Clamp chroma texcoords to avoid sampling alignment padding

This resolves #885 without the massive perf hit on lower end Intel GPUs.

app/shaders/d3d11_genyuv_pixel.hlsl

@@ -1,6 +1,6 @@
 #include "d3d11_video_pixel_start.hlsli"
 
-cbuffer CSC_CONST_BUF : register(b0)
+cbuffer CSC_CONST_BUF : register(b1)
 {
     min16float3x3 cscMatrix;
     min16float3 offsets;

app/shaders/d3d11_video_pixel_end.hlsli

@@ -1,7 +1,8 @@
 min16float4 main(ShaderInput input) : SV_TARGET
 {
+    // Clamp the chrominance texcoords to avoid sampling the row of texels adjacent to the alignment padding
     min16float3 yuv = min16float3(luminancePlane.Sample(theSampler, input.tex),
-                                  chrominancePlane.Sample(theSampler, input.tex));
+                                  chrominancePlane.Sample(theSampler, min(input.tex, chromaTexMax.rg)));
 
     // Subtract the YUV offset for limited vs full range
     yuv -= offsets;

app/shaders/d3d11_video_pixel_start.hlsli

@@ -7,3 +7,8 @@ struct ShaderInput
     float4 pos : SV_POSITION;
     float2 tex : TEXCOORD0;
 };
+
+cbuffer ChromaLimitBuf : register(b0)
+{
+    min16float3 chromaTexMax;
+};

app/streaming/video/ffmpeg-renderers/d3d11va.cpp

@@ -728,7 +728,7 @@ void D3D11VARenderer::bindColorConversion(AVFrame* frame)
         ID3D11Buffer* constantBuffer;
         HRESULT hr = m_Device->CreateBuffer(&constDesc, &constData, &constantBuffer);
         if (SUCCEEDED(hr)) {
-            m_DeviceContext->PSSetConstantBuffers(0, 1, &constantBuffer);
+            m_DeviceContext->PSSetConstantBuffers(1, 1, &constantBuffer);
             constantBuffer->Release();
         }
         else {
@@ -1313,6 +1313,39 @@ bool D3D11VARenderer::setupRenderingResources()
         }
     }
 
+    // Create our fixed constant buffer to limit chroma texcoords and avoid sampling from alignment texels.
+    {
+        D3D11_BUFFER_DESC constDesc = {};
+        constDesc.ByteWidth = sizeof(CSC_CONST_BUF);
+        constDesc.Usage = D3D11_USAGE_IMMUTABLE;
+        constDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER;
+        constDesc.CPUAccessFlags = 0;
+        constDesc.MiscFlags = 0;
+
+        int alignedWidth = FFALIGN(m_DecoderParams.width, m_TextureAlignment);
+        int alignedHeight = FFALIGN(m_DecoderParams.height, m_TextureAlignment);
+
+        float chromaUVMax[3] = {};
+        chromaUVMax[0] = m_DecoderParams.width != alignedWidth ? ((float)(m_DecoderParams.width - 1) / alignedWidth) : 1.0f;
+        chromaUVMax[1] = m_DecoderParams.height != alignedHeight ? ((float)(m_DecoderParams.height - 1) / alignedHeight) : 1.0f;
+
+        D3D11_SUBRESOURCE_DATA constData = {};
+        constData.pSysMem = chromaUVMax;
+
+        ID3D11Buffer* constantBuffer;
+        HRESULT hr = m_Device->CreateBuffer(&constDesc, &constData, &constantBuffer);
+        if (SUCCEEDED(hr)) {
+            m_DeviceContext->PSSetConstantBuffers(0, 1, &constantBuffer);
+            constantBuffer->Release();
+        }
+        else {
+            SDL_LogError(SDL_LOG_CATEGORY_APPLICATION,
+                         "ID3D11Device::CreateBuffer() failed: %x",
+                         hr);
+            return false;
+        }
+    }
+
     // Create our blend state
     {
         D3D11_BLEND_DESC blendDesc = {};