16bit processing

.gitea/workflows/ci.yml

@@ -31,14 +31,6 @@ jobs:
       - name: Configure Debug
         shell: powershell
         run: |
-          $gpudirectDir = "${{ vars.GPUDIRECT_DIR }}"
-          if ([string]::IsNullOrWhiteSpace($gpudirectDir)) {
-            $gpudirectDir = $env:GPUDIRECT_DIR
-          }
-          if ([string]::IsNullOrWhiteSpace($gpudirectDir)) {
-            $gpudirectDir = Join-Path $PWD "3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect"
-          }
-
           $slangRoot = "${{ vars.SLANG_ROOT }}"
           if ([string]::IsNullOrWhiteSpace($slangRoot)) {
             $slangRoot = $env:SLANG_ROOT
@@ -48,9 +40,6 @@ jobs:
           }
 
           $requiredFiles = @(
-            (Join-Path $gpudirectDir "include\DVPAPI.h"),
-            (Join-Path $gpudirectDir "lib\x64\dvp.lib"),
-            (Join-Path $gpudirectDir "bin\x64\dvp.dll"),
             (Join-Path $slangRoot "bin\slangc.exe"),
             (Join-Path $slangRoot "bin\slang-compiler.dll"),
             (Join-Path $slangRoot "bin\slang-glslang.dll"),
@@ -59,13 +48,12 @@ jobs:
 
           $missingFiles = @($requiredFiles | Where-Object { -not (Test-Path -LiteralPath $_) })
           if ($missingFiles.Count -gt 0) {
-            Write-Error "Missing native third-party dependencies. Set Gitea repository variables GPUDIRECT_DIR and SLANG_ROOT, or pre-populate the repo-local 3rdParty folder on the Windows runner. Missing: $($missingFiles -join ', ')"
+            Write-Error "Missing native third-party dependencies. Set Gitea repository variable SLANG_ROOT, or pre-populate the repo-local 3rdParty folder on the Windows runner. Missing: $($missingFiles -join ', ')"
             exit 1
           }
 
-          Write-Host "Using GPUDIRECT_DIR=$gpudirectDir"
           Write-Host "Using SLANG_ROOT=$slangRoot"
-          cmake --preset vs2022-x64-debug -DGPUDIRECT_DIR="$gpudirectDir" -DSLANG_ROOT="$slangRoot"
+          cmake --preset vs2022-x64-debug -DSLANG_ROOT="$slangRoot"
 
       - name: Build Debug
         shell: powershell
@@ -122,14 +110,6 @@ jobs:
       - name: Configure Release
         shell: powershell
         run: |
-          $gpudirectDir = "${{ vars.GPUDIRECT_DIR }}"
-          if ([string]::IsNullOrWhiteSpace($gpudirectDir)) {
-            $gpudirectDir = $env:GPUDIRECT_DIR
-          }
-          if ([string]::IsNullOrWhiteSpace($gpudirectDir)) {
-            $gpudirectDir = Join-Path $PWD "3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect"
-          }
-
           $slangRoot = "${{ vars.SLANG_ROOT }}"
           if ([string]::IsNullOrWhiteSpace($slangRoot)) {
             $slangRoot = $env:SLANG_ROOT
@@ -139,9 +119,6 @@ jobs:
           }
 
           $requiredFiles = @(
-            (Join-Path $gpudirectDir "include\DVPAPI.h"),
-            (Join-Path $gpudirectDir "lib\x64\dvp.lib"),
-            (Join-Path $gpudirectDir "bin\x64\dvp.dll"),
             (Join-Path $slangRoot "bin\slangc.exe"),
             (Join-Path $slangRoot "bin\slang-compiler.dll"),
             (Join-Path $slangRoot "bin\slang-glslang.dll"),
@@ -150,13 +127,12 @@ jobs:
 
           $missingFiles = @($requiredFiles | Where-Object { -not (Test-Path -LiteralPath $_) })
           if ($missingFiles.Count -gt 0) {
-            Write-Error "Missing native third-party dependencies. Set Gitea repository variables GPUDIRECT_DIR and SLANG_ROOT, or pre-populate the repo-local 3rdParty folder on the Windows runner. Missing: $($missingFiles -join ', ')"
+            Write-Error "Missing native third-party dependencies. Set Gitea repository variable SLANG_ROOT, or pre-populate the repo-local 3rdParty folder on the Windows runner. Missing: $($missingFiles -join ', ')"
             exit 1
           }
 
-          Write-Host "Using GPUDIRECT_DIR=$gpudirectDir"
           Write-Host "Using SLANG_ROOT=$slangRoot"
-          cmake --preset vs2022-x64-release -DGPUDIRECT_DIR="$gpudirectDir" -DSLANG_ROOT="$slangRoot"
+          cmake --preset vs2022-x64-release -DSLANG_ROOT="$slangRoot"
 
       - name: Build Release
         shell: powershell

CMakeLists.txt

@@ -7,17 +7,12 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 
 set(APP_DIR "${CMAKE_CURRENT_SOURCE_DIR}/apps/LoopThroughWithOpenGLCompositing")
-set(GPUDIRECT_DIR "${CMAKE_CURRENT_SOURCE_DIR}/3rdParty/Blackmagic DeckLink SDK 16.0/Win/Samples/NVIDIA_GPUDirect" CACHE PATH "Path to the NVIDIA_GPUDirect sample directory from the Blackmagic DeckLink SDK")
 set(SLANG_ROOT "${CMAKE_CURRENT_SOURCE_DIR}/3rdParty/slang-2026.8-windows-x86_64" CACHE PATH "Path to a Slang binary release containing bin/slangc.exe")
 
 if(NOT EXISTS "${APP_DIR}/LoopThroughWithOpenGLCompositing.cpp")
 	message(FATAL_ERROR "Imported app sources were not found under ${APP_DIR}")
 endif()
 
-if(NOT EXISTS "${GPUDIRECT_DIR}/lib/x64/dvp.lib")
-	message(FATAL_ERROR "NVIDIA GPUDirect library not found under ${GPUDIRECT_DIR}")
-endif()
-
 set(SLANG_RUNTIME_FILES
 	"${SLANG_ROOT}/bin/slangc.exe"
 	"${SLANG_ROOT}/bin/slang-compiler.dll"
@@ -51,6 +46,8 @@ set(APP_SOURCES
 	"${APP_DIR}/decklink/DeckLinkFrameTransfer.h"
 	"${APP_DIR}/decklink/DeckLinkSession.cpp"
 	"${APP_DIR}/decklink/DeckLinkSession.h"
+	"${APP_DIR}/decklink/VideoIOFormat.cpp"
+	"${APP_DIR}/decklink/VideoIOFormat.h"
 	"${APP_DIR}/gl/GLExtensions.cpp"
 	"${APP_DIR}/gl/GLExtensions.h"
 	"${APP_DIR}/gl/GlobalParamsBuffer.cpp"
@@ -83,8 +80,6 @@ set(APP_SOURCES
 	"${APP_DIR}/gl/TextureAssetLoader.h"
 	"${APP_DIR}/gl/TemporalHistoryBuffers.cpp"
 	"${APP_DIR}/gl/TemporalHistoryBuffers.h"
-	"${APP_DIR}/gl/VideoFrameTransfer.cpp"
-	"${APP_DIR}/gl/VideoFrameTransfer.h"
 	"${APP_DIR}/LoopThroughWithOpenGLCompositing.cpp"
 	"${APP_DIR}/LoopThroughWithOpenGLCompositing.h"
 	"${APP_DIR}/LoopThroughWithOpenGLCompositing.rc"
@@ -119,15 +114,9 @@ target_include_directories(LoopThroughWithOpenGLCompositing PRIVATE
 	"${APP_DIR}/platform"
 	"${APP_DIR}/runtime"
 	"${APP_DIR}/shader"
-	"${GPUDIRECT_DIR}/include"
-)
-
-target_link_directories(LoopThroughWithOpenGLCompositing PRIVATE
-	"${GPUDIRECT_DIR}/lib/x64"
 )
 
 target_link_libraries(LoopThroughWithOpenGLCompositing PRIVATE
-	dvp.lib
 	opengl32
 	glu32
 	Ws2_32
@@ -250,20 +239,26 @@ endif()
 
 add_test(NAME OscServerTests COMMAND OscServerTests)
 
-add_custom_command(TARGET LoopThroughWithOpenGLCompositing POST_BUILD
+add_executable(VideoIOFormatTests
-	COMMAND ${CMAKE_COMMAND} -E copy_if_different
+	"${APP_DIR}/decklink/VideoIOFormat.cpp"
-		"${GPUDIRECT_DIR}/bin/x64/dvp.dll"
+	"${CMAKE_CURRENT_SOURCE_DIR}/tests/VideoIOFormatTests.cpp"
-		"$<TARGET_FILE_DIR:LoopThroughWithOpenGLCompositing>/dvp.dll"
 )
 
+target_include_directories(VideoIOFormatTests PRIVATE
+	"${APP_DIR}"
+	"${APP_DIR}/decklink"
+)
+
+if(MSVC)
+	target_compile_options(VideoIOFormatTests PRIVATE /W3)
+endif()
+
+add_test(NAME VideoIOFormatTests COMMAND VideoIOFormatTests)
+
 install(TARGETS LoopThroughWithOpenGLCompositing
 	RUNTIME DESTINATION "."
 )
 
-install(FILES "${GPUDIRECT_DIR}/bin/x64/dvp.dll"
-	DESTINATION "."
-)
-
 install(FILES ${SLANG_RUNTIME_FILES}
 	DESTINATION "3rdParty/slang/bin"
 )

README.md

@@ -20,23 +20,9 @@ The app loads shader packages from `shaders/`, compiles Slang to GLSL at runtime
 - Windows with Visual Studio 2022 C++ tooling.
 - CMake 3.24 or newer.
 - Node.js and npm for the control UI.
-- Blackmagic DeckLink SDK 16.0 with the NVIDIA GPUDirect sample files available locally.
+- Blackmagic Desktop Video drivers and a DeckLink device.
 - Slang binary release with `slangc.exe`, `slang-compiler.dll`, `slang-glslang.dll`, and `LICENSE`.
 
-The Blackmagic/GPUDirect SDK should not be committed to this repository. `CMakeLists.txt` exposes `GPUDIRECT_DIR` as a cache path so local machines and CI runners can point at their installed SDK location.
-
-Default expected SDK path:
-
-```text
-3rdParty/Blackmagic DeckLink SDK 16.0/Win/Samples/NVIDIA_GPUDirect
-```
-
-Override example:
-
-```powershell
-cmake --preset vs2022-x64-debug -DGPUDIRECT_DIR="D:/SDKs/Blackmagic DeckLink SDK 16.0/Win/Samples/NVIDIA_GPUDirect"
-```
-
 Default expected Slang path:
 
 ```text
@@ -87,7 +73,6 @@ The package folder will contain:
 ```text
 dist/VideoShader/
   LoopThroughWithOpenGLCompositing.exe
-  dvp.dll
   config/
   shaders/
   3rdParty/slang/bin/
@@ -227,9 +212,8 @@ The Gitea workflow expects two act runners:
 - `windows-2022`: builds the native app and runs native tests.
 - `ubuntu-latest`: installs UI dependencies and runs the Vite build.
 
-The Windows jobs validate native third-party dependencies before configuring CMake. Because `3rdParty/` is ignored, configure these paths on the runner or in Gitea repository variables:
+The Windows jobs validate native third-party dependencies before configuring CMake. Because `3rdParty/` is ignored, configure this path on the runner or in a Gitea repository variable:
 
-- `GPUDIRECT_DIR`: path to `Blackmagic DeckLink SDK 16.0/Win/Samples/NVIDIA_GPUDirect`.
 - `SLANG_ROOT`: path to the Slang binary release folder containing `bin/slangc.exe`.
 
 The Windows runner also needs the Visual Studio ATL component installed. In Visual Studio Build Tools 2022, add `C++ ATL for latest v143 build tools (x86 & x64)`, component ID `Microsoft.VisualStudio.Component.VC.ATL`.
@@ -237,11 +221,10 @@ The Windows runner also needs the Visual Studio ATL component installed. In Visu
 Example runner paths:
 
 ```text
-D:\SDKs\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect
 D:\SDKs\slang-2026.8-windows-x86_64
 ```
 
-If neither variable is set, the workflow falls back to the repo-local defaults under `3rdParty/`.
+If `SLANG_ROOT` is not set, the workflow falls back to the repo-local default under `3rdParty/`.
 
 ## Still Todo
 
@@ -261,4 +244,5 @@ If neither variable is set, the workflow falls back to the repo-local defaults u
 - compute shaders or a small 1x1 or nx1 RGBA16f render target for abritary data store
 - allow shaders to read other shaders data store based on name? or putput over OSC
 - Mipmappong needed?
-- unwrap a fish eyelens and mirror it and map it to equirectangulr for environmnet map purposes
+- unwrap a fish eyelens and mirror it and map it to equirectangulr for environmnet map purposes
+- add a random flaot between 0-1 that changes each time the application loads for shaders to use

SHADER_CONTRACT.md

@@ -172,6 +172,13 @@ Fields:
 - `sourceHistoryLength`: number of usable source-history frames currently available.
 - `temporalHistoryLength`: number of usable temporal frames currently available for this layer.
 
+Color/precision notes:
+
+- `context.sourceColor`, `sampleVideo()`, and temporal history samples are display-referred Rec.709-like RGB, not linear-light RGB.
+- The host prefers 10-bit DeckLink YUV capture and output when the card/mode supports it, with automatic 8-bit fallback.
+- Internal decoded, layer, composite, output, and temporal render targets are 16-bit floating point, so gradients and LUT work have more headroom than the packed DeckLink byte formats.
+- Do not add extra Rec.709 or linear conversions unless the shader intentionally documents that behavior.
+
 ## Helper Functions
 
 The wrapper provides:

apps/LoopThroughWithOpenGLCompositing/LoopThroughWithOpenGLCompositing.vcxproj

@@ -89,7 +89,7 @@
   <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
     <ClCompile>
       <Optimization>Disabled</Optimization>
-      <AdditionalIncludeDirectories>.;control;decklink;gl;..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>.;control;decklink;gl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions>WIN32;_DEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <BasicRuntimeChecks>EnableFastChecks</BasicRuntimeChecks>
       <RuntimeLibrary>MultiThreadedDebugDLL</RuntimeLibrary>
@@ -99,15 +99,11 @@
       <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
-      <AdditionalDependencies>dvp.lib;opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalDependencies>opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
-      <AdditionalLibraryDirectories>..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\lib\win32;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
       <GenerateDebugInformation>true</GenerateDebugInformation>
       <SubSystem>Windows</SubSystem>
       <TargetMachine>MachineX86</TargetMachine>
     </Link>
-    <PostBuildEvent>
-      <Command>copy /y "..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\bin\$(Platform)\dvp.dll" "$(TargetDir)"</Command>
-    </PostBuildEvent>
   </ItemDefinitionGroup>
   <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
     <Midl>
@@ -115,7 +111,7 @@
     </Midl>
     <ClCompile>
       <Optimization>Disabled</Optimization>
-      <AdditionalIncludeDirectories>.;control;decklink;gl;..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>.;control;decklink;gl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions>WIN32;_DEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <BasicRuntimeChecks>EnableFastChecks</BasicRuntimeChecks>
       <RuntimeLibrary>MultiThreadedDebugDLL</RuntimeLibrary>
@@ -125,21 +121,17 @@
       <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
-      <AdditionalDependencies>dvp.lib;opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalDependencies>opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
-      <AdditionalLibraryDirectories>..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\lib\x64;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
       <GenerateDebugInformation>true</GenerateDebugInformation>
       <SubSystem>Windows</SubSystem>
       <TargetMachine>MachineX64</TargetMachine>
     </Link>
-    <PostBuildEvent>
-      <Command>copy /y "..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\bin\$(Platform)\dvp.dll" "$(TargetDir)"</Command>
-    </PostBuildEvent>
   </ItemDefinitionGroup>
   <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
     <ClCompile>
       <Optimization>MaxSpeed</Optimization>
       <IntrinsicFunctions>true</IntrinsicFunctions>
-      <AdditionalIncludeDirectories>.;control;decklink;gl;..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>.;control;decklink;gl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions>WIN32;NDEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <RuntimeLibrary>MultiThreadedDLL</RuntimeLibrary>
       <FunctionLevelLinking>true</FunctionLevelLinking>
@@ -149,18 +141,13 @@
       <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
-      <AdditionalDependencies>dvp.lib;opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalDependencies>opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
-      <AdditionalLibraryDirectories>..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\lib\win32;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
       <GenerateDebugInformation>true</GenerateDebugInformation>
       <SubSystem>Windows</SubSystem>
       <OptimizeReferences>true</OptimizeReferences>
       <EnableCOMDATFolding>true</EnableCOMDATFolding>
       <TargetMachine>MachineX86</TargetMachine>
     </Link>
-    <PostBuildEvent>
-      <Message>Copy nececssary DLLs to target directory</Message>
-      <Command>copy /y "..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\bin\$(Platform)\dvp.dll" "$(TargetDir)"</Command>
-    </PostBuildEvent>
   </ItemDefinitionGroup>
   <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
     <Midl>
@@ -169,7 +156,7 @@
     <ClCompile>
       <Optimization>MaxSpeed</Optimization>
       <IntrinsicFunctions>true</IntrinsicFunctions>
-      <AdditionalIncludeDirectories>.;control;decklink;gl;..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>.;control;decklink;gl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
       <PreprocessorDefinitions>WIN32;NDEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <RuntimeLibrary>MultiThreadedDLL</RuntimeLibrary>
       <FunctionLevelLinking>true</FunctionLevelLinking>
@@ -179,17 +166,13 @@
       <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
-      <AdditionalDependencies>dvp.lib;opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalDependencies>opengl32.lib;Glu32.lib;%(AdditionalDependencies)</AdditionalDependencies>
-      <AdditionalLibraryDirectories>..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\lib\x64;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
       <GenerateDebugInformation>true</GenerateDebugInformation>
       <SubSystem>Windows</SubSystem>
       <OptimizeReferences>true</OptimizeReferences>
       <EnableCOMDATFolding>true</EnableCOMDATFolding>
       <TargetMachine>MachineX64</TargetMachine>
     </Link>
-    <PostBuildEvent>
-      <Command>copy /y "..\..\3rdParty\Blackmagic DeckLink SDK 16.0\Win\Samples\NVIDIA_GPUDirect\bin\$(Platform)\dvp.dll" "$(TargetDir)"</Command>
-    </PostBuildEvent>
   </ItemDefinitionGroup>
   <ItemGroup>
     <ClCompile Include="gl\GLExtensions.cpp" />
@@ -206,10 +189,10 @@
       <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">Create</PrecompiledHeader>
       <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Release|x64'">Create</PrecompiledHeader>
     </ClCompile>
-    <ClCompile Include="gl\VideoFrameTransfer.cpp" />
     <ClCompile Include="DeckLinkAPI_i.c" />
     <ClCompile Include="control\RuntimeServices.cpp" />
     <ClCompile Include="decklink\DeckLinkSession.cpp" />
+    <ClCompile Include="decklink\VideoIOFormat.cpp" />
     <ClCompile Include="runtime\RuntimeClock.cpp" />
   </ItemGroup>
   <ItemGroup>
@@ -224,9 +207,9 @@
     <ClInclude Include="resource.h" />
     <ClInclude Include="stdafx.h" />
     <ClInclude Include="targetver.h" />
-    <ClInclude Include="gl\VideoFrameTransfer.h" />
     <ClInclude Include="control\RuntimeServices.h" />
     <ClInclude Include="decklink\DeckLinkSession.h" />
+    <ClInclude Include="decklink\VideoIOFormat.h" />
     <ClInclude Include="runtime\RuntimeClock.h" />
   </ItemGroup>
   <ItemGroup>

apps/LoopThroughWithOpenGLCompositing/LoopThroughWithOpenGLCompositing.vcxproj.filters

@@ -45,9 +45,6 @@
     <ClCompile Include="stdafx.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
-    <ClCompile Include="gl\VideoFrameTransfer.cpp">
-      <Filter>Source Files</Filter>
-    </ClCompile>
     <ClCompile Include="DeckLinkAPI_i.c">
       <Filter>DeckLink API</Filter>
     </ClCompile>
@@ -57,6 +54,9 @@
     <ClCompile Include="decklink\DeckLinkSession.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="decklink\VideoIOFormat.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
     <ClCompile Include="runtime\RuntimeClock.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
@@ -95,15 +95,15 @@
     <ClInclude Include="targetver.h">
       <Filter>Header Files</Filter>
     </ClInclude>
-    <ClInclude Include="gl\VideoFrameTransfer.h">
-      <Filter>Header Files</Filter>
-    </ClInclude>
     <ClInclude Include="control\RuntimeServices.h">
       <Filter>Header Files</Filter>
     </ClInclude>
     <ClInclude Include="decklink\DeckLinkSession.h">
       <Filter>Header Files</Filter>
     </ClInclude>
+    <ClInclude Include="decklink\VideoIOFormat.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
     <ClInclude Include="runtime\RuntimeClock.h">
       <Filter>Header Files</Filter>
     </ClInclude>

apps/LoopThroughWithOpenGLCompositing/decklink/DeckLinkFrameTransfer.cpp

@@ -2,311 +2,6 @@
 
 #include "OpenGLComposite.h"
 
-#include <initguid.h>
-
-DEFINE_GUID(IID_PinnedMemoryAllocator,
-	0xddf921a6, 0x279d, 0x4dcd, 0x86, 0x26, 0x75, 0x7f, 0x58, 0xa8, 0xc4, 0x35);
-
-////////////////////////////////////////////
-// PinnedMemoryAllocator
-////////////////////////////////////////////
-
-// PinnedMemoryAllocator implements the IDeckLinkVideoBufferAllocator interface to be used instead of the
-// built-in buffer allocator.
-//
-// For this sample application a custom buffer allocator is used to ensure each address
-// of buffer memory is aligned on a 4kB boundary required by the OpenGL pinned memory extension.
-// If the pinned memory extension is not available, this allocator will still be used and
-// demonstrates how to cache buffer allocations for efficiency.
-//
-// The frame cache delays the releasing of buffers until the cache fills up, thereby avoiding an
-// allocate plus pin operation for every frame, followed by an unpin and deallocate on every frame.
-PinnedMemoryAllocator::PinnedMemoryAllocator(HDC hdc, HGLRC hglrc, VideoFrameTransfer::Direction direction, unsigned cacheSize, unsigned bufferSize) :
-	mHGLDC(hdc),
-	mHGLRC(hglrc),
-	mRefCount(1),
-	mDirection(direction),
-	mBufferSize(bufferSize),
-	mFrameCacheSize(cacheSize)
-{
-}
-
-PinnedMemoryAllocator::~PinnedMemoryAllocator()
-{
-	// Cleanup any unused buffers that remain in the cache
-	while (!mFrameCache.empty())
-	{
-		unPinAddress(mFrameCache.back());
-		VirtualFree(mFrameCache.back(), 0, MEM_RELEASE);
-		mFrameCache.pop_back();
-	}
-
-	for (auto iter = mFrameTransfer.begin(); iter != mFrameTransfer.end(); ++iter)
-		delete iter->second;
-	mFrameTransfer.clear();
-}
-
-bool PinnedMemoryAllocator::transferFrame(void* address, GLuint gpuTexture)
-{
-	if (mFrameTransfer.count(address) == 0)
-	{
-		// VideoFrameTransfer prepares and pins address
-		mFrameTransfer[address] = new VideoFrameTransfer(mBufferSize, address, mDirection);
-	}
-
-	return mFrameTransfer[address]->performFrameTransfer();
-}
-
-void PinnedMemoryAllocator::waitForTransferComplete(void* address)
-{
-	if (mFrameTransfer.count(address))
-		mFrameTransfer[address]->waitForTransferComplete();
-}
-
-void PinnedMemoryAllocator::unPinAddress(void* address)
-{
-	// un-pin address only if it has been pinned for transfer
-	if (mFrameTransfer.count(address) > 0)
-	{
-		wglMakeCurrent(mHGLDC, mHGLRC);
-		mFrameTransfer.erase(address);
-		wglMakeCurrent(NULL, NULL);
-	}
-}
-
-HRESULT STDMETHODCALLTYPE PinnedMemoryAllocator::QueryInterface(REFIID iid, LPVOID* ppv)
-{
-	if (!ppv)
-		return E_POINTER;
-
-	if (iid == IID_IUnknown || iid == IID_PinnedMemoryAllocator)
-	{
-		*ppv = this;
-	}
-	else if (iid == IID_IDeckLinkVideoBufferAllocator)
-	{
-		*ppv = static_cast<IDeckLinkVideoBufferAllocator*>(this);
-	}
-	else
-	{
-		*ppv = nullptr;
-		return E_NOINTERFACE;
-	}
-	AddRef();
-	return S_OK;
-}
-
-ULONG STDMETHODCALLTYPE PinnedMemoryAllocator::AddRef(void)
-{
-	return ++mRefCount;
-}
-
-ULONG STDMETHODCALLTYPE PinnedMemoryAllocator::Release(void)
-{
-	int newCount = --mRefCount;
-	if (newCount == 0)
-		delete this;
-	return newCount;
-}
-
-HRESULT STDMETHODCALLTYPE PinnedMemoryAllocator::AllocateVideoBuffer(IDeckLinkVideoBuffer** allocatedBuffer)
-{
-	std::shared_ptr<void> sharedMemBuffer;
-
-	// Manage caching of allocated buffers via shared_ptr deleter.
-	auto deleter = [this](void* buffer) mutable {
-		if (mFrameCache.size() < mFrameCacheSize)
-		{
-			mFrameCache.push_back(buffer);
-		}
-		else
-		{
-			// No room left in cache, so un-pin (if it was pinned) and free this buffer
-			unPinAddress(buffer);
-			VirtualFree(buffer, 0, MEM_RELEASE);
-		}
-		// We AddRef this class once the deleter is used because this class owns the mem
-		Release();
-	};
-
-	if (mFrameCache.empty())
-	{
-		// Allocate memory on a page boundary
-		void* memBuffer = VirtualAlloc(NULL, mBufferSize, MEM_COMMIT | MEM_RESERVE | MEM_WRITE_WATCH, PAGE_READWRITE);
-		if (!memBuffer)
-			return E_OUTOFMEMORY;
-
-		sharedMemBuffer = std::shared_ptr<void>(memBuffer, deleter);
-	}
-	else
-	{
-		// Re-use most recently released address
-		sharedMemBuffer = std::shared_ptr<void>(mFrameCache.back(), deleter);
-		mFrameCache.pop_back();
-	}
-
-	// This class owns the mem so the buffer we return needs to AddRef() this, and Release() in the deleter
-	AddRef();
-
-	*allocatedBuffer = new DeckLinkVideoBuffer(sharedMemBuffer, this);
-	return S_OK;
-}
-
-////////////////////////////////////////////
-// InputAllocatorPool Class
-////////////////////////////////////////////
-InputAllocatorPool::InputAllocatorPool(HDC hdc, HGLRC hglrc)
-{
-	mHDC = hdc;
-	mHGLRC = hglrc;
-}
-
-HRESULT InputAllocatorPool::QueryInterface(REFIID iid, void** ppv)
-{
-	if (!ppv)
-		return E_POINTER;
-
-	if (iid == IID_IUnknown)
-	{
-		*ppv = this;
-	}
-	else if (iid == IID_IDeckLinkVideoBufferAllocatorProvider)
-	{
-		*ppv = static_cast<IDeckLinkVideoBufferAllocatorProvider*>(this);
-	}
-	else
-	{
-		*ppv = nullptr;
-		return E_NOINTERFACE;
-	}
-	AddRef();
-	return S_OK;
-}
-
-ULONG InputAllocatorPool::AddRef(void)
-{
-	return ++mRefCount;
-}
-
-ULONG InputAllocatorPool::Release(void)
-{
-	int newCount = --mRefCount;
-	if (newCount == 0)
-		delete this;
-	return newCount;
-}
-
-HRESULT InputAllocatorPool::GetVideoBufferAllocator(
-	/* [in] */ unsigned int bufferSize,
-	/* [in] */ unsigned int,
-	/* [in] */ unsigned int,
-	/* [in] */ unsigned int,
-	/* [in] */ BMDPixelFormat,
-	/* [out] */ IDeckLinkVideoBufferAllocator** allocator)
-{
-	if (!allocator)
-		return E_POINTER;
-
-	auto existing = mAllocatorBySize.find(bufferSize);
-	if (existing != mAllocatorBySize.end())
-	{
-		*allocator = &*existing->second;
-		(*allocator)->AddRef();
-		return S_OK;
-	}
-
-	CComPtr<PinnedMemoryAllocator> newAllocator;
-	newAllocator.Attach(new (std::nothrow) PinnedMemoryAllocator(mHDC, mHGLRC, VideoFrameTransfer::CPUtoGPU, 3, bufferSize));
-	if (!newAllocator)
-		return E_OUTOFMEMORY;
-
-	mAllocatorBySize.emplace(std::make_pair(bufferSize, newAllocator));
-	*allocator = newAllocator.Detach();
-	return S_OK;
-}
-
-////////////////////////////////////////////
-// DeckLink Video Buffer Class
-////////////////////////////////////////////
-DeckLinkVideoBuffer::DeckLinkVideoBuffer(std::shared_ptr<void>& buffer, PinnedMemoryAllocator* parent) :
-	mParentAllocator(parent),
-	mRefCount(1),
-	mBuffer(buffer)
-{
-}
-
-HRESULT STDMETHODCALLTYPE DeckLinkVideoBuffer::QueryInterface(REFIID riid, void** ppvObject)
-{
-	HRESULT result = S_OK;
-
-	if (ppvObject == nullptr)
-		return E_POINTER;
-
-	if (riid == IID_IUnknown)
-	{
-		*ppvObject = this;
-		AddRef();
-	}
-	else if (riid == IID_IDeckLinkVideoBuffer)
-	{
-		*ppvObject = static_cast<IDeckLinkVideoBuffer*>(this);
-		AddRef();
-	}
-	else if (riid == IID_PinnedMemoryAllocator)
-	{
-		result = mParentAllocator->QueryInterface(riid, ppvObject);
-	}
-	else
-	{
-		*ppvObject = nullptr;
-		result = E_NOINTERFACE;
-	}
-
-	return result;
-}
-
-ULONG STDMETHODCALLTYPE DeckLinkVideoBuffer::AddRef()
-{
-	return ++mRefCount;
-}
-
-ULONG STDMETHODCALLTYPE DeckLinkVideoBuffer::Release()
-{
-	int newValue = --mRefCount;
-	if (newValue == 0)
-		delete this;
-
-	return newValue;
-}
-
-HRESULT STDMETHODCALLTYPE DeckLinkVideoBuffer::GetBytes(void** buffer)
-{
-	if (buffer == nullptr)
-		return E_POINTER;
-
-	*buffer = mBuffer.get();
-	return S_OK;
-}
-
-HRESULT STDMETHODCALLTYPE DeckLinkVideoBuffer::GetSize(uint64_t* size)
-{
-	if (size == nullptr)
-		return E_POINTER;
-
-	*size = mParentAllocator->bufferSize();
-	return S_OK;
-}
-
-HRESULT STDMETHODCALLTYPE DeckLinkVideoBuffer::StartAccess(BMDBufferAccessFlags)
-{
-	return S_OK;
-}
-
-HRESULT STDMETHODCALLTYPE DeckLinkVideoBuffer::EndAccess(BMDBufferAccessFlags)
-{
-	return S_OK;
-}
-
 ////////////////////////////////////////////
 // DeckLink Capture Delegate Class
 ////////////////////////////////////////////

apps/LoopThroughWithOpenGLCompositing/decklink/DeckLinkFrameTransfer.h

@@ -1,111 +1,13 @@
 #pragma once
 
 #include <windows.h>
-#include <gl/gl.h>
 
-#include <atlbase.h>
 #include <atomic>
-#include <map>
-#include <memory>
-#include <vector>
 
 #include "DeckLinkAPI_h.h"
-#include "VideoFrameTransfer.h"
-
-extern "C" const IID IID_PinnedMemoryAllocator;
 
 class OpenGLComposite;
 
-////////////////////////////////////////////
-// PinnedMemoryAllocator
-////////////////////////////////////////////
-class PinnedMemoryAllocator : public IDeckLinkVideoBufferAllocator
-{
-public:
-	PinnedMemoryAllocator(HDC hdc, HGLRC hglrc, VideoFrameTransfer::Direction direction, unsigned cacheSize, unsigned bufferSize);
-	virtual ~PinnedMemoryAllocator();
-
-	bool transferFrame(void* address, GLuint gpuTexture);
-	void waitForTransferComplete(void* address);
-	unsigned bufferSize() { return mBufferSize; }
-
-	// IUnknown methods
-	virtual HRESULT STDMETHODCALLTYPE QueryInterface(REFIID iid, LPVOID* ppv) override;
-	virtual ULONG STDMETHODCALLTYPE AddRef(void) override;
-	virtual ULONG STDMETHODCALLTYPE Release(void) override;
-
-	// IDeckLinkVideoBufferAllocator methods
-	virtual HRESULT STDMETHODCALLTYPE AllocateVideoBuffer(IDeckLinkVideoBuffer** allocatedBuffer) override;
-
-private:
-	void unPinAddress(void* address);
-
-private:
-	HDC mHGLDC;
-	HGLRC mHGLRC;
-	std::atomic<ULONG> mRefCount;
-	VideoFrameTransfer::Direction mDirection;
-	std::map<void*, VideoFrameTransfer*> mFrameTransfer;
-	unsigned mBufferSize;
-	std::vector<void*> mFrameCache;
-	unsigned mFrameCacheSize;
-};
-
-////////////////////////////////////////////
-// InputAllocatorPool
-////////////////////////////////////////////
-class InputAllocatorPool : public IDeckLinkVideoBufferAllocatorProvider
-{
-public:
-	InputAllocatorPool(HDC hdc, HGLRC hglrc);
-
-	// IUnknown interface
-	ULONG STDMETHODCALLTYPE AddRef() override;
-	ULONG STDMETHODCALLTYPE Release() override;
-	HRESULT STDMETHODCALLTYPE QueryInterface(REFIID iid, void** ppv) override;
-
-	// IDeckLinkVideoBufferAllocatorProvider interface
-	HRESULT STDMETHODCALLTYPE GetVideoBufferAllocator(
-		/* [in] */ unsigned int bufferSize,
-		/* [in] */ unsigned int width,
-		/* [in] */ unsigned int height,
-		/* [in] */ unsigned int rowBytes,
-		/* [in] */ BMDPixelFormat pixelFormat,
-		/* [out] */ IDeckLinkVideoBufferAllocator** allocator) override;
-
-private:
-	std::atomic<ULONG> mRefCount;
-	std::map<unsigned int, CComPtr<PinnedMemoryAllocator> > mAllocatorBySize;
-	HDC mHDC;
-	HGLRC mHGLRC;
-};
-
-////////////////////////////////////////////
-// DeckLinkVideoBuffer
-////////////////////////////////////////////
-class DeckLinkVideoBuffer : public IDeckLinkVideoBuffer
-{
-public:
-	explicit DeckLinkVideoBuffer(std::shared_ptr<void>& buffer, PinnedMemoryAllocator* parent);
-	virtual ~DeckLinkVideoBuffer() = default;
-
-	// IUnknown interface
-	virtual HRESULT STDMETHODCALLTYPE QueryInterface(REFIID riid, void** ppvObject) override;
-	virtual ULONG STDMETHODCALLTYPE AddRef(void) override;
-	virtual ULONG STDMETHODCALLTYPE Release(void) override;
-
-	// IDeckLinkVideoBuffer interface
-	virtual HRESULT STDMETHODCALLTYPE GetBytes(void** buffer) override;
-	virtual HRESULT STDMETHODCALLTYPE GetSize(uint64_t* size) override;
-	virtual HRESULT STDMETHODCALLTYPE StartAccess(BMDBufferAccessFlags flags) override;
-	virtual HRESULT STDMETHODCALLTYPE EndAccess(BMDBufferAccessFlags flags) override;
-
-private:
-	CComPtr<PinnedMemoryAllocator> mParentAllocator;
-	std::atomic<ULONG> mRefCount;
-	std::shared_ptr<void> mBuffer;
-};
-
 ////////////////////////////////////////////
 // Capture Delegate Class
 ////////////////////////////////////////////

apps/LoopThroughWithOpenGLCompositing/decklink/DeckLinkSession.cpp

@@ -6,6 +6,7 @@
 #include <cstdio>
 #include <cstring>
 #include <new>
+#include <sstream>
 #include <vector>
 
 namespace
@@ -25,6 +26,42 @@ std::string BstrToUtf8(BSTR value)
 
 	return std::string(utf8Name.data());
 }
+
+bool InputSupportsFormat(IDeckLinkInput* input, BMDDisplayMode displayMode, BMDPixelFormat pixelFormat)
+{
+	if (input == nullptr)
+		return false;
+
+	BOOL supported = FALSE;
+	BMDDisplayMode actualMode = bmdModeUnknown;
+	const HRESULT result = input->DoesSupportVideoMode(
+		bmdVideoConnectionUnspecified,
+		displayMode,
+		pixelFormat,
+		bmdNoVideoInputConversion,
+		bmdSupportedVideoModeDefault,
+		&actualMode,
+		&supported);
+	return result == S_OK && supported != FALSE;
+}
+
+bool OutputSupportsFormat(IDeckLinkOutput* output, BMDDisplayMode displayMode, BMDPixelFormat pixelFormat)
+{
+	if (output == nullptr)
+		return false;
+
+	BOOL supported = FALSE;
+	BMDDisplayMode actualMode = bmdModeUnknown;
+	const HRESULT result = output->DoesSupportVideoMode(
+		bmdVideoConnectionUnspecified,
+		displayMode,
+		pixelFormat,
+		bmdNoVideoOutputConversion,
+		bmdSupportedVideoModeDefault,
+		&actualMode,
+		&supported);
+	return result == S_OK && supported != FALSE;
+}
 }
 
 DeckLinkSession::~DeckLinkSession()
@@ -52,8 +89,6 @@ void DeckLinkSession::ReleaseResources()
 	playoutDelegate.Release();
 	outputVideoFrameQueue.clear();
 	output.Release();
-
-	playoutAllocator.Release();
 }
 
 bool DeckLinkSession::DiscoverDevicesAndModes(const VideoFormatSelection& videoModes, std::string& error)
@@ -173,11 +208,75 @@ bool DeckLinkSession::DiscoverDevicesAndModes(const VideoFormatSelection& videoM
 		inputDisplayModeName = "No input - black frame";
 	outputMode->GetFrameRate(&frameDuration, &frameTimescale);
 
+	inputFrameRowBytes = inputFrameSize.width * 2u;
+	outputFrameRowBytes = outputFrameSize.width * 4u;
+	captureTextureWidth = inputFrameSize.width / 2u;
+	outputPackTextureWidth = outputFrameSize.width;
+
+	return true;
+}
+
+bool DeckLinkSession::SelectPreferredFormats(const VideoFormatSelection& videoModes, std::string& error)
+{
+	if (!output)
+	{
+		error = "Expected an Output DeckLink device";
+		return false;
+	}
+
+	formatStatusMessage.clear();
+
+	const bool inputTenBitSupported = input != nullptr && InputSupportsFormat(input, videoModes.input.displayMode, bmdFormat10BitYUV);
+	inputPixelFormat = input != nullptr ? ChoosePreferredVideoIOFormat(inputTenBitSupported) : VideoIOPixelFormat::Uyvy8;
+	if (input != nullptr && !inputTenBitSupported)
+		formatStatusMessage += "DeckLink input does not report 10-bit YUV support for the configured mode; using 8-bit capture. ";
+
+	const bool outputTenBitSupported = OutputSupportsFormat(output, videoModes.output.displayMode, bmdFormat10BitYUV);
+	outputPixelFormat = ChoosePreferredVideoIOFormat(outputTenBitSupported);
+	if (!outputTenBitSupported)
+		formatStatusMessage += "DeckLink output does not report 10-bit YUV support for the configured mode; using 8-bit BGRA output. ";
+
+	int deckLinkOutputRowBytes = 0;
+	if (output->RowBytesForPixelFormat(OutputIsTenBit() ? bmdFormat10BitYUV : bmdFormat8BitBGRA, outputFrameSize.width, &deckLinkOutputRowBytes) != S_OK)
+	{
+		error = "DeckLink output setup failed while calculating output row bytes.";
+		return false;
+	}
+	outputFrameRowBytes = static_cast<unsigned>(deckLinkOutputRowBytes);
+	outputPackTextureWidth = OutputIsTenBit()
+		? PackedTextureWidthFromRowBytes(outputFrameRowBytes)
+		: outputFrameSize.width;
+
+	if (InputIsTenBit())
+	{
+		int deckLinkInputRowBytes = 0;
+		if (output->RowBytesForPixelFormat(bmdFormat10BitYUV, inputFrameSize.width, &deckLinkInputRowBytes) == S_OK)
+			inputFrameRowBytes = static_cast<unsigned>(deckLinkInputRowBytes);
+		else
+			inputFrameRowBytes = MinimumV210RowBytes(inputFrameSize.width);
+	}
+	else
+	{
+		inputFrameRowBytes = inputFrameSize.width * 2u;
+	}
+	captureTextureWidth = InputIsTenBit()
+		? PackedTextureWidthFromRowBytes(inputFrameRowBytes)
+		: inputFrameSize.width / 2u;
+
+	std::ostringstream status;
+	status << "DeckLink formats: capture " << (input ? VideoIOPixelFormatName(inputPixelFormat) : "none")
+		<< ", output " << (OutputIsTenBit() ? "10-bit YUV v210" : "8-bit BGRA") << ".";
+	if (!formatStatusMessage.empty())
+		status << " " << formatStatusMessage;
+	formatStatusMessage = status.str();
 	return true;
 }
 
 bool DeckLinkSession::ConfigureInput(OpenGLComposite* owner, HDC hdc, HGLRC hglrc, const VideoFormat& inputVideoMode, std::string& error)
 {
+	(void)hdc;
+	(void)hglrc;
+
 	if (!input)
 	{
 		hasNoInputSource = true;
@@ -185,10 +284,26 @@ bool DeckLinkSession::ConfigureInput(OpenGLComposite* owner, HDC hdc, HGLRC hglr
 		return true;
 	}
 
-	CComPtr<IDeckLinkVideoBufferAllocatorProvider> captureAllocator(new (std::nothrow) InputAllocatorPool(hdc, hglrc));
+	const BMDPixelFormat deckLinkInputPixelFormat = DeckLinkPixelFormatForVideoIO(inputPixelFormat);
-
+	if (input->EnableVideoInput(inputVideoMode.displayMode, deckLinkInputPixelFormat, bmdVideoInputFlagDefault) != S_OK)
-	if (input->EnableVideoInputWithAllocatorProvider(inputVideoMode.displayMode, bmdFormat8BitYUV, bmdVideoInputFlagDefault, captureAllocator) != S_OK)
 	{
+		if (inputPixelFormat == VideoIOPixelFormat::V210)
+		{
+			OutputDebugStringA("DeckLink 10-bit input could not be enabled; falling back to 8-bit capture.\n");
+			inputPixelFormat = VideoIOPixelFormat::Uyvy8;
+			inputFrameRowBytes = inputFrameSize.width * 2u;
+			captureTextureWidth = inputFrameSize.width / 2u;
+			if (input->EnableVideoInput(inputVideoMode.displayMode, bmdFormat8BitYUV, bmdVideoInputFlagDefault) == S_OK)
+			{
+				std::ostringstream status;
+				status << "DeckLink formats: capture " << VideoIOPixelFormatName(inputPixelFormat)
+					<< ", output " << (OutputIsTenBit() ? "10-bit YUV v210" : "8-bit BGRA")
+					<< ". DeckLink 10-bit input enable failed; using 8-bit capture.";
+				formatStatusMessage = status.str();
+				goto input_enabled;
+			}
+		}
+
 		OutputDebugStringA("DeckLink input could not be enabled; continuing in output-only black-frame mode.\n");
 		input.Release();
 		hasNoInputSource = true;
@@ -196,6 +311,7 @@ bool DeckLinkSession::ConfigureInput(OpenGLComposite* owner, HDC hdc, HGLRC hglr
 		return true;
 	}
 
+input_enabled:
 	captureDelegate.Attach(new (std::nothrow) CaptureDelegate(owner));
 	if (captureDelegate == nullptr)
 	{
@@ -213,19 +329,8 @@ bool DeckLinkSession::ConfigureInput(OpenGLComposite* owner, HDC hdc, HGLRC hglr
 
 bool DeckLinkSession::ConfigureOutput(OpenGLComposite* owner, HDC hdc, HGLRC hglrc, const VideoFormat& outputVideoMode, bool externalKeyingEnabled, std::string& error)
 {
-	int outputFrameRowBytes = 0;
+	(void)hdc;
-	if (output->RowBytesForPixelFormat(bmdFormat8BitBGRA, outputFrameSize.width, &outputFrameRowBytes) != S_OK)
+	(void)hglrc;
-	{
-		error = "DeckLink output setup failed while calculating BGRA row bytes.";
-		return false;
-	}
-
-	playoutAllocator.Attach(new (std::nothrow) PinnedMemoryAllocator(hdc, hglrc, VideoFrameTransfer::GPUtoCPU, 1, outputFrameRowBytes * outputFrameSize.height));
-	if (playoutAllocator == nullptr)
-	{
-		error = "DeckLink output setup failed while creating the playout allocator.";
-		return false;
-	}
 
 	if (output->EnableVideoOutput(outputVideoMode.displayMode, bmdVideoOutputFlagDefault) != S_OK)
 	{
@@ -264,15 +369,9 @@ bool DeckLinkSession::ConfigureOutput(OpenGLComposite* owner, HDC hdc, HGLRC hgl
 	for (int i = 0; i < 10; i++)
 	{
 		CComPtr<IDeckLinkMutableVideoFrame> outputFrame;
-		CComPtr<IDeckLinkVideoBuffer> outputFrameBuffer;
 
-		if (playoutAllocator->AllocateVideoBuffer(&outputFrameBuffer) != S_OK)
+		const BMDPixelFormat deckLinkOutputPixelFormat = OutputIsTenBit() ? bmdFormat10BitYUV : bmdFormat8BitBGRA;
-		{
+		if (output->CreateVideoFrame(outputFrameSize.width, outputFrameSize.height, outputFrameRowBytes, deckLinkOutputPixelFormat, bmdFrameFlagFlipVertical, &outputFrame) != S_OK)
-			error = "DeckLink output setup failed while allocating an output frame buffer.";
-			return false;
-		}
-
-		if (output->CreateVideoFrameWithBuffer(outputFrameSize.width, outputFrameSize.height, outputFrameRowBytes, bmdFormat8BitBGRA, bmdFrameFlagFlipVertical, outputFrameBuffer, &outputFrame) != S_OK)
 		{
 			error = "DeckLink output setup failed while creating an output video frame.";
 			return false;
@@ -294,6 +393,9 @@ bool DeckLinkSession::ConfigureOutput(OpenGLComposite* owner, HDC hdc, HGLRC hgl
 		return false;
 	}
 
+	if (!formatStatusMessage.empty())
+		statusMessage = statusMessage.empty() ? formatStatusMessage : formatStatusMessage + " " + statusMessage;
+
 	return true;
 }
 
@@ -312,17 +414,6 @@ IDeckLinkMutableVideoFrame* DeckLinkSession::RotateOutputFrame()
 	return outputVideoFrame.p;
 }
 
-bool DeckLinkSession::TransferPlayoutFrame(void* address, GLuint outputTexture)
-{
-	return playoutAllocator != nullptr && playoutAllocator->transferFrame(address, outputTexture);
-}
-
-void DeckLinkSession::WaitForPlayoutTransferComplete(void* address)
-{
-	if (playoutAllocator != nullptr)
-		playoutAllocator->waitForTransferComplete(address);
-}
-
 void DeckLinkSession::AccountForCompletionResult(BMDOutputFrameCompletionResult completionResult)
 {
 	if (completionResult == bmdOutputFrameDisplayedLate || completionResult == bmdOutputFrameDropped)

apps/LoopThroughWithOpenGLCompositing/decklink/DeckLinkSession.h

@@ -3,6 +3,7 @@
 #include "DeckLinkAPI_h.h"
 #include "DeckLinkDisplayMode.h"
 #include "DeckLinkFrameTransfer.h"
+#include "VideoIOFormat.h"
 
 #include <atlbase.h>
 #include <deque>
@@ -18,6 +19,7 @@ public:
 
 	void ReleaseResources();
 	bool DiscoverDevicesAndModes(const VideoFormatSelection& videoModes, std::string& error);
+	bool SelectPreferredFormats(const VideoFormatSelection& videoModes, std::string& error);
 	bool ConfigureInput(OpenGLComposite* owner, HDC hdc, HGLRC hglrc, const VideoFormat& inputVideoMode, std::string& error);
 	bool ConfigureOutput(OpenGLComposite* owner, HDC hdc, HGLRC hglrc, const VideoFormat& outputVideoMode, bool externalKeyingEnabled, std::string& error);
 	bool Start();
@@ -33,6 +35,15 @@ public:
 	unsigned InputFrameHeight() const { return inputFrameSize.height; }
 	unsigned OutputFrameWidth() const { return outputFrameSize.width; }
 	unsigned OutputFrameHeight() const { return outputFrameSize.height; }
+	VideoIOPixelFormat InputPixelFormat() const { return inputPixelFormat; }
+	VideoIOPixelFormat OutputPixelFormat() const { return outputPixelFormat; }
+	bool InputIsTenBit() const { return VideoIOPixelFormatIsTenBit(inputPixelFormat); }
+	bool OutputIsTenBit() const { return VideoIOPixelFormatIsTenBit(outputPixelFormat); }
+	unsigned InputFrameRowBytes() const { return inputFrameRowBytes; }
+	unsigned OutputFrameRowBytes() const { return outputFrameRowBytes; }
+	unsigned CaptureTextureWidth() const { return captureTextureWidth; }
+	unsigned OutputPackTextureWidth() const { return outputPackTextureWidth; }
+	const std::string& FormatStatusMessage() const { return formatStatusMessage; }
 	const std::string& InputDisplayModeName() const { return inputDisplayModeName; }
 	const std::string& OutputModelName() const { return outputModelName; }
 	bool SupportsInternalKeying() const { return supportsInternalKeying; }
@@ -43,8 +54,6 @@ public:
 	void SetStatusMessage(const std::string& message) { statusMessage = message; }
 	double FrameBudgetMilliseconds() const;
 	IDeckLinkMutableVideoFrame* RotateOutputFrame();
-	bool TransferPlayoutFrame(void* address, GLuint outputTexture);
-	void WaitForPlayoutTransferComplete(void* address);
 	void AccountForCompletionResult(BMDOutputFrameCompletionResult completionResult);
 	bool ScheduleOutputFrame(IDeckLinkMutableVideoFrame* outputVideoFrame);
 
@@ -55,12 +64,17 @@ private:
 	CComPtr<IDeckLinkOutput>				output;
 	CComPtr<IDeckLinkKeyer>					keyer;
 	std::deque<CComPtr<IDeckLinkMutableVideoFrame>> outputVideoFrameQueue;
-	CComPtr<PinnedMemoryAllocator>			playoutAllocator;
 	BMDTimeValue							frameDuration = 0;
 	BMDTimeScale							frameTimescale = 0;
 	unsigned								totalPlayoutFrames = 0;
 	FrameSize								inputFrameSize;
 	FrameSize								outputFrameSize;
+	VideoIOPixelFormat						inputPixelFormat = VideoIOPixelFormat::Uyvy8;
+	VideoIOPixelFormat						outputPixelFormat = VideoIOPixelFormat::Uyvy8;
+	unsigned								inputFrameRowBytes = 0;
+	unsigned								outputFrameRowBytes = 0;
+	unsigned								captureTextureWidth = 0;
+	unsigned								outputPackTextureWidth = 0;
 	std::string								inputDisplayModeName = "1080p59.94";
 	std::string								outputDisplayModeName = "1080p59.94";
 	bool									hasNoInputSource = true;
@@ -70,4 +84,5 @@ private:
 	bool									keyerInterfaceAvailable = false;
 	bool									externalKeyingActive = false;
 	std::string								statusMessage;
+	std::string								formatStatusMessage;
 };

apps/LoopThroughWithOpenGLCompositing/decklink/VideoIOFormat.cpp

@@ -0,0 +1,139 @@
+#include "VideoIOFormat.h"
+
+#include <algorithm>
+#include <cmath>
+
+#ifdef min
+#undef min
+#endif
+#ifdef max
+#undef max
+#endif
+
+namespace
+{
+uint16_t Clamp10(int value, int minimum, int maximum)
+{
+	return static_cast<uint16_t>(std::max(minimum, std::min(maximum, value)));
+}
+
+uint32_t MakeV210Word(uint16_t a, uint16_t b, uint16_t c)
+{
+	return (static_cast<uint32_t>(a) & 0x3ffu)
+		| ((static_cast<uint32_t>(b) & 0x3ffu) << 10)
+		| ((static_cast<uint32_t>(c) & 0x3ffu) << 20);
+}
+
+void StoreWord(std::array<uint8_t, 16>& bytes, std::size_t wordIndex, uint32_t word)
+{
+	const std::size_t offset = wordIndex * 4;
+	bytes[offset + 0] = static_cast<uint8_t>(word & 0xffu);
+	bytes[offset + 1] = static_cast<uint8_t>((word >> 8) & 0xffu);
+	bytes[offset + 2] = static_cast<uint8_t>((word >> 16) & 0xffu);
+	bytes[offset + 3] = static_cast<uint8_t>((word >> 24) & 0xffu);
+}
+
+uint32_t LoadWord(const std::array<uint8_t, 16>& bytes, std::size_t wordIndex)
+{
+	const std::size_t offset = wordIndex * 4;
+	return static_cast<uint32_t>(bytes[offset + 0])
+		| (static_cast<uint32_t>(bytes[offset + 1]) << 8)
+		| (static_cast<uint32_t>(bytes[offset + 2]) << 16)
+		| (static_cast<uint32_t>(bytes[offset + 3]) << 24);
+}
+
+uint16_t Component(uint32_t word, unsigned index)
+{
+	return static_cast<uint16_t>((word >> (index * 10)) & 0x3ffu);
+}
+}
+
+const char* VideoIOPixelFormatName(VideoIOPixelFormat format)
+{
+	return format == VideoIOPixelFormat::V210 ? "10-bit YUV v210" : "8-bit YUV UYVY";
+}
+
+bool VideoIOPixelFormatIsTenBit(VideoIOPixelFormat format)
+{
+	return format == VideoIOPixelFormat::V210;
+}
+
+BMDPixelFormat DeckLinkPixelFormatForVideoIO(VideoIOPixelFormat format)
+{
+	return format == VideoIOPixelFormat::V210 ? bmdFormat10BitYUV : bmdFormat8BitYUV;
+}
+
+VideoIOPixelFormat VideoIOPixelFormatFromDeckLink(BMDPixelFormat format)
+{
+	return format == bmdFormat10BitYUV ? VideoIOPixelFormat::V210 : VideoIOPixelFormat::Uyvy8;
+}
+
+VideoIOPixelFormat ChoosePreferredVideoIOFormat(bool tenBitSupported)
+{
+	return tenBitSupported ? VideoIOPixelFormat::V210 : VideoIOPixelFormat::Uyvy8;
+}
+
+unsigned PackedTextureWidthFromRowBytes(unsigned rowBytes)
+{
+	return (rowBytes + 3u) / 4u;
+}
+
+unsigned MinimumV210RowBytes(unsigned frameWidth)
+{
+	return ((frameWidth + 5u) / 6u) * 16u;
+}
+
+unsigned ActiveV210WordsForWidth(unsigned frameWidth)
+{
+	return ((frameWidth + 5u) / 6u) * 4u;
+}
+
+V210CodeValues Rec709RgbToLegalV210(float red, float green, float blue)
+{
+	red = std::max(0.0f, std::min(1.0f, red));
+	green = std::max(0.0f, std::min(1.0f, green));
+	blue = std::max(0.0f, std::min(1.0f, blue));
+
+	const float y = 0.2126f * red + 0.7152f * green + 0.0722f * blue;
+	const float cb = (blue - y) / 1.8556f + 0.5f;
+	const float cr = (red - y) / 1.5748f + 0.5f;
+
+	V210CodeValues values;
+	values.y = Clamp10(static_cast<int>(std::lround(64.0f + y * 876.0f)), 64, 940);
+	values.cb = Clamp10(static_cast<int>(std::lround(64.0f + cb * 896.0f)), 64, 960);
+	values.cr = Clamp10(static_cast<int>(std::lround(64.0f + cr * 896.0f)), 64, 960);
+	return values;
+}
+
+std::array<uint8_t, 16> PackV210Block(const V210SixPixelBlock& block)
+{
+	std::array<uint8_t, 16> bytes = {};
+	StoreWord(bytes, 0, MakeV210Word(block.cb[0], block.y[0], block.cr[0]));
+	StoreWord(bytes, 1, MakeV210Word(block.y[1], block.cb[1], block.y[2]));
+	StoreWord(bytes, 2, MakeV210Word(block.cr[1], block.y[3], block.cb[2]));
+	StoreWord(bytes, 3, MakeV210Word(block.y[4], block.cr[2], block.y[5]));
+	return bytes;
+}
+
+V210SixPixelBlock UnpackV210Block(const std::array<uint8_t, 16>& bytes)
+{
+	const uint32_t word0 = LoadWord(bytes, 0);
+	const uint32_t word1 = LoadWord(bytes, 1);
+	const uint32_t word2 = LoadWord(bytes, 2);
+	const uint32_t word3 = LoadWord(bytes, 3);
+
+	V210SixPixelBlock block;
+	block.cb[0] = Component(word0, 0);
+	block.y[0] = Component(word0, 1);
+	block.cr[0] = Component(word0, 2);
+	block.y[1] = Component(word1, 0);
+	block.cb[1] = Component(word1, 1);
+	block.y[2] = Component(word1, 2);
+	block.cr[1] = Component(word2, 0);
+	block.y[3] = Component(word2, 1);
+	block.cb[2] = Component(word2, 2);
+	block.y[4] = Component(word3, 0);
+	block.cr[2] = Component(word3, 1);
+	block.y[5] = Component(word3, 2);
+	return block;
+}

apps/LoopThroughWithOpenGLCompositing/decklink/VideoIOFormat.h

@@ -0,0 +1,38 @@
+#pragma once
+
+#include "DeckLinkAPI_h.h"
+
+#include <array>
+#include <cstdint>
+
+enum class VideoIOPixelFormat
+{
+	Uyvy8,
+	V210
+};
+
+struct V210CodeValues
+{
+	uint16_t y = 64;
+	uint16_t cb = 512;
+	uint16_t cr = 512;
+};
+
+struct V210SixPixelBlock
+{
+	std::array<uint16_t, 6> y = {};
+	std::array<uint16_t, 3> cb = {};
+	std::array<uint16_t, 3> cr = {};
+};
+
+const char* VideoIOPixelFormatName(VideoIOPixelFormat format);
+bool VideoIOPixelFormatIsTenBit(VideoIOPixelFormat format);
+BMDPixelFormat DeckLinkPixelFormatForVideoIO(VideoIOPixelFormat format);
+VideoIOPixelFormat VideoIOPixelFormatFromDeckLink(BMDPixelFormat format);
+VideoIOPixelFormat ChoosePreferredVideoIOFormat(bool tenBitSupported);
+unsigned PackedTextureWidthFromRowBytes(unsigned rowBytes);
+unsigned MinimumV210RowBytes(unsigned frameWidth);
+unsigned ActiveV210WordsForWidth(unsigned frameWidth);
+V210CodeValues Rec709RgbToLegalV210(float red, float green, float blue);
+std::array<uint8_t, 16> PackV210Block(const V210SixPixelBlock& block);
+V210SixPixelBlock UnpackV210Block(const std::array<uint8_t, 16>& bytes);

apps/LoopThroughWithOpenGLCompositing/gl/GLExtensions.h

@@ -60,6 +60,9 @@
 #define GL_DYNAMIC_DRAW							0x88E8
 #define GL_UNIFORM_BUFFER						0x8A11
 #define GL_RGBA8								0x8058
+#define GL_RGBA16F								0x881A
+#define GL_TEXTURE0								0x84C0
+#define GL_ACTIVE_TEXTURE						0x84E0
 #define GL_ARRAY_BUFFER							0x8892
 #define GL_PIXEL_PACK_BUFFER					0x88EB
 #define GL_PIXEL_UNPACK_BUFFER					0x88EC

apps/LoopThroughWithOpenGLCompositing/gl/GlShaderSources.cpp

@@ -16,6 +16,7 @@ const char* kDecodeFragmentShaderSource =
 	"layout(binding = 2) uniform sampler2D uPackedVideoInput;\n"
 	"uniform vec2 uPackedVideoResolution;\n"
 	"uniform vec2 uDecodedVideoResolution;\n"
+	"uniform int uInputPixelFormat;\n"
 	"in vec2 vTexCoord;\n"
 	"layout(location = 0) out vec4 fragColor;\n"
 	"vec4 rec709YCbCr2rgba(float Y, float Cb, float Cr, float a)\n"
@@ -25,14 +26,116 @@ const char* kDecodeFragmentShaderSource =
 	"	Cr = (Cr * 256.0 - 16.0) / 224.0 - 0.5;\n"
 	"	return vec4(Y + 1.5748 * Cr, Y - 0.1873 * Cb - 0.4681 * Cr, Y + 1.8556 * Cb, a);\n"
 	"}\n"
+	"vec4 rec709YCbCr10_2rgba(float Y, float Cb, float Cr, float a)\n"
+	"{\n"
+	"	Y = (Y - 64.0) / 876.0;\n"
+	"	Cb = (Cb - 64.0) / 896.0 - 0.5;\n"
+	"	Cr = (Cr - 64.0) / 896.0 - 0.5;\n"
+	"	return vec4(Y + 1.5748 * Cr, Y - 0.1873 * Cb - 0.4681 * Cr, Y + 1.8556 * Cb, a);\n"
+	"}\n"
+	"uint loadV210Word(ivec2 coord)\n"
+	"{\n"
+	"	vec4 b = round(texelFetch(uPackedVideoInput, coord, 0) * 255.0);\n"
+	"	return uint(b.r) | (uint(b.g) << 8) | (uint(b.b) << 16) | (uint(b.a) << 24);\n"
+	"}\n"
+	"float v210Component(uint word, int index)\n"
+	"{\n"
+	"	return float((word >> uint(index * 10)) & 1023u);\n"
+	"}\n"
+	"vec4 decodeUyvy8(ivec2 outputCoord, ivec2 packedSize)\n"
+	"{\n"
+	"	ivec2 packedCoord = ivec2(clamp(outputCoord.x / 2, 0, packedSize.x - 1), clamp(outputCoord.y, 0, packedSize.y - 1));\n"
+	"	vec4 macroPixel = texelFetch(uPackedVideoInput, packedCoord, 0);\n"
+	"	float ySample = (outputCoord.x & 1) != 0 ? macroPixel.a : macroPixel.g;\n"
+	"	return rec709YCbCr2rgba(ySample, macroPixel.b, macroPixel.r, 1.0);\n"
+	"}\n"
+	"vec4 decodeV210(ivec2 outputCoord, ivec2 packedSize)\n"
+	"{\n"
+	"	int group = outputCoord.x / 6;\n"
+	"	int pixel = outputCoord.x - group * 6;\n"
+	"	int wordBase = group * 4;\n"
+	"	ivec2 rowBase = ivec2(wordBase, clamp(outputCoord.y, 0, packedSize.y - 1));\n"
+	"	uint w0 = loadV210Word(ivec2(min(rowBase.x + 0, packedSize.x - 1), rowBase.y));\n"
+	"	uint w1 = loadV210Word(ivec2(min(rowBase.x + 1, packedSize.x - 1), rowBase.y));\n"
+	"	uint w2 = loadV210Word(ivec2(min(rowBase.x + 2, packedSize.x - 1), rowBase.y));\n"
+	"	uint w3 = loadV210Word(ivec2(min(rowBase.x + 3, packedSize.x - 1), rowBase.y));\n"
+	"	float y0 = v210Component(w0, 1);\n"
+	"	float y1 = v210Component(w1, 0);\n"
+	"	float y2 = v210Component(w1, 2);\n"
+	"	float y3 = v210Component(w2, 1);\n"
+	"	float y4 = v210Component(w3, 0);\n"
+	"	float y5 = v210Component(w3, 2);\n"
+	"	float cb0 = v210Component(w0, 0);\n"
+	"	float cr0 = v210Component(w0, 2);\n"
+	"	float cb2 = v210Component(w1, 1);\n"
+	"	float cr2 = v210Component(w2, 0);\n"
+	"	float cb4 = v210Component(w2, 2);\n"
+	"	float cr4 = v210Component(w3, 1);\n"
+	"	float ySample = pixel == 0 ? y0 : pixel == 1 ? y1 : pixel == 2 ? y2 : pixel == 3 ? y3 : pixel == 4 ? y4 : y5;\n"
+	"	float cbSample = pixel < 2 ? cb0 : pixel < 4 ? cb2 : cb4;\n"
+	"	float crSample = pixel < 2 ? cr0 : pixel < 4 ? cr2 : cr4;\n"
+	"	return rec709YCbCr10_2rgba(ySample, cbSample, crSample, 1.0);\n"
+	"}\n"
 	"void main()\n"
 	"{\n"
 	"	vec2 correctedUv = vec2(vTexCoord.x, 1.0 - vTexCoord.y);\n"
 	"	ivec2 decodedSize = ivec2(max(uDecodedVideoResolution, vec2(1.0, 1.0)));\n"
 	"	ivec2 outputCoord = clamp(ivec2(correctedUv * vec2(decodedSize)), ivec2(0, 0), decodedSize - ivec2(1, 1));\n"
 	"	ivec2 packedSize = ivec2(max(uPackedVideoResolution, vec2(1.0, 1.0)));\n"
-	"	ivec2 packedCoord = ivec2(clamp(outputCoord.x / 2, 0, packedSize.x - 1), clamp(outputCoord.y, 0, packedSize.y - 1));\n"
+	"	fragColor = uInputPixelFormat == 1 ? decodeV210(outputCoord, packedSize) : decodeUyvy8(outputCoord, packedSize);\n"
-	"	vec4 macroPixel = texelFetch(uPackedVideoInput, packedCoord, 0);\n"
+	"}\n";
-	"	float ySample = (outputCoord.x & 1) != 0 ? macroPixel.a : macroPixel.g;\n"
+
-	"	fragColor = rec709YCbCr2rgba(ySample, macroPixel.b, macroPixel.r, 1.0);\n"
+const char* kOutputPackFragmentShaderSource =
+	"#version 430 core\n"
+	"layout(binding = 0) uniform sampler2D uOutputRgb;\n"
+	"uniform vec2 uOutputVideoResolution;\n"
+	"uniform float uActiveV210Words;\n"
+	"in vec2 vTexCoord;\n"
+	"layout(location = 0) out vec4 fragColor;\n"
+	"vec3 rgbAt(int x, int y)\n"
+	"{\n"
+	"	ivec2 size = ivec2(max(uOutputVideoResolution, vec2(1.0, 1.0)));\n"
+	"	return clamp(texelFetch(uOutputRgb, ivec2(clamp(x, 0, size.x - 1), clamp(y, 0, size.y - 1)), 0).rgb, vec3(0.0), vec3(1.0));\n"
+	"}\n"
+	"vec3 rgbToLegalYcbcr10(vec3 rgb)\n"
+	"{\n"
+	"	float y = dot(rgb, vec3(0.2126, 0.7152, 0.0722));\n"
+	"	float cb = (rgb.b - y) / 1.8556 + 0.5;\n"
+	"	float cr = (rgb.r - y) / 1.5748 + 0.5;\n"
+	"	return vec3(clamp(round(64.0 + y * 876.0), 64.0, 940.0), clamp(round(64.0 + cb * 896.0), 64.0, 960.0), clamp(round(64.0 + cr * 896.0), 64.0, 960.0));\n"
+	"}\n"
+	"uint makeWord(float a, float b, float c)\n"
+	"{\n"
+	"	return (uint(a) & 1023u) | ((uint(b) & 1023u) << 10) | ((uint(c) & 1023u) << 20);\n"
+	"}\n"
+	"vec4 wordToBytes(uint word)\n"
+	"{\n"
+	"	return vec4(float(word & 255u), float((word >> 8) & 255u), float((word >> 16) & 255u), float((word >> 24) & 255u)) / 255.0;\n"
+	"}\n"
+	"void main()\n"
+	"{\n"
+	"	ivec2 outCoord = ivec2(gl_FragCoord.xy);\n"
+	"	if (float(outCoord.x) >= uActiveV210Words)\n"
+	"	{\n"
+	"		fragColor = vec4(0.0);\n"
+	"		return;\n"
+	"	}\n"
+	"	int group = outCoord.x / 4;\n"
+	"	int wordIndex = outCoord.x - group * 4;\n"
+	"	int pixelBase = group * 6;\n"
+	"	int y = outCoord.y;\n"
+	"	vec3 c0 = rgbToLegalYcbcr10(rgbAt(pixelBase + 0, y));\n"
+	"	vec3 c1 = rgbToLegalYcbcr10(rgbAt(pixelBase + 1, y));\n"
+	"	vec3 c2 = rgbToLegalYcbcr10(rgbAt(pixelBase + 2, y));\n"
+	"	vec3 c3 = rgbToLegalYcbcr10(rgbAt(pixelBase + 3, y));\n"
+	"	vec3 c4 = rgbToLegalYcbcr10(rgbAt(pixelBase + 4, y));\n"
+	"	vec3 c5 = rgbToLegalYcbcr10(rgbAt(pixelBase + 5, y));\n"
+	"	float cb0 = round((c0.y + c1.y) * 0.5);\n"
+	"	float cr0 = round((c0.z + c1.z) * 0.5);\n"
+	"	float cb2 = round((c2.y + c3.y) * 0.5);\n"
+	"	float cr2 = round((c2.z + c3.z) * 0.5);\n"
+	"	float cb4 = round((c4.y + c5.y) * 0.5);\n"
+	"	float cr4 = round((c4.z + c5.z) * 0.5);\n"
+	"	uint word = wordIndex == 0 ? makeWord(cb0, c0.x, cr0) : wordIndex == 1 ? makeWord(c1.x, cb2, c2.x) : wordIndex == 2 ? makeWord(cr2, c3.x, cb4) : makeWord(c4.x, cr4, c5.x);\n"
+	"	fragColor = wordToBytes(word);\n"
 	"}\n";

apps/LoopThroughWithOpenGLCompositing/gl/GlShaderSources.h

@@ -2,3 +2,4 @@
 
 extern const char* kFullscreenTriangleVertexShaderSource;
 extern const char* kDecodeFragmentShaderSource;
+extern const char* kOutputPackFragmentShaderSource;

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLComposite.cpp

@@ -1,5 +1,4 @@
 #include "DeckLinkDisplayMode.h"
-#include "DeckLinkFrameTransfer.h"
 #include "DeckLinkSession.h"
 #include "OpenGLComposite.h"
 #include "GLExtensions.h"
@@ -87,17 +86,14 @@ bool OpenGLComposite::InitDeckLink()
 		MessageBoxA(NULL, initFailureReason.c_str(), title, MB_OK | MB_ICONERROR);
 		return false;
 	}
+	if (!mDeckLink->SelectPreferredFormats(videoModes, initFailureReason))
+		goto error;
 
 	if (! CheckOpenGLExtensions())
 	{
 		initFailureReason = "OpenGL extension checks failed.";
 		goto error;
 	}
-	if (mDeckLink->InputOutputDimensionsDiffer())
-	{
-		mRenderer->SetFastTransferAvailable(false);
-		OutputDebugStringA("Input/output dimensions differ; using regular OpenGL transfer fallback instead of fast transfer.\n");
-	}
 
 	if (! InitOpenGLState())
 	{
@@ -115,16 +111,6 @@ bool OpenGLComposite::InitDeckLink()
 	else
 		resizeWindow(mDeckLink->OutputFrameWidth() / 2, mDeckLink->OutputFrameHeight() / 2);
 
-	if (mRenderer->FastTransferAvailable())
-	{
-		// Initialize fast video frame transfers
-		if (! VideoFrameTransfer::initialize(mDeckLink->InputFrameWidth(), mDeckLink->InputFrameHeight(), mRenderer->CaptureTexture(), mRenderer->OutputTexture()))
-		{
-			MessageBox(NULL, _T("Cannot initialize video transfers."), _T("VideoFrameTransfer error."), MB_OK);
-			goto error;
-		}
-	}
-
 	if (!mDeckLink->ConfigureInput(this, hGLDC, hGLRC, videoModes.input, initFailureReason))
 	{
 		goto error;
@@ -222,6 +208,11 @@ bool OpenGLComposite::InitOpenGLState()
 		MessageBoxA(NULL, compilerErrorMessage, "OpenGL decode shader failed to load or compile", MB_OK);
 		return false;
 	}
+	if (!mShaderPrograms->CompileOutputPackShader(sizeof(compilerErrorMessage), compilerErrorMessage))
+	{
+		MessageBoxA(NULL, compilerErrorMessage, "OpenGL output pack shader failed to load or compile", MB_OK);
+		return false;
+	}
 
 	if (!mShaderPrograms->CompileLayerPrograms(mDeckLink->InputFrameWidth(), mDeckLink->InputFrameHeight(), sizeof(compilerErrorMessage), compilerErrorMessage))
 	{
@@ -233,7 +224,14 @@ bool OpenGLComposite::InitOpenGLState()
 	mShaderPrograms->ResetTemporalHistoryState();
 
 	std::string rendererError;
-	if (!mRenderer->InitializeResources(mDeckLink->InputFrameWidth(), mDeckLink->InputFrameHeight(), mDeckLink->OutputFrameWidth(), mDeckLink->OutputFrameHeight(), rendererError))
+	if (!mRenderer->InitializeResources(
+		mDeckLink->InputFrameWidth(),
+		mDeckLink->InputFrameHeight(),
+		mDeckLink->CaptureTextureWidth(),
+		mDeckLink->OutputFrameWidth(),
+		mDeckLink->OutputFrameHeight(),
+		mDeckLink->OutputPackTextureWidth(),
+		rendererError))
 	{
 		MessageBoxA(NULL, rendererError.c_str(), "OpenGL initialization error.", MB_OK);
 		return false;
@@ -315,6 +313,8 @@ void OpenGLComposite::renderEffect()
 		layerStates,
 		mDeckLink->InputFrameWidth(),
 		mDeckLink->InputFrameHeight(),
+		mDeckLink->CaptureTextureWidth(),
+		mDeckLink->InputPixelFormat(),
 		historyCap,
 		[this](const RuntimeRenderState& state, LayerProgram::TextBinding& textBinding, std::string& error) {
 			return mShaderPrograms->UpdateTextBindingTexture(state, textBinding, error);
@@ -392,11 +392,6 @@ void OpenGLComposite::resetTemporalHistoryState()
 
 bool OpenGLComposite::CheckOpenGLExtensions()
 {
-	mRenderer->SetFastTransferAvailable(VideoFrameTransfer::checkFastMemoryTransferAvailable());
-
-	if (!mRenderer->FastTransferAvailable())
-		OutputDebugStringA("Fast memory transfer extension not available, using regular OpenGL transfer fallback instead\n");
-
 	return true;
 }
 

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLDeckLinkBridge.cpp

@@ -1,12 +1,9 @@
 #include "OpenGLDeckLinkBridge.h"
 
-#include "DeckLinkFrameTransfer.h"
 #include "DeckLinkSession.h"
 #include "OpenGLRenderer.h"
 #include "RuntimeHost.h"
-#include "VideoFrameTransfer.h"
 
-#include <atlbase.h>
 #include <chrono>
 #include <gl/gl.h>
 
@@ -87,29 +84,20 @@ void OpenGLDeckLinkBridge::VideoFrameArrived(IDeckLinkVideoInputFrame* inputFram
 
 	wglMakeCurrent(mHdc, mHglrc);		// make OpenGL context current in this thread
 
-	if (mRenderer.FastTransferAvailable())
+	glPixelStorei(GL_UNPACK_ALIGNMENT, 4);
-	{
+	glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
-		CComQIPtr<PinnedMemoryAllocator, &IID_PinnedMemoryAllocator> allocator(inputFrameBuffer);
+	glBindBuffer(GL_PIXEL_UNPACK_BUFFER, mRenderer.TextureUploadBuffer());
-		if (!allocator || !allocator->transferFrame(videoPixels, mRenderer.CaptureTexture()))
+	glBufferData(GL_PIXEL_UNPACK_BUFFER, textureSize, videoPixels, GL_DYNAMIC_DRAW);
-			OutputDebugStringA("Capture: transferFrame() failed\n");
+	glBindTexture(GL_TEXTURE_2D, mRenderer.CaptureTexture());
 
-		allocator->waitForTransferComplete(videoPixels);
+	// NULL for last arg indicates use current GL_PIXEL_UNPACK_BUFFER target as texture data.
-	}
+	if (mDeckLink.InputPixelFormat() == VideoIOPixelFormat::V210)
+		glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, mDeckLink.CaptureTextureWidth(), mDeckLink.InputFrameHeight(), GL_RGBA, GL_UNSIGNED_BYTE, NULL);
 	else
-	{
+		glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, mDeckLink.CaptureTextureWidth(), mDeckLink.InputFrameHeight(), GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, NULL);
-		// Use a straightforward texture buffer
-		glPixelStorei(GL_UNPACK_ALIGNMENT, 4);
-		glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
-		glBindBuffer(GL_PIXEL_UNPACK_BUFFER, mRenderer.UnpinnedTextureBuffer());
-		glBufferData(GL_PIXEL_UNPACK_BUFFER, textureSize, videoPixels, GL_DYNAMIC_DRAW);
-		glBindTexture(GL_TEXTURE_2D, mRenderer.CaptureTexture());
 
-		// NULL for last arg indicates use current GL_PIXEL_UNPACK_BUFFER target as texture data
+	glBindTexture(GL_TEXTURE_2D, 0);
-		glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, mDeckLink.InputFrameWidth() / 2, mDeckLink.InputFrameHeight(), GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, NULL);
+	glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
-
-		glBindTexture(GL_TEXTURE_2D, 0);
-		glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
-	}
 
 	wglMakeCurrent(NULL, NULL);
 
@@ -135,17 +123,35 @@ void OpenGLDeckLinkBridge::PlayoutFrameCompleted(IDeckLinkVideoFrame* completedF
 
 	// Draw the effect output to the off-screen framebuffer.
 	const auto renderStartTime = std::chrono::steady_clock::now();
-	if (mRenderer.FastTransferAvailable())
-		VideoFrameTransfer::beginTextureInUse(VideoFrameTransfer::GPUtoCPU);
 	glBindFramebuffer(GL_FRAMEBUFFER, mRenderer.CompositeFramebuffer());
 	mRenderEffect();
 	glBindFramebuffer(GL_READ_FRAMEBUFFER, mRenderer.CompositeFramebuffer());
 	glBindFramebuffer(GL_DRAW_FRAMEBUFFER, mRenderer.OutputFramebuffer());
 	glBlitFramebuffer(0, 0, mDeckLink.InputFrameWidth(), mDeckLink.InputFrameHeight(), 0, 0, mDeckLink.OutputFrameWidth(), mDeckLink.OutputFrameHeight(), GL_COLOR_BUFFER_BIT, GL_LINEAR);
 	glBindFramebuffer(GL_FRAMEBUFFER, mRenderer.OutputFramebuffer());
+	if (mDeckLink.OutputIsTenBit())
+	{
+		glBindFramebuffer(GL_FRAMEBUFFER, mRenderer.OutputPackFramebuffer());
+		glViewport(0, 0, mDeckLink.OutputPackTextureWidth(), mDeckLink.OutputFrameHeight());
+		glDisable(GL_SCISSOR_TEST);
+		glDisable(GL_BLEND);
+		glDisable(GL_DEPTH_TEST);
+		glActiveTexture(GL_TEXTURE0);
+		glBindTexture(GL_TEXTURE_2D, mRenderer.OutputTexture());
+		glBindVertexArray(mRenderer.FullscreenVertexArray());
+		glUseProgram(mRenderer.OutputPackProgram());
+		const GLint outputResolutionLocation = glGetUniformLocation(mRenderer.OutputPackProgram(), "uOutputVideoResolution");
+		const GLint activeWordsLocation = glGetUniformLocation(mRenderer.OutputPackProgram(), "uActiveV210Words");
+		if (outputResolutionLocation >= 0)
+			glUniform2f(outputResolutionLocation, static_cast<float>(mDeckLink.OutputFrameWidth()), static_cast<float>(mDeckLink.OutputFrameHeight()));
+		if (activeWordsLocation >= 0)
+			glUniform1f(activeWordsLocation, static_cast<float>(ActiveV210WordsForWidth(mDeckLink.OutputFrameWidth())));
+		glDrawArrays(GL_TRIANGLES, 0, 3);
+		glUseProgram(0);
+		glBindVertexArray(0);
+		glBindTexture(GL_TEXTURE_2D, 0);
+	}
 	glFlush();
-	if (mRenderer.FastTransferAvailable())
-		VideoFrameTransfer::endTextureInUse(VideoFrameTransfer::GPUtoCPU);
 	const auto renderEndTime = std::chrono::steady_clock::now();
 	const double frameBudgetMilliseconds = mDeckLink.FrameBudgetMilliseconds();
 	const double renderMilliseconds = std::chrono::duration_cast<std::chrono::duration<double, std::milli>>(renderEndTime - renderStartTime).count();
@@ -171,23 +177,19 @@ void OpenGLDeckLinkBridge::PlayoutFrameCompleted(IDeckLinkVideoFrame* completedF
 	void*	pFrame;
 	outputVideoFrameBuffer->GetBytes(&pFrame);
 
-	if (mRenderer.FastTransferAvailable())
+	glPixelStorei(GL_PACK_ALIGNMENT, 4);
+	glPixelStorei(GL_PACK_ROW_LENGTH, 0);
+	if (mDeckLink.OutputIsTenBit())
 	{
-		if (!mDeckLink.TransferPlayoutFrame(pFrame, mRenderer.OutputTexture()))
+		glBindFramebuffer(GL_READ_FRAMEBUFFER, mRenderer.OutputPackFramebuffer());
-			OutputDebugStringA("Playback: transferFrame() failed\n");
+		glReadPixels(0, 0, mDeckLink.OutputPackTextureWidth(), mDeckLink.OutputFrameHeight(), GL_RGBA, GL_UNSIGNED_BYTE, pFrame);
-
-		// Wait for transfer to system memory to complete
-		mDeckLink.WaitForPlayoutTransferComplete(pFrame);
-		mPaint();
 	}
 	else
 	{
-		glPixelStorei(GL_PACK_ALIGNMENT, 4);
-		glPixelStorei(GL_PACK_ROW_LENGTH, 0);
 		glBindFramebuffer(GL_READ_FRAMEBUFFER, mRenderer.OutputFramebuffer());
 		glReadPixels(0, 0, mDeckLink.OutputFrameWidth(), mDeckLink.OutputFrameHeight(), GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, pFrame);
-		mPaint();
 	}
+	mPaint();
 
 	outputVideoFrameBuffer->EndAccess(bmdBufferAccessWrite);
 	outputVideoFrameBuffer->Release();

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLRenderPass.cpp

@@ -1,7 +1,6 @@
 #include "OpenGLRenderPass.h"
 
 #include "GlRenderConstants.h"
-#include "VideoFrameTransfer.h"
 
 OpenGLRenderPass::OpenGLRenderPass(OpenGLRenderer& renderer) :
 	mRenderer(renderer)
@@ -13,22 +12,18 @@ void OpenGLRenderPass::Render(
 	const std::vector<RuntimeRenderState>& layerStates,
 	unsigned inputFrameWidth,
 	unsigned inputFrameHeight,
+	unsigned captureTextureWidth,
+	VideoIOPixelFormat inputPixelFormat,
 	unsigned historyCap,
 	const TextBindingUpdater& updateTextBinding,
 	const GlobalParamsUpdater& updateGlobalParams)
 {
-	if (hasInputSource && mRenderer.FastTransferAvailable())
-	{
-		// Signal that the capture texture is about to be sampled into the composite framebuffer.
-		VideoFrameTransfer::beginTextureInUse(VideoFrameTransfer::CPUtoGPU);
-	}
-
 	glDisable(GL_SCISSOR_TEST);
 	glDisable(GL_BLEND);
 	glDisable(GL_DEPTH_TEST);
 	if (hasInputSource)
 	{
-		RenderDecodePass(inputFrameWidth, inputFrameHeight);
+		RenderDecodePass(inputFrameWidth, inputFrameHeight, captureTextureWidth, inputPixelFormat);
 	}
 	else
 	{
@@ -72,12 +67,9 @@ void OpenGLRenderPass::Render(
 	}
 
 	mRenderer.TemporalHistory().PushSourceFramebuffer(mRenderer.DecodeFramebuffer(), inputFrameWidth, inputFrameHeight);
-
-	if (hasInputSource && mRenderer.FastTransferAvailable())
-		VideoFrameTransfer::endTextureInUse(VideoFrameTransfer::CPUtoGPU);
 }
 
-void OpenGLRenderPass::RenderDecodePass(unsigned inputFrameWidth, unsigned inputFrameHeight)
+void OpenGLRenderPass::RenderDecodePass(unsigned inputFrameWidth, unsigned inputFrameHeight, unsigned captureTextureWidth, VideoIOPixelFormat inputPixelFormat)
 {
 	glBindFramebuffer(GL_FRAMEBUFFER, mRenderer.DecodeFramebuffer());
 	glViewport(0, 0, inputFrameWidth, inputFrameHeight);
@@ -89,10 +81,13 @@ void OpenGLRenderPass::RenderDecodePass(unsigned inputFrameWidth, unsigned input
 
 	const GLint packedResolutionLocation = glGetUniformLocation(mRenderer.DecodeProgram(), "uPackedVideoResolution");
 	const GLint decodedResolutionLocation = glGetUniformLocation(mRenderer.DecodeProgram(), "uDecodedVideoResolution");
+	const GLint inputPixelFormatLocation = glGetUniformLocation(mRenderer.DecodeProgram(), "uInputPixelFormat");
 	if (packedResolutionLocation >= 0)
-		glUniform2f(packedResolutionLocation, static_cast<float>(inputFrameWidth / 2), static_cast<float>(inputFrameHeight));
+		glUniform2f(packedResolutionLocation, static_cast<float>(captureTextureWidth), static_cast<float>(inputFrameHeight));
 	if (decodedResolutionLocation >= 0)
 		glUniform2f(decodedResolutionLocation, static_cast<float>(inputFrameWidth), static_cast<float>(inputFrameHeight));
+	if (inputPixelFormatLocation >= 0)
+		glUniform1i(inputPixelFormatLocation, inputPixelFormat == VideoIOPixelFormat::V210 ? 1 : 0);
 
 	glDrawArrays(GL_TRIANGLES, 0, 3);
 

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLRenderPass.h

@@ -2,6 +2,7 @@
 
 #include "OpenGLRenderer.h"
 #include "ShaderTypes.h"
+#include "VideoIOFormat.h"
 
 #include <functional>
 #include <string>
@@ -21,12 +22,14 @@ public:
 		const std::vector<RuntimeRenderState>& layerStates,
 		unsigned inputFrameWidth,
 		unsigned inputFrameHeight,
+		unsigned captureTextureWidth,
+		VideoIOPixelFormat inputPixelFormat,
 		unsigned historyCap,
 		const TextBindingUpdater& updateTextBinding,
 		const GlobalParamsUpdater& updateGlobalParams);
 
 private:
-	void RenderDecodePass(unsigned inputFrameWidth, unsigned inputFrameHeight);
+	void RenderDecodePass(unsigned inputFrameWidth, unsigned inputFrameHeight, unsigned captureTextureWidth, VideoIOPixelFormat inputPixelFormat);
 	void RenderShaderProgram(
 		GLuint sourceTexture,
 		GLuint destinationFrameBuffer,

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLRenderer.cpp

@@ -4,43 +4,52 @@
 
 namespace
 {
-	void ConfigureFrameTexture(unsigned width, unsigned height)
+	void ConfigureByteFrameTexture(unsigned width, unsigned height)
 	{
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-		glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, width, height, 0, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, NULL);
+		glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, width, height, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
+	}
+
+	void ConfigureDisplayFrameTexture(unsigned width, unsigned height)
+	{
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+		glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA16F, width, height, 0, GL_RGBA, GL_FLOAT, NULL);
 	}
 }
 
-bool OpenGLRenderer::InitializeResources(unsigned inputFrameWidth, unsigned inputFrameHeight, unsigned outputFrameWidth, unsigned outputFrameHeight, std::string& error)
+bool OpenGLRenderer::InitializeResources(unsigned inputFrameWidth, unsigned inputFrameHeight, unsigned captureTextureWidth, unsigned outputFrameWidth, unsigned outputFrameHeight, unsigned outputPackTextureWidth, std::string& error)
 {
 	glClearColor(0.0f, 0.0f, 0.0f, 0.5f);
 	glDisable(GL_DEPTH_TEST);
 
-	if (!mFastTransferExtensionAvailable)
+	glGenBuffers(1, &mTextureUploadBuffer);
-		glGenBuffers(1, &mUnpinnedTextureBuffer);
 
 	glGenTextures(1, &mCaptureTexture);
 	glBindTexture(GL_TEXTURE_2D, mCaptureTexture);
-	ConfigureFrameTexture(inputFrameWidth / 2, inputFrameHeight);
+	ConfigureByteFrameTexture(captureTextureWidth, inputFrameHeight);
 	glBindTexture(GL_TEXTURE_2D, 0);
 
 	glGenTextures(1, &mDecodedTexture);
 	glBindTexture(GL_TEXTURE_2D, mDecodedTexture);
-	ConfigureFrameTexture(inputFrameWidth, inputFrameHeight);
+	ConfigureDisplayFrameTexture(inputFrameWidth, inputFrameHeight);
 	glBindTexture(GL_TEXTURE_2D, 0);
 
 	glGenTextures(1, &mLayerTempTexture);
 	glBindTexture(GL_TEXTURE_2D, mLayerTempTexture);
-	ConfigureFrameTexture(inputFrameWidth, inputFrameHeight);
+	ConfigureDisplayFrameTexture(inputFrameWidth, inputFrameHeight);
 	glBindTexture(GL_TEXTURE_2D, 0);
 
 	glGenFramebuffers(1, &mDecodeFrameBuf);
 	glGenFramebuffers(1, &mLayerTempFrameBuf);
 	glGenFramebuffers(1, &mIdFrameBuf);
 	glGenFramebuffers(1, &mOutputFrameBuf);
+	glGenFramebuffers(1, &mOutputPackFrameBuf);
 	glGenRenderbuffers(1, &mIdColorBuf);
 	glGenRenderbuffers(1, &mIdDepthBuf);
 	glGenVertexArrays(1, &mFullscreenVAO);
@@ -65,7 +74,7 @@ bool OpenGLRenderer::InitializeResources(unsigned inputFrameWidth, unsigned inpu
 	glBindFramebuffer(GL_FRAMEBUFFER, mIdFrameBuf);
 	glGenTextures(1, &mFBOTexture);
 	glBindTexture(GL_TEXTURE_2D, mFBOTexture);
-	ConfigureFrameTexture(inputFrameWidth, inputFrameHeight);
+	ConfigureDisplayFrameTexture(inputFrameWidth, inputFrameHeight);
 
 	glBindRenderbuffer(GL_RENDERBUFFER, mIdDepthBuf);
 	glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT24, inputFrameWidth, inputFrameHeight);
@@ -79,7 +88,7 @@ bool OpenGLRenderer::InitializeResources(unsigned inputFrameWidth, unsigned inpu
 
 	glGenTextures(1, &mOutputTexture);
 	glBindTexture(GL_TEXTURE_2D, mOutputTexture);
-	ConfigureFrameTexture(outputFrameWidth, outputFrameHeight);
+	ConfigureDisplayFrameTexture(outputFrameWidth, outputFrameHeight);
 
 	glBindFramebuffer(GL_FRAMEBUFFER, mOutputFrameBuf);
 	glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, mOutputTexture, 0);
@@ -89,6 +98,18 @@ bool OpenGLRenderer::InitializeResources(unsigned inputFrameWidth, unsigned inpu
 		return false;
 	}
 
+	glGenTextures(1, &mOutputPackTexture);
+	glBindTexture(GL_TEXTURE_2D, mOutputPackTexture);
+	ConfigureByteFrameTexture(outputPackTextureWidth, outputFrameHeight);
+
+	glBindFramebuffer(GL_FRAMEBUFFER, mOutputPackFrameBuf);
+	glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, mOutputPackTexture, 0);
+	if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE)
+	{
+		error = "Cannot initialize output pack framebuffer.";
+		return false;
+	}
+
 	glBindTexture(GL_TEXTURE_2D, 0);
 	glBindRenderbuffer(GL_RENDERBUFFER, 0);
 	glBindFramebuffer(GL_FRAMEBUFFER, 0);
@@ -109,6 +130,13 @@ void OpenGLRenderer::SetDecodeShaderProgram(GLuint program, GLuint vertexShader,
 	mDecodeFragmentShader = fragmentShader;
 }
 
+void OpenGLRenderer::SetOutputPackShaderProgram(GLuint program, GLuint vertexShader, GLuint fragmentShader)
+{
+	mOutputPackProgram = program;
+	mOutputPackVertexShader = vertexShader;
+	mOutputPackFragmentShader = fragmentShader;
+}
+
 void OpenGLRenderer::ResizeView(int width, int height)
 {
 	mViewWidth = width;
@@ -166,6 +194,8 @@ void OpenGLRenderer::DestroyResources()
 		glDeleteFramebuffers(1, &mIdFrameBuf);
 	if (mOutputFrameBuf != 0)
 		glDeleteFramebuffers(1, &mOutputFrameBuf);
+	if (mOutputPackFrameBuf != 0)
+		glDeleteFramebuffers(1, &mOutputPackFrameBuf);
 	if (mIdColorBuf != 0)
 		glDeleteRenderbuffers(1, &mIdColorBuf);
 	if (mIdDepthBuf != 0)
@@ -180,8 +210,10 @@ void OpenGLRenderer::DestroyResources()
 		glDeleteTextures(1, &mFBOTexture);
 	if (mOutputTexture != 0)
 		glDeleteTextures(1, &mOutputTexture);
-	if (mUnpinnedTextureBuffer != 0)
+	if (mOutputPackTexture != 0)
-		glDeleteBuffers(1, &mUnpinnedTextureBuffer);
+		glDeleteTextures(1, &mOutputPackTexture);
+	if (mTextureUploadBuffer != 0)
+		glDeleteBuffers(1, &mTextureUploadBuffer);
 
 	mFullscreenVAO = 0;
 	mGlobalParamsUBO = 0;
@@ -189,6 +221,7 @@ void OpenGLRenderer::DestroyResources()
 	mLayerTempFrameBuf = 0;
 	mIdFrameBuf = 0;
 	mOutputFrameBuf = 0;
+	mOutputPackFrameBuf = 0;
 	mIdColorBuf = 0;
 	mIdDepthBuf = 0;
 	mCaptureTexture = 0;
@@ -196,12 +229,14 @@ void OpenGLRenderer::DestroyResources()
 	mLayerTempTexture = 0;
 	mFBOTexture = 0;
 	mOutputTexture = 0;
-	mUnpinnedTextureBuffer = 0;
+	mOutputPackTexture = 0;
+	mTextureUploadBuffer = 0;
 	mGlobalParamsUBOSize = 0;
 
 	mTemporalHistory.DestroyResources();
 	DestroyLayerPrograms();
 	DestroyDecodeShaderProgram();
+	DestroyOutputPackShaderProgram();
 }
 
 void OpenGLRenderer::DestroySingleLayerProgram(LayerProgram& layerProgram)
@@ -272,3 +307,24 @@ void OpenGLRenderer::DestroyDecodeShaderProgram()
 		mDecodeVertexShader = 0;
 	}
 }
+
+void OpenGLRenderer::DestroyOutputPackShaderProgram()
+{
+	if (mOutputPackProgram != 0)
+	{
+		glDeleteProgram(mOutputPackProgram);
+		mOutputPackProgram = 0;
+	}
+
+	if (mOutputPackFragmentShader != 0)
+	{
+		glDeleteShader(mOutputPackFragmentShader);
+		mOutputPackFragmentShader = 0;
+	}
+
+	if (mOutputPackVertexShader != 0)
+	{
+		glDeleteShader(mOutputPackVertexShader);
+		mOutputPackVertexShader = 0;
+	}
+}

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLRenderer.h

@@ -44,21 +44,22 @@ public:
 		std::vector<TextBinding> textBindings;
 	};
 
-	bool FastTransferAvailable() const { return mFastTransferExtensionAvailable; }
-	void SetFastTransferAvailable(bool available) { mFastTransferExtensionAvailable = available; }
 	GLuint CaptureTexture() const { return mCaptureTexture; }
 	GLuint DecodedTexture() const { return mDecodedTexture; }
 	GLuint LayerTempTexture() const { return mLayerTempTexture; }
 	GLuint CompositeTexture() const { return mFBOTexture; }
 	GLuint OutputTexture() const { return mOutputTexture; }
-	GLuint UnpinnedTextureBuffer() const { return mUnpinnedTextureBuffer; }
+	GLuint OutputPackTexture() const { return mOutputPackTexture; }
+	GLuint TextureUploadBuffer() const { return mTextureUploadBuffer; }
 	GLuint DecodeFramebuffer() const { return mDecodeFrameBuf; }
 	GLuint LayerTempFramebuffer() const { return mLayerTempFrameBuf; }
 	GLuint CompositeFramebuffer() const { return mIdFrameBuf; }
 	GLuint OutputFramebuffer() const { return mOutputFrameBuf; }
+	GLuint OutputPackFramebuffer() const { return mOutputPackFrameBuf; }
 	GLuint FullscreenVertexArray() const { return mFullscreenVAO; }
 	GLuint GlobalParamsUBO() const { return mGlobalParamsUBO; }
 	GLuint DecodeProgram() const { return mDecodeProgram; }
+	GLuint OutputPackProgram() const { return mOutputPackProgram; }
 	GLsizeiptr GlobalParamsUBOSize() const { return mGlobalParamsUBOSize; }
 	void SetGlobalParamsUBOSize(GLsizeiptr size) { mGlobalParamsUBOSize = size; }
 	void ReplaceLayerPrograms(std::vector<LayerProgram>& newPrograms) { mLayerPrograms.swap(newPrograms); }
@@ -67,26 +68,29 @@ public:
 	TemporalHistoryBuffers& TemporalHistory() { return mTemporalHistory; }
 	const TemporalHistoryBuffers& TemporalHistory() const { return mTemporalHistory; }
 	void SetDecodeShaderProgram(GLuint program, GLuint vertexShader, GLuint fragmentShader);
-	bool InitializeResources(unsigned inputFrameWidth, unsigned inputFrameHeight, unsigned outputFrameWidth, unsigned outputFrameHeight, std::string& error);
+	void SetOutputPackShaderProgram(GLuint program, GLuint vertexShader, GLuint fragmentShader);
+	bool InitializeResources(unsigned inputFrameWidth, unsigned inputFrameHeight, unsigned captureTextureWidth, unsigned outputFrameWidth, unsigned outputFrameHeight, unsigned outputPackTextureWidth, std::string& error);
 	void ResizeView(int width, int height);
 	void PresentToWindow(HDC hdc, unsigned outputFrameWidth, unsigned outputFrameHeight);
 	void DestroyResources();
 	void DestroySingleLayerProgram(LayerProgram& layerProgram);
 	void DestroyLayerPrograms();
 	void DestroyDecodeShaderProgram();
+	void DestroyOutputPackShaderProgram();
 
 private:
-	bool mFastTransferExtensionAvailable = false;
 	GLuint mCaptureTexture = 0;
 	GLuint mDecodedTexture = 0;
 	GLuint mLayerTempTexture = 0;
 	GLuint mFBOTexture = 0;
 	GLuint mOutputTexture = 0;
-	GLuint mUnpinnedTextureBuffer = 0;
+	GLuint mOutputPackTexture = 0;
+	GLuint mTextureUploadBuffer = 0;
 	GLuint mDecodeFrameBuf = 0;
 	GLuint mLayerTempFrameBuf = 0;
 	GLuint mIdFrameBuf = 0;
 	GLuint mOutputFrameBuf = 0;
+	GLuint mOutputPackFrameBuf = 0;
 	GLuint mIdColorBuf = 0;
 	GLuint mIdDepthBuf = 0;
 	GLuint mFullscreenVAO = 0;
@@ -94,6 +98,9 @@ private:
 	GLuint mDecodeProgram = 0;
 	GLuint mDecodeVertexShader = 0;
 	GLuint mDecodeFragmentShader = 0;
+	GLuint mOutputPackProgram = 0;
+	GLuint mOutputPackVertexShader = 0;
+	GLuint mOutputPackFragmentShader = 0;
 	GLsizeiptr mGlobalParamsUBOSize = 0;
 	int mViewWidth = 0;
 	int mViewHeight = 0;

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLShaderPrograms.cpp

@@ -115,6 +115,11 @@ bool OpenGLShaderPrograms::CompileDecodeShader(int errorMessageSize, char* error
 	return mCompiler.CompileDecodeShader(errorMessageSize, errorMessage);
 }
 
+bool OpenGLShaderPrograms::CompileOutputPackShader(int errorMessageSize, char* errorMessage)
+{
+	return mCompiler.CompileOutputPackShader(errorMessageSize, errorMessage);
+}
+
 void OpenGLShaderPrograms::DestroySingleLayerProgram(LayerProgram& layerProgram)
 {
 	mRenderer.DestroySingleLayerProgram(layerProgram);

apps/LoopThroughWithOpenGLCompositing/gl/OpenGLShaderPrograms.h

@@ -20,6 +20,7 @@ public:
 	bool CompileLayerPrograms(unsigned inputFrameWidth, unsigned inputFrameHeight, int errorMessageSize, char* errorMessage);
 	bool CommitPreparedLayerPrograms(const PreparedShaderBuild& preparedBuild, unsigned inputFrameWidth, unsigned inputFrameHeight, int errorMessageSize, char* errorMessage);
 	bool CompileDecodeShader(int errorMessageSize, char* errorMessage);
+	bool CompileOutputPackShader(int errorMessageSize, char* errorMessage);
 	void DestroyLayerPrograms();
 	void DestroySingleLayerProgram(LayerProgram& layerProgram);
 	void DestroyDecodeShaderProgram();

apps/LoopThroughWithOpenGLCompositing/gl/ShaderProgramCompiler.cpp

@@ -192,3 +192,53 @@ bool ShaderProgramCompiler::CompileDecodeShader(int errorMessageSize, char* erro
 	mRenderer.SetDecodeShaderProgram(newProgram.release(), newVertexShader.release(), newFragmentShader.release());
 	return true;
 }
+
+bool ShaderProgramCompiler::CompileOutputPackShader(int errorMessageSize, char* errorMessage)
+{
+	GLsizei errorBufferSize = 0;
+	GLint compileResult = GL_FALSE;
+	GLint linkResult = GL_FALSE;
+	const char* vertexSource = kFullscreenTriangleVertexShaderSource;
+	const char* fragmentSource = kOutputPackFragmentShaderSource;
+
+	ScopedGlShader newVertexShader(glCreateShader(GL_VERTEX_SHADER));
+	glShaderSource(newVertexShader.get(), 1, (const GLchar**)&vertexSource, NULL);
+	glCompileShader(newVertexShader.get());
+	glGetShaderiv(newVertexShader.get(), GL_COMPILE_STATUS, &compileResult);
+	if (compileResult == GL_FALSE)
+	{
+		glGetShaderInfoLog(newVertexShader.get(), errorMessageSize, &errorBufferSize, errorMessage);
+		return false;
+	}
+
+	ScopedGlShader newFragmentShader(glCreateShader(GL_FRAGMENT_SHADER));
+	glShaderSource(newFragmentShader.get(), 1, (const GLchar**)&fragmentSource, NULL);
+	glCompileShader(newFragmentShader.get());
+	glGetShaderiv(newFragmentShader.get(), GL_COMPILE_STATUS, &compileResult);
+	if (compileResult == GL_FALSE)
+	{
+		glGetShaderInfoLog(newFragmentShader.get(), errorMessageSize, &errorBufferSize, errorMessage);
+		return false;
+	}
+
+	ScopedGlProgram newProgram(glCreateProgram());
+	glAttachShader(newProgram.get(), newVertexShader.get());
+	glAttachShader(newProgram.get(), newFragmentShader.get());
+	glLinkProgram(newProgram.get());
+	glGetProgramiv(newProgram.get(), GL_LINK_STATUS, &linkResult);
+	if (linkResult == GL_FALSE)
+	{
+		glGetProgramInfoLog(newProgram.get(), errorMessageSize, &errorBufferSize, errorMessage);
+		return false;
+	}
+
+	glUseProgram(newProgram.get());
+	const GLint outputSamplerLocation = glGetUniformLocation(newProgram.get(), "uOutputRgb");
+	if (outputSamplerLocation >= 0)
+		glUniform1i(outputSamplerLocation, 0);
+	glUseProgram(0);
+
+	mRenderer.DestroyOutputPackShaderProgram();
+	mRenderer.SetOutputPackShaderProgram(newProgram.release(), newVertexShader.release(), newFragmentShader.release());
+	return true;
+}

apps/LoopThroughWithOpenGLCompositing/gl/ShaderProgramCompiler.h

@@ -16,6 +16,7 @@ public:
 	bool CompileLayerProgram(const RuntimeRenderState& state, LayerProgram& layerProgram, int errorMessageSize, char* errorMessage);
 	bool CompilePreparedLayerProgram(const RuntimeRenderState& state, const std::string& fragmentShaderSource, LayerProgram& layerProgram, int errorMessageSize, char* errorMessage);
 	bool CompileDecodeShader(int errorMessageSize, char* errorMessage);
+	bool CompileOutputPackShader(int errorMessageSize, char* errorMessage);
 
 private:
 	OpenGLRenderer& mRenderer;

apps/LoopThroughWithOpenGLCompositing/gl/TemporalHistoryBuffers.cpp

@@ -105,7 +105,7 @@ bool TemporalHistoryBuffers::CreateRing(Ring& ring, unsigned effectiveLength, Te
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-		glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, frameWidth, frameHeight, 0, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, NULL);
+		glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA16F, frameWidth, frameHeight, 0, GL_RGBA, GL_FLOAT, NULL);
 
 		glGenFramebuffers(1, &slot.framebuffer);
 		glBindFramebuffer(GL_FRAMEBUFFER, slot.framebuffer);

apps/LoopThroughWithOpenGLCompositing/gl/TextureAssetLoader.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include "GLExtensions.h"
 #include "ShaderTypes.h"
 
 #include <windows.h>

apps/LoopThroughWithOpenGLCompositing/gl/VideoFrameTransfer.cpp

@@ -1,377 +0,0 @@
-/* -LICENSE-START-
- ** Copyright (c) 2012 Blackmagic Design
- **  
- ** Permission is hereby granted, free of charge, to any person or organization 
- ** obtaining a copy of the software and accompanying documentation (the 
- ** "Software") to use, reproduce, display, distribute, sub-license, execute, 
- ** and transmit the Software, and to prepare derivative works of the Software, 
- ** and to permit third-parties to whom the Software is furnished to do so, in 
- ** accordance with:
- ** 
- ** (1) if the Software is obtained from Blackmagic Design, the End User License 
- ** Agreement for the Software Development Kit ("EULA") available at 
- ** https://www.blackmagicdesign.com/EULA/DeckLinkSDK; or
- ** 
- ** (2) if the Software is obtained from any third party, such licensing terms 
- ** as notified by that third party,
- ** 
- ** and all subject to the following:
- ** 
- ** (3) the copyright notices in the Software and this entire statement, 
- ** including the above license grant, this restriction and the following 
- ** disclaimer, must be included in all copies of the Software, in whole or in 
- ** part, and all derivative works of the Software, unless such copies or 
- ** derivative works are solely in the form of machine-executable object code 
- ** generated by a source language processor.
- ** 
- ** (4) THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 
- ** OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 
- ** FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT 
- ** SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE 
- ** FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE, 
- ** ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 
- ** DEALINGS IN THE SOFTWARE.
- ** 
- ** A copy of the Software is available free of charge at 
- ** https://www.blackmagicdesign.com/desktopvideo_sdk under the EULA.
- ** 
- ** -LICENSE-END-
- */
-
-#include "VideoFrameTransfer.h"
-#include "NativeHandles.h"
-
-
-#define DVP_CHECK(cmd) {						\
-    DVPStatus hr = (cmd);						\
-    if (DVP_STATUS_OK != hr) {                  \
-		OutputDebugStringA( #cmd " failed\n" );   \
-		ExitProcess(hr);						\
-    }                                           \
-}
-
-
-// Initialise static members
-bool								VideoFrameTransfer::mInitialized = false;
-bool								VideoFrameTransfer::mUseDvp = false;
-unsigned							VideoFrameTransfer::mWidth = 0;
-unsigned							VideoFrameTransfer::mHeight = 0;
-GLuint								VideoFrameTransfer::mCaptureTexture = 0;
-
-// NVIDIA specific static members
-DVPBufferHandle						VideoFrameTransfer::mDvpCaptureTextureHandle = 0;
-DVPBufferHandle						VideoFrameTransfer::mDvpPlaybackTextureHandle = 0;
-uint32_t							VideoFrameTransfer::mBufferAddrAlignment = 0;
-uint32_t							VideoFrameTransfer::mBufferGpuStrideAlignment = 0;
-uint32_t							VideoFrameTransfer::mSemaphoreAddrAlignment = 0;
-uint32_t							VideoFrameTransfer::mSemaphoreAllocSize = 0;
-uint32_t							VideoFrameTransfer::mSemaphorePayloadOffset = 0;
-uint32_t							VideoFrameTransfer::mSemaphorePayloadSize = 0;
-
-
-bool VideoFrameTransfer::isNvidiaDvpAvailable()
-{
-	// Look for supported graphics boards
-	const GLubyte* renderer = glGetString(GL_RENDERER);
-	if (renderer == NULL)
-		return false;
-
-	bool hasDvp = (strstr((char*)renderer, "Quadro") != NULL);
-	return hasDvp;
-}
-
-bool VideoFrameTransfer::isAMDPinnedMemoryAvailable()
-{
-	// GL_AMD_pinned_memory presence indicates GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD buffer target is supported
-	const GLubyte* strExt = glGetString(GL_EXTENSIONS);
-	if (strExt == NULL)
-	{
-		// In a core profile context GL_EXTENSIONS is no longer queryable via glGetString().
-		// Treat this as "extension unavailable" for now; the fast-transfer path is optional.
-		return false;
-	}
-
-	bool hasAMDPinned = (strstr((char*)strExt, "GL_AMD_pinned_memory") != NULL);
-	return hasAMDPinned;
-}
-
-bool VideoFrameTransfer::checkFastMemoryTransferAvailable()
-{
-	return (isNvidiaDvpAvailable() || isAMDPinnedMemoryAvailable());
-}
-
-bool VideoFrameTransfer::initialize(unsigned width, unsigned height, GLuint captureTexture, GLuint playbackTexture)
-{
-	if (mInitialized)
-		return false;
-
-	bool hasDvp = isNvidiaDvpAvailable();
-	bool hasAMDPinned = isAMDPinnedMemoryAvailable();
-
-	if (!hasDvp && !hasAMDPinned)
-		return false;
-
-	mUseDvp = hasDvp;
-	mWidth = width;
-	mHeight = height;
-	mCaptureTexture = captureTexture;
-
-	if (! initializeMemoryLocking(mWidth * mHeight * 4))		// BGRA uses 4 bytes per pixel
-		return false;
-
-	if (mUseDvp)
-	{
-		// DVP initialisation
-		DVP_CHECK(dvpInitGLContext(DVP_DEVICE_FLAGS_SHARE_APP_CONTEXT));
-		DVP_CHECK(dvpGetRequiredConstantsGLCtx(	&mBufferAddrAlignment, &mBufferGpuStrideAlignment,
-												&mSemaphoreAddrAlignment, &mSemaphoreAllocSize,
-												&mSemaphorePayloadOffset, &mSemaphorePayloadSize));
-	
-		// Register textures with DVP
-		DVP_CHECK(dvpCreateGPUTextureGL(captureTexture, &mDvpCaptureTextureHandle));
-		DVP_CHECK(dvpCreateGPUTextureGL(playbackTexture, &mDvpPlaybackTextureHandle));
-	}
-
-	mInitialized = true;
-
-	return true;
-}
-
-bool VideoFrameTransfer::initializeMemoryLocking(unsigned memSize)
-{
-	// Increase the process working set size to allow pinning of memory.
-    static SIZE_T	dwMin = 0, dwMax = 0;
-    UniqueHandle processHandle(OpenProcess(PROCESS_QUERY_INFORMATION | PROCESS_SET_QUOTA, FALSE, GetCurrentProcessId()));
-    if (!processHandle.valid())
-		return false;
-
-    // Retrieve the working set size of the process.
-    if (!dwMin && !GetProcessWorkingSetSize(processHandle.get(), &dwMin, &dwMax))
-		return false;
-
-	// Allow for 80 frames to be locked
-	BOOL res = SetProcessWorkingSetSize(processHandle.get(), memSize * 80 + dwMin, memSize * 80 + (dwMax-dwMin));
-    if (!res)
-		return false;
-
-	return true;
-}
-
-// SyncInfo sets up a semaphore which is shared between the GPU and CPU and used to
-// synchronise access to DVP buffers.
-struct SyncInfo
-{
-	SyncInfo(uint32_t semaphoreAllocSize, uint32_t semaphoreAddrAlignment);
-	~SyncInfo();
-
-	volatile uint32_t*	mSem; 
-	volatile uint32_t	mReleaseValue; 
-	volatile uint32_t	mAcquireValue;
-	DVPSyncObjectHandle	mDvpSync;
-};
-
-SyncInfo::SyncInfo(uint32_t semaphoreAllocSize, uint32_t semaphoreAddrAlignment)
-{
-	mSem = (uint32_t*)_aligned_malloc(semaphoreAllocSize, semaphoreAddrAlignment);
-    
-	// Initialise
-	mSem[0] = 0;
-	mReleaseValue = 0;
-	mAcquireValue = 0;
-
-	// Setup DVP sync object and import it
-	DVPSyncObjectDesc syncObjectDesc;
-	syncObjectDesc.externalClientWaitFunc = NULL;
-	syncObjectDesc.sem = (uint32_t*)mSem;
-
-	DVP_CHECK(dvpImportSyncObject(&syncObjectDesc, &mDvpSync));
-}
-
-SyncInfo::~SyncInfo()
-{
-	DVP_CHECK(dvpFreeSyncObject(mDvpSync));
-	_aligned_free((void*)mSem);
-}
-
-VideoFrameTransfer::VideoFrameTransfer(unsigned long memSize, void* address, Direction direction) :
-	mBuffer(address),
-	mMemSize(memSize),
-	mDirection(direction),
-	mExtSync(NULL),
-	mGpuSync(NULL),
-	mDvpSysMemHandle(0),
-	mBufferHandle(0)
-{
-	if (mUseDvp)
-	{
-		// Pin the memory
-		if (! VirtualLock(mBuffer, mMemSize))
-			throw std::runtime_error("Error pinning memory with VirtualLock");
-
-		// Create necessary sysmem and gpu sync objects
-		mExtSync = new SyncInfo(mSemaphoreAllocSize, mSemaphoreAddrAlignment);
-		mGpuSync = new SyncInfo(mSemaphoreAllocSize, mSemaphoreAddrAlignment);
-
-		// Register system memory buffers with DVP
-		DVPSysmemBufferDesc sysMemBuffersDesc;
-		sysMemBuffersDesc.width		= mWidth;
-		sysMemBuffersDesc.height	= mHeight;
-		sysMemBuffersDesc.stride	= mWidth * 4;
-		sysMemBuffersDesc.format	= DVP_BGRA;
-		sysMemBuffersDesc.type		= DVP_UNSIGNED_BYTE;
-		sysMemBuffersDesc.size		= mMemSize;
-		sysMemBuffersDesc.bufAddr	= mBuffer;
-
-		if (mDirection == CPUtoGPU)
-		{
-			// A UYVY 4:2:2 frame is transferred to the GPU, rather than RGB 4:4:4, so width is halved
-			sysMemBuffersDesc.width /= 2;
-			sysMemBuffersDesc.stride /= 2;
-		}
-
-		DVP_CHECK(dvpCreateBuffer(&sysMemBuffersDesc, &mDvpSysMemHandle));
-		DVP_CHECK(dvpBindToGLCtx(mDvpSysMemHandle));
-	}
-	else
-	{
-		// Create an OpenGL buffer handle to use for pinned memory
-		GLuint bufferHandle;
-		glGenBuffers(1, &bufferHandle);
-
-		// Pin memory by binding buffer to special AMD target.
-		glBindBuffer(GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD, bufferHandle);
-
-		// glBufferData() sets up the address so any OpenGL operation on this buffer will use system memory directly
-		// (assumes address is aligned to 4k boundary).
-		glBufferData(GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD, mMemSize, address, GL_STREAM_DRAW);
-		GLenum result = glGetError();
-		if (result != GL_NO_ERROR)
-		{
-			throw std::runtime_error("Error pinning memory with glBufferData(GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD, ...)");
-		}
-		glBindBuffer(GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD, 0);		// Unbind buffer to target
-
-		mBufferHandle = bufferHandle;
-	}
-}
-
-VideoFrameTransfer::~VideoFrameTransfer()
-{
-	if (mUseDvp)
-	{
-		DVP_CHECK(dvpUnbindFromGLCtx(mDvpSysMemHandle));
-		DVP_CHECK(dvpDestroyBuffer(mDvpSysMemHandle));
-
-		delete mExtSync;
-		delete mGpuSync;
-
-		VirtualUnlock(mBuffer, mMemSize);
-	}
-	else
-	{
-		// The buffer is un-pinned by the GPU when the buffer is deleted
-		glDeleteBuffers(1, &mBufferHandle);
-	}
-}
-
-bool VideoFrameTransfer::performFrameTransfer()
-{
-	if (mUseDvp)
-	{
-		// NVIDIA DVP transfers
-		DVPStatus status;
-
-		mGpuSync->mReleaseValue++;
-
-		dvpBegin();
-		if (mDirection == CPUtoGPU)
-		{
-			// Copy from system memory to GPU texture
-			dvpMapBufferWaitDVP(mDvpCaptureTextureHandle);
-			status = dvpMemcpyLined(	mDvpSysMemHandle, mExtSync->mDvpSync, mExtSync->mAcquireValue, DVP_TIMEOUT_IGNORED,
-										mDvpCaptureTextureHandle, mGpuSync->mDvpSync, mGpuSync->mReleaseValue, 0, mHeight);
-			dvpMapBufferEndDVP(mDvpCaptureTextureHandle);
-		}
-		else
-		{
-			// Copy from GPU texture to system memory
-			dvpMapBufferWaitDVP(mDvpPlaybackTextureHandle);
-			status = dvpMemcpyLined(	mDvpPlaybackTextureHandle, mExtSync->mDvpSync, mExtSync->mReleaseValue, DVP_TIMEOUT_IGNORED,
-										mDvpSysMemHandle, mGpuSync->mDvpSync, mGpuSync->mReleaseValue, 0, mHeight);
-			dvpMapBufferEndDVP(mDvpPlaybackTextureHandle);
-		}
-		dvpEnd();
-
-		return (status == DVP_STATUS_OK);
-	}
-	else
-	{
-		// AMD pinned memory transfers
-		if (mDirection == CPUtoGPU)
-		{
-			glEnable(GL_TEXTURE_2D);
-
-			// Use a pinned buffer for the GL_PIXEL_UNPACK_BUFFER target
-			glBindBuffer(GL_PIXEL_UNPACK_BUFFER, mBufferHandle);
-			glBindTexture(GL_TEXTURE_2D, mCaptureTexture);
-
-			// NULL for last arg indicates use current GL_PIXEL_UNPACK_BUFFER target as texture data
-			glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, mWidth/2, mHeight, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, NULL);
-
-			// Ensure pinned texture has been transferred to GPU before we draw with it
-			GLsync fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
-			glClientWaitSync(fence, GL_SYNC_FLUSH_COMMANDS_BIT, 40 * 1000 * 1000);	// timeout in nanosec
-			glDeleteSync(fence);
-
-			glBindTexture(GL_TEXTURE_2D, 0);
-			glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
-			glDisable(GL_TEXTURE_2D);
-		}
-		else
-		{
-			// Use a PIXEL PACK BUFFER to read back pixels
-			glBindBuffer(GL_PIXEL_PACK_BUFFER, mBufferHandle);
-			glReadPixels(0, 0, mWidth, mHeight, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, NULL);
-
-			// Ensure GPU has processed all commands in the pipeline up to this point, before memory is read by the CPU
-			GLsync fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
-			glClientWaitSync(fence, GL_SYNC_FLUSH_COMMANDS_BIT, 40 * 1000 * 1000);	// timeout in nanosec
-			glDeleteSync(fence);
-		}
-
-		return (glGetError() == GL_NO_ERROR);
-	}
-}
-
-void VideoFrameTransfer::waitForTransferComplete()
-{
-	if (!mUseDvp)
-		return;
-
-	// Block until buffer has completely transferred between GPU and CPU buffer
-	dvpBegin();
-	dvpSyncObjClientWaitComplete(mGpuSync->mDvpSync, DVP_TIMEOUT_IGNORED);
-	dvpEnd();
-}
-
-void VideoFrameTransfer::beginTextureInUse(Direction direction)
-{
-	if (!mUseDvp)
-		return;
-
-	if (direction == CPUtoGPU)
-		dvpMapBufferWaitAPI(mDvpCaptureTextureHandle);
-	else
-		dvpMapBufferWaitAPI(mDvpPlaybackTextureHandle);
-}
-
-void VideoFrameTransfer::endTextureInUse(Direction direction)
-{
-	if (!mUseDvp)
-		return;
-
-	if (direction == CPUtoGPU)
-		dvpMapBufferEndAPI(mDvpCaptureTextureHandle);
-	else
-		dvpMapBufferEndAPI(mDvpPlaybackTextureHandle);
-}

apps/LoopThroughWithOpenGLCompositing/gl/VideoFrameTransfer.h

@@ -1,109 +0,0 @@
-/* -LICENSE-START-
- ** Copyright (c) 2012 Blackmagic Design
- **  
- ** Permission is hereby granted, free of charge, to any person or organization 
- ** obtaining a copy of the software and accompanying documentation (the 
- ** "Software") to use, reproduce, display, distribute, sub-license, execute, 
- ** and transmit the Software, and to prepare derivative works of the Software, 
- ** and to permit third-parties to whom the Software is furnished to do so, in 
- ** accordance with:
- ** 
- ** (1) if the Software is obtained from Blackmagic Design, the End User License 
- ** Agreement for the Software Development Kit ("EULA") available at 
- ** https://www.blackmagicdesign.com/EULA/DeckLinkSDK; or
- ** 
- ** (2) if the Software is obtained from any third party, such licensing terms 
- ** as notified by that third party,
- ** 
- ** and all subject to the following:
- ** 
- ** (3) the copyright notices in the Software and this entire statement, 
- ** including the above license grant, this restriction and the following 
- ** disclaimer, must be included in all copies of the Software, in whole or in 
- ** part, and all derivative works of the Software, unless such copies or 
- ** derivative works are solely in the form of machine-executable object code 
- ** generated by a source language processor.
- ** 
- ** (4) THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 
- ** OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 
- ** FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT 
- ** SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE 
- ** FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE, 
- ** ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 
- ** DEALINGS IN THE SOFTWARE.
- ** 
- ** A copy of the Software is available free of charge at 
- ** https://www.blackmagicdesign.com/desktopvideo_sdk under the EULA.
- ** 
- ** -LICENSE-END-
- */
-#ifndef __VIDEO_FRAME_TRANSFER_H__
-#define __VIDEO_FRAME_TRANSFER_H__
-
-#include "GLExtensions.h"
-#include <stdexcept>
-#include <map>
-
-// NVIDIA GPU Direct For Video with OpenGL requires the following two headers.
-// See the NVIDIA website to check if your graphics card is supported.
-#include <DVPAPI.h>
-#include <dvpapi_gl.h>
-
-struct SyncInfo;
-
-
-// Class for performing efficient frame memory transfers between the CPU and GPU,
-// using NVIDIA and AMD extensions.
-class VideoFrameTransfer
-{
-public:
-	enum Direction
-	{
-		CPUtoGPU,
-		GPUtoCPU
-	};
-
-	VideoFrameTransfer(unsigned long memSize, void* address, Direction direction);
-	~VideoFrameTransfer();
-
-	static bool checkFastMemoryTransferAvailable();
-	static bool initialize(unsigned width, unsigned height, GLuint captureTexture, GLuint playbackTexture);
-	static void beginTextureInUse(Direction direction);
-	static void endTextureInUse(Direction direction);
-	
-	bool performFrameTransfer();
-	void waitForTransferComplete();
-
-private:
-	static bool isNvidiaDvpAvailable();
-	static bool isAMDPinnedMemoryAvailable();
-	static bool initializeMemoryLocking(unsigned memSize);
-
-	void*						mBuffer;
-	unsigned long				mMemSize;
-	Direction					mDirection;
-	static bool					mInitialized;
-	static bool					mUseDvp;
-	static unsigned				mWidth;
-	static unsigned				mHeight;
-	static GLuint				mCaptureTexture;
-
-	// NVIDIA GPU Direct for Video support
-	SyncInfo*					mExtSync;
-	SyncInfo*					mGpuSync;
-	DVPBufferHandle				mDvpSysMemHandle;
-
-	static DVPBufferHandle		mDvpCaptureTextureHandle;
-	static DVPBufferHandle		mDvpPlaybackTextureHandle;
-	static uint32_t				mBufferAddrAlignment;
-	static uint32_t				mBufferGpuStrideAlignment;
-	static uint32_t				mSemaphoreAddrAlignment;
-	static uint32_t				mSemaphoreAllocSize;
-	static uint32_t				mSemaphorePayloadOffset;
-	static uint32_t				mSemaphorePayloadSize;
-
-	// GPU buffer bound to the target GL_EXTERNAL_VIRTUAL_MEMORY_BUFFER_AMD for pinned memory
-	GLuint						mBufferHandle;
-};
-
-#endif

tests/VideoIOFormatTests.cpp

@@ -0,0 +1,79 @@
+#include "VideoIOFormat.h"
+
+#include <cmath>
+#include <iostream>
+
+namespace
+{
+int gFailures = 0;
+
+void Expect(bool condition, const char* message)
+{
+	if (condition)
+		return;
+
+	std::cerr << "FAIL: " << message << "\n";
+	++gFailures;
+}
+
+void TestPreferredFormatSelection()
+{
+	Expect(ChoosePreferredVideoIOFormat(true) == VideoIOPixelFormat::V210, "10-bit is preferred when supported");
+	Expect(ChoosePreferredVideoIOFormat(false) == VideoIOPixelFormat::Uyvy8, "8-bit is used as fallback");
+	Expect(DeckLinkPixelFormatForVideoIO(VideoIOPixelFormat::V210) == bmdFormat10BitYUV, "v210 maps to DeckLink 10-bit YUV");
+	Expect(DeckLinkPixelFormatForVideoIO(VideoIOPixelFormat::Uyvy8) == bmdFormat8BitYUV, "UYVY maps to DeckLink 8-bit YUV");
+}
+
+void TestRowByteHelpers()
+{
+	Expect(MinimumV210RowBytes(1920) == 5120, "1920-wide v210 active row bytes");
+	Expect(MinimumV210RowBytes(1280) == 3424, "1280-wide v210 active row bytes rounds up to six-pixel group");
+	Expect(MinimumV210RowBytes(3840) == 10240, "3840-wide v210 active row bytes");
+	Expect(PackedTextureWidthFromRowBytes(5120) == 1280, "packed texture width is row bytes divided into RGBA byte texels");
+	Expect(ActiveV210WordsForWidth(1920) == 1280, "active v210 words match 1920 width");
+}
+
+void TestV210PackUnpack()
+{
+	V210SixPixelBlock input;
+	input.y = { 64, 128, 256, 512, 768, 940 };
+	input.cb = { 64, 512, 960 };
+	input.cr = { 960, 512, 64 };
+
+	const V210SixPixelBlock output = UnpackV210Block(PackV210Block(input));
+	Expect(output.y == input.y, "v210 luma survives pack/unpack");
+	Expect(output.cb == input.cb, "v210 Cb survives pack/unpack");
+	Expect(output.cr == input.cr, "v210 Cr survives pack/unpack");
+}
+
+void TestRec709LegalRanges()
+{
+	const V210CodeValues black = Rec709RgbToLegalV210(0.0f, 0.0f, 0.0f);
+	const V210CodeValues grey = Rec709RgbToLegalV210(0.5f, 0.5f, 0.5f);
+	const V210CodeValues white = Rec709RgbToLegalV210(1.0f, 1.0f, 1.0f);
+
+	Expect(black.y == 64, "black maps to legal-range 10-bit luma minimum");
+	Expect(white.y == 940, "white maps to legal-range 10-bit luma maximum");
+	Expect(std::abs(static_cast<int>(grey.y) - 502) <= 1, "middle grey maps near legal-range midpoint");
+	Expect(black.cb == 512 && black.cr == 512, "black keeps neutral chroma");
+	Expect(grey.cb == 512 && grey.cr == 512, "grey keeps neutral chroma");
+	Expect(white.cb == 512 && white.cr == 512, "white keeps neutral chroma");
+}
+}
+
+int main()
+{
+	TestPreferredFormatSelection();
+	TestRowByteHelpers();
+	TestV210PackUnpack();
+	TestRec709LegalRanges();
+
+	if (gFailures != 0)
+	{
+		std::cerr << gFailures << " VideoIOFormat test failure(s).\n";
+		return 1;
+	}
+
+	std::cout << "VideoIOFormat tests passed.\n";
+	return 0;
+}