAnyFalse(XMVector4EqualIntR(V1, V2)); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4Greater +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] > V2.vector4_f32[0]) && (V1.vector4_f32[1] > V2.vector4_f32[1]) && (V1.vector4_f32[2] > V2.vector4_f32[2]) && (V1.vector4_f32[3] > V2.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgtq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + return ((_mm_movemask_ps(vTemp) == 0x0f) != 0); +#else + return XMComparisonAllTrue(XMVector4GreaterR(V1, V2)); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector4GreaterR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t CR = 0; + if (V1.vector4_f32[0] > V2.vector4_f32[0] && + V1.vector4_f32[1] > V2.vector4_f32[1] && + V1.vector4_f32[2] > V2.vector4_f32[2] && + V1.vector4_f32[3] > V2.vector4_f32[3]) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (V1.vector4_f32[0] <= V2.vector4_f32[0] && + V1.vector4_f32[1] <= V2.vector4_f32[1] && + V1.vector4_f32[2] <= V2.vector4_f32[2] && + V1.vector4_f32[3] <= V2.vector4_f32[3]) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgtq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + uint32_t CR = 0; + XMVECTOR vTemp = _mm_cmpgt_ps(V1, V2); + int iTest = _mm_movemask_ps(vTemp); + if (iTest == 0xf) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4GreaterOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] >= V2.vector4_f32[0]) && (V1.vector4_f32[1] >= V2.vector4_f32[1]) && (V1.vector4_f32[2] >= V2.vector4_f32[2]) && (V1.vector4_f32[3] >= V2.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgeq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + return ((_mm_movemask_ps(vTemp) == 0x0f) != 0); +#else + return XMComparisonAllTrue(XMVector4GreaterOrEqualR(V1, V2)); +#endif +} + +//------------------------------------------------------------------------------ + +inline uint32_t XM_CALLCONV XMVector4GreaterOrEqualR +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + uint32_t CR = 0; + if ((V1.vector4_f32[0] >= V2.vector4_f32[0]) && + (V1.vector4_f32[1] >= V2.vector4_f32[1]) && + (V1.vector4_f32[2] >= V2.vector4_f32[2]) && + (V1.vector4_f32[3] >= V2.vector4_f32[3])) + { + CR = XM_CRMASK_CR6TRUE; + } + else if ((V1.vector4_f32[0] < V2.vector4_f32[0]) && + (V1.vector4_f32[1] < V2.vector4_f32[1]) && + (V1.vector4_f32[2] < V2.vector4_f32[2]) && + (V1.vector4_f32[3] < V2.vector4_f32[3])) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcgeq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + uint32_t r = vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1); + + uint32_t CR = 0; + if (r == 0xFFFFFFFFU) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!r) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#elif defined(_XM_SSE_INTRINSICS_) + uint32_t CR = 0; + XMVECTOR vTemp = _mm_cmpge_ps(V1, V2); + int iTest = _mm_movemask_ps(vTemp); + if (iTest == 0x0f) + { + CR = XM_CRMASK_CR6TRUE; + } + else if (!iTest) + { + CR = XM_CRMASK_CR6FALSE; + } + return CR; +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4Less +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] < V2.vector4_f32[0]) && (V1.vector4_f32[1] < V2.vector4_f32[1]) && (V1.vector4_f32[2] < V2.vector4_f32[2]) && (V1.vector4_f32[3] < V2.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcltq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmplt_ps(V1, V2); + return ((_mm_movemask_ps(vTemp) == 0x0f) != 0); +#else + return XMComparisonAllTrue(XMVector4GreaterR(V2, V1)); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4LessOrEqual +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V1.vector4_f32[0] <= V2.vector4_f32[0]) && (V1.vector4_f32[1] <= V2.vector4_f32[1]) && (V1.vector4_f32[2] <= V2.vector4_f32[2]) && (V1.vector4_f32[3] <= V2.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vResult = vcleq_f32(V1, V2); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vResult)), vget_high_u8(vreinterpretq_u8_u32(vResult))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp = _mm_cmple_ps(V1, V2); + return ((_mm_movemask_ps(vTemp) == 0x0f) != 0); +#else + return XMComparisonAllTrue(XMVector4GreaterOrEqualR(V2, V1)); +#endif +} + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4InBounds +( + FXMVECTOR V, + FXMVECTOR Bounds +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (((V.vector4_f32[0] <= Bounds.vector4_f32[0] && V.vector4_f32[0] >= -Bounds.vector4_f32[0]) && + (V.vector4_f32[1] <= Bounds.vector4_f32[1] && V.vector4_f32[1] >= -Bounds.vector4_f32[1]) && + (V.vector4_f32[2] <= Bounds.vector4_f32[2] && V.vector4_f32[2] >= -Bounds.vector4_f32[2]) && + (V.vector4_f32[3] <= Bounds.vector4_f32[3] && V.vector4_f32[3] >= -Bounds.vector4_f32[3])) != 0); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Test if less than or equal + uint32x4_t ivTemp1 = vcleq_f32(V, Bounds); + // Negate the bounds + float32x4_t vTemp2 = vnegq_f32(Bounds); + // Test if greater or equal (Reversed) + uint32x4_t ivTemp2 = vcleq_f32(vTemp2, V); + // Blend answers + ivTemp1 = vandq_u32(ivTemp1, ivTemp2); + // in bounds? + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(ivTemp1)), vget_high_u8(vreinterpretq_u8_u32(ivTemp1))); + uint16x4x2_t vTemp3 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp3.val[1]), 1) == 0xFFFFFFFFU); +#elif defined(_XM_SSE_INTRINSICS_) + // Test if less than or equal + XMVECTOR vTemp1 = _mm_cmple_ps(V, Bounds); + // Negate the bounds + XMVECTOR vTemp2 = _mm_mul_ps(Bounds, g_XMNegativeOne); + // Test if greater or equal (Reversed) + vTemp2 = _mm_cmple_ps(vTemp2, V); + // Blend answers + vTemp1 = _mm_and_ps(vTemp1, vTemp2); + // All in bounds? + return ((_mm_movemask_ps(vTemp1) == 0x0f) != 0); +#else + return XMComparisonAllInBounds(XMVector4InBoundsR(V, Bounds)); +#endif +} + +//------------------------------------------------------------------------------ + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(push) +#pragma float_control(precise, on) +#endif + +inline bool XM_CALLCONV XMVector4IsNaN(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + return (XMISNAN(V.vector4_f32[0]) || + XMISNAN(V.vector4_f32[1]) || + XMISNAN(V.vector4_f32[2]) || + XMISNAN(V.vector4_f32[3])); +#elif defined(_XM_ARM_NEON_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + return isnan(vgetq_lane_f32(V, 0)) || isnan(vgetq_lane_f32(V, 1)) || isnan(vgetq_lane_f32(V, 2)) || isnan(vgetq_lane_f32(V, 3)); +#else +// Test against itself. NaN is always not equal + uint32x4_t vTempNan = vceqq_f32(V, V); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vTempNan)), vget_high_u8(vreinterpretq_u8_u32(vTempNan))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + // If any are NaN, the mask is zero + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) != 0xFFFFFFFFU); +#endif +#elif defined(_XM_SSE_INTRINSICS_) +#if defined(__clang__) && defined(__FINITE_MATH_ONLY__) + XM_ALIGNED_DATA(16) float tmp[4]; + _mm_store_ps(tmp, V); + return isnan(tmp[0]) || isnan(tmp[1]) || isnan(tmp[2]) || isnan(tmp[3]); +#else +// Test against itself. NaN is always not equal + XMVECTOR vTempNan = _mm_cmpneq_ps(V, V); + // If any are NaN, the mask is non-zero + return (_mm_movemask_ps(vTempNan) != 0); +#endif +#endif +} + +#if !defined(_XM_NO_INTRINSICS_) && defined(_MSC_VER) && !defined(__INTEL_COMPILER) +#pragma float_control(pop) +#endif + +//------------------------------------------------------------------------------ + +inline bool XM_CALLCONV XMVector4IsInfinite(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + return (XMISINF(V.vector4_f32[0]) || + XMISINF(V.vector4_f32[1]) || + XMISINF(V.vector4_f32[2]) || + XMISINF(V.vector4_f32[3])); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Mask off the sign bit + uint32x4_t vTempInf = vandq_u32(vreinterpretq_u32_f32(V), g_XMAbsMask); + // Compare to infinity + vTempInf = vceqq_f32(vreinterpretq_f32_u32(vTempInf), g_XMInfinity); + // If any are infinity, the signs are true. + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(vTempInf)), vget_high_u8(vreinterpretq_u8_u32(vTempInf))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + return (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) != 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Mask off the sign bit + XMVECTOR vTemp = _mm_and_ps(V, g_XMAbsMask); + // Compare to infinity + vTemp = _mm_cmpeq_ps(vTemp, g_XMInfinity); + // If any are infinity, the signs are true. + return (_mm_movemask_ps(vTemp) != 0); +#endif +} + +//------------------------------------------------------------------------------ +// Computation operations +//------------------------------------------------------------------------------ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Dot +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result; + Result.f[0] = + Result.f[1] = + Result.f[2] = + Result.f[3] = V1.vector4_f32[0] * V2.vector4_f32[0] + V1.vector4_f32[1] * V2.vector4_f32[1] + V1.vector4_f32[2] * V2.vector4_f32[2] + V1.vector4_f32[3] * V2.vector4_f32[3]; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vTemp = vmulq_f32(V1, V2); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + return vcombine_f32(v1, v1); +#elif defined(_XM_SSE4_INTRINSICS_) + return _mm_dp_ps(V1, V2, 0xff); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vTemp = _mm_mul_ps(V1, V2); + vTemp = _mm_hadd_ps(vTemp, vTemp); + return _mm_hadd_ps(vTemp, vTemp); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vTemp2 = V2; + XMVECTOR vTemp = _mm_mul_ps(V1, vTemp2); + vTemp2 = _mm_shuffle_ps(vTemp2, vTemp, _MM_SHUFFLE(1, 0, 0, 0)); // Copy X to the Z position and Y to the W position + vTemp2 = _mm_add_ps(vTemp2, vTemp); // Add Z = X+Z; W = Y+W; + vTemp = _mm_shuffle_ps(vTemp, vTemp2, _MM_SHUFFLE(0, 3, 0, 0)); // Copy W to the Z position + vTemp = _mm_add_ps(vTemp, vTemp2); // Add Z and W together + return XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(2, 2, 2, 2)); // Splat Z and return +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Cross +( + FXMVECTOR V1, + FXMVECTOR V2, + FXMVECTOR V3 +) noexcept +{ + // [ ((v2.z*v3.w-v2.w*v3.z)*v1.y)-((v2.y*v3.w-v2.w*v3.y)*v1.z)+((v2.y*v3.z-v2.z*v3.y)*v1.w), + // ((v2.w*v3.z-v2.z*v3.w)*v1.x)-((v2.w*v3.x-v2.x*v3.w)*v1.z)+((v2.z*v3.x-v2.x*v3.z)*v1.w), + // ((v2.y*v3.w-v2.w*v3.y)*v1.x)-((v2.x*v3.w-v2.w*v3.x)*v1.y)+((v2.x*v3.y-v2.y*v3.x)*v1.w), + // ((v2.z*v3.y-v2.y*v3.z)*v1.x)-((v2.z*v3.x-v2.x*v3.z)*v1.y)+((v2.y*v3.x-v2.x*v3.y)*v1.z) ] + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + (((V2.vector4_f32[2] * V3.vector4_f32[3]) - (V2.vector4_f32[3] * V3.vector4_f32[2])) * V1.vector4_f32[1]) - (((V2.vector4_f32[1] * V3.vector4_f32[3]) - (V2.vector4_f32[3] * V3.vector4_f32[1])) * V1.vector4_f32[2]) + (((V2.vector4_f32[1] * V3.vector4_f32[2]) - (V2.vector4_f32[2] * V3.vector4_f32[1])) * V1.vector4_f32[3]), + (((V2.vector4_f32[3] * V3.vector4_f32[2]) - (V2.vector4_f32[2] * V3.vector4_f32[3])) * V1.vector4_f32[0]) - (((V2.vector4_f32[3] * V3.vector4_f32[0]) - (V2.vector4_f32[0] * V3.vector4_f32[3])) * V1.vector4_f32[2]) + (((V2.vector4_f32[2] * V3.vector4_f32[0]) - (V2.vector4_f32[0] * V3.vector4_f32[2])) * V1.vector4_f32[3]), + (((V2.vector4_f32[1] * V3.vector4_f32[3]) - (V2.vector4_f32[3] * V3.vector4_f32[1])) * V1.vector4_f32[0]) - (((V2.vector4_f32[0] * V3.vector4_f32[3]) - (V2.vector4_f32[3] * V3.vector4_f32[0])) * V1.vector4_f32[1]) + (((V2.vector4_f32[0] * V3.vector4_f32[1]) - (V2.vector4_f32[1] * V3.vector4_f32[0])) * V1.vector4_f32[3]), + (((V2.vector4_f32[2] * V3.vector4_f32[1]) - (V2.vector4_f32[1] * V3.vector4_f32[2])) * V1.vector4_f32[0]) - (((V2.vector4_f32[2] * V3.vector4_f32[0]) - (V2.vector4_f32[0] * V3.vector4_f32[2])) * V1.vector4_f32[1]) + (((V2.vector4_f32[1] * V3.vector4_f32[0]) - (V2.vector4_f32[0] * V3.vector4_f32[1])) * V1.vector4_f32[2]), + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + const uint32x2_t select = vget_low_u32(g_XMMaskX); + + // Term1: V2zwyz * V3wzwy + const float32x2_t v2xy = vget_low_f32(V2); + const float32x2_t v2zw = vget_high_f32(V2); + const float32x2_t v2yx = vrev64_f32(v2xy); + const float32x2_t v2wz = vrev64_f32(v2zw); + const float32x2_t v2yz = vbsl_f32(select, v2yx, v2wz); + + const float32x2_t v3zw = vget_high_f32(V3); + const float32x2_t v3wz = vrev64_f32(v3zw); + const float32x2_t v3xy = vget_low_f32(V3); + const float32x2_t v3wy = vbsl_f32(select, v3wz, v3xy); + + float32x4_t vTemp1 = vcombine_f32(v2zw, v2yz); + float32x4_t vTemp2 = vcombine_f32(v3wz, v3wy); + XMVECTOR vResult = vmulq_f32(vTemp1, vTemp2); + + // - V2wzwy * V3zwyz + const float32x2_t v2wy = vbsl_f32(select, v2wz, v2xy); + + const float32x2_t v3yx = vrev64_f32(v3xy); + const float32x2_t v3yz = vbsl_f32(select, v3yx, v3wz); + + vTemp1 = vcombine_f32(v2wz, v2wy); + vTemp2 = vcombine_f32(v3zw, v3yz); + vResult = vmlsq_f32(vResult, vTemp1, vTemp2); + + // term1 * V1yxxx + const float32x2_t v1xy = vget_low_f32(V1); + const float32x2_t v1yx = vrev64_f32(v1xy); + + vTemp1 = vcombine_f32(v1yx, vdup_lane_f32(v1yx, 1)); + vResult = vmulq_f32(vResult, vTemp1); + + // Term2: V2ywxz * V3wxwx + const float32x2_t v2yw = vrev64_f32(v2wy); + const float32x2_t v2xz = vbsl_f32(select, v2xy, v2wz); + + const float32x2_t v3wx = vbsl_f32(select, v3wz, v3yx); + + vTemp1 = vcombine_f32(v2yw, v2xz); + vTemp2 = vcombine_f32(v3wx, v3wx); + float32x4_t vTerm = vmulq_f32(vTemp1, vTemp2); + + // - V2wxwx * V3ywxz + const float32x2_t v2wx = vbsl_f32(select, v2wz, v2yx); + + const float32x2_t v3yw = vrev64_f32(v3wy); + const float32x2_t v3xz = vbsl_f32(select, v3xy, v3wz); + + vTemp1 = vcombine_f32(v2wx, v2wx); + vTemp2 = vcombine_f32(v3yw, v3xz); + vTerm = vmlsq_f32(vTerm, vTemp1, vTemp2); + + // vResult - term2 * V1zzyy + const float32x2_t v1zw = vget_high_f32(V1); + + vTemp1 = vcombine_f32(vdup_lane_f32(v1zw, 0), vdup_lane_f32(v1yx, 0)); + vResult = vmlsq_f32(vResult, vTerm, vTemp1); + + // Term3: V2yzxy * V3zxyx + const float32x2_t v3zx = vrev64_f32(v3xz); + + vTemp1 = vcombine_f32(v2yz, v2xy); + vTemp2 = vcombine_f32(v3zx, v3yx); + vTerm = vmulq_f32(vTemp1, vTemp2); + + // - V2zxyx * V3yzxy + const float32x2_t v2zx = vrev64_f32(v2xz); + + vTemp1 = vcombine_f32(v2zx, v2yx); + vTemp2 = vcombine_f32(v3yz, v3xy); + vTerm = vmlsq_f32(vTerm, vTemp1, vTemp2); + + // vResult + term3 * V1wwwz + const float32x2_t v1wz = vrev64_f32(v1zw); + + vTemp1 = vcombine_f32(vdup_lane_f32(v1wz, 0), v1wz); + return vmlaq_f32(vResult, vTerm, vTemp1); +#elif defined(_XM_SSE_INTRINSICS_) + // V2zwyz * V3wzwy + XMVECTOR vResult = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 1, 3, 2)); + XMVECTOR vTemp3 = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 3, 2, 3)); + vResult = _mm_mul_ps(vResult, vTemp3); + // - V2wzwy * V3zwyz + XMVECTOR vTemp2 = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 3, 2, 3)); + vTemp3 = XM_PERMUTE_PS(vTemp3, _MM_SHUFFLE(1, 3, 0, 1)); + vResult = XM_FNMADD_PS(vTemp2, vTemp3, vResult); + // term1 * V1yxxx + XMVECTOR vTemp1 = XM_PERMUTE_PS(V1, _MM_SHUFFLE(0, 0, 0, 1)); + vResult = _mm_mul_ps(vResult, vTemp1); + + // V2ywxz * V3wxwx + vTemp2 = XM_PERMUTE_PS(V2, _MM_SHUFFLE(2, 0, 3, 1)); + vTemp3 = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 3, 0, 3)); + vTemp3 = _mm_mul_ps(vTemp3, vTemp2); + // - V2wxwx * V3ywxz + vTemp2 = XM_PERMUTE_PS(vTemp2, _MM_SHUFFLE(2, 1, 2, 1)); + vTemp1 = XM_PERMUTE_PS(V3, _MM_SHUFFLE(2, 0, 3, 1)); + vTemp3 = XM_FNMADD_PS(vTemp2, vTemp1, vTemp3); + // vResult - temp * V1zzyy + vTemp1 = XM_PERMUTE_PS(V1, _MM_SHUFFLE(1, 1, 2, 2)); + vResult = XM_FNMADD_PS(vTemp1, vTemp3, vResult); + + // V2yzxy * V3zxyx + vTemp2 = XM_PERMUTE_PS(V2, _MM_SHUFFLE(1, 0, 2, 1)); + vTemp3 = XM_PERMUTE_PS(V3, _MM_SHUFFLE(0, 1, 0, 2)); + vTemp3 = _mm_mul_ps(vTemp3, vTemp2); + // - V2zxyx * V3yzxy + vTemp2 = XM_PERMUTE_PS(vTemp2, _MM_SHUFFLE(2, 0, 2, 1)); + vTemp1 = XM_PERMUTE_PS(V3, _MM_SHUFFLE(1, 0, 2, 1)); + vTemp3 = XM_FNMADD_PS(vTemp1, vTemp2, vTemp3); + // vResult + term * V1wwwz + vTemp1 = XM_PERMUTE_PS(V1, _MM_SHUFFLE(2, 3, 3, 3)); + vResult = XM_FMADD_PS(vTemp3, vTemp1, vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4LengthSq(FXMVECTOR V) noexcept +{ + return XMVector4Dot(V, V); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4ReciprocalLengthEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector4LengthSq(V); + Result = XMVectorReciprocalSqrtEst(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot4 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + // Reciprocal sqrt (estimate) + v2 = vrsqrte_f32(v1); + return vcombine_f32(v2, v2); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0xff); + return _mm_rsqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_rsqrt_ps(vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and w + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(3, 2, 3, 2)); + // x+z, y+w + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // x+z,x+z,x+z,y+w + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 0, 0, 0)); + // ??,??,y+w,y+w + vTemp = _mm_shuffle_ps(vTemp, vLengthSq, _MM_SHUFFLE(3, 3, 0, 0)); + // ??,??,x+z+y+w,?? + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // Splat the length + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 2, 2, 2)); + // Get the reciprocal + vLengthSq = _mm_rsqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4ReciprocalLength(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector4LengthSq(V); + Result = XMVectorReciprocalSqrt(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot4 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + // Reciprocal sqrt + float32x2_t S0 = vrsqrte_f32(v1); + float32x2_t P0 = vmul_f32(v1, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(v1, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t Result = vmul_f32(S1, R1); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0xff); + XMVECTOR vLengthSq = _mm_sqrt_ps(vTemp); + return _mm_div_ps(g_XMOne, vLengthSq); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ps(vLengthSq); + vLengthSq = _mm_div_ps(g_XMOne, vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and w + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(3, 2, 3, 2)); + // x+z, y+w + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // x+z,x+z,x+z,y+w + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 0, 0, 0)); + // ??,??,y+w,y+w + vTemp = _mm_shuffle_ps(vTemp, vLengthSq, _MM_SHUFFLE(3, 3, 0, 0)); + // ??,??,x+z+y+w,?? + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // Splat the length + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 2, 2, 2)); + // Get the reciprocal + vLengthSq = _mm_sqrt_ps(vLengthSq); + // Accurate! + vLengthSq = _mm_div_ps(g_XMOne, vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4LengthEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector4LengthSq(V); + Result = XMVectorSqrtEst(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot4 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + const float32x2_t zero = vdup_n_f32(0); + uint32x2_t VEqualsZero = vceq_f32(v1, zero); + // Sqrt (estimate) + float32x2_t Result = vrsqrte_f32(v1); + Result = vmul_f32(v1, Result); + Result = vbsl_f32(VEqualsZero, zero, Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0xff); + return _mm_sqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and w + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(3, 2, 3, 2)); + // x+z, y+w + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // x+z,x+z,x+z,y+w + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 0, 0, 0)); + // ??,??,y+w,y+w + vTemp = _mm_shuffle_ps(vTemp, vLengthSq, _MM_SHUFFLE(3, 3, 0, 0)); + // ??,??,x+z+y+w,?? + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // Splat the length + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 2, 2, 2)); + // Get the length + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Length(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + + Result = XMVector4LengthSq(V); + Result = XMVectorSqrt(Result); + + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot4 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + const float32x2_t zero = vdup_n_f32(0); + uint32x2_t VEqualsZero = vceq_f32(v1, zero); + // Sqrt + float32x2_t S0 = vrsqrte_f32(v1); + float32x2_t P0 = vmul_f32(v1, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(v1, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + float32x2_t Result = vmul_f32(S1, R1); + Result = vmul_f32(v1, Result); + Result = vbsl_f32(VEqualsZero, zero, Result); + return vcombine_f32(Result, Result); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0xff); + return _mm_sqrt_ps(vTemp); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and w + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(3, 2, 3, 2)); + // x+z, y+w + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // x+z,x+z,x+z,y+w + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 0, 0, 0)); + // ??,??,y+w,y+w + vTemp = _mm_shuffle_ps(vTemp, vLengthSq, _MM_SHUFFLE(3, 3, 0, 0)); + // ??,??,x+z+y+w,?? + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // Splat the length + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 2, 2, 2)); + // Get the length + vLengthSq = _mm_sqrt_ps(vLengthSq); + return vLengthSq; +#endif +} + +//------------------------------------------------------------------------------ +// XMVector4NormalizeEst uses a reciprocal estimate and +// returns QNaN on zero and infinite vectors. + +inline XMVECTOR XM_CALLCONV XMVector4NormalizeEst(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR Result; + Result = XMVector4ReciprocalLength(V); + Result = XMVectorMultiply(V, Result); + return Result; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot4 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + // Reciprocal sqrt (estimate) + v2 = vrsqrte_f32(v1); + // Normalize + return vmulq_f32(V, vcombine_f32(v2, v2)); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vTemp = _mm_dp_ps(V, V, 0xff); + XMVECTOR vResult = _mm_rsqrt_ps(vTemp); + return _mm_mul_ps(vResult, V); +#elif defined(_XM_SSE3_INTRINSICS_) + XMVECTOR vDot = _mm_mul_ps(V, V); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_hadd_ps(vDot, vDot); + vDot = _mm_rsqrt_ps(vDot); + vDot = _mm_mul_ps(vDot, V); + return vDot; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and w + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(3, 2, 3, 2)); + // x+z, y+w + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // x+z,x+z,x+z,y+w + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 0, 0, 0)); + // ??,??,y+w,y+w + vTemp = _mm_shuffle_ps(vTemp, vLengthSq, _MM_SHUFFLE(3, 3, 0, 0)); + // ??,??,x+z+y+w,?? + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // Splat the length + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 2, 2, 2)); + // Get the reciprocal + XMVECTOR vResult = _mm_rsqrt_ps(vLengthSq); + // Reciprocal mul to perform the normalization + vResult = _mm_mul_ps(vResult, V); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Normalize(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + float fLength; + XMVECTOR vResult; + + vResult = XMVector4Length(V); + fLength = vResult.vector4_f32[0]; + + // Prevent divide by zero + if (fLength > 0) + { + fLength = 1.0f / fLength; + } + + vResult.vector4_f32[0] = V.vector4_f32[0] * fLength; + vResult.vector4_f32[1] = V.vector4_f32[1] * fLength; + vResult.vector4_f32[2] = V.vector4_f32[2] * fLength; + vResult.vector4_f32[3] = V.vector4_f32[3] * fLength; + return vResult; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + // Dot4 + float32x4_t vTemp = vmulq_f32(V, V); + float32x2_t v1 = vget_low_f32(vTemp); + float32x2_t v2 = vget_high_f32(vTemp); + v1 = vadd_f32(v1, v2); + v1 = vpadd_f32(v1, v1); + uint32x2_t VEqualsZero = vceq_f32(v1, vdup_n_f32(0)); + uint32x2_t VEqualsInf = vceq_f32(v1, vget_low_f32(g_XMInfinity)); + // Reciprocal sqrt (2 iterations of Newton-Raphson) + float32x2_t S0 = vrsqrte_f32(v1); + float32x2_t P0 = vmul_f32(v1, S0); + float32x2_t R0 = vrsqrts_f32(P0, S0); + float32x2_t S1 = vmul_f32(S0, R0); + float32x2_t P1 = vmul_f32(v1, S1); + float32x2_t R1 = vrsqrts_f32(P1, S1); + v2 = vmul_f32(S1, R1); + // Normalize + XMVECTOR vResult = vmulq_f32(V, vcombine_f32(v2, v2)); + vResult = vbslq_f32(vcombine_u32(VEqualsZero, VEqualsZero), vdupq_n_f32(0), vResult); + return vbslq_f32(vcombine_u32(VEqualsInf, VEqualsInf), g_XMQNaN, vResult); +#elif defined(_XM_SSE4_INTRINSICS_) + XMVECTOR vLengthSq = _mm_dp_ps(V, V, 0xff); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Divide to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#elif defined(_XM_SSE3_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + vLengthSq = _mm_hadd_ps(vLengthSq, vLengthSq); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Divide to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + // Perform the dot product on x,y,z and w + XMVECTOR vLengthSq = _mm_mul_ps(V, V); + // vTemp has z and w + XMVECTOR vTemp = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(3, 2, 3, 2)); + // x+z, y+w + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // x+z,x+z,x+z,y+w + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(1, 0, 0, 0)); + // ??,??,y+w,y+w + vTemp = _mm_shuffle_ps(vTemp, vLengthSq, _MM_SHUFFLE(3, 3, 0, 0)); + // ??,??,x+z+y+w,?? + vLengthSq = _mm_add_ps(vLengthSq, vTemp); + // Splat the length + vLengthSq = XM_PERMUTE_PS(vLengthSq, _MM_SHUFFLE(2, 2, 2, 2)); + // Prepare for the division + XMVECTOR vResult = _mm_sqrt_ps(vLengthSq); + // Create zero with a single instruction + XMVECTOR vZeroMask = _mm_setzero_ps(); + // Test for a divide by zero (Must be FP to detect -0.0) + vZeroMask = _mm_cmpneq_ps(vZeroMask, vResult); + // Failsafe on zero (Or epsilon) length planes + // If the length is infinity, set the elements to zero + vLengthSq = _mm_cmpneq_ps(vLengthSq, g_XMInfinity); + // Divide to perform the normalization + vResult = _mm_div_ps(V, vResult); + // Any that are infinity, set to zero + vResult = _mm_and_ps(vResult, vZeroMask); + // Select qnan or result based on infinite length + XMVECTOR vTemp1 = _mm_andnot_ps(vLengthSq, g_XMQNaN); + XMVECTOR vTemp2 = _mm_and_ps(vResult, vLengthSq); + vResult = _mm_or_ps(vTemp1, vTemp2); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4ClampLength +( + FXMVECTOR V, + float LengthMin, + float LengthMax +) noexcept +{ + XMVECTOR ClampMax = XMVectorReplicate(LengthMax); + XMVECTOR ClampMin = XMVectorReplicate(LengthMin); + + return XMVector4ClampLengthV(V, ClampMin, ClampMax); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4ClampLengthV +( + FXMVECTOR V, + FXMVECTOR LengthMin, + FXMVECTOR LengthMax +) noexcept +{ + assert((XMVectorGetY(LengthMin) == XMVectorGetX(LengthMin)) && (XMVectorGetZ(LengthMin) == XMVectorGetX(LengthMin)) && (XMVectorGetW(LengthMin) == XMVectorGetX(LengthMin))); + assert((XMVectorGetY(LengthMax) == XMVectorGetX(LengthMax)) && (XMVectorGetZ(LengthMax) == XMVectorGetX(LengthMax)) && (XMVectorGetW(LengthMax) == XMVectorGetX(LengthMax))); + assert(XMVector4GreaterOrEqual(LengthMin, XMVectorZero())); + assert(XMVector4GreaterOrEqual(LengthMax, XMVectorZero())); + assert(XMVector4GreaterOrEqual(LengthMax, LengthMin)); + + XMVECTOR LengthSq = XMVector4LengthSq(V); + + const XMVECTOR Zero = XMVectorZero(); + + XMVECTOR RcpLength = XMVectorReciprocalSqrt(LengthSq); + + XMVECTOR InfiniteLength = XMVectorEqualInt(LengthSq, g_XMInfinity.v); + XMVECTOR ZeroLength = XMVectorEqual(LengthSq, Zero); + + XMVECTOR Normal = XMVectorMultiply(V, RcpLength); + + XMVECTOR Length = XMVectorMultiply(LengthSq, RcpLength); + + XMVECTOR Select = XMVectorEqualInt(InfiniteLength, ZeroLength); + Length = XMVectorSelect(LengthSq, Length, Select); + Normal = XMVectorSelect(LengthSq, Normal, Select); + + XMVECTOR ControlMax = XMVectorGreater(Length, LengthMax); + XMVECTOR ControlMin = XMVectorLess(Length, LengthMin); + + XMVECTOR ClampLength = XMVectorSelect(Length, LengthMax, ControlMax); + ClampLength = XMVectorSelect(ClampLength, LengthMin, ControlMin); + + XMVECTOR Result = XMVectorMultiply(Normal, ClampLength); + + // Preserve the original vector (with no precision loss) if the length falls within the given range + XMVECTOR Control = XMVectorEqualInt(ControlMax, ControlMin); + Result = XMVectorSelect(Result, V, Control); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Reflect +( + FXMVECTOR Incident, + FXMVECTOR Normal +) noexcept +{ + // Result = Incident - (2 * dot(Incident, Normal)) * Normal + + XMVECTOR Result = XMVector4Dot(Incident, Normal); + Result = XMVectorAdd(Result, Result); + Result = XMVectorNegativeMultiplySubtract(Result, Normal, Incident); + + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Refract +( + FXMVECTOR Incident, + FXMVECTOR Normal, + float RefractionIndex +) noexcept +{ + XMVECTOR Index = XMVectorReplicate(RefractionIndex); + return XMVector4RefractV(Incident, Normal, Index); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4RefractV +( + FXMVECTOR Incident, + FXMVECTOR Normal, + FXMVECTOR RefractionIndex +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR IDotN; + XMVECTOR R; + const XMVECTOR Zero = XMVectorZero(); + + // Result = RefractionIndex * Incident - Normal * (RefractionIndex * dot(Incident, Normal) + + // sqrt(1 - RefractionIndex * RefractionIndex * (1 - dot(Incident, Normal) * dot(Incident, Normal)))) + + IDotN = XMVector4Dot(Incident, Normal); + + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + R = XMVectorNegativeMultiplySubtract(IDotN, IDotN, g_XMOne.v); + R = XMVectorMultiply(R, RefractionIndex); + R = XMVectorNegativeMultiplySubtract(R, RefractionIndex, g_XMOne.v); + + if (XMVector4LessOrEqual(R, Zero)) + { + // Total internal reflection + return Zero; + } + else + { + XMVECTOR Result; + + // R = RefractionIndex * IDotN + sqrt(R) + R = XMVectorSqrt(R); + R = XMVectorMultiplyAdd(RefractionIndex, IDotN, R); + + // Result = RefractionIndex * Incident - Normal * R + Result = XMVectorMultiply(RefractionIndex, Incident); + Result = XMVectorNegativeMultiplySubtract(Normal, R, Result); + + return Result; + } + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + XMVECTOR IDotN = XMVector4Dot(Incident, Normal); + + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + float32x4_t R = vmlsq_f32(g_XMOne, IDotN, IDotN); + R = vmulq_f32(R, RefractionIndex); + R = vmlsq_f32(g_XMOne, R, RefractionIndex); + + uint32x4_t isrzero = vcleq_f32(R, g_XMZero); + uint8x8x2_t vTemp = vzip_u8(vget_low_u8(vreinterpretq_u8_u32(isrzero)), vget_high_u8(vreinterpretq_u8_u32(isrzero))); + uint16x4x2_t vTemp2 = vzip_u16(vreinterpret_u16_u8(vTemp.val[0]), vreinterpret_u16_u8(vTemp.val[1])); + + float32x4_t vResult; + if (vget_lane_u32(vreinterpret_u32_u16(vTemp2.val[1]), 1) == 0xFFFFFFFFU) + { + // Total internal reflection + vResult = g_XMZero; + } + else + { + // Sqrt(R) + float32x4_t S0 = vrsqrteq_f32(R); + float32x4_t P0 = vmulq_f32(R, S0); + float32x4_t R0 = vrsqrtsq_f32(P0, S0); + float32x4_t S1 = vmulq_f32(S0, R0); + float32x4_t P1 = vmulq_f32(R, S1); + float32x4_t R1 = vrsqrtsq_f32(P1, S1); + float32x4_t S2 = vmulq_f32(S1, R1); + R = vmulq_f32(R, S2); + // R = RefractionIndex * IDotN + sqrt(R) + R = vmlaq_f32(R, RefractionIndex, IDotN); + // Result = RefractionIndex * Incident - Normal * R + vResult = vmulq_f32(RefractionIndex, Incident); + vResult = vmlsq_f32(vResult, R, Normal); + } + return vResult; +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR IDotN = XMVector4Dot(Incident, Normal); + + // R = 1.0f - RefractionIndex * RefractionIndex * (1.0f - IDotN * IDotN) + XMVECTOR R = XM_FNMADD_PS(IDotN, IDotN, g_XMOne); + XMVECTOR R2 = _mm_mul_ps(RefractionIndex, RefractionIndex); + R = XM_FNMADD_PS(R, R2, g_XMOne); + + XMVECTOR vResult = _mm_cmple_ps(R, g_XMZero); + if (_mm_movemask_ps(vResult) == 0x0f) + { + // Total internal reflection + vResult = g_XMZero; + } + else + { + // R = RefractionIndex * IDotN + sqrt(R) + R = _mm_sqrt_ps(R); + R = XM_FMADD_PS(RefractionIndex, IDotN, R); + // Result = RefractionIndex * Incident - Normal * R + vResult = _mm_mul_ps(RefractionIndex, Incident); + vResult = XM_FNMADD_PS(R, Normal, vResult); + } + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Orthogonal(FXMVECTOR V) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + XMVECTORF32 Result = { { { + V.vector4_f32[2], + V.vector4_f32[3], + -V.vector4_f32[0], + -V.vector4_f32[1] + } } }; + return Result.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Negate = { { { 1.f, 1.f, -1.f, -1.f } } }; + + float32x4_t Result = vcombine_f32(vget_high_f32(V), vget_low_f32(V)); + return vmulq_f32(Result, Negate); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 FlipZW = { { { 1.0f, 1.0f, -1.0f, -1.0f } } }; + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 0, 3, 2)); + vResult = _mm_mul_ps(vResult, FlipZW); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4AngleBetweenNormalsEst +( + FXMVECTOR N1, + FXMVECTOR N2 +) noexcept +{ + XMVECTOR Result = XMVector4Dot(N1, N2); + Result = XMVectorClamp(Result, g_XMNegativeOne.v, g_XMOne.v); + Result = XMVectorACosEst(Result); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4AngleBetweenNormals +( + FXMVECTOR N1, + FXMVECTOR N2 +) noexcept +{ + XMVECTOR Result = XMVector4Dot(N1, N2); + Result = XMVectorClamp(Result, g_XMNegativeOne.v, g_XMOne.v); + Result = XMVectorACos(Result); + return Result; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4AngleBetweenVectors +( + FXMVECTOR V1, + FXMVECTOR V2 +) noexcept +{ + XMVECTOR L1 = XMVector4ReciprocalLength(V1); + XMVECTOR L2 = XMVector4ReciprocalLength(V2); + + XMVECTOR Dot = XMVector4Dot(V1, V2); + + L1 = XMVectorMultiply(L1, L2); + + XMVECTOR CosAngle = XMVectorMultiply(Dot, L1); + CosAngle = XMVectorClamp(CosAngle, g_XMNegativeOne.v, g_XMOne.v); + + return XMVectorACos(CosAngle); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV XMVector4Transform +( + FXMVECTOR V, + FXMMATRIX M +) noexcept +{ +#if defined(_XM_NO_INTRINSICS_) + + float fX = (M.m[0][0] * V.vector4_f32[0]) + (M.m[1][0] * V.vector4_f32[1]) + (M.m[2][0] * V.vector4_f32[2]) + (M.m[3][0] * V.vector4_f32[3]); + float fY = (M.m[0][1] * V.vector4_f32[0]) + (M.m[1][1] * V.vector4_f32[1]) + (M.m[2][1] * V.vector4_f32[2]) + (M.m[3][1] * V.vector4_f32[3]); + float fZ = (M.m[0][2] * V.vector4_f32[0]) + (M.m[1][2] * V.vector4_f32[1]) + (M.m[2][2] * V.vector4_f32[2]) + (M.m[3][2] * V.vector4_f32[3]); + float fW = (M.m[0][3] * V.vector4_f32[0]) + (M.m[1][3] * V.vector4_f32[1]) + (M.m[2][3] * V.vector4_f32[2]) + (M.m[3][3] * V.vector4_f32[3]); + XMVECTORF32 vResult = { { { fX, fY, fZ, fW } } }; + return vResult.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x2_t VL = vget_low_f32(V); + XMVECTOR vResult = vmulq_lane_f32(M.r[0], VL, 0); // X + vResult = vmlaq_lane_f32(vResult, M.r[1], VL, 1); // Y + float32x2_t VH = vget_high_f32(V); + vResult = vmlaq_lane_f32(vResult, M.r[2], VH, 0); // Z + return vmlaq_lane_f32(vResult, M.r[3], VH, 1); // W +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); // W + vResult = _mm_mul_ps(vResult, M.r[3]); + XMVECTOR vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); // Z + vResult = XM_FMADD_PS(vTemp, M.r[2], vResult); + vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); // Y + vResult = XM_FMADD_PS(vTemp, M.r[1], vResult); + vTemp = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); // X + vResult = XM_FMADD_PS(vTemp, M.r[0], vResult); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMFLOAT4* XM_CALLCONV XMVector4TransformStream +( + XMFLOAT4* pOutputStream, + size_t OutputStride, + const XMFLOAT4* pInputStream, + size_t InputStride, + size_t VectorCount, + FXMMATRIX M +) noexcept +{ + assert(pOutputStream != nullptr); + assert(pInputStream != nullptr); + + assert(InputStride >= sizeof(XMFLOAT4)); + _Analysis_assume_(InputStride >= sizeof(XMFLOAT4)); + + assert(OutputStride >= sizeof(XMFLOAT4)); + _Analysis_assume_(OutputStride >= sizeof(XMFLOAT4)); + +#if defined(_XM_NO_INTRINSICS_) + + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + for (size_t i = 0; i < VectorCount; i++) + { + XMVECTOR V = XMLoadFloat4(reinterpret_cast(pInputVector)); + XMVECTOR W = XMVectorSplatW(V); + XMVECTOR Z = XMVectorSplatZ(V); + XMVECTOR Y = XMVectorSplatY(V); + XMVECTOR X = XMVectorSplatX(V); + + XMVECTOR Result = XMVectorMultiply(W, row3); + Result = XMVectorMultiplyAdd(Z, row2, Result); + Result = XMVectorMultiplyAdd(Y, row1, Result); + Result = XMVectorMultiplyAdd(X, row0, Result); + + #ifdef _PREFAST_ + #pragma prefast(push) + #pragma prefast(disable : 26015, "PREfast noise: Esp:1307" ) + #endif + + XMStoreFloat4(reinterpret_cast(pOutputVector), Result); + + #ifdef _PREFAST_ + #pragma prefast(pop) + #endif + + pInputVector += InputStride; + pOutputVector += OutputStride; + } + + return pOutputStream; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + size_t i = 0; + size_t four = VectorCount >> 2; + if (four > 0) + { + if ((InputStride == sizeof(XMFLOAT4)) && (OutputStride == sizeof(XMFLOAT4))) + { + for (size_t j = 0; j < four; ++j) + { + float32x4x4_t V = vld4q_f32(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT4) * 4; + + float32x2_t r = vget_low_f32(row0); + XMVECTOR vResult0 = vmulq_lane_f32(V.val[0], r, 0); // Ax + XMVECTOR vResult1 = vmulq_lane_f32(V.val[0], r, 1); // Bx + + XM_PREFETCH(pInputVector); + + r = vget_high_f32(row0); + XMVECTOR vResult2 = vmulq_lane_f32(V.val[0], r, 0); // Cx + XMVECTOR vResult3 = vmulq_lane_f32(V.val[0], r, 1); // Dx + + XM_PREFETCH(pInputVector + XM_CACHE_LINE_SIZE); + + r = vget_low_f32(row1); + vResult0 = vmlaq_lane_f32(vResult0, V.val[1], r, 0); // Ax+Ey + vResult1 = vmlaq_lane_f32(vResult1, V.val[1], r, 1); // Bx+Fy + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 2)); + + r = vget_high_f32(row1); + vResult2 = vmlaq_lane_f32(vResult2, V.val[1], r, 0); // Cx+Gy + vResult3 = vmlaq_lane_f32(vResult3, V.val[1], r, 1); // Dx+Hy + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 3)); + + r = vget_low_f32(row2); + vResult0 = vmlaq_lane_f32(vResult0, V.val[2], r, 0); // Ax+Ey+Iz + vResult1 = vmlaq_lane_f32(vResult1, V.val[2], r, 1); // Bx+Fy+Jz + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 4)); + + r = vget_high_f32(row2); + vResult2 = vmlaq_lane_f32(vResult2, V.val[2], r, 0); // Cx+Gy+Kz + vResult3 = vmlaq_lane_f32(vResult3, V.val[2], r, 1); // Dx+Hy+Lz + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 5)); + + r = vget_low_f32(row3); + vResult0 = vmlaq_lane_f32(vResult0, V.val[3], r, 0); // Ax+Ey+Iz+Mw + vResult1 = vmlaq_lane_f32(vResult1, V.val[3], r, 1); // Bx+Fy+Jz+Nw + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 6)); + + r = vget_high_f32(row3); + vResult2 = vmlaq_lane_f32(vResult2, V.val[3], r, 0); // Cx+Gy+Kz+Ow + vResult3 = vmlaq_lane_f32(vResult3, V.val[3], r, 1); // Dx+Hy+Lz+Pw + + XM_PREFETCH(pInputVector + (XM_CACHE_LINE_SIZE * 7)); + + V.val[0] = vResult0; + V.val[1] = vResult1; + V.val[2] = vResult2; + V.val[3] = vResult3; + + vst4q_f32(reinterpret_cast(pOutputVector), V); + pOutputVector += sizeof(XMFLOAT4) * 4; + + i += 4; + } + } + } + + for (; i < VectorCount; i++) + { + XMVECTOR V = vld1q_f32(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + float32x2_t VL = vget_low_f32(V); + XMVECTOR vResult = vmulq_lane_f32(row0, VL, 0); // X + vResult = vmlaq_lane_f32(vResult, row1, VL, 1); // Y + float32x2_t VH = vget_high_f32(V); + vResult = vmlaq_lane_f32(vResult, row2, VH, 0); // Z + vResult = vmlaq_lane_f32(vResult, row3, VH, 1); // W + + vst1q_f32(reinterpret_cast(pOutputVector), vResult); + pOutputVector += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_AVX2_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t two = VectorCount >> 1; + if (two > 0) + { + __m256 row0 = _mm256_broadcast_ps(&M.r[0]); + __m256 row1 = _mm256_broadcast_ps(&M.r[1]); + __m256 row2 = _mm256_broadcast_ps(&M.r[2]); + __m256 row3 = _mm256_broadcast_ps(&M.r[3]); + + if (InputStride == sizeof(XMFLOAT4)) + { + if (OutputStride == sizeof(XMFLOAT4)) + { + if (!(reinterpret_cast(pOutputStream) & 0x1F)) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < two; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT4) * 2; + + __m256 vTempX = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 vTempY = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 vTempZ = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 vTempW = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm256_mul_ps(vTempX, row0); + vTempY = _mm256_mul_ps(vTempY, row1); + vTempZ = _mm256_fmadd_ps(vTempZ, row2, vTempX); + vTempW = _mm256_fmadd_ps(vTempW, row3, vTempY); + vTempX = _mm256_add_ps(vTempZ, vTempW); + + XM256_STREAM_PS(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += sizeof(XMFLOAT4) * 2; + + i += 2; + } + } + else + { + // Packed input, packed output + for (size_t j = 0; j < two; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT4) * 2; + + __m256 vTempX = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 vTempY = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 vTempZ = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 vTempW = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm256_mul_ps(vTempX, row0); + vTempY = _mm256_mul_ps(vTempY, row1); + vTempZ = _mm256_fmadd_ps(vTempZ, row2, vTempX); + vTempW = _mm256_fmadd_ps(vTempW, row3, vTempY); + vTempX = _mm256_add_ps(vTempZ, vTempW); + + _mm256_storeu_ps(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += sizeof(XMFLOAT4) * 2; + + i += 2; + } + } + } + else + { + // Packed input, unpacked output + for (size_t j = 0; j < two; ++j) + { + __m256 VV = _mm256_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += sizeof(XMFLOAT4) * 2; + + __m256 vTempX = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(0, 0, 0, 0)); + __m256 vTempY = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(1, 1, 1, 1)); + __m256 vTempZ = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(2, 2, 2, 2)); + __m256 vTempW = _mm256_shuffle_ps(VV, VV, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm256_mul_ps(vTempX, row0); + vTempY = _mm256_mul_ps(vTempY, row1); + vTempZ = _mm256_fmadd_ps(vTempZ, row2, vTempX); + vTempW = _mm256_fmadd_ps(vTempW, row3, vTempY); + vTempX = _mm256_add_ps(vTempZ, vTempW); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), _mm256_castps256_ps128(vTempX)); + pOutputVector += OutputStride; + + _mm_storeu_ps(reinterpret_cast(pOutputVector), _mm256_extractf128_ps(vTempX, 1)); + pOutputVector += OutputStride; + i += 2; + } + } + } + } + + if (i < VectorCount) + { + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + for (; i < VectorCount; i++) + { + __m128 V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vTempX = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vTempY = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vTempZ = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR vTempW = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm_mul_ps(vTempX, row0); + vTempY = _mm_mul_ps(vTempY, row1); + vTempZ = XM_FMADD_PS(vTempZ, row2, vTempX); + vTempW = XM_FMADD_PS(vTempW, row3, vTempY); + vTempX = _mm_add_ps(vTempZ, vTempW); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += OutputStride; + } + } + + XM_SFENCE(); + + return pOutputStream; +#elif defined(_XM_SSE_INTRINSICS_) + auto pInputVector = reinterpret_cast(pInputStream); + auto pOutputVector = reinterpret_cast(pOutputStream); + + const XMVECTOR row0 = M.r[0]; + const XMVECTOR row1 = M.r[1]; + const XMVECTOR row2 = M.r[2]; + const XMVECTOR row3 = M.r[3]; + + if (!(reinterpret_cast(pOutputStream) & 0xF) && !(OutputStride & 0xF)) + { + if (!(reinterpret_cast(pInputStream) & 0xF) && !(InputStride & 0xF)) + { + // Aligned input, aligned output + for (size_t i = 0; i < VectorCount; i++) + { + __m128 V = _mm_load_ps(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vTempX = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vTempY = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vTempZ = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR vTempW = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm_mul_ps(vTempX, row0); + vTempY = _mm_mul_ps(vTempY, row1); + vTempZ = XM_FMADD_PS(vTempZ, row2, vTempX); + vTempW = XM_FMADD_PS(vTempW, row3, vTempY); + vTempX = _mm_add_ps(vTempZ, vTempW); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += OutputStride; + } + } + else + { + // Unaligned input, aligned output + for (size_t i = 0; i < VectorCount; i++) + { + __m128 V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vTempX = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vTempY = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vTempZ = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR vTempW = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm_mul_ps(vTempX, row0); + vTempY = _mm_mul_ps(vTempY, row1); + vTempZ = XM_FMADD_PS(vTempZ, row2, vTempX); + vTempW = XM_FMADD_PS(vTempW, row3, vTempY); + vTempX = _mm_add_ps(vTempZ, vTempW); + + XM_STREAM_PS(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += OutputStride; + } + } + } + else + { + if (!(reinterpret_cast(pInputStream) & 0xF) && !(InputStride & 0xF)) + { + // Aligned input, unaligned output + for (size_t i = 0; i < VectorCount; i++) + { + __m128 V = _mm_load_ps(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vTempX = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vTempY = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vTempZ = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR vTempW = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm_mul_ps(vTempX, row0); + vTempY = _mm_mul_ps(vTempY, row1); + vTempZ = XM_FMADD_PS(vTempZ, row2, vTempX); + vTempW = XM_FMADD_PS(vTempW, row3, vTempY); + vTempX = _mm_add_ps(vTempZ, vTempW); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += OutputStride; + } + } + else + { + // Unaligned input, unaligned output + for (size_t i = 0; i < VectorCount; i++) + { + __m128 V = _mm_loadu_ps(reinterpret_cast(pInputVector)); + pInputVector += InputStride; + + XMVECTOR vTempX = XM_PERMUTE_PS(V, _MM_SHUFFLE(0, 0, 0, 0)); + XMVECTOR vTempY = XM_PERMUTE_PS(V, _MM_SHUFFLE(1, 1, 1, 1)); + XMVECTOR vTempZ = XM_PERMUTE_PS(V, _MM_SHUFFLE(2, 2, 2, 2)); + XMVECTOR vTempW = XM_PERMUTE_PS(V, _MM_SHUFFLE(3, 3, 3, 3)); + + vTempX = _mm_mul_ps(vTempX, row0); + vTempY = _mm_mul_ps(vTempY, row1); + vTempZ = XM_FMADD_PS(vTempZ, row2, vTempX); + vTempW = XM_FMADD_PS(vTempW, row3, vTempY); + vTempX = _mm_add_ps(vTempZ, vTempW); + + _mm_storeu_ps(reinterpret_cast(pOutputVector), vTempX); + pOutputVector += OutputStride; + } + } + } + + XM_SFENCE(); + + return pOutputStream; +#endif +} + +/**************************************************************************** + * + * XMVECTOR operators + * + ****************************************************************************/ + +#ifndef _XM_NO_XMVECTOR_OVERLOADS_ + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator+ (FXMVECTOR V) noexcept +{ + return V; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator- (FXMVECTOR V) noexcept +{ + return XMVectorNegate(V); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR& XM_CALLCONV operator+= +( + XMVECTOR& V1, + FXMVECTOR V2 + ) noexcept +{ + V1 = XMVectorAdd(V1, V2); + return V1; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR& XM_CALLCONV operator-= +( + XMVECTOR& V1, + FXMVECTOR V2 + ) noexcept +{ + V1 = XMVectorSubtract(V1, V2); + return V1; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR& XM_CALLCONV operator*= +( + XMVECTOR& V1, + FXMVECTOR V2 + ) noexcept +{ + V1 = XMVectorMultiply(V1, V2); + return V1; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR& XM_CALLCONV operator/= +( + XMVECTOR& V1, + FXMVECTOR V2 + ) noexcept +{ + V1 = XMVectorDivide(V1, V2); + return V1; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR& operator*= +( + XMVECTOR& V, + const float S + ) noexcept +{ + V = XMVectorScale(V, S); + return V; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR& operator/= +( + XMVECTOR& V, + const float S + ) noexcept +{ + XMVECTOR vS = XMVectorReplicate(S); + V = XMVectorDivide(V, vS); + return V; +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator+ +( + FXMVECTOR V1, + FXMVECTOR V2 + ) noexcept +{ + return XMVectorAdd(V1, V2); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator- +( + FXMVECTOR V1, + FXMVECTOR V2 + ) noexcept +{ + return XMVectorSubtract(V1, V2); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator* +( + FXMVECTOR V1, + FXMVECTOR V2 + ) noexcept +{ + return XMVectorMultiply(V1, V2); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator/ +( + FXMVECTOR V1, + FXMVECTOR V2 + ) noexcept +{ + return XMVectorDivide(V1, V2); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator* +( + FXMVECTOR V, + const float S + ) noexcept +{ + return XMVectorScale(V, S); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator/ +( + FXMVECTOR V, + const float S + ) noexcept +{ + XMVECTOR vS = XMVectorReplicate(S); + return XMVectorDivide(V, vS); +} + +//------------------------------------------------------------------------------ + +inline XMVECTOR XM_CALLCONV operator* +( + float S, + FXMVECTOR V + ) noexcept +{ + return XMVectorScale(V, S); +} + +#endif /* !_XM_NO_XMVECTOR_OVERLOADS_ */ + +#if defined(_XM_NO_INTRINSICS_) +#undef XMISNAN +#undef XMISINF +#endif + +#if defined(_XM_SSE_INTRINSICS_) +#undef XM3UNPACK3INTO4 +#undef XM3PACK4INTO3 +#endif + diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXPackedVector.h b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXPackedVector.h new file mode 100644 index 00000000..8af22683 --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXPackedVector.h @@ -0,0 +1,1252 @@ +//------------------------------------------------------------------------------------- +// DirectXPackedVector.h -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +#include "DirectXMath.h" + +namespace DirectX +{ + + namespace PackedVector + { + #ifdef _MSC_VER + #pragma warning(push) + #pragma warning(disable:4201 4365 4324 4996) + // C4201: nonstandard extension used + // C4365: Off by default noise + // C4324: alignment padding warnings + // C4996: deprecation warnings + #endif + + #ifdef __clang__ + #pragma clang diagnostic push + #pragma clang diagnostic ignored "-Wgnu-anonymous-struct" + #pragma clang diagnostic ignored "-Wnested-anon-types" + #endif + + //------------------------------------------------------------------------------ + // ARGB Color; 8-8-8-8 bit unsigned normalized integer components packed into + // a 32 bit integer. The normalized color is packed into 32 bits using 8 bit + // unsigned, normalized integers for the alpha, red, green, and blue components. + // The alpha component is stored in the most significant bits and the blue + // component in the least significant bits (A8R8G8B8): + // [32] aaaaaaaa rrrrrrrr gggggggg bbbbbbbb [0] + struct XMCOLOR + { + union + { + struct + { + uint8_t b; // Blue: 0/255 to 255/255 + uint8_t g; // Green: 0/255 to 255/255 + uint8_t r; // Red: 0/255 to 255/255 + uint8_t a; // Alpha: 0/255 to 255/255 + }; + uint32_t c; + }; + + XMCOLOR() = default; + + XMCOLOR(const XMCOLOR&) = default; + XMCOLOR& operator=(const XMCOLOR&) = default; + + XMCOLOR(XMCOLOR&&) = default; + XMCOLOR& operator=(XMCOLOR&&) = default; + + constexpr XMCOLOR(uint32_t Color) noexcept : c(Color) {} + XMCOLOR(float _r, float _g, float _b, float _a) noexcept; + explicit XMCOLOR(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return c; } + + XMCOLOR& operator= (const uint32_t Color) noexcept { c = Color; return *this; } + }; + + //------------------------------------------------------------------------------ + // 16 bit floating point number consisting of a sign bit, a 5 bit biased + // exponent, and a 10 bit mantissa + using HALF = uint16_t; + + //------------------------------------------------------------------------------ + // 2D Vector; 16 bit floating point components + struct XMHALF2 + { + union + { + struct + { + HALF x; + HALF y; + }; + uint32_t v; + }; + + XMHALF2() = default; + + XMHALF2(const XMHALF2&) = default; + XMHALF2& operator=(const XMHALF2&) = default; + + XMHALF2(XMHALF2&&) = default; + XMHALF2& operator=(XMHALF2&&) = default; + + explicit constexpr XMHALF2(uint32_t Packed) noexcept : v(Packed) {} + constexpr XMHALF2(HALF _x, HALF _y) noexcept : x(_x), y(_y) {} + explicit XMHALF2(_In_reads_(2) const HALF* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMHALF2(float _x, float _y) noexcept; + explicit XMHALF2(_In_reads_(2) const float* pArray) noexcept; + + XMHALF2& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 2D Vector; 16 bit signed normalized integer components + struct XMSHORTN2 + { + union + { + struct + { + int16_t x; + int16_t y; + }; + uint32_t v; + }; + + XMSHORTN2() = default; + + XMSHORTN2(const XMSHORTN2&) = default; + XMSHORTN2& operator=(const XMSHORTN2&) = default; + + XMSHORTN2(XMSHORTN2&&) = default; + XMSHORTN2& operator=(XMSHORTN2&&) = default; + + explicit constexpr XMSHORTN2(uint32_t Packed) noexcept : v(Packed) {} + constexpr XMSHORTN2(int16_t _x, int16_t _y) noexcept : x(_x), y(_y) {} + explicit XMSHORTN2(_In_reads_(2) const int16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMSHORTN2(float _x, float _y) noexcept; + explicit XMSHORTN2(_In_reads_(2) const float* pArray) noexcept; + + XMSHORTN2& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 2D Vector; 16 bit signed integer components + struct XMSHORT2 + { + union + { + struct + { + int16_t x; + int16_t y; + }; + uint32_t v; + }; + + XMSHORT2() = default; + + XMSHORT2(const XMSHORT2&) = default; + XMSHORT2& operator=(const XMSHORT2&) = default; + + XMSHORT2(XMSHORT2&&) = default; + XMSHORT2& operator=(XMSHORT2&&) = default; + + explicit constexpr XMSHORT2(uint32_t Packed) noexcept : v(Packed) {} + constexpr XMSHORT2(int16_t _x, int16_t _y) noexcept : x(_x), y(_y) {} + explicit XMSHORT2(_In_reads_(2) const int16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMSHORT2(float _x, float _y) noexcept; + explicit XMSHORT2(_In_reads_(2) const float* pArray) noexcept; + + XMSHORT2& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 2D Vector; 16 bit unsigned normalized integer components + struct XMUSHORTN2 + { + union + { + struct + { + uint16_t x; + uint16_t y; + }; + uint32_t v; + }; + + XMUSHORTN2() = default; + + XMUSHORTN2(const XMUSHORTN2&) = default; + XMUSHORTN2& operator=(const XMUSHORTN2&) = default; + + XMUSHORTN2(XMUSHORTN2&&) = default; + XMUSHORTN2& operator=(XMUSHORTN2&&) = default; + + explicit constexpr XMUSHORTN2(uint32_t Packed) noexcept : v(Packed) {} + constexpr XMUSHORTN2(uint16_t _x, uint16_t _y) noexcept : x(_x), y(_y) {} + explicit XMUSHORTN2(_In_reads_(2) const uint16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMUSHORTN2(float _x, float _y) noexcept; + explicit XMUSHORTN2(_In_reads_(2) const float* pArray) noexcept; + + XMUSHORTN2& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 2D Vector; 16 bit unsigned integer components + struct XMUSHORT2 + { + union + { + struct + { + uint16_t x; + uint16_t y; + }; + uint32_t v; + }; + + XMUSHORT2() = default; + + XMUSHORT2(const XMUSHORT2&) = default; + XMUSHORT2& operator=(const XMUSHORT2&) = default; + + XMUSHORT2(XMUSHORT2&&) = default; + XMUSHORT2& operator=(XMUSHORT2&&) = default; + + explicit constexpr XMUSHORT2(uint32_t Packed) noexcept : v(Packed) {} + constexpr XMUSHORT2(uint16_t _x, uint16_t _y) noexcept : x(_x), y(_y) {} + explicit XMUSHORT2(_In_reads_(2) const uint16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMUSHORT2(float _x, float _y) noexcept; + explicit XMUSHORT2(_In_reads_(2) const float* pArray) noexcept; + + XMUSHORT2& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 2D Vector; 8 bit signed normalized integer components + struct XMBYTEN2 + { + union + { + struct + { + int8_t x; + int8_t y; + }; + uint16_t v; + }; + + XMBYTEN2() = default; + + XMBYTEN2(const XMBYTEN2&) = default; + XMBYTEN2& operator=(const XMBYTEN2&) = default; + + XMBYTEN2(XMBYTEN2&&) = default; + XMBYTEN2& operator=(XMBYTEN2&&) = default; + + explicit constexpr XMBYTEN2(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMBYTEN2(int8_t _x, int8_t _y) noexcept : x(_x), y(_y) {} + explicit XMBYTEN2(_In_reads_(2) const int8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMBYTEN2(float _x, float _y) noexcept; + explicit XMBYTEN2(_In_reads_(2) const float* pArray) noexcept; + + XMBYTEN2& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + // 2D Vector; 8 bit signed integer components + struct XMBYTE2 + { + union + { + struct + { + int8_t x; + int8_t y; + }; + uint16_t v; + }; + + XMBYTE2() = default; + + XMBYTE2(const XMBYTE2&) = default; + XMBYTE2& operator=(const XMBYTE2&) = default; + + XMBYTE2(XMBYTE2&&) = default; + XMBYTE2& operator=(XMBYTE2&&) = default; + + explicit constexpr XMBYTE2(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMBYTE2(int8_t _x, int8_t _y) noexcept : x(_x), y(_y) {} + explicit XMBYTE2(_In_reads_(2) const int8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMBYTE2(float _x, float _y) noexcept; + explicit XMBYTE2(_In_reads_(2) const float* pArray) noexcept; + + XMBYTE2& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + // 2D Vector; 8 bit unsigned normalized integer components + struct XMUBYTEN2 + { + union + { + struct + { + uint8_t x; + uint8_t y; + }; + uint16_t v; + }; + + XMUBYTEN2() = default; + + XMUBYTEN2(const XMUBYTEN2&) = default; + XMUBYTEN2& operator=(const XMUBYTEN2&) = default; + + XMUBYTEN2(XMUBYTEN2&&) = default; + XMUBYTEN2& operator=(XMUBYTEN2&&) = default; + + explicit constexpr XMUBYTEN2(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMUBYTEN2(uint8_t _x, uint8_t _y) noexcept : x(_x), y(_y) {} + explicit XMUBYTEN2(_In_reads_(2) const uint8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMUBYTEN2(float _x, float _y) noexcept; + explicit XMUBYTEN2(_In_reads_(2) const float* pArray) noexcept; + + XMUBYTEN2& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + // 2D Vector; 8 bit unsigned integer components + struct XMUBYTE2 + { + union + { + struct + { + uint8_t x; + uint8_t y; + }; + uint16_t v; + }; + + XMUBYTE2() = default; + + XMUBYTE2(const XMUBYTE2&) = default; + XMUBYTE2& operator=(const XMUBYTE2&) = default; + + XMUBYTE2(XMUBYTE2&&) = default; + XMUBYTE2& operator=(XMUBYTE2&&) = default; + + explicit constexpr XMUBYTE2(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMUBYTE2(uint8_t _x, uint8_t _y) noexcept : x(_x), y(_y) {} + explicit XMUBYTE2(_In_reads_(2) const uint8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]) {} + XMUBYTE2(float _x, float _y) noexcept; + explicit XMUBYTE2(_In_reads_(2) const float* pArray) noexcept; + + XMUBYTE2& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 3D vector: 5/6/5 unsigned integer components + struct XMU565 + { + union + { + struct + { + uint16_t x : 5; // 0 to 31 + uint16_t y : 6; // 0 to 63 + uint16_t z : 5; // 0 to 31 + }; + uint16_t v; + }; + + XMU565() = default; + + XMU565(const XMU565&) = default; + XMU565& operator=(const XMU565&) = default; + + XMU565(XMU565&&) = default; + XMU565& operator=(XMU565&&) = default; + + explicit constexpr XMU565(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMU565(uint8_t _x, uint8_t _y, uint8_t _z) noexcept : x(_x), y(_y), z(_z) {} + explicit XMU565(_In_reads_(3) const uint8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]) {} + XMU565(float _x, float _y, float _z) noexcept; + explicit XMU565(_In_reads_(3) const float* pArray) noexcept; + + operator uint16_t () const noexcept { return v; } + + XMU565& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 3D vector: 11/11/10 floating-point components + // The 3D vector is packed into 32 bits as follows: a 5-bit biased exponent + // and 6-bit mantissa for x component, a 5-bit biased exponent and + // 6-bit mantissa for y component, a 5-bit biased exponent and a 5-bit + // mantissa for z. The z component is stored in the most significant bits + // and the x component in the least significant bits. No sign bits so + // all partial-precision numbers are positive. + // (Z10Y11X11): [32] ZZZZZzzz zzzYYYYY yyyyyyXX XXXxxxxx [0] + struct XMFLOAT3PK + { + union + { + struct + { + uint32_t xm : 6; // x-mantissa + uint32_t xe : 5; // x-exponent + uint32_t ym : 6; // y-mantissa + uint32_t ye : 5; // y-exponent + uint32_t zm : 5; // z-mantissa + uint32_t ze : 5; // z-exponent + }; + uint32_t v; + }; + + XMFLOAT3PK() = default; + + XMFLOAT3PK(const XMFLOAT3PK&) = default; + XMFLOAT3PK& operator=(const XMFLOAT3PK&) = default; + + XMFLOAT3PK(XMFLOAT3PK&&) = default; + XMFLOAT3PK& operator=(XMFLOAT3PK&&) = default; + + explicit constexpr XMFLOAT3PK(uint32_t Packed) noexcept : v(Packed) {} + XMFLOAT3PK(float _x, float _y, float _z) noexcept; + explicit XMFLOAT3PK(_In_reads_(3) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMFLOAT3PK& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 3D vector: 9/9/9 floating-point components with shared 5-bit exponent + // The 3D vector is packed into 32 bits as follows: a 5-bit biased exponent + // with 9-bit mantissa for the x, y, and z component. The shared exponent + // is stored in the most significant bits and the x component mantissa is in + // the least significant bits. No sign bits so all partial-precision numbers + // are positive. + // (E5Z9Y9X9): [32] EEEEEzzz zzzzzzyy yyyyyyyx xxxxxxxx [0] + struct XMFLOAT3SE + { + union + { + struct + { + uint32_t xm : 9; // x-mantissa + uint32_t ym : 9; // y-mantissa + uint32_t zm : 9; // z-mantissa + uint32_t e : 5; // shared exponent + }; + uint32_t v; + }; + + XMFLOAT3SE() = default; + + XMFLOAT3SE(const XMFLOAT3SE&) = default; + XMFLOAT3SE& operator=(const XMFLOAT3SE&) = default; + + XMFLOAT3SE(XMFLOAT3SE&&) = default; + XMFLOAT3SE& operator=(XMFLOAT3SE&&) = default; + + explicit constexpr XMFLOAT3SE(uint32_t Packed) noexcept : v(Packed) {} + XMFLOAT3SE(float _x, float _y, float _z) noexcept; + explicit XMFLOAT3SE(_In_reads_(3) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMFLOAT3SE& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 4D Vector; 16 bit floating point components + struct XMHALF4 + { + union + { + struct + { + HALF x; + HALF y; + HALF z; + HALF w; + }; + uint64_t v; + }; + + XMHALF4() = default; + + XMHALF4(const XMHALF4&) = default; + XMHALF4& operator=(const XMHALF4&) = default; + + XMHALF4(XMHALF4&&) = default; + XMHALF4& operator=(XMHALF4&&) = default; + + explicit constexpr XMHALF4(uint64_t Packed) noexcept : v(Packed) {} + constexpr XMHALF4(HALF _x, HALF _y, HALF _z, HALF _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMHALF4(_In_reads_(4) const HALF* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMHALF4(float _x, float _y, float _z, float _w) noexcept; + explicit XMHALF4(_In_reads_(4) const float* pArray) noexcept; + + XMHALF4& operator= (uint64_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 4D Vector; 16 bit signed normalized integer components + struct XMSHORTN4 + { + union + { + struct + { + int16_t x; + int16_t y; + int16_t z; + int16_t w; + }; + uint64_t v; + }; + + XMSHORTN4() = default; + + XMSHORTN4(const XMSHORTN4&) = default; + XMSHORTN4& operator=(const XMSHORTN4&) = default; + + XMSHORTN4(XMSHORTN4&&) = default; + XMSHORTN4& operator=(XMSHORTN4&&) = default; + + explicit constexpr XMSHORTN4(uint64_t Packed) noexcept : v(Packed) {} + constexpr XMSHORTN4(int16_t _x, int16_t _y, int16_t _z, int16_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMSHORTN4(_In_reads_(4) const int16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMSHORTN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMSHORTN4(_In_reads_(4) const float* pArray) noexcept; + + XMSHORTN4& operator= (uint64_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 16 bit signed integer components + struct XMSHORT4 + { + union + { + struct + { + int16_t x; + int16_t y; + int16_t z; + int16_t w; + }; + uint64_t v; + }; + + XMSHORT4() = default; + + XMSHORT4(const XMSHORT4&) = default; + XMSHORT4& operator=(const XMSHORT4&) = default; + + XMSHORT4(XMSHORT4&&) = default; + XMSHORT4& operator=(XMSHORT4&&) = default; + + explicit constexpr XMSHORT4(uint64_t Packed) noexcept : v(Packed) {} + constexpr XMSHORT4(int16_t _x, int16_t _y, int16_t _z, int16_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMSHORT4(_In_reads_(4) const int16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMSHORT4(float _x, float _y, float _z, float _w) noexcept; + explicit XMSHORT4(_In_reads_(4) const float* pArray) noexcept; + + XMSHORT4& operator= (uint64_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 16 bit unsigned normalized integer components + struct XMUSHORTN4 + { + union + { + struct + { + uint16_t x; + uint16_t y; + uint16_t z; + uint16_t w; + }; + uint64_t v; + }; + + XMUSHORTN4() = default; + + XMUSHORTN4(const XMUSHORTN4&) = default; + XMUSHORTN4& operator=(const XMUSHORTN4&) = default; + + XMUSHORTN4(XMUSHORTN4&&) = default; + XMUSHORTN4& operator=(XMUSHORTN4&&) = default; + + explicit constexpr XMUSHORTN4(uint64_t Packed) noexcept : v(Packed) {} + constexpr XMUSHORTN4(uint16_t _x, uint16_t _y, uint16_t _z, uint16_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMUSHORTN4(_In_reads_(4) const uint16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMUSHORTN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUSHORTN4(_In_reads_(4) const float* pArray) noexcept; + + XMUSHORTN4& operator= (uint64_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 16 bit unsigned integer components + struct XMUSHORT4 + { + union + { + struct + { + uint16_t x; + uint16_t y; + uint16_t z; + uint16_t w; + }; + uint64_t v; + }; + + XMUSHORT4() = default; + + XMUSHORT4(const XMUSHORT4&) = default; + XMUSHORT4& operator=(const XMUSHORT4&) = default; + + XMUSHORT4(XMUSHORT4&&) = default; + XMUSHORT4& operator=(XMUSHORT4&&) = default; + + explicit constexpr XMUSHORT4(uint64_t Packed) noexcept : v(Packed) {} + constexpr XMUSHORT4(uint16_t _x, uint16_t _y, uint16_t _z, uint16_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMUSHORT4(_In_reads_(4) const uint16_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMUSHORT4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUSHORT4(_In_reads_(4) const float* pArray) noexcept; + + XMUSHORT4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 4D Vector; 10-10-10-2 bit normalized components packed into a 32 bit integer + // The normalized 4D Vector is packed into 32 bits as follows: a 2 bit unsigned, + // normalized integer for the w component and 10 bit signed, normalized + // integers for the z, y, and x components. The w component is stored in the + // most significant bits and the x component in the least significant bits + // (W2Z10Y10X10): [32] wwzzzzzz zzzzyyyy yyyyyyxx xxxxxxxx [0] + struct XMXDECN4 + { + union + { + struct + { + int32_t x : 10; // -511/511 to 511/511 + int32_t y : 10; // -511/511 to 511/511 + int32_t z : 10; // -511/511 to 511/511 + uint32_t w : 2; // 0/3 to 3/3 + }; + uint32_t v; + }; + + XMXDECN4() = default; + + XMXDECN4(const XMXDECN4&) = default; + XMXDECN4& operator=(const XMXDECN4&) = default; + + XMXDECN4(XMXDECN4&&) = default; + XMXDECN4& operator=(XMXDECN4&&) = default; + + explicit constexpr XMXDECN4(uint32_t Packed) : v(Packed) {} + XMXDECN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMXDECN4(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMXDECN4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 10-10-10-2 bit components packed into a 32 bit integer + // The normalized 4D Vector is packed into 32 bits as follows: a 2 bit unsigned + // integer for the w component and 10 bit signed integers for the + // z, y, and x components. The w component is stored in the + // most significant bits and the x component in the least significant bits + // (W2Z10Y10X10): [32] wwzzzzzz zzzzyyyy yyyyyyxx xxxxxxxx [0] + struct XM_DEPRECATED XMXDEC4 + { + union + { + struct + { + int32_t x : 10; // -511 to 511 + int32_t y : 10; // -511 to 511 + int32_t z : 10; // -511 to 511 + uint32_t w : 2; // 0 to 3 + }; + uint32_t v; + }; + + XMXDEC4() = default; + + XMXDEC4(const XMXDEC4&) = default; + XMXDEC4& operator=(const XMXDEC4&) = default; + + XMXDEC4(XMXDEC4&&) = default; + XMXDEC4& operator=(XMXDEC4&&) = default; + + explicit constexpr XMXDEC4(uint32_t Packed) noexcept : v(Packed) {} + XMXDEC4(float _x, float _y, float _z, float _w) noexcept; + explicit XMXDEC4(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMXDEC4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 10-10-10-2 bit normalized components packed into a 32 bit integer + // The normalized 4D Vector is packed into 32 bits as follows: a 2 bit signed, + // normalized integer for the w component and 10 bit signed, normalized + // integers for the z, y, and x components. The w component is stored in the + // most significant bits and the x component in the least significant bits + // (W2Z10Y10X10): [32] wwzzzzzz zzzzyyyy yyyyyyxx xxxxxxxx [0] + struct XM_DEPRECATED XMDECN4 + { + union + { + struct + { + int32_t x : 10; // -511/511 to 511/511 + int32_t y : 10; // -511/511 to 511/511 + int32_t z : 10; // -511/511 to 511/511 + int32_t w : 2; // -1/1 to 1/1 + }; + uint32_t v; + }; + + XMDECN4() = default; + + XMDECN4(const XMDECN4&) = default; + XMDECN4& operator=(const XMDECN4&) = default; + + XMDECN4(XMDECN4&&) = default; + XMDECN4& operator=(XMDECN4&&) = default; + + explicit constexpr XMDECN4(uint32_t Packed) noexcept : v(Packed) {} + XMDECN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMDECN4(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMDECN4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 10-10-10-2 bit components packed into a 32 bit integer + // The 4D Vector is packed into 32 bits as follows: a 2 bit signed, + // integer for the w component and 10 bit signed integers for the + // z, y, and x components. The w component is stored in the + // most significant bits and the x component in the least significant bits + // (W2Z10Y10X10): [32] wwzzzzzz zzzzyyyy yyyyyyxx xxxxxxxx [0] + struct XM_DEPRECATED XMDEC4 + { + union + { + struct + { + int32_t x : 10; // -511 to 511 + int32_t y : 10; // -511 to 511 + int32_t z : 10; // -511 to 511 + int32_t w : 2; // -1 to 1 + }; + uint32_t v; + }; + + XMDEC4() = default; + + XMDEC4(const XMDEC4&) = default; + XMDEC4& operator=(const XMDEC4&) = default; + + XMDEC4(XMDEC4&&) = default; + XMDEC4& operator=(XMDEC4&&) = default; + + explicit constexpr XMDEC4(uint32_t Packed) noexcept : v(Packed) {} + XMDEC4(float _x, float _y, float _z, float _w) noexcept; + explicit XMDEC4(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMDEC4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 10-10-10-2 bit normalized components packed into a 32 bit integer + // The normalized 4D Vector is packed into 32 bits as follows: a 2 bit unsigned, + // normalized integer for the w component and 10 bit unsigned, normalized + // integers for the z, y, and x components. The w component is stored in the + // most significant bits and the x component in the least significant bits + // (W2Z10Y10X10): [32] wwzzzzzz zzzzyyyy yyyyyyxx xxxxxxxx [0] + struct XMUDECN4 + { + union + { + struct + { + uint32_t x : 10; // 0/1023 to 1023/1023 + uint32_t y : 10; // 0/1023 to 1023/1023 + uint32_t z : 10; // 0/1023 to 1023/1023 + uint32_t w : 2; // 0/3 to 3/3 + }; + uint32_t v; + }; + + XMUDECN4() = default; + + XMUDECN4(const XMUDECN4&) = default; + XMUDECN4& operator=(const XMUDECN4&) = default; + + XMUDECN4(XMUDECN4&&) = default; + XMUDECN4& operator=(XMUDECN4&&) = default; + + explicit constexpr XMUDECN4(uint32_t Packed) noexcept : v(Packed) {} + XMUDECN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUDECN4(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMUDECN4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 10-10-10-2 bit components packed into a 32 bit integer + // The 4D Vector is packed into 32 bits as follows: a 2 bit unsigned, + // integer for the w component and 10 bit unsigned integers + // for the z, y, and x components. The w component is stored in the + // most significant bits and the x component in the least significant bits + // (W2Z10Y10X10): [32] wwzzzzzz zzzzyyyy yyyyyyxx xxxxxxxx [0] + struct XMUDEC4 + { + union + { + struct + { + uint32_t x : 10; // 0 to 1023 + uint32_t y : 10; // 0 to 1023 + uint32_t z : 10; // 0 to 1023 + uint32_t w : 2; // 0 to 3 + }; + uint32_t v; + }; + + XMUDEC4() = default; + + XMUDEC4(const XMUDEC4&) = default; + XMUDEC4& operator=(const XMUDEC4&) = default; + + XMUDEC4(XMUDEC4&&) = default; + XMUDEC4& operator=(XMUDEC4&&) = default; + + explicit constexpr XMUDEC4(uint32_t Packed) noexcept : v(Packed) {} + XMUDEC4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUDEC4(_In_reads_(4) const float* pArray) noexcept; + + operator uint32_t () const noexcept { return v; } + + XMUDEC4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 4D Vector; 8 bit signed normalized integer components + struct XMBYTEN4 + { + union + { + struct + { + int8_t x; + int8_t y; + int8_t z; + int8_t w; + }; + uint32_t v; + }; + + XMBYTEN4() = default; + + XMBYTEN4(const XMBYTEN4&) = default; + XMBYTEN4& operator=(const XMBYTEN4&) = default; + + XMBYTEN4(XMBYTEN4&&) = default; + XMBYTEN4& operator=(XMBYTEN4&&) = default; + + constexpr XMBYTEN4(int8_t _x, int8_t _y, int8_t _z, int8_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit constexpr XMBYTEN4(uint32_t Packed) noexcept : v(Packed) {} + explicit XMBYTEN4(_In_reads_(4) const int8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMBYTEN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMBYTEN4(_In_reads_(4) const float* pArray) noexcept; + + XMBYTEN4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 8 bit signed integer components + struct XMBYTE4 + { + union + { + struct + { + int8_t x; + int8_t y; + int8_t z; + int8_t w; + }; + uint32_t v; + }; + + XMBYTE4() = default; + + XMBYTE4(const XMBYTE4&) = default; + XMBYTE4& operator=(const XMBYTE4&) = default; + + XMBYTE4(XMBYTE4&&) = default; + XMBYTE4& operator=(XMBYTE4&&) = default; + + constexpr XMBYTE4(int8_t _x, int8_t _y, int8_t _z, int8_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit constexpr XMBYTE4(uint32_t Packed) noexcept : v(Packed) {} + explicit XMBYTE4(_In_reads_(4) const int8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMBYTE4(float _x, float _y, float _z, float _w) noexcept; + explicit XMBYTE4(_In_reads_(4) const float* pArray) noexcept; + + XMBYTE4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 8 bit unsigned normalized integer components + struct XMUBYTEN4 + { + union + { + struct + { + uint8_t x; + uint8_t y; + uint8_t z; + uint8_t w; + }; + uint32_t v; + }; + + XMUBYTEN4() = default; + + XMUBYTEN4(const XMUBYTEN4&) = default; + XMUBYTEN4& operator=(const XMUBYTEN4&) = default; + + XMUBYTEN4(XMUBYTEN4&&) = default; + XMUBYTEN4& operator=(XMUBYTEN4&&) = default; + + constexpr XMUBYTEN4(uint8_t _x, uint8_t _y, uint8_t _z, uint8_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit constexpr XMUBYTEN4(uint32_t Packed) noexcept : v(Packed) {} + explicit XMUBYTEN4(_In_reads_(4) const uint8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMUBYTEN4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUBYTEN4(_In_reads_(4) const float* pArray) noexcept; + + XMUBYTEN4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + // 4D Vector; 8 bit unsigned integer components + struct XMUBYTE4 + { + union + { + struct + { + uint8_t x; + uint8_t y; + uint8_t z; + uint8_t w; + }; + uint32_t v; + }; + + XMUBYTE4() = default; + + XMUBYTE4(const XMUBYTE4&) = default; + XMUBYTE4& operator=(const XMUBYTE4&) = default; + + XMUBYTE4(XMUBYTE4&&) = default; + XMUBYTE4& operator=(XMUBYTE4&&) = default; + + constexpr XMUBYTE4(uint8_t _x, uint8_t _y, uint8_t _z, uint8_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit constexpr XMUBYTE4(uint32_t Packed) noexcept : v(Packed) {} + explicit XMUBYTE4(_In_reads_(4) const uint8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMUBYTE4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUBYTE4(_In_reads_(4) const float* pArray) noexcept; + + XMUBYTE4& operator= (uint32_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 4D vector; 4 bit unsigned integer components + struct XMUNIBBLE4 + { + union + { + struct + { + uint16_t x : 4; // 0 to 15 + uint16_t y : 4; // 0 to 15 + uint16_t z : 4; // 0 to 15 + uint16_t w : 4; // 0 to 15 + }; + uint16_t v; + }; + + XMUNIBBLE4() = default; + + XMUNIBBLE4(const XMUNIBBLE4&) = default; + XMUNIBBLE4& operator=(const XMUNIBBLE4&) = default; + + XMUNIBBLE4(XMUNIBBLE4&&) = default; + XMUNIBBLE4& operator=(XMUNIBBLE4&&) = default; + + explicit constexpr XMUNIBBLE4(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMUNIBBLE4(uint8_t _x, uint8_t _y, uint8_t _z, uint8_t _w) noexcept : x(_x), y(_y), z(_z), w(_w) {} + explicit XMUNIBBLE4(_In_reads_(4) const uint8_t* pArray) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(pArray[3]) {} + XMUNIBBLE4(float _x, float _y, float _z, float _w) noexcept; + explicit XMUNIBBLE4(_In_reads_(4) const float* pArray) noexcept; + + operator uint16_t () const noexcept { return v; } + + XMUNIBBLE4& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + //------------------------------------------------------------------------------ + // 4D vector: 5/5/5/1 unsigned integer components + struct XMU555 + { + union + { + struct + { + uint16_t x : 5; // 0 to 31 + uint16_t y : 5; // 0 to 31 + uint16_t z : 5; // 0 to 31 + uint16_t w : 1; // 0 or 1 + }; + uint16_t v; + }; + + XMU555() = default; + + XMU555(const XMU555&) = default; + XMU555& operator=(const XMU555&) = default; + + XMU555(XMU555&&) = default; + XMU555& operator=(XMU555&&) = default; + + explicit constexpr XMU555(uint16_t Packed) noexcept : v(Packed) {} + constexpr XMU555(uint8_t _x, uint8_t _y, uint8_t _z, bool _w) noexcept : x(_x), y(_y), z(_z), w(_w ? 0x1 : 0) {} + XMU555(_In_reads_(3) const uint8_t* pArray, _In_ bool _w) noexcept : x(pArray[0]), y(pArray[1]), z(pArray[2]), w(_w ? 0x1 : 0) {} + XMU555(float _x, float _y, float _z, bool _w) noexcept; + XMU555(_In_reads_(3) const float* pArray, _In_ bool _w) noexcept; + + operator uint16_t () const noexcept { return v; } + + XMU555& operator= (uint16_t Packed) noexcept { v = Packed; return *this; } + }; + + #ifdef __clang__ + #pragma clang diagnostic pop + #endif + #ifdef _MSC_VER + #pragma warning(pop) + #endif + + /**************************************************************************** + * + * Data conversion operations + * + ****************************************************************************/ + + float XMConvertHalfToFloat(HALF Value) noexcept; + float* XMConvertHalfToFloatStream(_Out_writes_bytes_(sizeof(float) + OutputStride * (HalfCount - 1)) float* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(HALF) + InputStride * (HalfCount - 1)) const HALF* pInputStream, + _In_ size_t InputStride, _In_ size_t HalfCount) noexcept; + HALF XMConvertFloatToHalf(float Value) noexcept; + HALF* XMConvertFloatToHalfStream(_Out_writes_bytes_(sizeof(HALF) + OutputStride * (FloatCount - 1)) HALF* pOutputStream, + _In_ size_t OutputStride, + _In_reads_bytes_(sizeof(float) + InputStride * (FloatCount - 1)) const float* pInputStream, + _In_ size_t InputStride, _In_ size_t FloatCount) noexcept; + + /**************************************************************************** + * + * Load operations + * + ****************************************************************************/ + + XMVECTOR XM_CALLCONV XMLoadColor(_In_ const XMCOLOR* pSource) noexcept; + + XMVECTOR XM_CALLCONV XMLoadHalf2(_In_ const XMHALF2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadShortN2(_In_ const XMSHORTN2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadShort2(_In_ const XMSHORT2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUShortN2(_In_ const XMUSHORTN2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUShort2(_In_ const XMUSHORT2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadByteN2(_In_ const XMBYTEN2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadByte2(_In_ const XMBYTE2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUByteN2(_In_ const XMUBYTEN2* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUByte2(_In_ const XMUBYTE2* pSource) noexcept; + + XMVECTOR XM_CALLCONV XMLoadU565(_In_ const XMU565* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat3PK(_In_ const XMFLOAT3PK* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadFloat3SE(_In_ const XMFLOAT3SE* pSource) noexcept; + + XMVECTOR XM_CALLCONV XMLoadHalf4(_In_ const XMHALF4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadShortN4(_In_ const XMSHORTN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadShort4(_In_ const XMSHORT4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUShortN4(_In_ const XMUSHORTN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUShort4(_In_ const XMUSHORT4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadXDecN4(_In_ const XMXDECN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUDecN4(_In_ const XMUDECN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUDecN4_XR(_In_ const XMUDECN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUDec4(_In_ const XMUDEC4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadByteN4(_In_ const XMBYTEN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadByte4(_In_ const XMBYTE4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUByteN4(_In_ const XMUBYTEN4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUByte4(_In_ const XMUBYTE4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadUNibble4(_In_ const XMUNIBBLE4* pSource) noexcept; + XMVECTOR XM_CALLCONV XMLoadU555(_In_ const XMU555* pSource) noexcept; + + #ifdef _MSC_VER + #pragma warning(push) + #pragma warning(disable : 4996) + #endif + + #ifdef __clang__ + #pragma clang diagnostic push + #pragma clang diagnostic ignored "-Wdeprecated-declarations" + #endif + + #ifdef __GNUC__ + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wdeprecated-declarations" + #endif + + XM_DEPRECATED + XMVECTOR XM_CALLCONV XMLoadDecN4(_In_ const XMDECN4* pSource) noexcept; + + XM_DEPRECATED + XMVECTOR XM_CALLCONV XMLoadDec4(_In_ const XMDEC4* pSource) noexcept; + + XM_DEPRECATED + XMVECTOR XM_CALLCONV XMLoadXDec4(_In_ const XMXDEC4* pSource) noexcept; + + #ifdef __GNUC__ + #pragma GCC diagnostic pop + #endif + #ifdef __clang__ + #pragma clang diagnostic pop + #endif + #ifdef _MSC_VER + #pragma warning(pop) + #endif + + /**************************************************************************** + * + * Store operations + * + ****************************************************************************/ + + void XM_CALLCONV XMStoreColor(_Out_ XMCOLOR* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreHalf2(_Out_ XMHALF2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreShortN2(_Out_ XMSHORTN2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreShort2(_Out_ XMSHORT2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUShortN2(_Out_ XMUSHORTN2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUShort2(_Out_ XMUSHORT2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreByteN2(_Out_ XMBYTEN2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreByte2(_Out_ XMBYTE2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUByteN2(_Out_ XMUBYTEN2* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUByte2(_Out_ XMUBYTE2* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreU565(_Out_ XMU565* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat3PK(_Out_ XMFLOAT3PK* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreFloat3SE(_Out_ XMFLOAT3SE* pDestination, _In_ FXMVECTOR V) noexcept; + + void XM_CALLCONV XMStoreHalf4(_Out_ XMHALF4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreShortN4(_Out_ XMSHORTN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreShort4(_Out_ XMSHORT4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUShortN4(_Out_ XMUSHORTN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUShort4(_Out_ XMUSHORT4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreXDecN4(_Out_ XMXDECN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUDecN4(_Out_ XMUDECN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUDecN4_XR(_Out_ XMUDECN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUDec4(_Out_ XMUDEC4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreByteN4(_Out_ XMBYTEN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreByte4(_Out_ XMBYTE4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUByteN4(_Out_ XMUBYTEN4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUByte4(_Out_ XMUBYTE4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreUNibble4(_Out_ XMUNIBBLE4* pDestination, _In_ FXMVECTOR V) noexcept; + void XM_CALLCONV XMStoreU555(_Out_ XMU555* pDestination, _In_ FXMVECTOR V) noexcept; + + #ifdef _MSC_VER + #pragma warning(push) + #pragma warning(disable : 4996) + #endif + + #ifdef __clang__ + #pragma clang diagnostic push + #pragma clang diagnostic ignored "-Wdeprecated-declarations" + #endif + + #ifdef __GNUC__ + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wdeprecated-declarations" + #endif + + XM_DEPRECATED + void XM_CALLCONV XMStoreDecN4(_Out_ XMDECN4* pDestination, _In_ FXMVECTOR V) noexcept; + + XM_DEPRECATED + void XM_CALLCONV XMStoreDec4(_Out_ XMDEC4* pDestination, _In_ FXMVECTOR V) noexcept; + + XM_DEPRECATED + void XM_CALLCONV XMStoreXDec4(_Out_ XMXDEC4* pDestination, _In_ FXMVECTOR V) noexcept; + + #ifdef __GNUC__ + #pragma GCC diagnostic pop + #endif + #ifdef __clang__ + #pragma clang diagnostic pop + #endif + #ifdef _MSC_VER + #pragma warning(pop) + #endif + + /**************************************************************************** + * + * Implementation + * + ****************************************************************************/ + #ifdef _MSC_VER + #pragma warning(push) + #pragma warning(disable:4068 4214 4204 4365 4616 6001 6101) + #endif + + #ifdef _PREFAST_ + #pragma prefast(push) + #pragma prefast(disable : 25000, "FXMVECTOR is 16 bytes") + #pragma prefast(disable : 26495, "Union initialization confuses /analyze") + #endif + + #ifdef __clang__ + #pragma clang diagnostic push + #pragma clang diagnostic ignored "-Wunknown-warning-option" + #pragma clang diagnostic ignored "-Wunsafe-buffer-usage" + #endif + + #include "DirectXPackedVector.inl" + + #ifdef __clang__ + #pragma clang diagnostic pop + #endif + #ifdef _PREFAST_ + #pragma prefast(pop) + #endif + #ifdef _MSC_VER + #pragma warning(pop) + #endif + } // namespace PackedVector + +} // namespace DirectX diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXPackedVector.inl b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXPackedVector.inl new file mode 100644 index 00000000..be99cecc --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/4JLibs/impls/Windows_Libs/Render/vendor/DirectXMath/DirectXPackedVector.inl @@ -0,0 +1,4628 @@ +//------------------------------------------------------------------------------------- +// DirectXPackedVector.inl -- SIMD C++ Math library +// +// Copyright (c) Microsoft Corporation. +// Licensed under the MIT License. +// +// https://go.microsoft.com/fwlink/?LinkID=615560 +//------------------------------------------------------------------------------------- + +#pragma once + +/**************************************************************************** + * + * Data conversion + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline float XMConvertHalfToFloat(HALF Value) noexcept +{ +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + __m128i V1 = _mm_cvtsi32_si128(static_cast(Value)); + __m128 V2 = _mm_cvtph_ps(V1); + return _mm_cvtss_f32(V2); +#elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__) && !defined(_XM_NO_INTRINSICS_) && (!defined(__GNUC__) || (__ARM_FP & 2)) + uint16x4_t vHalf = vdup_n_u16(Value); + float32x4_t vFloat = vcvt_f32_f16(vreinterpret_f16_u16(vHalf)); + return vgetq_lane_f32(vFloat, 0); +#else + auto Mantissa = static_cast(Value & 0x03FF); + + uint32_t Exponent = (Value & 0x7C00); + if (Exponent == 0x7C00) // INF/NAN + { + Exponent = 0x8f; + } + else if (Exponent != 0) // The value is normalized + { + Exponent = static_cast((static_cast(Value) >> 10) & 0x1F); + } + else if (Mantissa != 0) // The value is denormalized + { + // Normalize the value in the resulting float + Exponent = 1; + + do + { + Exponent--; + Mantissa <<= 1; + } + while ((Mantissa & 0x0400) == 0); + + Mantissa &= 0x03FF; + } + else // The value is zero + { + Exponent = static_cast(-112); + } + + uint32_t Result = + ((static_cast(Value) & 0x8000) << 16) // Sign + | ((Exponent + 112) << 23) // Exponent + | (Mantissa << 13); // Mantissa + + return reinterpret_cast(&Result)[0]; +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable : 26015 26019, "PREfast noise: Esp:1307" ) +#endif + +_Use_decl_annotations_ +inline float* XMConvertHalfToFloatStream +( + float* pOutputStream, + size_t OutputStride, + const HALF* pInputStream, + size_t InputStride, + size_t HalfCount +) noexcept +{ + assert(pOutputStream); + assert(pInputStream); + + assert(InputStride >= sizeof(HALF)); + _Analysis_assume_(InputStride >= sizeof(HALF)); + + assert(OutputStride >= sizeof(float)); + _Analysis_assume_(OutputStride >= sizeof(float)); + +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + auto pHalf = reinterpret_cast(pInputStream); + auto pFloat = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = HalfCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(HALF)) + { + if (OutputStride == sizeof(float)) + { + if ((reinterpret_cast(pFloat) & 0xF) == 0) + { + // Packed input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + __m128i HV = _mm_loadl_epi64(reinterpret_cast(pHalf)); + pHalf += InputStride * 4; + + __m128 FV = _mm_cvtph_ps(HV); + + XM_STREAM_PS(reinterpret_cast(pFloat), FV); + pFloat += OutputStride * 4; + i += 4; + } + } + else + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + __m128i HV = _mm_loadl_epi64(reinterpret_cast(pHalf)); + pHalf += InputStride * 4; + + __m128 FV = _mm_cvtph_ps(HV); + + _mm_storeu_ps(reinterpret_cast(pFloat), FV); + pFloat += OutputStride * 4; + i += 4; + } + } + } + else + { + // Packed input, scattered output + for (size_t j = 0; j < four; ++j) + { + __m128i HV = _mm_loadl_epi64(reinterpret_cast(pHalf)); + pHalf += InputStride * 4; + + __m128 FV = _mm_cvtph_ps(HV); + + _mm_store_ss(reinterpret_cast(pFloat), FV); + pFloat += OutputStride; + *reinterpret_cast(pFloat) = _mm_extract_ps(FV, 1); + pFloat += OutputStride; + *reinterpret_cast(pFloat) = _mm_extract_ps(FV, 2); + pFloat += OutputStride; + *reinterpret_cast(pFloat) = _mm_extract_ps(FV, 3); + pFloat += OutputStride; + i += 4; + } + } + } + else if (OutputStride == sizeof(float)) + { + if ((reinterpret_cast(pFloat) & 0xF) == 0) + { + // Scattered input, aligned & packed output + for (size_t j = 0; j < four; ++j) + { + uint16_t H1 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H2 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H3 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H4 = *reinterpret_cast(pHalf); + pHalf += InputStride; + + __m128i HV = _mm_setzero_si128(); + HV = _mm_insert_epi16(HV, H1, 0); + HV = _mm_insert_epi16(HV, H2, 1); + HV = _mm_insert_epi16(HV, H3, 2); + HV = _mm_insert_epi16(HV, H4, 3); + __m128 FV = _mm_cvtph_ps(HV); + + XM_STREAM_PS(reinterpret_cast(pFloat), FV); + pFloat += OutputStride * 4; + i += 4; + } + } + else + { + // Scattered input, packed output + for (size_t j = 0; j < four; ++j) + { + uint16_t H1 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H2 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H3 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H4 = *reinterpret_cast(pHalf); + pHalf += InputStride; + + __m128i HV = _mm_setzero_si128(); + HV = _mm_insert_epi16(HV, H1, 0); + HV = _mm_insert_epi16(HV, H2, 1); + HV = _mm_insert_epi16(HV, H3, 2); + HV = _mm_insert_epi16(HV, H4, 3); + __m128 FV = _mm_cvtph_ps(HV); + + _mm_storeu_ps(reinterpret_cast(pFloat), FV); + pFloat += OutputStride * 4; + i += 4; + } + } + } + else + { + // Scattered input, scattered output + for (size_t j = 0; j < four; ++j) + { + uint16_t H1 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H2 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H3 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H4 = *reinterpret_cast(pHalf); + pHalf += InputStride; + + __m128i HV = _mm_setzero_si128(); + HV = _mm_insert_epi16(HV, H1, 0); + HV = _mm_insert_epi16(HV, H2, 1); + HV = _mm_insert_epi16(HV, H3, 2); + HV = _mm_insert_epi16(HV, H4, 3); + __m128 FV = _mm_cvtph_ps(HV); + + _mm_store_ss(reinterpret_cast(pFloat), FV); + pFloat += OutputStride; + *reinterpret_cast(pFloat) = _mm_extract_ps(FV, 1); + pFloat += OutputStride; + *reinterpret_cast(pFloat) = _mm_extract_ps(FV, 2); + pFloat += OutputStride; + *reinterpret_cast(pFloat) = _mm_extract_ps(FV, 3); + pFloat += OutputStride; + i += 4; + } + } + } + + for (; i < HalfCount; ++i) + { + *reinterpret_cast(pFloat) = XMConvertHalfToFloat(reinterpret_cast(pHalf)[0]); + pHalf += InputStride; + pFloat += OutputStride; + } + + XM_SFENCE(); + + return pOutputStream; +#elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) ||__aarch64__) && !defined(_XM_NO_INTRINSICS_) && (!defined(__GNUC__) || (__ARM_FP & 2)) + auto pHalf = reinterpret_cast(pInputStream); + auto pFloat = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = HalfCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(HALF)) + { + if (OutputStride == sizeof(float)) + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + uint16x4_t vHalf = vld1_u16(reinterpret_cast(pHalf)); + pHalf += InputStride * 4; + + float32x4_t vFloat = vcvt_f32_f16(vreinterpret_f16_u16(vHalf)); + + vst1q_f32(reinterpret_cast(pFloat), vFloat); + pFloat += OutputStride * 4; + i += 4; + } + } + else + { + // Packed input, scattered output + for (size_t j = 0; j < four; ++j) + { + uint16x4_t vHalf = vld1_u16(reinterpret_cast(pHalf)); + pHalf += InputStride * 4; + + float32x4_t vFloat = vcvt_f32_f16(vreinterpret_f16_u16(vHalf)); + + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 0); + pFloat += OutputStride; + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 1); + pFloat += OutputStride; + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 2); + pFloat += OutputStride; + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 3); + pFloat += OutputStride; + i += 4; + } + } + } + else if (OutputStride == sizeof(float)) + { + // Scattered input, packed output + for (size_t j = 0; j < four; ++j) + { + uint16_t H1 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H2 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H3 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H4 = *reinterpret_cast(pHalf); + pHalf += InputStride; + + uint64_t iHalf = uint64_t(H1) | (uint64_t(H2) << 16) | (uint64_t(H3) << 32) | (uint64_t(H4) << 48); + uint16x4_t vHalf = vcreate_u16(iHalf); + + float32x4_t vFloat = vcvt_f32_f16(vreinterpret_f16_u16(vHalf)); + + vst1q_f32(reinterpret_cast(pFloat), vFloat); + pFloat += OutputStride * 4; + i += 4; + } + } + else + { + // Scattered input, scattered output + for (size_t j = 0; j < four; ++j) + { + uint16_t H1 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H2 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H3 = *reinterpret_cast(pHalf); + pHalf += InputStride; + uint16_t H4 = *reinterpret_cast(pHalf); + pHalf += InputStride; + + uint64_t iHalf = uint64_t(H1) | (uint64_t(H2) << 16) | (uint64_t(H3) << 32) | (uint64_t(H4) << 48); + uint16x4_t vHalf = vcreate_u16(iHalf); + + float32x4_t vFloat = vcvt_f32_f16(vreinterpret_f16_u16(vHalf)); + + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 0); + pFloat += OutputStride; + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 1); + pFloat += OutputStride; + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 2); + pFloat += OutputStride; + vst1q_lane_f32(reinterpret_cast(pFloat), vFloat, 3); + pFloat += OutputStride; + i += 4; + } + } + } + + for (; i < HalfCount; ++i) + { + *reinterpret_cast(pFloat) = XMConvertHalfToFloat(reinterpret_cast(pHalf)[0]); + pHalf += InputStride; + pFloat += OutputStride; + } + + return pOutputStream; +#else + auto pHalf = reinterpret_cast(pInputStream); + auto pFloat = reinterpret_cast(pOutputStream); + + for (size_t i = 0; i < HalfCount; i++) + { + *reinterpret_cast(pFloat) = XMConvertHalfToFloat(reinterpret_cast(pHalf)[0]); + pHalf += InputStride; + pFloat += OutputStride; + } + + return pOutputStream; +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ + +inline HALF XMConvertFloatToHalf(float Value) noexcept +{ +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + __m128 V1 = _mm_set_ss(Value); + __m128i V2 = _mm_cvtps_ph(V1, _MM_FROUND_TO_NEAREST_INT); + return static_cast(_mm_extract_epi16(V2, 0)); +#elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__) && !defined(_XM_NO_INTRINSICS_) && (!defined(__GNUC__) || (__ARM_FP & 2)) + float32x4_t vFloat = vdupq_n_f32(Value); + float16x4_t vHalf = vcvt_f16_f32(vFloat); + return vget_lane_u16(vreinterpret_u16_f16(vHalf), 0); +#else + uint32_t Result; + + auto IValue = reinterpret_cast(&Value)[0]; + uint32_t Sign = (IValue & 0x80000000U) >> 16U; + IValue = IValue & 0x7FFFFFFFU; // Hack off the sign + if (IValue >= 0x47800000 /*e+16*/) + { + // The number is too large to be represented as a half. Return infinity or NaN + Result = 0x7C00U | ((IValue > 0x7F800000) ? (0x200 | ((IValue >> 13U) & 0x3FFU)) : 0U); + } + else if (IValue <= 0x33000000U /*e-25*/) + { + Result = 0; + } + else if (IValue < 0x38800000U /*e-14*/) + { + // The number is too small to be represented as a normalized half. + // Convert it to a denormalized value. + uint32_t Shift = 125U - (IValue >> 23U); + IValue = 0x800000U | (IValue & 0x7FFFFFU); + Result = IValue >> (Shift + 1); + uint32_t s = (IValue & ((1U << Shift) - 1)) != 0; + Result += (Result | s) & ((IValue >> Shift) & 1U); + } + else + { + // Rebias the exponent to represent the value as a normalized half. + IValue += 0xC8000000U; + Result = ((IValue + 0x0FFFU + ((IValue >> 13U) & 1U)) >> 13U) & 0x7FFFU; + } + return static_cast(Result | Sign); +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline HALF* XMConvertFloatToHalfStream +( + HALF* pOutputStream, + size_t OutputStride, + const float* pInputStream, + size_t InputStride, + size_t FloatCount +) noexcept +{ + assert(pOutputStream); + assert(pInputStream); + + assert(InputStride >= sizeof(float)); + _Analysis_assume_(InputStride >= sizeof(float)); + + assert(OutputStride >= sizeof(HALF)); + _Analysis_assume_(OutputStride >= sizeof(HALF)); + +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + auto pFloat = reinterpret_cast(pInputStream); + auto pHalf = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = FloatCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(float)) + { + if (OutputStride == sizeof(HALF)) + { + if ((reinterpret_cast(pFloat) & 0xF) == 0) + { + // Aligned and packed input, packed output + for (size_t j = 0; j < four; ++j) + { + __m128 FV = _mm_load_ps(reinterpret_cast(pFloat)); + pFloat += InputStride * 4; + + __m128i HV = _mm_cvtps_ph(FV, _MM_FROUND_TO_NEAREST_INT); + + _mm_storel_epi64(reinterpret_cast<__m128i*>(pHalf), HV); + pHalf += OutputStride * 4; + i += 4; + } + } + else + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + __m128 FV = _mm_loadu_ps(reinterpret_cast(pFloat)); + pFloat += InputStride * 4; + + __m128i HV = _mm_cvtps_ph(FV, _MM_FROUND_TO_NEAREST_INT); + + _mm_storel_epi64(reinterpret_cast<__m128i*>(pHalf), HV); + pHalf += OutputStride * 4; + i += 4; + } + } + } + else + { + if ((reinterpret_cast(pFloat) & 0xF) == 0) + { + // Aligned & packed input, scattered output + for (size_t j = 0; j < four; ++j) + { + __m128 FV = _mm_load_ps(reinterpret_cast(pFloat)); + pFloat += InputStride * 4; + + __m128i HV = _mm_cvtps_ph(FV, _MM_FROUND_TO_NEAREST_INT); + + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 0)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 1)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 2)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 3)); + pHalf += OutputStride; + i += 4; + } + } + else + { + // Packed input, scattered output + for (size_t j = 0; j < four; ++j) + { + __m128 FV = _mm_loadu_ps(reinterpret_cast(pFloat)); + pFloat += InputStride * 4; + + __m128i HV = _mm_cvtps_ph(FV, _MM_FROUND_TO_NEAREST_INT); + + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 0)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 1)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 2)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 3)); + pHalf += OutputStride; + i += 4; + } + } + } + } + else if (OutputStride == sizeof(HALF)) + { + // Scattered input, packed output + for (size_t j = 0; j < four; ++j) + { + __m128 FV1 = _mm_load_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV2 = _mm_broadcast_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV3 = _mm_broadcast_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV4 = _mm_broadcast_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV = _mm_blend_ps(FV1, FV2, 0x2); + __m128 FT = _mm_blend_ps(FV3, FV4, 0x8); + FV = _mm_blend_ps(FV, FT, 0xC); + + __m128i HV = _mm_cvtps_ph(FV, _MM_FROUND_TO_NEAREST_INT); + + _mm_storel_epi64(reinterpret_cast<__m128i*>(pHalf), HV); + pHalf += OutputStride * 4; + i += 4; + } + } + else + { + // Scattered input, scattered output + for (size_t j = 0; j < four; ++j) + { + __m128 FV1 = _mm_load_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV2 = _mm_broadcast_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV3 = _mm_broadcast_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV4 = _mm_broadcast_ss(reinterpret_cast(pFloat)); + pFloat += InputStride; + + __m128 FV = _mm_blend_ps(FV1, FV2, 0x2); + __m128 FT = _mm_blend_ps(FV3, FV4, 0x8); + FV = _mm_blend_ps(FV, FT, 0xC); + + __m128i HV = _mm_cvtps_ph(FV, _MM_FROUND_TO_NEAREST_INT); + + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 0)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 1)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 2)); + pHalf += OutputStride; + *reinterpret_cast(pHalf) = static_cast(_mm_extract_epi16(HV, 3)); + pHalf += OutputStride; + i += 4; + } + } + } + + for (; i < FloatCount; ++i) + { + *reinterpret_cast(pHalf) = XMConvertFloatToHalf(reinterpret_cast(pFloat)[0]); + pFloat += InputStride; + pHalf += OutputStride; + } + + return pOutputStream; +#elif defined(_XM_ARM_NEON_INTRINSICS_) && (defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__) && !defined(_XM_NO_INTRINSICS_) && (!defined(__GNUC__) || (__ARM_FP & 2)) + auto pFloat = reinterpret_cast(pInputStream); + auto pHalf = reinterpret_cast(pOutputStream); + + size_t i = 0; + size_t four = FloatCount >> 2; + if (four > 0) + { + if (InputStride == sizeof(float)) + { + if (OutputStride == sizeof(HALF)) + { + // Packed input, packed output + for (size_t j = 0; j < four; ++j) + { + float32x4_t vFloat = vld1q_f32(reinterpret_cast(pFloat)); + pFloat += InputStride * 4; + + uint16x4_t vHalf = vreinterpret_u16_f16(vcvt_f16_f32(vFloat)); + + vst1_u16(reinterpret_cast(pHalf), vHalf); + pHalf += OutputStride * 4; + i += 4; + } + } + else + { + // Packed input, scattered output + for (size_t j = 0; j < four; ++j) + { + float32x4_t vFloat = vld1q_f32(reinterpret_cast(pFloat)); + pFloat += InputStride * 4; + + uint16x4_t vHalf = vreinterpret_u16_f16(vcvt_f16_f32(vFloat)); + + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 0); + pHalf += OutputStride; + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 1); + pHalf += OutputStride; + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 2); + pHalf += OutputStride; + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 3); + pHalf += OutputStride; + i += 4; + } + } + } + else if (OutputStride == sizeof(HALF)) + { + // Scattered input, packed output + for (size_t j = 0; j < four; ++j) + { + float32x4_t vFloat = vdupq_n_f32(0); + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 0); + pFloat += InputStride; + + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 1); + pFloat += InputStride; + + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 2); + pFloat += InputStride; + + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 3); + pFloat += InputStride; + + uint16x4_t vHalf = vreinterpret_u16_f16(vcvt_f16_f32(vFloat)); + + vst1_u16(reinterpret_cast(pHalf), vHalf); + pHalf += OutputStride * 4; + i += 4; + } + } + else + { + // Scattered input, scattered output + for (size_t j = 0; j < four; ++j) + { + float32x4_t vFloat = vdupq_n_f32(0); + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 0); + pFloat += InputStride; + + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 1); + pFloat += InputStride; + + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 2); + pFloat += InputStride; + + vFloat = vld1q_lane_f32(reinterpret_cast(pFloat), vFloat, 3); + pFloat += InputStride; + + uint16x4_t vHalf = vreinterpret_u16_f16(vcvt_f16_f32(vFloat)); + + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 0); + pHalf += OutputStride; + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 1); + pHalf += OutputStride; + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 2); + pHalf += OutputStride; + vst1_lane_u16(reinterpret_cast(pHalf), vHalf, 3); + pHalf += OutputStride; + i += 4; + } + } + } + + for (; i < FloatCount; ++i) + { + *reinterpret_cast(pHalf) = XMConvertFloatToHalf(reinterpret_cast(pFloat)[0]); + pFloat += InputStride; + pHalf += OutputStride; + } + + return pOutputStream; +#else + auto pFloat = reinterpret_cast(pInputStream); + auto pHalf = reinterpret_cast(pOutputStream); + + for (size_t i = 0; i < FloatCount; i++) + { + *reinterpret_cast(pHalf) = XMConvertFloatToHalf(reinterpret_cast(pFloat)[0]); + pFloat += InputStride; + pHalf += OutputStride; + } + return pOutputStream; +#endif // !_XM_F16C_INTRINSICS_ +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +/**************************************************************************** + * + * Vector and matrix load operations + * + ****************************************************************************/ + +#ifdef _PREFAST_ +#pragma prefast(push) +#pragma prefast(disable:28931, "PREfast noise: Esp:1266") +#endif + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadColor(const XMCOLOR* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + // int32_t -> Float conversions are done in one instruction. + // uint32_t -> Float calls a runtime function. Keep in int32_t + auto iColor = static_cast(pSource->c); + XMVECTORF32 vColor = { { { + static_cast((iColor >> 16) & 0xFF)* (1.0f / 255.0f), + static_cast((iColor >> 8) & 0xFF)* (1.0f / 255.0f), + static_cast(iColor & 0xFF)* (1.0f / 255.0f), + static_cast((iColor >> 24) & 0xFF)* (1.0f / 255.0f) + } } }; + return vColor.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32_t bgra = pSource->c; + uint32_t rgba = (bgra & 0xFF00FF00) | ((bgra >> 16) & 0xFF) | ((bgra << 16) & 0xFF0000); + uint32x2_t vInt8 = vdup_n_u32(rgba); + uint16x8_t vInt16 = vmovl_u8(vreinterpret_u8_u32(vInt8)); + uint32x4_t vInt = vmovl_u16(vget_low_u16(vInt16)); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_n_f32(R, 1.0f / 255.0f); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the color in all four entries + __m128i vInt = _mm_set1_epi32(static_cast(pSource->c)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vInt = _mm_and_si128(vInt, g_XMMaskA8R8G8B8); + // a is unsigned! Flip the bit to convert the order to signed + vInt = _mm_xor_si128(vInt, g_XMFlipA8R8G8B8); + // Convert to floating point numbers + XMVECTOR vTemp = _mm_cvtepi32_ps(vInt); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMFixAA8R8G8B8); + // Convert 0-255 to 0.0f-1.0f + return _mm_mul_ps(vTemp, g_XMNormalizeA8R8G8B8); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadHalf2(const XMHALF2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + __m128 V = _mm_load_ss(reinterpret_cast(pSource)); + return _mm_cvtph_ps(_mm_castps_si128(V)); +#else + XMVECTORF32 vResult = { { { + XMConvertHalfToFloat(pSource->x), + XMConvertHalfToFloat(pSource->y), + 0.0f, + 0.0f + } } }; + return vResult.v; +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadShortN2(const XMSHORTN2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + (pSource->x == -32768) ? -1.f : (static_cast(pSource->x)* (1.0f / 32767.0f)), + (pSource->y == -32768) ? -1.f : (static_cast(pSource->y)* (1.0f / 32767.0f)), + 0.0f, + 0.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt16 = vld1_dup_u32(reinterpret_cast(pSource)); + int32x4_t vInt = vmovl_s16(vreinterpret_s16_u32(vInt16)); + vInt = vandq_s32(vInt, g_XMMaskXY); + float32x4_t R = vcvtq_f32_s32(vInt); + R = vmulq_n_f32(R, 1.0f / 32767.0f); + return vmaxq_f32(R, vdupq_n_f32(-1.f)); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the two shorts in all four entries (WORD alignment okay, + // DWORD alignment preferred) + __m128 vTemp = _mm_load_ps1(reinterpret_cast(&pSource->x)); + // Mask x&0xFFFF, y&0xFFFF0000,z&0,w&0 + vTemp = _mm_and_ps(vTemp, g_XMMaskX16Y16); + // x needs to be sign extended + vTemp = _mm_xor_ps(vTemp, g_XMFlipX16Y16); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x - 0x8000 to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMFixX16Y16); + // Convert -1.0f - 1.0f + vTemp = _mm_mul_ps(vTemp, g_XMNormalizeX16Y16); + // Clamp result (for case of -32768) + return _mm_max_ps(vTemp, g_XMNegativeOne); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadShort2(const XMSHORT2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + 0.f, + 0.f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt16 = vld1_dup_u32(reinterpret_cast(pSource)); + int32x4_t vInt = vmovl_s16(vreinterpret_s16_u32(vInt16)); + vInt = vandq_s32(vInt, g_XMMaskXY); + return vcvtq_f32_s32(vInt); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the two shorts in all four entries (WORD alignment okay, + // DWORD alignment preferred) + __m128 vTemp = _mm_load_ps1(reinterpret_cast(&pSource->x)); + // Mask x&0xFFFF, y&0xFFFF0000,z&0,w&0 + vTemp = _mm_and_ps(vTemp, g_XMMaskX16Y16); + // x needs to be sign extended + vTemp = _mm_xor_ps(vTemp, g_XMFlipX16Y16); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x - 0x8000 to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMFixX16Y16); + // Y is 65536 too large + return _mm_mul_ps(vTemp, g_XMFixupY16); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUShortN2(const XMUSHORTN2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x) / 65535.0f, + static_cast(pSource->y) / 65535.0f, + 0.f, + 0.f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt16 = vld1_dup_u32(reinterpret_cast(pSource)); + uint32x4_t vInt = vmovl_u16(vreinterpret_u16_u32(vInt16)); + vInt = vandq_u32(vInt, g_XMMaskXY); + float32x4_t R = vcvtq_f32_u32(vInt); + R = vmulq_n_f32(R, 1.0f / 65535.0f); + return vmaxq_f32(R, vdupq_n_f32(-1.f)); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 FixupY16 = { { { 1.0f / 65535.0f, 1.0f / (65535.0f * 65536.0f), 0.0f, 0.0f } } }; + static const XMVECTORF32 FixaddY16 = { { { 0, 32768.0f * 65536.0f, 0, 0 } } }; + // Splat the two shorts in all four entries (WORD alignment okay, + // DWORD alignment preferred) + __m128 vTemp = _mm_load_ps1(reinterpret_cast(&pSource->x)); + // Mask x&0xFFFF, y&0xFFFF0000,z&0,w&0 + vTemp = _mm_and_ps(vTemp, g_XMMaskX16Y16); + // y needs to be sign flipped + vTemp = _mm_xor_ps(vTemp, g_XMFlipY); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // y + 0x8000 to undo the signed order. + vTemp = _mm_add_ps(vTemp, FixaddY16); + // Y is 65536 times too large + vTemp = _mm_mul_ps(vTemp, FixupY16); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUShort2(const XMUSHORT2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + 0.f, + 0.f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt16 = vld1_dup_u32(reinterpret_cast(pSource)); + uint32x4_t vInt = vmovl_u16(vreinterpret_u16_u32(vInt16)); + vInt = vandq_u32(vInt, g_XMMaskXY); + return vcvtq_f32_u32(vInt); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 FixaddY16 = { { { 0, 32768.0f, 0, 0 } } }; + // Splat the two shorts in all four entries (WORD alignment okay, + // DWORD alignment preferred) + __m128 vTemp = _mm_load_ps1(reinterpret_cast(&pSource->x)); + // Mask x&0xFFFF, y&0xFFFF0000,z&0,w&0 + vTemp = _mm_and_ps(vTemp, g_XMMaskX16Y16); + // y needs to be sign flipped + vTemp = _mm_xor_ps(vTemp, g_XMFlipY); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // Y is 65536 times too large + vTemp = _mm_mul_ps(vTemp, g_XMFixupY16); + // y + 0x8000 to undo the signed order. + vTemp = _mm_add_ps(vTemp, FixaddY16); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadByteN2(const XMBYTEN2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + (pSource->x == -128) ? -1.f : (static_cast(pSource->x)* (1.0f / 127.0f)), + (pSource->y == -128) ? -1.f : (static_cast(pSource->y)* (1.0f / 127.0f)), + 0.0f, + 0.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint16x4_t vInt8 = vld1_dup_u16(reinterpret_cast(pSource)); + int16x8_t vInt16 = vmovl_s8(vreinterpret_s8_u16(vInt8)); + int32x4_t vInt = vmovl_s16(vget_low_s16(vInt16)); + vInt = vandq_s32(vInt, g_XMMaskXY); + float32x4_t R = vcvtq_f32_s32(vInt); + R = vmulq_n_f32(R, 1.0f / 127.0f); + return vmaxq_f32(R, vdupq_n_f32(-1.f)); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f / 127.0f, 1.0f / (127.0f * 256.0f), 0, 0 } } }; + static const XMVECTORU32 Mask = { { { 0xFF, 0xFF00, 0, 0 } } }; + // Splat the color in all four entries (x,z,y,w) + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vTemp = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask + vTemp = _mm_and_ps(vTemp, Mask); + // x,y and z are unsigned! Flip the bits to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMXorByte4); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x, y and z - 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddByte4); + // Fix y, z and w because they are too large + vTemp = _mm_mul_ps(vTemp, Scale); + // Clamp result (for case of -128) + return _mm_max_ps(vTemp, g_XMNegativeOne); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadByte2(const XMBYTE2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + 0.0f, + 0.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint16x4_t vInt8 = vld1_dup_u16(reinterpret_cast(pSource)); + int16x8_t vInt16 = vmovl_s8(vreinterpret_s8_u16(vInt8)); + int32x4_t vInt = vmovl_s16(vget_low_s16(vInt16)); + vInt = vandq_s32(vInt, g_XMMaskXY); + return vcvtq_f32_s32(vInt); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f, 1.0f / 256.0f, 1.0f / 65536.0f, 1.0f / (65536.0f * 256.0f) } } }; + static const XMVECTORU32 Mask = { { { 0xFF, 0xFF00, 0, 0 } } }; + // Splat the color in all four entries (x,z,y,w) + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vTemp = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask + vTemp = _mm_and_ps(vTemp, Mask); + // x,y and z are unsigned! Flip the bits to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMXorByte4); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x, y and z - 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddByte4); + // Fix y, z and w because they are too large + return _mm_mul_ps(vTemp, Scale); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUByteN2(const XMUBYTEN2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x)* (1.0f / 255.0f), + static_cast(pSource->y)* (1.0f / 255.0f), + 0.0f, + 0.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint16x4_t vInt8 = vld1_dup_u16(reinterpret_cast(pSource)); + uint16x8_t vInt16 = vmovl_u8(vreinterpret_u8_u16(vInt8)); + uint32x4_t vInt = vmovl_u16(vget_low_u16(vInt16)); + vInt = vandq_u32(vInt, g_XMMaskXY); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_n_f32(R, 1.0f / 255.0f); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f / 255.0f, 1.0f / (255.0f * 256.0f), 0, 0 } } }; + static const XMVECTORU32 Mask = { { { 0xFF, 0xFF00, 0, 0 } } }; + // Splat the color in all four entries (x,z,y,w) + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vTemp = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask + vTemp = _mm_and_ps(vTemp, Mask); + // w is signed! Flip the bits to convert the order to unsigned + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // w + 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Fix y, z and w because they are too large + return _mm_mul_ps(vTemp, Scale); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUByte2(const XMUBYTE2* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + 0.0f, + 0.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint16x4_t vInt8 = vld1_dup_u16(reinterpret_cast(pSource)); + uint16x8_t vInt16 = vmovl_u8(vreinterpret_u8_u16(vInt8)); + uint32x4_t vInt = vmovl_u16(vget_low_u16(vInt16)); + vInt = vandq_u32(vInt, g_XMMaskXY); + return vcvtq_f32_u32(vInt); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f, 1.0f / 256.0f, 0, 0 } } }; + static const XMVECTORU32 Mask = { { { 0xFF, 0xFF00, 0, 0 } } }; + // Splat the color in all four entries (x,z,y,w) + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vTemp = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask + vTemp = _mm_and_ps(vTemp, Mask); + // w is signed! Flip the bits to convert the order to unsigned + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // w + 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Fix y, z and w because they are too large + return _mm_mul_ps(vTemp, Scale); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadU565(const XMU565* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + float(pSource->v & 0x1F), + float((pSource->v >> 5) & 0x3F), + float((pSource->v >> 11) & 0x1F), + 0.f, + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORI32 U565And = { { { 0x1F, 0x3F << 5, 0x1F << 11, 0 } } }; + static const XMVECTORF32 U565Mul = { { { 1.0f, 1.0f / 32.0f, 1.0f / 2048.f, 0 } } }; + uint16x4_t vInt16 = vld1_dup_u16(reinterpret_cast(pSource)); + uint32x4_t vInt = vmovl_u16(vInt16); + vInt = vandq_u32(vInt, U565And); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_f32(R, U565Mul); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORI32 U565And = { { { 0x1F, 0x3F << 5, 0x1F << 11, 0 } } }; + static const XMVECTORF32 U565Mul = { { { 1.0f, 1.0f / 32.0f, 1.0f / 2048.f, 0 } } }; + // Get the 16 bit value and splat it + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vResult = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask off x, y and z + vResult = _mm_and_ps(vResult, U565And); + // Convert to float + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Normalize x, y, and z + vResult = _mm_mul_ps(vResult, U565Mul); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat3PK(const XMFLOAT3PK* pSource) noexcept +{ + assert(pSource); + + XM_ALIGNED_DATA(16) uint32_t Result[4]; + uint32_t Mantissa; + uint32_t Exponent; + + // X Channel (6-bit mantissa) + Mantissa = pSource->xm; + + if (pSource->xe == 0x1f) // INF or NAN + { + Result[0] = static_cast(0x7f800000 | (static_cast(pSource->xm) << 17)); + } + else + { + if (pSource->xe != 0) // The value is normalized + { + Exponent = pSource->xe; + } + else if (Mantissa != 0) // The value is denormalized + { + // Normalize the value in the resulting float + Exponent = 1; + + do + { + Exponent--; + Mantissa <<= 1; + } + while ((Mantissa & 0x40) == 0); + + Mantissa &= 0x3F; + } + else // The value is zero + { + Exponent = static_cast(-112); + } + + Result[0] = ((Exponent + 112) << 23) | (Mantissa << 17); + } + + // Y Channel (6-bit mantissa) + Mantissa = pSource->ym; + + if (pSource->ye == 0x1f) // INF or NAN + { + Result[1] = static_cast(0x7f800000 | (static_cast(pSource->ym) << 17)); + } + else + { + if (pSource->ye != 0) // The value is normalized + { + Exponent = pSource->ye; + } + else if (Mantissa != 0) // The value is denormalized + { + // Normalize the value in the resulting float + Exponent = 1; + + do + { + Exponent--; + Mantissa <<= 1; + } + while ((Mantissa & 0x40) == 0); + + Mantissa &= 0x3F; + } + else // The value is zero + { + Exponent = static_cast(-112); + } + + Result[1] = ((Exponent + 112) << 23) | (Mantissa << 17); + } + + // Z Channel (5-bit mantissa) + Mantissa = pSource->zm; + + if (pSource->ze == 0x1f) // INF or NAN + { + Result[2] = static_cast(0x7f800000 | (static_cast(pSource->zm) << 17)); + } + else + { + if (pSource->ze != 0) // The value is normalized + { + Exponent = pSource->ze; + } + else if (Mantissa != 0) // The value is denormalized + { + // Normalize the value in the resulting float + Exponent = 1; + + do + { + Exponent--; + Mantissa <<= 1; + } + while ((Mantissa & 0x20) == 0); + + Mantissa &= 0x1F; + } + else // The value is zero + { + Exponent = static_cast(-112); + } + + Result[2] = ((Exponent + 112) << 23) | (Mantissa << 18); + } + + return XMLoadFloat3A(reinterpret_cast(&Result)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadFloat3SE(const XMFLOAT3SE* pSource) noexcept +{ + assert(pSource); + +#if defined(_XM_NO_INTRINSICS_) + + union { float f; int32_t i; } fi; + fi.i = 0x33800000 + (pSource->e << 23); + float Scale = fi.f; + + XMVECTORF32 v = { { { + Scale * float(pSource->xm), + Scale * float(pSource->ym), + Scale * float(pSource->zm), + 1.0f } } }; + return v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + + // Build scale factor from shared exponent + union { float f; int32_t i; } fi; + fi.i = 0x33800000 + (pSource->e << 23); + + // Extract 9-bit mantissas into vector lanes + uint32x4_t mantissas = vdupq_n_u32(0); + mantissas = vsetq_lane_u32(pSource->xm, mantissas, 0); + mantissas = vsetq_lane_u32(pSource->ym, mantissas, 1); + mantissas = vsetq_lane_u32(pSource->zm, mantissas, 2); + + // Convert to float, scale, and set w = 1.0f + float32x4_t result = vmulq_n_f32(vcvtq_f32_u32(mantissas), fi.f); + return vsetq_lane_f32(1.0f, result, 3); + +#elif defined(_XM_SSE_INTRINSICS_) + + // Build scale factor from shared exponent + union { float f; int32_t i; } fi; + fi.i = 0x33800000 + (pSource->e << 23); + + // Extract 9-bit mantissas, convert to float, and scale + __m128i mantissas = _mm_set_epi32( + 0, + static_cast(pSource->zm), + static_cast(pSource->ym), + static_cast(pSource->xm)); + __m128 result = _mm_mul_ps(_mm_cvtepi32_ps(mantissas), _mm_set1_ps(fi.f)); + + // Set w = 1.0f (w lane is +0.0f so bitwise OR inserts 1.0f cleanly) + return _mm_or_ps(result, g_XMIdentityR3); + +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadHalf4(const XMHALF4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + __m128i V = _mm_loadl_epi64(reinterpret_cast(pSource)); + return _mm_cvtph_ps(V); +#else + XMVECTORF32 vResult = { { { + XMConvertHalfToFloat(pSource->x), + XMConvertHalfToFloat(pSource->y), + XMConvertHalfToFloat(pSource->z), + XMConvertHalfToFloat(pSource->w) + } } }; + return vResult.v; +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadShortN4(const XMSHORTN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + (pSource->x == -32768) ? -1.f : (static_cast(pSource->x)* (1.0f / 32767.0f)), + (pSource->y == -32768) ? -1.f : (static_cast(pSource->y)* (1.0f / 32767.0f)), + (pSource->z == -32768) ? -1.f : (static_cast(pSource->z)* (1.0f / 32767.0f)), + (pSource->w == -32768) ? -1.f : (static_cast(pSource->w)* (1.0f / 32767.0f)) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int16x4_t vInt = vld1_s16(reinterpret_cast(pSource)); + int32x4_t V = vmovl_s16(vInt); + float32x4_t vResult = vcvtq_f32_s32(V); + vResult = vmulq_n_f32(vResult, 1.0f / 32767.0f); + return vmaxq_f32(vResult, vdupq_n_f32(-1.f)); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the color in all four entries (x,z,y,w) + __m128d vIntd = _mm_load1_pd(reinterpret_cast(&pSource->x)); + // Shift x&0ffff,z&0xffff,y&0xffff0000,w&0xffff0000 + __m128 vTemp = _mm_and_ps(_mm_castpd_ps(vIntd), g_XMMaskX16Y16Z16W16); + // x and z are unsigned! Flip the bits to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMFlipX16Y16Z16W16); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x and z - 0x8000 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMFixX16Y16Z16W16); + // Convert to -1.0f - 1.0f + vTemp = _mm_mul_ps(vTemp, g_XMNormalizeX16Y16Z16W16); + // Very important! The entries are x,z,y,w, flip it to x,y,z,w + vTemp = XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 1, 2, 0)); + // Clamp result (for case of -32768) + return _mm_max_ps(vTemp, g_XMNegativeOne); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadShort4(const XMSHORT4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + static_cast(pSource->z), + static_cast(pSource->w) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + int16x4_t vInt = vld1_s16(reinterpret_cast(pSource)); + int32x4_t V = vmovl_s16(vInt); + return vcvtq_f32_s32(V); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the color in all four entries (x,z,y,w) + __m128d vIntd = _mm_load1_pd(reinterpret_cast(&pSource->x)); + // Shift x&0ffff,z&0xffff,y&0xffff0000,w&0xffff0000 + __m128 vTemp = _mm_and_ps(_mm_castpd_ps(vIntd), g_XMMaskX16Y16Z16W16); + // x and z are unsigned! Flip the bits to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMFlipX16Y16Z16W16); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x and z - 0x8000 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMFixX16Y16Z16W16); + // Fix y and w because they are 65536 too large + vTemp = _mm_mul_ps(vTemp, g_XMFixupY16W16); + // Very important! The entries are x,z,y,w, flip it to x,y,z,w + return XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 1, 2, 0)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUShortN4(const XMUSHORTN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x) / 65535.0f, + static_cast(pSource->y) / 65535.0f, + static_cast(pSource->z) / 65535.0f, + static_cast(pSource->w) / 65535.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint16x4_t vInt = vld1_u16(reinterpret_cast(pSource)); + uint32x4_t V = vmovl_u16(vInt); + float32x4_t vResult = vcvtq_f32_u32(V); + return vmulq_n_f32(vResult, 1.0f / 65535.0f); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 FixupY16W16 = { { { 1.0f / 65535.0f, 1.0f / 65535.0f, 1.0f / (65535.0f * 65536.0f), 1.0f / (65535.0f * 65536.0f) } } }; + static const XMVECTORF32 FixaddY16W16 = { { { 0, 0, 32768.0f * 65536.0f, 32768.0f * 65536.0f } } }; + // Splat the color in all four entries (x,z,y,w) + __m128d vIntd = _mm_load1_pd(reinterpret_cast(&pSource->x)); + // Shift x&0ffff,z&0xffff,y&0xffff0000,w&0xffff0000 + __m128 vTemp = _mm_and_ps(_mm_castpd_ps(vIntd), g_XMMaskX16Y16Z16W16); + // y and w are signed! Flip the bits to convert the order to unsigned + vTemp = _mm_xor_ps(vTemp, g_XMFlipZW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // y and w + 0x8000 to complete the conversion + vTemp = _mm_add_ps(vTemp, FixaddY16W16); + // Fix y and w because they are 65536 too large + vTemp = _mm_mul_ps(vTemp, FixupY16W16); + // Very important! The entries are x,z,y,w, flip it to x,y,z,w + return XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 1, 2, 0)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUShort4(const XMUSHORT4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + static_cast(pSource->z), + static_cast(pSource->w) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint16x4_t vInt = vld1_u16(reinterpret_cast(pSource)); + uint32x4_t V = vmovl_u16(vInt); + return vcvtq_f32_u32(V); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 FixaddY16W16 = { { { 0, 0, 32768.0f, 32768.0f } } }; + // Splat the color in all four entries (x,z,y,w) + __m128d vIntd = _mm_load1_pd(reinterpret_cast(&pSource->x)); + // Shift x&0ffff,z&0xffff,y&0xffff0000,w&0xffff0000 + __m128 vTemp = _mm_and_ps(_mm_castpd_ps(vIntd), g_XMMaskX16Y16Z16W16); + // y and w are signed! Flip the bits to convert the order to unsigned + vTemp = _mm_xor_ps(vTemp, g_XMFlipZW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // Fix y and w because they are 65536 too large + vTemp = _mm_mul_ps(vTemp, g_XMFixupY16W16); + // y and w + 0x8000 to complete the conversion + vTemp = _mm_add_ps(vTemp, FixaddY16W16); + // Very important! The entries are x,z,y,w, flip it to x,y,z,w + return XM_PERMUTE_PS(vTemp, _MM_SHUFFLE(3, 1, 2, 0)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadXDecN4(const XMXDECN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + static const uint32_t SignExtend[] = { 0x00000000, 0xFFFFFC00 }; + + uint32_t ElementX = pSource->v & 0x3FF; + uint32_t ElementY = (pSource->v >> 10) & 0x3FF; + uint32_t ElementZ = (pSource->v >> 20) & 0x3FF; + + XMVECTORF32 vResult = { { { + (ElementX == 0x200) ? -1.f : (static_cast(static_cast(ElementX | SignExtend[ElementX >> 9])) / 511.0f), + (ElementY == 0x200) ? -1.f : (static_cast(static_cast(ElementY | SignExtend[ElementY >> 9])) / 511.0f), + (ElementZ == 0x200) ? -1.f : (static_cast(static_cast(ElementZ | SignExtend[ElementZ >> 9])) / 511.0f), + static_cast(pSource->v >> 30) / 3.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskA2B10G10R10); + vInt = veorq_u32(vInt, g_XMFlipA2B10G10R10); + float32x4_t R = vcvtq_f32_s32(vreinterpretq_s32_u32(vInt)); + R = vaddq_f32(R, g_XMFixAA2B10G10R10); + R = vmulq_f32(R, g_XMNormalizeA2B10G10R10); + return vmaxq_f32(R, vdupq_n_f32(-1.0f)); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the color in all four entries + __m128 vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskA2B10G10R10); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMFlipA2B10G10R10); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMFixAA2B10G10R10); + // Convert 0-255 to 0.0f-1.0f + vTemp = _mm_mul_ps(vTemp, g_XMNormalizeA2B10G10R10); + // Clamp result (for case of -512) + return _mm_max_ps(vTemp, g_XMNegativeOne); +#endif +} + +//------------------------------------------------------------------------------ +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4996) +// C4996: ignore deprecation warning +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wdeprecated-declarations" +#endif + +#ifdef __GNUC__ +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" +#endif + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadXDec4(const XMXDEC4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + static const uint32_t SignExtend[] = { 0x00000000, 0xFFFFFC00 }; + + uint32_t ElementX = pSource->v & 0x3FF; + uint32_t ElementY = (pSource->v >> 10) & 0x3FF; + uint32_t ElementZ = (pSource->v >> 20) & 0x3FF; + + XMVECTORF32 vResult = { { { + static_cast(static_cast(ElementX | SignExtend[ElementX >> 9])), + static_cast(static_cast(ElementY | SignExtend[ElementY >> 9])), + static_cast(static_cast(ElementZ | SignExtend[ElementZ >> 9])), + static_cast(pSource->v >> 30) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORU32 XDec4Xor = { { { 0x200, 0x200 << 10, 0x200 << 20, 0x80000000 } } }; + static const XMVECTORF32 XDec4Add = { { { -512.0f, -512.0f * 1024.0f, -512.0f * 1024.0f * 1024.0f, 32768 * 65536.0f } } }; + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskDec4); + vInt = veorq_u32(vInt, XDec4Xor); + float32x4_t R = vcvtq_f32_s32(vreinterpretq_s32_u32(vInt)); + R = vaddq_f32(R, XDec4Add); + return vmulq_f32(R, g_XMMulDec4); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORU32 XDec4Xor = { { { 0x200, 0x200 << 10, 0x200 << 20, 0x80000000 } } }; + static const XMVECTORF32 XDec4Add = { { { -512.0f, -512.0f * 1024.0f, -512.0f * 1024.0f * 1024.0f, 32768 * 65536.0f } } }; + // Splat the color in all four entries + XMVECTOR vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskDec4); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, XDec4Xor); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, XDec4Add); + // Convert 0-255 to 0.0f-1.0f + vTemp = _mm_mul_ps(vTemp, g_XMMulDec4); + return vTemp; +#endif +} + +#ifdef __GNUC__ +#pragma GCC diagnostic pop +#endif +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUDecN4(const XMUDECN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + uint32_t ElementX = pSource->v & 0x3FF; + uint32_t ElementY = (pSource->v >> 10) & 0x3FF; + uint32_t ElementZ = (pSource->v >> 20) & 0x3FF; + + XMVECTORF32 vResult = { { { + static_cast(ElementX) / 1023.0f, + static_cast(ElementY) / 1023.0f, + static_cast(ElementZ) / 1023.0f, + static_cast(pSource->v >> 30) / 3.0f + } } }; + return vResult.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 UDecN4Mul = { { { 1.0f / 1023.0f, 1.0f / (1023.0f * 1024.0f), 1.0f / (1023.0f * 1024.0f * 1024.0f), 1.0f / (3.0f * 1024.0f * 1024.0f * 1024.0f) } } }; + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskDec4); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_f32(R, UDecN4Mul); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 UDecN4Mul = { { { 1.0f / 1023.0f, 1.0f / (1023.0f * 1024.0f), 1.0f / (1023.0f * 1024.0f * 1024.0f), 1.0f / (3.0f * 1024.0f * 1024.0f * 1024.0f) } } }; + // Splat the color in all four entries + XMVECTOR vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskDec4); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Convert 0-255 to 0.0f-1.0f + vTemp = _mm_mul_ps(vTemp, UDecN4Mul); + return vTemp; +#endif +} + + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUDecN4_XR(const XMUDECN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + + int32_t ElementX = pSource->v & 0x3FF; + int32_t ElementY = (pSource->v >> 10) & 0x3FF; + int32_t ElementZ = (pSource->v >> 20) & 0x3FF; + + XMVECTORF32 vResult = { { { + static_cast(ElementX - 0x180) / 510.0f, + static_cast(ElementY - 0x180) / 510.0f, + static_cast(ElementZ - 0x180) / 510.0f, + static_cast(pSource->v >> 30) / 3.0f + } } }; + + return vResult.v; + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 XRMul = { { { 1.0f / 510.0f, 1.0f / (510.0f * 1024.0f), 1.0f / (510.0f * 1024.0f * 1024.0f), 1.0f / (3.0f * 1024.0f * 1024.0f * 1024.0f) } } }; + static const XMVECTORI32 XRBias = { { { 0x180, 0x180 * 1024, 0x180 * 1024 * 1024, 0 } } }; + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskDec4); + int32x4_t vTemp = vsubq_s32(vreinterpretq_s32_u32(vInt), XRBias); + vTemp = veorq_s32(vTemp, g_XMFlipW); + float32x4_t R = vcvtq_f32_s32(vTemp); + R = vaddq_f32(R, g_XMAddUDec4); + return vmulq_f32(R, XRMul); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 XRMul = { { { 1.0f / 510.0f, 1.0f / (510.0f * 1024.0f), 1.0f / (510.0f * 1024.0f * 1024.0f), 1.0f / (3.0f * 1024.0f * 1024.0f * 1024.0f) } } }; + static const XMVECTORI32 XRBias = { { { 0x180, 0x180 * 1024, 0x180 * 1024 * 1024, 0 } } }; + // Splat the color in all four entries + XMVECTOR vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Mask channels + vTemp = _mm_and_ps(vTemp, g_XMMaskDec4); + // Subtract bias + vTemp = _mm_castsi128_ps(_mm_sub_epi32(_mm_castps_si128(vTemp), XRBias)); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Convert to 0.0f-1.0f + return _mm_mul_ps(vTemp, XRMul); +#endif +} + + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUDec4(const XMUDEC4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + uint32_t ElementX = pSource->v & 0x3FF; + uint32_t ElementY = (pSource->v >> 10) & 0x3FF; + uint32_t ElementZ = (pSource->v >> 20) & 0x3FF; + + XMVECTORF32 vResult = { { { + static_cast(ElementX), + static_cast(ElementY), + static_cast(ElementZ), + static_cast(pSource->v >> 30) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskDec4); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_f32(R, g_XMMulDec4); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the color in all four entries + XMVECTOR vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskDec4); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Convert 0-255 to 0.0f-1.0f + vTemp = _mm_mul_ps(vTemp, g_XMMulDec4); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4996) +// C4996: ignore deprecation warning +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wdeprecated-declarations" +#endif + +#ifdef __GNUC__ +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" +#endif + +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadDecN4(const XMDECN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + static const uint32_t SignExtend[] = { 0x00000000, 0xFFFFFC00 }; + static const uint32_t SignExtendW[] = { 0x00000000, 0xFFFFFFFC }; + + uint32_t ElementX = pSource->v & 0x3FF; + uint32_t ElementY = (pSource->v >> 10) & 0x3FF; + uint32_t ElementZ = (pSource->v >> 20) & 0x3FF; + uint32_t ElementW = pSource->v >> 30; + + XMVECTORF32 vResult = { { { + (ElementX == 0x200) ? -1.f : (static_cast(static_cast(ElementX | SignExtend[ElementX >> 9])) / 511.0f), + (ElementY == 0x200) ? -1.f : (static_cast(static_cast(ElementY | SignExtend[ElementY >> 9])) / 511.0f), + (ElementZ == 0x200) ? -1.f : (static_cast(static_cast(ElementZ | SignExtend[ElementZ >> 9])) / 511.0f), + (ElementW == 0x2) ? -1.f : static_cast(static_cast(ElementW | SignExtendW[(ElementW >> 1) & 1])) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 DecN4Mul = { { { 1.0f / 511.0f, 1.0f / (511.0f * 1024.0f), 1.0f / (511.0f * 1024.0f * 1024.0f), 1.0f / (1024.0f * 1024.0f * 1024.0f) } } }; + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskDec4); + vInt = veorq_u32(vInt, g_XMXorDec4); + float32x4_t R = vcvtq_f32_s32(vreinterpretq_s32_u32(vInt)); + R = vaddq_f32(R, g_XMAddDec4); + R = vmulq_f32(R, DecN4Mul); + return vmaxq_f32(R, vdupq_n_f32(-1.0f)); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 DecN4Mul = { { { 1.0f / 511.0f, 1.0f / (511.0f * 1024.0f), 1.0f / (511.0f * 1024.0f * 1024.0f), 1.0f / (1024.0f * 1024.0f * 1024.0f) } } }; + // Splat the color in all four entries + XMVECTOR vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskDec4); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMXorDec4); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMAddDec4); + // Convert 0-255 to 0.0f-1.0f + vTemp = _mm_mul_ps(vTemp, DecN4Mul); + // Clamp result (for case of -512/-1) + return _mm_max_ps(vTemp, g_XMNegativeOne); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadDec4(const XMDEC4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + static const uint32_t SignExtend[] = { 0x00000000, 0xFFFFFC00 }; + static const uint32_t SignExtendW[] = { 0x00000000, 0xFFFFFFFC }; + + uint32_t ElementX = pSource->v & 0x3FF; + uint32_t ElementY = (pSource->v >> 10) & 0x3FF; + uint32_t ElementZ = (pSource->v >> 20) & 0x3FF; + uint32_t ElementW = pSource->v >> 30; + + XMVECTORF32 vResult = { { { + static_cast(static_cast(ElementX | SignExtend[ElementX >> 9])), + static_cast(static_cast(ElementY | SignExtend[ElementY >> 9])), + static_cast(static_cast(ElementZ | SignExtend[ElementZ >> 9])), + static_cast(static_cast(ElementW | SignExtendW[ElementW >> 1])) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x4_t vInt = vld1q_dup_u32(reinterpret_cast(pSource)); + vInt = vandq_u32(vInt, g_XMMaskDec4); + vInt = veorq_u32(vInt, g_XMXorDec4); + float32x4_t R = vcvtq_f32_s32(vreinterpretq_s32_u32(vInt)); + R = vaddq_f32(R, g_XMAddDec4); + return vmulq_f32(R, g_XMMulDec4); +#elif defined(_XM_SSE_INTRINSICS_) + // Splat the color in all four entries + XMVECTOR vTemp = _mm_load_ps1(reinterpret_cast(&pSource->v)); + // Shift R&0xFF0000, G&0xFF00, B&0xFF, A&0xFF000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskDec4); + // a is unsigned! Flip the bit to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMXorDec4); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // RGB + 0, A + 0x80000000.f to undo the signed order. + vTemp = _mm_add_ps(vTemp, g_XMAddDec4); + // Convert 0-255 to 0.0f-1.0f + vTemp = _mm_mul_ps(vTemp, g_XMMulDec4); + return vTemp; +#endif +} + +#ifdef __GNUC__ +#pragma GCC diagnostic pop +#endif +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUByteN4(const XMUBYTEN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x) / 255.0f, + static_cast(pSource->y) / 255.0f, + static_cast(pSource->z) / 255.0f, + static_cast(pSource->w) / 255.0f + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt8 = vld1_dup_u32(reinterpret_cast(pSource)); + uint16x8_t vInt16 = vmovl_u8(vreinterpret_u8_u32(vInt8)); + uint32x4_t vInt = vmovl_u16(vget_low_u16(vInt16)); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_n_f32(R, 1.0f / 255.0f); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 LoadUByteN4Mul = { { { 1.0f / 255.0f, 1.0f / (255.0f * 256.0f), 1.0f / (255.0f * 65536.0f), 1.0f / (255.0f * 65536.0f * 256.0f) } } }; + // Splat the color in all four entries (x,z,y,w) + XMVECTOR vTemp = _mm_load1_ps(reinterpret_cast(&pSource->x)); + // Mask x&0ff,y&0xff00,z&0xff0000,w&0xff000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskByte4); + // w is signed! Flip the bits to convert the order to unsigned + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // w + 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Fix y, z and w because they are too large + vTemp = _mm_mul_ps(vTemp, LoadUByteN4Mul); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUByte4(const XMUBYTE4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + static_cast(pSource->z), + static_cast(pSource->w) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt8 = vld1_dup_u32(reinterpret_cast(pSource)); + uint16x8_t vInt16 = vmovl_u8(vreinterpret_u8_u32(vInt8)); + uint32x4_t vInt = vmovl_u16(vget_low_u16(vInt16)); + return vcvtq_f32_u32(vInt); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 LoadUByte4Mul = { { { 1.0f, 1.0f / 256.0f, 1.0f / 65536.0f, 1.0f / (65536.0f * 256.0f) } } }; + // Splat the color in all four entries (x,z,y,w) + XMVECTOR vTemp = _mm_load1_ps(reinterpret_cast(&pSource->x)); + // Mask x&0ff,y&0xff00,z&0xff0000,w&0xff000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskByte4); + // w is signed! Flip the bits to convert the order to unsigned + vTemp = _mm_xor_ps(vTemp, g_XMFlipW); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // w + 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddUDec4); + // Fix y, z and w because they are too large + vTemp = _mm_mul_ps(vTemp, LoadUByte4Mul); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadByteN4(const XMBYTEN4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + (pSource->x == -128) ? -1.f : (static_cast(pSource->x) / 127.0f), + (pSource->y == -128) ? -1.f : (static_cast(pSource->y) / 127.0f), + (pSource->z == -128) ? -1.f : (static_cast(pSource->z) / 127.0f), + (pSource->w == -128) ? -1.f : (static_cast(pSource->w) / 127.0f) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt8 = vld1_dup_u32(reinterpret_cast(pSource)); + int16x8_t vInt16 = vmovl_s8(vreinterpret_s8_u32(vInt8)); + int32x4_t vInt = vmovl_s16(vget_low_s16(vInt16)); + float32x4_t R = vcvtq_f32_s32(vInt); + R = vmulq_n_f32(R, 1.0f / 127.0f); + return vmaxq_f32(R, vdupq_n_f32(-1.f)); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 LoadByteN4Mul = { { { 1.0f / 127.0f, 1.0f / (127.0f * 256.0f), 1.0f / (127.0f * 65536.0f), 1.0f / (127.0f * 65536.0f * 256.0f) } } }; + // Splat the color in all four entries (x,z,y,w) + XMVECTOR vTemp = _mm_load1_ps(reinterpret_cast(&pSource->x)); + // Mask x&0ff,y&0xff00,z&0xff0000,w&0xff000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskByte4); + // x,y and z are unsigned! Flip the bits to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMXorByte4); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x, y and z - 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddByte4); + // Fix y, z and w because they are too large + vTemp = _mm_mul_ps(vTemp, LoadByteN4Mul); + // Clamp result (for case of -128) + return _mm_max_ps(vTemp, g_XMNegativeOne); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadByte4(const XMBYTE4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + static_cast(pSource->x), + static_cast(pSource->y), + static_cast(pSource->z), + static_cast(pSource->w) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + uint32x2_t vInt8 = vld1_dup_u32(reinterpret_cast(pSource)); + int16x8_t vInt16 = vmovl_s8(vreinterpret_s8_u32(vInt8)); + int32x4_t vInt = vmovl_s16(vget_low_s16(vInt16)); + return vcvtq_f32_s32(vInt); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 LoadByte4Mul = { { { 1.0f, 1.0f / 256.0f, 1.0f / 65536.0f, 1.0f / (65536.0f * 256.0f) } } }; + // Splat the color in all four entries (x,z,y,w) + XMVECTOR vTemp = _mm_load1_ps(reinterpret_cast(&pSource->x)); + // Mask x&0ff,y&0xff00,z&0xff0000,w&0xff000000 + vTemp = _mm_and_ps(vTemp, g_XMMaskByte4); + // x,y and z are unsigned! Flip the bits to convert the order to signed + vTemp = _mm_xor_ps(vTemp, g_XMXorByte4); + // Convert to floating point numbers + vTemp = _mm_cvtepi32_ps(_mm_castps_si128(vTemp)); + // x, y and z - 0x80 to complete the conversion + vTemp = _mm_add_ps(vTemp, g_XMAddByte4); + // Fix y, z and w because they are too large + vTemp = _mm_mul_ps(vTemp, LoadByte4Mul); + return vTemp; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadUNibble4(const XMUNIBBLE4* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + float(pSource->v & 0xF), + float((pSource->v >> 4) & 0xF), + float((pSource->v >> 8) & 0xF), + float((pSource->v >> 12) & 0xF) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORI32 UNibble4And = { { { 0xF, 0xF0, 0xF00, 0xF000 } } }; + static const XMVECTORF32 UNibble4Mul = { { { 1.0f, 1.0f / 16.f, 1.0f / 256.f, 1.0f / 4096.f } } }; + uint16x4_t vInt16 = vld1_dup_u16(reinterpret_cast(pSource)); + uint32x4_t vInt = vmovl_u16(vInt16); + vInt = vandq_u32(vInt, UNibble4And); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_f32(R, UNibble4Mul); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORI32 UNibble4And = { { { 0xF, 0xF0, 0xF00, 0xF000 } } }; + static const XMVECTORF32 UNibble4Mul = { { { 1.0f, 1.0f / 16.f, 1.0f / 256.f, 1.0f / 4096.f } } }; + // Get the 16 bit value and splat it + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vResult = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask off x, y and z + vResult = _mm_and_ps(vResult, UNibble4And); + // Convert to float + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Normalize x, y, and z + vResult = _mm_mul_ps(vResult, UNibble4Mul); + return vResult; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMVECTOR XM_CALLCONV XMLoadU555(const XMU555* pSource) noexcept +{ + assert(pSource); +#if defined(_XM_NO_INTRINSICS_) + XMVECTORF32 vResult = { { { + float(pSource->v & 0x1F), + float((pSource->v >> 5) & 0x1F), + float((pSource->v >> 10) & 0x1F), + float((pSource->v >> 15) & 0x1) + } } }; + return vResult.v; +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORI32 U555And = { { { 0x1F, 0x1F << 5, 0x1F << 10, 0x8000 } } }; + static const XMVECTORF32 U555Mul = { { { 1.0f, 1.0f / 32.f, 1.0f / 1024.f, 1.0f / 32768.f } } }; + uint16x4_t vInt16 = vld1_dup_u16(reinterpret_cast(pSource)); + uint32x4_t vInt = vmovl_u16(vInt16); + vInt = vandq_u32(vInt, U555And); + float32x4_t R = vcvtq_f32_u32(vInt); + return vmulq_f32(R, U555Mul); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORI32 U555And = { { { 0x1F, 0x1F << 5, 0x1F << 10, 0x8000 } } }; + static const XMVECTORF32 U555Mul = { { { 1.0f, 1.0f / 32.f, 1.0f / 1024.f, 1.0f / 32768.f } } }; + // Get the 16bit value and splat it + __m128i vInt = XM_LOADU_SI16(&pSource->v); + XMVECTOR vResult = XM_PERMUTE_PS(_mm_castsi128_ps(vInt), _MM_SHUFFLE(0, 0, 0, 0)); + // Mask off x, y and z + vResult = _mm_and_ps(vResult, U555And); + // Convert to float + vResult = _mm_cvtepi32_ps(_mm_castps_si128(vResult)); + // Normalize x, y, and z + vResult = _mm_mul_ps(vResult, U555Mul); + return vResult; +#endif +} + +#ifdef _PREFAST_ +#pragma prefast(pop) +#endif + +/**************************************************************************** + * + * Vector and matrix store operations + * + ****************************************************************************/ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreColor +( + XMCOLOR* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorSaturate(V); + N = XMVectorMultiply(N, g_UByteMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->c = (static_cast(tmp.w) << 24) | + (static_cast(tmp.x) << 16) | + (static_cast(tmp.y) << 8) | + static_cast(tmp.z); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 255.0f); + R = XMVectorRound(R); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + uint8x8_t vInt8 = vqmovn_u16(vcombine_u16(vInt16, vInt16)); + uint32_t rgba = vget_lane_u32(vreinterpret_u32_u8(vInt8), 0); + pDestination->c = (rgba & 0xFF00FF00) | ((rgba >> 16) & 0xFF) | ((rgba << 16) & 0xFF0000); +#elif defined(_XM_SSE_INTRINSICS_) + // Set <0 to 0 + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + // Set>1 to 1 + vResult = _mm_min_ps(vResult, g_XMOne); + // Convert to 0-255 + vResult = _mm_mul_ps(vResult, g_UByteMax); + // Shuffle RGBA to ARGB + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(3, 0, 1, 2)); + // Convert to int + __m128i vInt = _mm_cvtps_epi32(vResult); + // Mash to shorts + vInt = _mm_packs_epi32(vInt, vInt); + // Mash to bytes + vInt = _mm_packus_epi16(vInt, vInt); + // Store the color + _mm_store_ss(reinterpret_cast(&pDestination->c), _mm_castsi128_ps(vInt)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreHalf2 +( + XMHALF2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + __m128i V1 = _mm_cvtps_ph(V, _MM_FROUND_TO_NEAREST_INT); + _mm_store_ss(reinterpret_cast(pDestination), _mm_castsi128_ps(V1)); +#else + pDestination->x = XMConvertFloatToHalf(XMVectorGetX(V)); + pDestination->y = XMConvertFloatToHalf(XMVectorGetY(V)); +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreShortN2 +( + XMSHORTN2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_XMNegativeOne.v, g_XMOne.v); + N = XMVectorMultiply(N, g_ShortMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(-1.f)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 32767.0f); + int32x4_t vInt32 = vcvtq_s32_f32(R); + int16x4_t vInt16 = vqmovn_s32(vInt32); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_s16(vInt16), 0); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_max_ps(V, g_XMNegativeOne); + vResult = _mm_min_ps(vResult, g_XMOne); + vResult = _mm_mul_ps(vResult, g_ShortMax); + __m128i vResulti = _mm_cvtps_epi32(vResult); + vResulti = _mm_packs_epi32(vResulti, vResulti); + _mm_store_ss(reinterpret_cast(&pDestination->x), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreShort2 +( + XMSHORT2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_ShortMin, g_ShortMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(-32767.f)); + R = vminq_f32(R, vdupq_n_f32(32767.0f)); + int32x4_t vInt32 = vcvtq_s32_f32(R); + int16x4_t vInt16 = vqmovn_s32(vInt32); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_s16(vInt16), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_ShortMin); + vResult = _mm_min_ps(vResult, g_ShortMax); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // Pack the ints into shorts + vInt = _mm_packs_epi32(vInt, vInt); + _mm_store_ss(reinterpret_cast(&pDestination->x), _mm_castsi128_ps(vInt)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUShortN2 +( + XMUSHORTN2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorSaturate(V); + N = XMVectorMultiplyAdd(N, g_UShortMax, g_XMOneHalf.v); + N = XMVectorTruncate(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0.f)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 65535.0f); + R = vaddq_f32(R, g_XMOneHalf); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_u16(vInt16), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_XMOne); + vResult = _mm_mul_ps(vResult, g_UShortMax); + vResult = _mm_add_ps(vResult, g_XMOneHalf); + // Convert to int + __m128i vInt = _mm_cvttps_epi32(vResult); + // Since the SSE pack instruction clamps using signed rules, + // manually extract the values to store them to memory + pDestination->x = static_cast(_mm_extract_epi16(vInt, 0)); + pDestination->y = static_cast(_mm_extract_epi16(vInt, 2)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUShort2 +( + XMUSHORT2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), g_UShortMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0.f)); + R = vminq_f32(R, vdupq_n_f32(65535.0f)); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_u16(vInt16), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_UShortMax); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // Since the SSE pack instruction clamps using signed rules, + // manually extract the values to store them to memory + pDestination->x = static_cast(_mm_extract_epi16(vInt, 0)); + pDestination->y = static_cast(_mm_extract_epi16(vInt, 2)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreByteN2 +( + XMBYTEN2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_XMNegativeOne.v, g_XMOne.v); + N = XMVectorMultiply(N, g_ByteMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(-1.f)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 127.0f); + int32x4_t vInt32 = vcvtq_s32_f32(R); + int16x4_t vInt16 = vqmovn_s32(vInt32); + int8x8_t vInt8 = vqmovn_s16(vcombine_s16(vInt16, vInt16)); + vst1_lane_u16(reinterpret_cast(pDestination), vreinterpret_u16_s8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMNegativeOne); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, g_ByteMax); + // Convert to int by rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + pDestination->v = static_cast(((static_cast(y) & 0xFF) << 8) | (static_cast(x) & 0xFF)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreByte2 +( + XMBYTE2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_ByteMin, g_ByteMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(-127.f)); + R = vminq_f32(R, vdupq_n_f32(127.0f)); + int32x4_t vInt32 = vcvtq_s32_f32(R); + int16x4_t vInt16 = vqmovn_s32(vInt32); + int8x8_t vInt8 = vqmovn_s16(vcombine_s16(vInt16, vInt16)); + vst1_lane_u16(reinterpret_cast(pDestination), vreinterpret_u16_s8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_ByteMin); + vResult = _mm_min_ps(vResult, g_ByteMax); + // Convert to int by rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + pDestination->v = static_cast(((static_cast(y) & 0xFF) << 8) | (static_cast(x) & 0xFF)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUByteN2 +( + XMUBYTEN2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorSaturate(V); + N = XMVectorMultiplyAdd(N, g_UByteMax, g_XMOneHalf.v); + N = XMVectorTruncate(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0.f)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 255.0f); + R = vaddq_f32(R, g_XMOneHalf); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + uint8x8_t vInt8 = vqmovn_u16(vcombine_u16(vInt16, vInt16)); + vst1_lane_u16(reinterpret_cast(pDestination), vreinterpret_u16_u8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, g_UByteMax); + vResult = _mm_add_ps(vResult, g_XMOneHalf); + // Convert to int + __m128i vInt = _mm_cvttps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + pDestination->v = static_cast(((static_cast(y) & 0xFF) << 8) | (static_cast(x) & 0xFF)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUByte2 +( + XMUBYTE2* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), g_UByteMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0.f)); + R = vminq_f32(R, vdupq_n_f32(255.0f)); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + uint8x8_t vInt8 = vqmovn_u16(vcombine_u16(vInt16, vInt16)); + vst1_lane_u16(reinterpret_cast(pDestination), vreinterpret_u16_u8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_UByteMax); + // Convert to int by rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + pDestination->v = static_cast(((static_cast(y) & 0xFF) << 8) | (static_cast(x) & 0xFF)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreU565 +( + XMU565* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 Max = { { { 31.0f, 63.0f, 31.0f, 0.0f } } }; + +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), Max.v); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + ((static_cast(tmp.z) & 0x1F) << 11) + | ((static_cast(tmp.y) & 0x3F) << 5) + | ((static_cast(tmp.x) & 0x1F))); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f, 32.f, 32.f * 64.f, 0.f } } }; + static const XMVECTORU32 Mask = { { { 0x1F, 0x3F << 5, 0x1F << 11, 0 } } }; + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0)); + vResult = vminq_f32(vResult, Max); + vResult = vmulq_f32(vResult, Scale); + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + vResulti = vandq_u32(vResulti, Mask); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vResulti); + uint32x2_t vhi = vget_high_u32(vResulti); + vTemp = vorr_u32(vTemp, vhi); + vTemp = vpadd_u32(vTemp, vTemp); + vst1_lane_u16(&pDestination->v, vreinterpret_u16_u32(vTemp), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, Max); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + auto z = static_cast(_mm_extract_epi16(vInt, 4)); + pDestination->v = static_cast( + ((static_cast(z) & 0x1F) << 11) + | ((static_cast(y) & 0x3F) << 5) + | ((static_cast(x) & 0x1F))); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3PK +( + XMFLOAT3PK* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + + XM_ALIGNED_DATA(16) uint32_t IValue[4]; + XMStoreFloat3A(reinterpret_cast(&IValue), V); + + uint32_t Result[3]; + + // X & Y Channels (5-bit exponent, 6-bit mantissa) + for (uint32_t j = 0; j < 2; ++j) + { + uint32_t Sign = IValue[j] & 0x80000000; + uint32_t I = IValue[j] & 0x7FFFFFFF; + + if ((I & 0x7F800000) == 0x7F800000) + { + // INF or NAN + Result[j] = 0x7C0U; + if ((I & 0x7FFFFF) != 0) + { + Result[j] = 0x7FFU; + } + else if (Sign) + { + // -INF is clamped to 0 since 3PK is positive only + Result[j] = 0; + } + } + else if (Sign || I < 0x35800000) + { + // 3PK is positive only, so clamp to zero + Result[j] = 0; + } + else if (I > 0x477E0000U) + { + // The number is too large to be represented as a float11, set to max + Result[j] = 0x7BFU; + } + else + { + if (I < 0x38800000U) + { + // The number is too small to be represented as a normalized float11 + // Convert it to a denormalized value. + uint32_t Shift = 113U - (I >> 23U); + I = (0x800000U | (I & 0x7FFFFFU)) >> Shift; + } + else + { + // Rebias the exponent to represent the value as a normalized float11 + I += 0xC8000000U; + } + + Result[j] = ((I + 0xFFFFU + ((I >> 17U) & 1U)) >> 17U) & 0x7ffU; + } + } + + // Z Channel (5-bit exponent, 5-bit mantissa) + uint32_t Sign = IValue[2] & 0x80000000; + uint32_t I = IValue[2] & 0x7FFFFFFF; + + if ((I & 0x7F800000) == 0x7F800000) + { + // INF or NAN + Result[2] = 0x3E0U; + if (I & 0x7FFFFF) + { + Result[2] = 0x3FFU; + } + else if (Sign || I < 0x36000000) + { + // -INF is clamped to 0 since 3PK is positive only + Result[2] = 0; + } + } + else if (Sign) + { + // 3PK is positive only, so clamp to zero + Result[2] = 0; + } + else if (I > 0x477C0000U) + { + // The number is too large to be represented as a float10, set to max + Result[2] = 0x3DFU; + } + else + { + if (I < 0x38800000U) + { + // The number is too small to be represented as a normalized float10 + // Convert it to a denormalized value. + uint32_t Shift = 113U - (I >> 23U); + I = (0x800000U | (I & 0x7FFFFFU)) >> Shift; + } + else + { + // Rebias the exponent to represent the value as a normalized float10 + I += 0xC8000000U; + } + + Result[2] = ((I + 0x1FFFFU + ((I >> 18U) & 1U)) >> 18U) & 0x3ffU; + } + + // Pack Result into memory + pDestination->v = (Result[0] & 0x7ff) + | ((Result[1] & 0x7ff) << 11) + | ((Result[2] & 0x3ff) << 22); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreFloat3SE +( + XMFLOAT3SE* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + +#if defined(_XM_NO_INTRINSICS_) + + XMFLOAT3A tmp; + XMStoreFloat3A(&tmp, V); + + static constexpr float maxf9 = float(0x1FF << 7); + static constexpr float minf9 = float(1.f / (1 << 16)); + + float x = (tmp.x >= 0.f) ? ((tmp.x > maxf9) ? maxf9 : tmp.x) : 0.f; + float y = (tmp.y >= 0.f) ? ((tmp.y > maxf9) ? maxf9 : tmp.y) : 0.f; + float z = (tmp.z >= 0.f) ? ((tmp.z > maxf9) ? maxf9 : tmp.z) : 0.f; + + const float max_xy = (x > y) ? x : y; + const float max_xyz = (max_xy > z) ? max_xy : z; + + const float maxColor = (max_xyz > minf9) ? max_xyz : minf9; + + union { float f; int32_t i; } fi; + fi.f = maxColor; + fi.i += 0x00004000; // round up leaving 9 bits in fraction (including assumed 1) + + auto exp = static_cast(fi.i) >> 23; + pDestination->e = exp - 0x6f; + + fi.i = static_cast(0x83000000 - (exp << 23)); + float ScaleR = fi.f; + + pDestination->xm = static_cast(MathInternal::round_to_nearest(x * ScaleR)); + pDestination->ym = static_cast(MathInternal::round_to_nearest(y * ScaleR)); + pDestination->zm = static_cast(MathInternal::round_to_nearest(z * ScaleR)); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + + static const XMVECTORF32 MaxFloat9 = { { { float(0x1FF << 7), float(0x1FF << 7), float(0x1FF << 7), float(0x1FF << 7) } } }; + static constexpr float minf9 = float(1.f / (1 << 16)); + + // Clamp to [0, maxf9] then zero w lane + float32x4_t clamped = vminq_f32(vmaxq_f32(V, vdupq_n_f32(0)), MaxFloat9); + clamped = vsetq_lane_f32(0.0f, clamped, 3); + + // Horizontal max of xyz for shared exponent +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + float maxVal = vmaxvq_f32(clamped); +#else + float32x2_t vlow = vget_low_f32(clamped); + float32x2_t vhigh = vget_high_f32(clamped); + float32x2_t maxPair = vpmax_f32(vlow, vhigh); + maxPair = vpmax_f32(maxPair, maxPair); + float maxVal = vget_lane_f32(maxPair, 0); +#endif + + if (maxVal < minf9) maxVal = minf9; + + // Compute shared exponent (inherently scalar) + union { float f; int32_t i; } fi; + fi.f = maxVal; + fi.i += 0x00004000; // round up leaving 9 bits in fraction (including assumed 1) + + auto exp = static_cast(fi.i) >> 23; + fi.i = static_cast(0x83000000 - (exp << 23)); + + // Scale all channels and convert to integer + float32x4_t scaled = vmulq_n_f32(clamped, fi.f); +#if defined(_M_ARM64) || defined(_M_HYBRID_X86_ARM64) || defined(_M_ARM64EC) || __aarch64__ + uint32x4_t ints = vcvtnq_u32_f32(scaled); +#else + scaled = vaddq_f32(scaled, vdupq_n_f32(0.5f)); + uint32x4_t ints = vcvtq_u32_f32(scaled); +#endif + + // Extract and pack into bitfields + pDestination->xm = vgetq_lane_u32(ints, 0) & 0x1FF; + pDestination->ym = vgetq_lane_u32(ints, 1) & 0x1FF; + pDestination->zm = vgetq_lane_u32(ints, 2) & 0x1FF; + pDestination->e = exp - 0x6f; + +#elif defined(_XM_SSE_INTRINSICS_) + + static const XMVECTORF32 MaxFloat9 = { { { float(0x1FF << 7), float(0x1FF << 7), float(0x1FF << 7), float(0x1FF << 7) } } }; + static constexpr float minf9 = float(1.f / (1 << 16)); + + // Clamp to [0, maxf9] then mask w to zero + __m128 clamped = _mm_min_ps(_mm_max_ps(V, _mm_setzero_ps()), MaxFloat9); + clamped = _mm_and_ps(clamped, g_XMMask3); + + // Horizontal max of xyz for shared exponent + __m128 maxV = clamped; + __m128 temp = XM_PERMUTE_PS(maxV, _MM_SHUFFLE(1, 1, 1, 1)); + maxV = _mm_max_ps(maxV, temp); + temp = XM_PERMUTE_PS(clamped, _MM_SHUFFLE(2, 2, 2, 2)); + maxV = _mm_max_ps(maxV, temp); + + // Ensure minimum threshold + maxV = _mm_max_ss(maxV, _mm_set_ss(minf9)); + + // Compute shared exponent (inherently scalar) + union { float f; int32_t i; } fi; + _mm_store_ss(&fi.f, maxV); + fi.i += 0x00004000; // round up leaving 9 bits in fraction (including assumed 1) + + auto exp = static_cast(fi.i) >> 23; + fi.i = static_cast(0x83000000 - (exp << 23)); + + // Scale all channels and round to nearest integer + __m128 scaled = _mm_mul_ps(clamped, _mm_set1_ps(fi.f)); + __m128i ints = _mm_cvtps_epi32(scaled); + + // Extract and pack into bitfields + XM_ALIGNED_DATA(16) uint32_t ivals[4]; + _mm_store_si128(reinterpret_cast<__m128i*>(ivals), ints); + + pDestination->xm = ivals[0] & 0x1FF; + pDestination->ym = ivals[1] & 0x1FF; + pDestination->zm = ivals[2] & 0x1FF; + pDestination->e = exp - 0x6f; +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreHalf4 +( + XMHALF4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_) + __m128i V1 = _mm_cvtps_ph(V, _MM_FROUND_TO_NEAREST_INT); + _mm_storel_epi64(reinterpret_cast<__m128i*>(pDestination), V1); +#else + XMFLOAT4A t; + XMStoreFloat4A(&t, V); + + pDestination->x = XMConvertFloatToHalf(t.x); + pDestination->y = XMConvertFloatToHalf(t.y); + pDestination->z = XMConvertFloatToHalf(t.z); + pDestination->w = XMConvertFloatToHalf(t.w); +#endif // !_XM_F16C_INTRINSICS_ +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreShortN4 +( + XMSHORTN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_XMNegativeOne.v, g_XMOne.v); + N = XMVectorMultiply(N, g_ShortMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(-1.f)); + vResult = vminq_f32(vResult, vdupq_n_f32(1.0f)); + vResult = vmulq_n_f32(vResult, 32767.0f); + int16x4_t vInt = vmovn_s32(vcvtq_s32_f32(vResult)); + vst1_s16(reinterpret_cast(pDestination), vInt); +#elif defined(_XM_SSE_INTRINSICS_) + XMVECTOR vResult = _mm_max_ps(V, g_XMNegativeOne); + vResult = _mm_min_ps(vResult, g_XMOne); + vResult = _mm_mul_ps(vResult, g_ShortMax); + __m128i vResulti = _mm_cvtps_epi32(vResult); + vResulti = _mm_packs_epi32(vResulti, vResulti); + _mm_store_sd(reinterpret_cast(&pDestination->x), _mm_castsi128_pd(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreShort4 +( + XMSHORT4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_ShortMin, g_ShortMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmaxq_f32(V, g_ShortMin); + vResult = vminq_f32(vResult, g_ShortMax); + int16x4_t vInt = vmovn_s32(vcvtq_s32_f32(vResult)); + vst1_s16(reinterpret_cast(pDestination), vInt); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_ShortMin); + vResult = _mm_min_ps(vResult, g_ShortMax); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // Pack the ints into shorts + vInt = _mm_packs_epi32(vInt, vInt); + _mm_store_sd(reinterpret_cast(&pDestination->x), _mm_castsi128_pd(vInt)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUShortN4 +( + XMUSHORTN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorSaturate(V); + N = XMVectorMultiplyAdd(N, g_UShortMax, g_XMOneHalf.v); + N = XMVectorTruncate(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0)); + vResult = vminq_f32(vResult, vdupq_n_f32(1.0f)); + vResult = vmulq_n_f32(vResult, 65535.0f); + vResult = vaddq_f32(vResult, g_XMOneHalf); + uint16x4_t vInt = vmovn_u32(vcvtq_u32_f32(vResult)); + vst1_u16(reinterpret_cast(pDestination), vInt); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_XMOne); + vResult = _mm_mul_ps(vResult, g_UShortMax); + vResult = _mm_add_ps(vResult, g_XMOneHalf); + // Convert to int + __m128i vInt = _mm_cvttps_epi32(vResult); + // Since the SSE pack instruction clamps using signed rules, + // manually extract the values to store them to memory + pDestination->x = static_cast(_mm_extract_epi16(vInt, 0)); + pDestination->y = static_cast(_mm_extract_epi16(vInt, 2)); + pDestination->z = static_cast(_mm_extract_epi16(vInt, 4)); + pDestination->w = static_cast(_mm_extract_epi16(vInt, 6)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUShort4 +( + XMUSHORT4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), g_UShortMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0)); + vResult = vminq_f32(vResult, g_UShortMax); + uint16x4_t vInt = vmovn_u32(vcvtq_u32_f32(vResult)); + vst1_u16(reinterpret_cast(pDestination), vInt); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_UShortMax); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // Since the SSE pack instruction clamps using signed rules, + // manually extract the values to store them to memory + pDestination->x = static_cast(_mm_extract_epi16(vInt, 0)); + pDestination->y = static_cast(_mm_extract_epi16(vInt, 2)); + pDestination->z = static_cast(_mm_extract_epi16(vInt, 4)); + pDestination->w = static_cast(_mm_extract_epi16(vInt, 6)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreXDecN4 +( + XMXDECN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 Min = { { { -1.0f, -1.0f, -1.0f, 0.0f } } }; + +#if defined(_XM_NO_INTRINSICS_) + + static const XMVECTORF32 Scale = { { { 511.0f, 511.0f, 511.0f, 3.0f } } }; + + XMVECTOR N = XMVectorClamp(V, Min.v, g_XMOne.v); + N = XMVectorMultiply(N, Scale.v); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | (static_cast(tmp.x) & 0x3FF)); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 511.0f, 511.0f * 1024.0f, 511.0f * 1048576.0f, 3.0f * 536870912.0f } } }; + static const XMVECTORI32 ScaleMask = { { { 0x3FF, 0x3FF << 10, 0x3FF << 20, 0x3 << 29 } } }; + float32x4_t vResult = vmaxq_f32(V, Min); + vResult = vminq_f32(vResult, vdupq_n_f32(1.0f)); + vResult = vmulq_f32(vResult, Scale); + int32x4_t vResulti = vcvtq_s32_f32(vResult); + vResulti = vandq_s32(vResulti, ScaleMask); + int32x4_t vResultw = vandq_s32(vResulti, g_XMMaskW); + vResulti = vaddq_s32(vResulti, vResultw); + // Do a horizontal or of all 4 entries + uint32x2_t vTemp = vget_low_u32(vreinterpretq_u32_s32(vResulti)); + uint32x2_t vhi = vget_high_u32(vreinterpretq_u32_s32(vResulti)); + vTemp = vorr_u32(vTemp, vhi); + vTemp = vpadd_u32(vTemp, vTemp); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 511.0f, 511.0f * 1024.0f, 511.0f * 1048576.0f, 3.0f * 536870912.0f } } }; + static const XMVECTORI32 ScaleMask = { { { 0x3FF, 0x3FF << 10, 0x3FF << 20, 0x3 << 29 } } }; + XMVECTOR vResult = _mm_max_ps(V, Min); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, Scale); + // Convert to int (W is unsigned) + __m128i vResulti = _mm_cvtps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, ScaleMask); + // To fix W, add itself to shift it up to <<30 instead of <<29 + __m128i vResultw = _mm_and_si128(vResulti, g_XMMaskW); + vResulti = _mm_add_epi32(vResulti, vResultw); + // Do a horizontal or of all 4 entries + vResult = XM_PERMUTE_PS(_mm_castsi128_ps(vResulti), _MM_SHUFFLE(0, 3, 2, 1)); + vResulti = _mm_or_si128(vResulti, _mm_castps_si128(vResult)); + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 3, 2, 1)); + vResulti = _mm_or_si128(vResulti, _mm_castps_si128(vResult)); + vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 3, 2, 1)); + vResulti = _mm_or_si128(vResulti, _mm_castps_si128(vResult)); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4996) +// C4996: ignore deprecation warning +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wdeprecated-declarations" +#endif + +#ifdef __GNUC__ +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" +#endif + +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreXDec4 +( + XMXDEC4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 MinXDec4 = { { { -511.0f, -511.0f, -511.0f, 0.0f } } }; + static const XMVECTORF32 MaxXDec4 = { { { 511.0f, 511.0f, 511.0f, 3.0f } } }; + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, MinXDec4, MaxXDec4); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | ((static_cast(tmp.x) & 0x3FF))); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 ScaleXDec4 = { { { 1.0f, 1024.0f / 2.0f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f / 2.0f } } }; + static const XMVECTORI32 MaskXDec4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + float32x4_t vResult = vmaxq_f32(V, MinXDec4); + vResult = vminq_f32(vResult, MaxXDec4); + vResult = vmulq_f32(vResult, ScaleXDec4); + int32x4_t vResulti = vcvtq_s32_f32(vResult); + vResulti = vandq_s32(vResulti, MaskXDec4); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vreinterpretq_u32_s32(vResulti)); + uint32x2_t vTemp2 = vget_high_u32(vreinterpretq_u32_s32(vResulti)); + vTemp = vorr_u32(vTemp, vTemp2); + // Perform a single bit left shift on y|w + vTemp2 = vdup_lane_u32(vTemp, 1); + vTemp2 = vadd_u32(vTemp2, vTemp2); + vTemp = vorr_u32(vTemp, vTemp2); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleXDec4 = { { { 1.0f, 1024.0f / 2.0f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f / 2.0f } } }; + static const XMVECTORI32 MaskXDec4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, MinXDec4); + vResult = _mm_min_ps(vResult, MaxXDec4); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleXDec4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskXDec4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // Perform a single bit left shift on y|w + vResulti2 = _mm_add_epi32(vResulti2, vResulti2); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +#ifdef __GNUC__ +#pragma GCC diagnostic pop +#endif +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUDecN4 +( + XMUDECN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + static const XMVECTORF32 Scale = { { { 1023.0f, 1023.0f, 1023.0f, 3.0f } } }; + + XMVECTOR N = XMVectorSaturate(V); + N = XMVectorMultiply(N, Scale.v); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | ((static_cast(tmp.x) & 0x3FF))); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 ScaleUDecN4 = { { { 1023.0f, 1023.0f * 1024.0f * 0.5f, 1023.0f * 1024.0f * 1024.0f, 3.0f * 1024.0f * 1024.0f * 1024.0f * 0.5f } } }; + static const XMVECTORI32 MaskUDecN4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0.f)); + vResult = vminq_f32(vResult, vdupq_n_f32(1.f)); + vResult = vmulq_f32(vResult, ScaleUDecN4); + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + vResulti = vandq_u32(vResulti, MaskUDecN4); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vResulti); + uint32x2_t vTemp2 = vget_high_u32(vResulti); + vTemp = vorr_u32(vTemp, vTemp2); + // Perform a single bit left shift on y|w + vTemp2 = vdup_lane_u32(vTemp, 1); + vTemp2 = vadd_u32(vTemp2, vTemp2); + vTemp = vorr_u32(vTemp, vTemp2); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleUDecN4 = { { { 1023.0f, 1023.0f * 1024.0f * 0.5f, 1023.0f * 1024.0f * 1024.0f, 3.0f * 1024.0f * 1024.0f * 1024.0f * 0.5f } } }; + static const XMVECTORI32 MaskUDecN4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleUDecN4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskUDecN4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // Perform a left shift by one bit on y|w + vResulti2 = _mm_add_epi32(vResulti2, vResulti2); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUDecN4_XR +( + XMUDECN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 Scale = { { { 510.0f, 510.0f, 510.0f, 3.0f } } }; + static const XMVECTORF32 Bias = { { { 384.0f, 384.0f, 384.0f, 0.0f } } }; + static const XMVECTORF32 C = { { { 1023.f, 1023.f, 1023.f, 3.f } } }; + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorMultiplyAdd(V, Scale, Bias); + N = XMVectorClamp(N, g_XMZero, C); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | ((static_cast(tmp.x) & 0x3FF))); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Shift = { { { 1.0f, 1024.0f * 0.5f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f * 0.5f } } }; + static const XMVECTORU32 MaskUDecN4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + float32x4_t vResult = vmlaq_f32(Bias, V, Scale); + vResult = vmaxq_f32(vResult, vdupq_n_f32(0.f)); + vResult = vminq_f32(vResult, C); + vResult = vmulq_f32(vResult, Shift); + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + vResulti = vandq_u32(vResulti, MaskUDecN4); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vResulti); + uint32x2_t vTemp2 = vget_high_u32(vResulti); + vTemp = vorr_u32(vTemp, vTemp2); + // Perform a single bit left shift on y|w + vTemp2 = vdup_lane_u32(vTemp, 1); + vTemp2 = vadd_u32(vTemp2, vTemp2); + vTemp = vorr_u32(vTemp, vTemp2); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 Shift = { { { 1.0f, 1024.0f * 0.5f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f * 0.5f } } }; + static const XMVECTORU32 MaskUDecN4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + // Scale & bias + XMVECTOR vResult = XM_FMADD_PS(V, Scale, Bias); + // Clamp to bounds + vResult = _mm_max_ps(vResult, g_XMZero); + vResult = _mm_min_ps(vResult, C); + // Scale by shift values + vResult = _mm_mul_ps(vResult, Shift); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskUDecN4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // Perform a left shift by one bit on y|w + vResulti2 = _mm_add_epi32(vResulti2, vResulti2); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUDec4 +( + XMUDEC4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 MaxUDec4 = { { { 1023.0f, 1023.0f, 1023.0f, 3.0f } } }; + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), MaxUDec4); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | ((static_cast(tmp.x) & 0x3FF))); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 ScaleUDec4 = { { { 1.0f, 1024.0f / 2.0f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f / 2.0f } } }; + static const XMVECTORI32 MaskUDec4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0.f)); + vResult = vminq_f32(vResult, MaxUDec4); + vResult = vmulq_f32(vResult, ScaleUDec4); + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + vResulti = vandq_u32(vResulti, MaskUDec4); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vResulti); + uint32x2_t vTemp2 = vget_high_u32(vResulti); + vTemp = vorr_u32(vTemp, vTemp2); + // Perform a single bit left shift on y|w + vTemp2 = vdup_lane_u32(vTemp, 1); + vTemp2 = vadd_u32(vTemp2, vTemp2); + vTemp = vorr_u32(vTemp, vTemp2); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleUDec4 = { { { 1.0f, 1024.0f / 2.0f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f / 2.0f } } }; + static const XMVECTORI32 MaskUDec4 = { { { 0x3FF, 0x3FF << (10 - 1), 0x3FF << 20, 0x3 << (30 - 1) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, MaxUDec4); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleUDec4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskUDec4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // Perform a left shift by one bit on y|w + vResulti2 = _mm_add_epi32(vResulti2, vResulti2); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4996) +// C4996: ignore deprecation warning +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wdeprecated-declarations" +#endif + +#ifdef __GNUC__ +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" +#endif + +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreDecN4 +( + XMDECN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + static const XMVECTORF32 Scale = { { { 511.0f, 511.0f, 511.0f, 1.0f } } }; + + XMVECTOR N = XMVectorClamp(V, g_XMNegativeOne.v, g_XMOne.v); + N = XMVectorMultiply(N, Scale.v); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | ((static_cast(tmp.x) & 0x3FF))); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 ScaleDecN4 = { { { 511.0f, 511.0f * 1024.0f, 511.0f * 1024.0f * 1024.0f, 1.0f * 1024.0f * 1024.0f * 1024.0f } } }; + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(-1.f)); + vResult = vminq_f32(vResult, vdupq_n_f32(1.f)); + vResult = vmulq_f32(vResult, ScaleDecN4); + int32x4_t vResulti = vcvtq_s32_f32(vResult); + vResulti = vandq_s32(vResulti, g_XMMaskDec4); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vreinterpretq_u32_s32(vResulti)); + uint32x2_t vhi = vget_high_u32(vreinterpretq_u32_s32(vResulti)); + vTemp = vorr_u32(vTemp, vhi); + vTemp = vpadd_u32(vTemp, vTemp); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleDecN4 = { { { 511.0f, 511.0f * 1024.0f, 511.0f * 1024.0f * 1024.0f, 1.0f * 1024.0f * 1024.0f * 1024.0f } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMNegativeOne); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleDecN4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, g_XMMaskDec4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreDec4 +( + XMDEC4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 MinDec4 = { { { -511.0f, -511.0f, -511.0f, -1.0f } } }; + static const XMVECTORF32 MaxDec4 = { { { 511.0f, 511.0f, 511.0f, 1.0f } } }; + +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, MinDec4, MaxDec4); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + (static_cast(tmp.w) << 30) + | ((static_cast(tmp.z) & 0x3FF) << 20) + | ((static_cast(tmp.y) & 0x3FF) << 10) + | ((static_cast(tmp.x) & 0x3FF))); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 ScaleDec4 = { { { 1.0f, 1024.0f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f } } }; + float32x4_t vResult = vmaxq_f32(V, MinDec4); + vResult = vminq_f32(vResult, MaxDec4); + vResult = vmulq_f32(vResult, ScaleDec4); + int32x4_t vResulti = vcvtq_s32_f32(vResult); + vResulti = vandq_s32(vResulti, g_XMMaskDec4); + // Do a horizontal or of all 4 entries + uint32x2_t vTemp = vget_low_u32(vreinterpretq_u32_s32(vResulti)); + uint32x2_t vhi = vget_high_u32(vreinterpretq_u32_s32(vResulti)); + vTemp = vorr_u32(vTemp, vhi); + vTemp = vpadd_u32(vTemp, vTemp); + vst1_lane_u32(&pDestination->v, vTemp, 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleDec4 = { { { 1.0f, 1024.0f, 1024.0f * 1024.0f, 1024.0f * 1024.0f * 1024.0f } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, MinDec4); + vResult = _mm_min_ps(vResult, MaxDec4); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleDec4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, g_XMMaskDec4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +#ifdef __GNUC__ +#pragma GCC diagnostic pop +#endif +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUByteN4 +( + XMUBYTEN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorSaturate(V); + N = XMVectorMultiply(N, g_UByteMax); + N = XMVectorTruncate(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 255.0f); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + uint8x8_t vInt8 = vqmovn_u16(vcombine_u16(vInt16, vInt16)); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_u8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleUByteN4 = { { { 255.0f, 255.0f * 256.0f * 0.5f, 255.0f * 256.0f * 256.0f, 255.0f * 256.0f * 256.0f * 256.0f * 0.5f } } }; + static const XMVECTORI32 MaskUByteN4 = { { { 0xFF, 0xFF << (8 - 1), 0xFF << 16, 0xFF << (24 - 1) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleUByteN4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskUByteN4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // Perform a single bit left shift to fix y|w + vResulti2 = _mm_add_epi32(vResulti2, vResulti2); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUByte4 +( + XMUBYTE4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), g_UByteMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(0)); + R = vminq_f32(R, vdupq_n_f32(255.0f)); + uint32x4_t vInt32 = vcvtq_u32_f32(R); + uint16x4_t vInt16 = vqmovn_u32(vInt32); + uint8x8_t vInt8 = vqmovn_u16(vcombine_u16(vInt16, vInt16)); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_u8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleUByte4 = { { { 1.0f, 256.0f * 0.5f, 256.0f * 256.0f, 256.0f * 256.0f * 256.0f * 0.5f } } }; + static const XMVECTORI32 MaskUByte4 = { { { 0xFF, 0xFF << (8 - 1), 0xFF << 16, 0xFF << (24 - 1) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, g_UByteMax); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleUByte4); + // Convert to int by rounding + __m128i vResulti = _mm_cvtps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskUByte4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // Perform a single bit left shift to fix y|w + vResulti2 = _mm_add_epi32(vResulti2, vResulti2); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreByteN4 +( + XMBYTEN4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_XMNegativeOne.v, g_XMOne.v); + N = XMVectorMultiply(N, g_ByteMax); + N = XMVectorTruncate(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(-1.f)); + R = vminq_f32(R, vdupq_n_f32(1.0f)); + R = vmulq_n_f32(R, 127.0f); + int32x4_t vInt32 = vcvtq_s32_f32(R); + int16x4_t vInt16 = vqmovn_s32(vInt32); + int8x8_t vInt8 = vqmovn_s16(vcombine_s16(vInt16, vInt16)); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_s8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleByteN4 = { { { 127.0f, 127.0f * 256.0f, 127.0f * 256.0f * 256.0f, 127.0f * 256.0f * 256.0f * 256.0f } } }; + static const XMVECTORI32 MaskByteN4 = { { { 0xFF, 0xFF << 8, 0xFF << 16, static_cast(0xFF000000) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_XMNegativeOne); + vResult = _mm_min_ps(vResult, g_XMOne); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleByteN4); + // Convert to int + __m128i vResulti = _mm_cvttps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskByteN4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreByte4 +( + XMBYTE4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, g_ByteMin, g_ByteMax); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->x = static_cast(tmp.x); + pDestination->y = static_cast(tmp.y); + pDestination->z = static_cast(tmp.z); + pDestination->w = static_cast(tmp.w); + +#elif defined(_XM_ARM_NEON_INTRINSICS_) + float32x4_t R = vmaxq_f32(V, vdupq_n_f32(-127.f)); + R = vminq_f32(R, vdupq_n_f32(127.f)); + int32x4_t vInt32 = vcvtq_s32_f32(R); + int16x4_t vInt16 = vqmovn_s32(vInt32); + int8x8_t vInt8 = vqmovn_s16(vcombine_s16(vInt16, vInt16)); + vst1_lane_u32(&pDestination->v, vreinterpret_u32_s8(vInt8), 0); +#elif defined(_XM_SSE_INTRINSICS_) + static const XMVECTORF32 ScaleByte4 = { { { 1.0f, 256.0f, 256.0f * 256.0f, 256.0f * 256.0f * 256.0f } } }; + static const XMVECTORI32 MaskByte4 = { { { 0xFF, 0xFF << 8, 0xFF << 16, static_cast(0xFF000000) } } }; + // Clamp to bounds + XMVECTOR vResult = _mm_max_ps(V, g_ByteMin); + vResult = _mm_min_ps(vResult, g_ByteMax); + // Scale by multiplication + vResult = _mm_mul_ps(vResult, ScaleByte4); + // Convert to int by rounding + __m128i vResulti = _mm_cvtps_epi32(vResult); + // Mask off any fraction + vResulti = _mm_and_si128(vResulti, MaskByte4); + // Do a horizontal or of 4 entries + __m128i vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(3, 2, 3, 2)); + // x = x|z, y = y|w + vResulti = _mm_or_si128(vResulti, vResulti2); + // Move Z to the x position + vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1)); + // i = x|y|z|w + vResulti = _mm_or_si128(vResulti, vResulti2); + _mm_store_ss(reinterpret_cast(&pDestination->v), _mm_castsi128_ps(vResulti)); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreUNibble4 +( + XMUNIBBLE4* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 Max = { { { 15.0f, 15.0f, 15.0f, 15.0f } } }; +#if defined(_XM_NO_INTRINSICS_) + + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), Max.v); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + ((static_cast(tmp.w) & 0xF) << 12) + | ((static_cast(tmp.z) & 0xF) << 8) + | ((static_cast(tmp.y) & 0xF) << 4) + | (static_cast(tmp.x) & 0xF)); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f, 16.f, 16.f * 16.f, 16.f * 16.f * 16.f } } }; + static const XMVECTORU32 Mask = { { { 0xF, 0xF << 4, 0xF << 8, 0xF << 12 } } }; + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0)); + vResult = vminq_f32(vResult, Max); + vResult = vmulq_f32(vResult, Scale); + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + vResulti = vandq_u32(vResulti, Mask); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vResulti); + uint32x2_t vhi = vget_high_u32(vResulti); + vTemp = vorr_u32(vTemp, vhi); + vTemp = vpadd_u32(vTemp, vTemp); + vst1_lane_u16(&pDestination->v, vreinterpret_u16_u32(vTemp), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, Max); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + auto z = static_cast(_mm_extract_epi16(vInt, 4)); + auto w = static_cast(_mm_extract_epi16(vInt, 6)); + pDestination->v = static_cast( + ((static_cast(w) & 0xF) << 12) + | ((static_cast(z) & 0xF) << 8) + | ((static_cast(y) & 0xF) << 4) + | ((static_cast(x) & 0xF))); +#endif +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline void XM_CALLCONV XMStoreU555 +( + XMU555* pDestination, + FXMVECTOR V +) noexcept +{ + assert(pDestination); + static const XMVECTORF32 Max = { { { 31.0f, 31.0f, 31.0f, 1.0f } } }; + +#if defined(_XM_NO_INTRINSICS_) + XMVECTOR N = XMVectorClamp(V, XMVectorZero(), Max.v); + N = XMVectorRound(N); + + XMFLOAT4A tmp; + XMStoreFloat4A(&tmp, N); + + pDestination->v = static_cast( + ((tmp.w > 0.f) ? 0x8000 : 0) + | ((static_cast(tmp.z) & 0x1F) << 10) + | ((static_cast(tmp.y) & 0x1F) << 5) + | (static_cast(tmp.x) & 0x1F)); +#elif defined(_XM_ARM_NEON_INTRINSICS_) + static const XMVECTORF32 Scale = { { { 1.0f, 32.f / 2.f, 32.f * 32.f, 32.f * 32.f * 32.f / 2.f } } }; + static const XMVECTORU32 Mask = { { { 0x1F, 0x1F << (5 - 1), 0x1F << 10, 0x1 << (15 - 1) } } }; + float32x4_t vResult = vmaxq_f32(V, vdupq_n_f32(0)); + vResult = vminq_f32(vResult, Max); + vResult = vmulq_f32(vResult, Scale); + uint32x4_t vResulti = vcvtq_u32_f32(vResult); + vResulti = vandq_u32(vResulti, Mask); + // Do a horizontal or of 4 entries + uint32x2_t vTemp = vget_low_u32(vResulti); + uint32x2_t vTemp2 = vget_high_u32(vResulti); + vTemp = vorr_u32(vTemp, vTemp2); + // Perform a single bit left shift on y|w + vTemp2 = vdup_lane_u32(vTemp, 1); + vTemp2 = vadd_u32(vTemp2, vTemp2); + vTemp = vorr_u32(vTemp, vTemp2); + vst1_lane_u16(&pDestination->v, vreinterpret_u16_u32(vTemp), 0); +#elif defined(_XM_SSE_INTRINSICS_) + // Bounds check + XMVECTOR vResult = _mm_max_ps(V, g_XMZero); + vResult = _mm_min_ps(vResult, Max); + // Convert to int with rounding + __m128i vInt = _mm_cvtps_epi32(vResult); + // No SSE operations will write to 16-bit values, so we have to extract them manually + auto x = static_cast(_mm_extract_epi16(vInt, 0)); + auto y = static_cast(_mm_extract_epi16(vInt, 2)); + auto z = static_cast(_mm_extract_epi16(vInt, 4)); + auto w = static_cast(_mm_extract_epi16(vInt, 6)); + pDestination->v = static_cast( + (static_cast(w) ? 0x8000 : 0) + | ((static_cast(z) & 0x1F) << 10) + | ((static_cast(y) & 0x1F) << 5) + | ((static_cast(x) & 0x1F))); +#endif +} + + +/**************************************************************************** + * + * XMCOLOR operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMCOLOR::XMCOLOR +( + float _r, + float _g, + float _b, + float _a +) noexcept +{ + XMStoreColor(this, XMVectorSet(_r, _g, _b, _a)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMCOLOR::XMCOLOR(const float* pArray) noexcept +{ + XMStoreColor(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMHALF2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMHALF2::XMHALF2 +( + float _x, + float _y +) noexcept +{ + x = XMConvertFloatToHalf(_x); + y = XMConvertFloatToHalf(_y); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMHALF2::XMHALF2(const float* pArray) noexcept +{ + assert(pArray != nullptr); + x = XMConvertFloatToHalf(pArray[0]); + y = XMConvertFloatToHalf(pArray[1]); +} + +/**************************************************************************** + * + * XMSHORTN2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMSHORTN2::XMSHORTN2 +( + float _x, + float _y +) noexcept +{ + XMStoreShortN2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMSHORTN2::XMSHORTN2(const float* pArray) noexcept +{ + XMStoreShortN2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMSHORT2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMSHORT2::XMSHORT2 +( + float _x, + float _y +) noexcept +{ + XMStoreShort2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMSHORT2::XMSHORT2(const float* pArray) noexcept +{ + XMStoreShort2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUSHORTN2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUSHORTN2::XMUSHORTN2 +( + float _x, + float _y +) noexcept +{ + XMStoreUShortN2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUSHORTN2::XMUSHORTN2(const float* pArray) noexcept +{ + XMStoreUShortN2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUSHORT2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUSHORT2::XMUSHORT2 +( + float _x, + float _y +) noexcept +{ + XMStoreUShort2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUSHORT2::XMUSHORT2(const float* pArray) noexcept +{ + XMStoreUShort2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMBYTEN2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMBYTEN2::XMBYTEN2 +( + float _x, + float _y +) noexcept +{ + XMStoreByteN2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMBYTEN2::XMBYTEN2(const float* pArray) noexcept +{ + XMStoreByteN2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMBYTE2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMBYTE2::XMBYTE2 +( + float _x, + float _y +) noexcept +{ + XMStoreByte2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMBYTE2::XMBYTE2(const float* pArray) noexcept +{ + XMStoreByte2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUBYTEN2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUBYTEN2::XMUBYTEN2 +( + float _x, + float _y +) noexcept +{ + XMStoreUByteN2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUBYTEN2::XMUBYTEN2(const float* pArray) noexcept +{ + XMStoreUByteN2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUBYTE2 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUBYTE2::XMUBYTE2 +( + float _x, + float _y +) noexcept +{ + XMStoreUByte2(this, XMVectorSet(_x, _y, 0.0f, 0.0f)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUBYTE2::XMUBYTE2(const float* pArray) noexcept +{ + XMStoreUByte2(this, XMLoadFloat2(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMU565 operators + * + ****************************************************************************/ + +inline XMU565::XMU565 +( + float _x, + float _y, + float _z +) noexcept +{ + XMStoreU565(this, XMVectorSet(_x, _y, _z, 0.0f)); +} + +_Use_decl_annotations_ +inline XMU565::XMU565(const float* pArray) noexcept +{ + XMStoreU565(this, XMLoadFloat3(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMFLOAT3PK operators + * + ****************************************************************************/ + +inline XMFLOAT3PK::XMFLOAT3PK +( + float _x, + float _y, + float _z +) noexcept +{ + XMStoreFloat3PK(this, XMVectorSet(_x, _y, _z, 0.0f)); +} + +_Use_decl_annotations_ +inline XMFLOAT3PK::XMFLOAT3PK(const float* pArray) noexcept +{ + XMStoreFloat3PK(this, XMLoadFloat3(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMFLOAT3SE operators + * + ****************************************************************************/ + +inline XMFLOAT3SE::XMFLOAT3SE +( + float _x, + float _y, + float _z +) noexcept +{ + XMStoreFloat3SE(this, XMVectorSet(_x, _y, _z, 0.0f)); +} + +_Use_decl_annotations_ +inline XMFLOAT3SE::XMFLOAT3SE(const float* pArray) noexcept +{ + XMStoreFloat3SE(this, XMLoadFloat3(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMHALF4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMHALF4::XMHALF4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + x = XMConvertFloatToHalf(_x); + y = XMConvertFloatToHalf(_y); + z = XMConvertFloatToHalf(_z); + w = XMConvertFloatToHalf(_w); +} + +//------------------------------------------------------------------------------ + +_Use_decl_annotations_ +inline XMHALF4::XMHALF4(const float* pArray) noexcept +{ + XMConvertFloatToHalfStream(&x, sizeof(HALF), pArray, sizeof(float), 4); +} + +/**************************************************************************** + * + * XMSHORTN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMSHORTN4::XMSHORTN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreShortN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMSHORTN4::XMSHORTN4(const float* pArray) noexcept +{ + XMStoreShortN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMSHORT4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMSHORT4::XMSHORT4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreShort4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMSHORT4::XMSHORT4(const float* pArray) noexcept +{ + XMStoreShort4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUSHORTN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUSHORTN4::XMUSHORTN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUShortN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUSHORTN4::XMUSHORTN4(const float* pArray) noexcept +{ + XMStoreUShortN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUSHORT4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUSHORT4::XMUSHORT4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUShort4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUSHORT4::XMUSHORT4(const float* pArray) noexcept +{ + XMStoreUShort4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMXDECN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMXDECN4::XMXDECN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreXDecN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMXDECN4::XMXDECN4(const float* pArray) noexcept +{ + XMStoreXDecN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMXDEC4 operators + * + ****************************************************************************/ +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4996) +// C4996: ignore deprecation warning +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Wdeprecated-declarations" +#endif + +#ifdef __GNUC__ +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" +#endif + +//------------------------------------------------------------------------------ + +inline XMXDEC4::XMXDEC4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreXDec4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMXDEC4::XMXDEC4(const float* pArray) noexcept +{ + XMStoreXDec4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMDECN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMDECN4::XMDECN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreDecN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMDECN4::XMDECN4(const float* pArray) noexcept +{ + XMStoreDecN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMDEC4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMDEC4::XMDEC4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreDec4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMDEC4::XMDEC4(const float* pArray) noexcept +{ + XMStoreDec4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +#ifdef __GNUC__ +#pragma GCC diagnostic pop +#endif +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +/**************************************************************************** + * + * XMUDECN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUDECN4::XMUDECN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUDecN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUDECN4::XMUDECN4(const float* pArray) noexcept +{ + XMStoreUDecN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUDEC4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUDEC4::XMUDEC4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUDec4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUDEC4::XMUDEC4(const float* pArray) noexcept +{ + XMStoreUDec4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMBYTEN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMBYTEN4::XMBYTEN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreByteN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMBYTEN4::XMBYTEN4(const float* pArray) noexcept +{ + XMStoreByteN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMBYTE4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMBYTE4::XMBYTE4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreByte4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMBYTE4::XMBYTE4(const float* pArray) noexcept +{ + XMStoreByte4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUBYTEN4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUBYTEN4::XMUBYTEN4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUByteN4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUBYTEN4::XMUBYTEN4(const float* pArray) noexcept +{ + XMStoreUByteN4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUBYTE4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUBYTE4::XMUBYTE4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUByte4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUBYTE4::XMUBYTE4(const float* pArray) noexcept +{ + XMStoreUByte4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMUNIBBLE4 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMUNIBBLE4::XMUNIBBLE4 +( + float _x, + float _y, + float _z, + float _w +) noexcept +{ + XMStoreUNibble4(this, XMVectorSet(_x, _y, _z, _w)); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMUNIBBLE4::XMUNIBBLE4(const float* pArray) noexcept +{ + XMStoreUNibble4(this, XMLoadFloat4(reinterpret_cast(pArray))); +} + +/**************************************************************************** + * + * XMU555 operators + * + ****************************************************************************/ + +//------------------------------------------------------------------------------ + +inline XMU555::XMU555 +( + float _x, + float _y, + float _z, + bool _w +) noexcept +{ + XMStoreU555(this, XMVectorSet(_x, _y, _z, ((_w) ? 1.0f : 0.0f))); +} + +//------------------------------------------------------------------------------ +_Use_decl_annotations_ +inline XMU555::XMU555 +( + const float* pArray, + bool _w +) noexcept +{ + XMVECTOR V = XMLoadFloat3(reinterpret_cast(pArray)); + XMStoreU555(this, XMVectorSetW(V, ((_w) ? 1.0f : 0.0f))); +} + diff --git a/Minecraft.Client/Windows64/Libs/4JLibs/inc/4J_Render.h b/Minecraft.Client/Windows64/Libs/4JLibs/inc/4J_Render.h index 7fe70415..f17edd12 100644 --- a/Minecraft.Client/Windows64/Libs/4JLibs/inc/4J_Render.h +++ b/Minecraft.Client/Windows64/Libs/4JLibs/inc/4J_Render.h @@ -40,11 +40,11 @@ public: } }; -using D3DXIMAGE_INFO = struct +typedef struct { int Width; int Height; -}; +} D3DXIMAGE_INFO; typedef struct _XSOCIAL_PREVIEWIMAGE { @@ -93,7 +93,7 @@ public: void EndConditionalRendering(); // Vertex data handling - using eVertexType = enum + enum eVertexType { VERTEX_TYPE_PF3_TF2_CB4_NB4_XW1, // Position 3 x float, texture 2 x float, colour 4 x byte, normal 4 x byte, padding 1 DWORD @@ -105,7 +105,7 @@ public: }; // Pixel shader - using ePixelShaderType = enum + enum ePixelShaderType { PIXEL_SHADER_TYPE_STANDARD, PIXEL_SHADER_TYPE_PROJECTION, @@ -113,7 +113,7 @@ public: PIXEL_SHADER_COUNT }; - using eViewportType = enum + enum eViewportType { VIEWPORT_TYPE_FULLSCREEN, VIEWPORT_TYPE_SPLIT_TOP, @@ -126,7 +126,7 @@ public: VIEWPORT_TYPE_QUADRANT_BOTTOM_RIGHT, }; - using ePrimitiveType = enum + enum ePrimitiveType { PRIMITIVE_TYPE_TRIANGLE_LIST, PRIMITIVE_TYPE_TRIANGLE_STRIP, @@ -155,7 +155,7 @@ public: void CBuffDeferredModeStart(); void CBuffDeferredModeEnd(); - using eTextureFormat = enum + enum eTextureFormat { TEXTURE_FORMAT_RxGyBzAw, // Normal 32-bit RGBA texture, 8 bits per component /* Don't think these are all directly available on D3D 11 - leaving for now diff --git a/Minecraft.Client/Windows64/Libs/Iggy/inc/rrCore.h b/Minecraft.Client/Windows64/Libs/Iggy/inc/rrCore.h index 17ebee3a..a500775e 100644 --- a/Minecraft.Client/Windows64/Libs/Iggy/inc/rrCore.h +++ b/Minecraft.Client/Windows64/Libs/Iggy/inc/rrCore.h @@ -1641,13 +1641,17 @@ RADDEFSTART void RR_BREAK( void ); #pragma aux RR_BREAK = "int 0x3"; + #elif defined(__RADWIN__) && defined(__GNUC__) + + #define RR_BREAK() __asm__ volatile("int3") + #elif defined(__RADWIN__) && defined(_MSC_VER) && _MSC_VER >= 1300 #define RR_BREAK __debugbreak #else - #define RR_BREAK() RAD_STATEMENT_WRAPPER( __asm {int 3} ) + #define RR_BREAK() __builtin_trap() #endif diff --git a/Minecraft.Client/Windows64/Libs/RmlUi/.appveyor.yml b/Minecraft.Client/Windows64/Libs/RmlUi/.appveyor.yml new file mode 100644 index 00000000..5b2afbb8 --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/RmlUi/.appveyor.yml @@ -0,0 +1,27 @@ +version: build.{build} + +image: Visual Studio 2017 + +environment: + FREETYPE_VER: 2.11.1 + VS_GENERATOR: Visual Studio 15 2017 Win64 + +install: +- cmd: |- + cd Dependencies + appveyor DownloadFile https://github.com/ubawurinna/freetype-windows-binaries/archive/refs/tags/v%FREETYPE_VER%.zip + unzip -o v%FREETYPE_VER%.zip + mv "freetype-windows-binaries-%FREETYPE_VER%/include" include + mv "freetype-windows-binaries-%FREETYPE_VER%/release dll/win64" lib + cd .. + + cmake -B Build -G "%VS_GENERATOR%" -DBUILD_SHARED_LIBS=ON -DRMLUI_WARNINGS_AS_ERRORS=ON + +build_script: +- cmd: |- + cmake --build Build --config Debug -- /verbosity:minimal /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" + cmake --build Build --config Release -- /verbosity:minimal /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll" + +after_build: + - cmd: |- + ls Build/*/*.dll -s -h -X diff --git a/Minecraft.Client/Windows64/Libs/RmlUi/.clang-format b/Minecraft.Client/Windows64/Libs/RmlUi/.clang-format new file mode 100644 index 00000000..a696f5ea --- /dev/null +++ b/Minecraft.Client/Windows64/Libs/RmlUi/.clang-format @@ -0,0 +1,69 @@ +AccessModifierOffset: -4 +AllowAllArgumentsOnNextLine: false +AllowAllParametersOfDeclarationOnNextLine: false +AlignAfterOpenBracket: DontAlign +AlignConsecutiveAssignments: false +AlignConsecutiveDeclarations: false +AlignEscapedNewlines: Left +AlignOperands: false +AlignTrailingComments: + Kind: Always + OverEmptyLines: 1 +AllowShortBlocksOnASingleLine: Empty +AllowShortCaseLabelsOnASingleLine: true +AllowShortFunctionsOnASingleLine: Inline +AlwaysBreakTemplateDeclarations: Yes +BinPackArguments: true +BinPackParameters: true +BreakBeforeBraces: Custom +BraceWrapping: + AfterCaseLabel: true + AfterControlStatement: true + AfterFunction: true + BeforeElse: true + SplitEmptyFunction: false + SplitEmptyRecord: false + SplitEmptyNamespace: false +BreakConstructorInitializers: AfterColon +ColumnLimit: 150 +CompactNamespaces: false +Cpp11BracedListStyle: true +FixNamespaceComments: true +IncludeBlocks: Merge +IncludeCategories: + - Regex: '^