Clean up copyvertex.h

projects/src/libGLESv2.vcxproj

@@ -206,6 +206,7 @@
     <None Include="..\..\src\angle.gyp"/>
     <None Include="..\..\src\libGLESv2\libGLESv2.def"/>
     <None Include="..\..\src\libGLESv2\renderer\loadimage.inl"/>
+    <None Include="..\..\src\libGLESv2\renderer\copyvertex.inl"/>
     <None Include="..\..\src\libGLESv2\renderer\generatemip.inl"/>
   </ItemGroup>
   <ItemGroup>

projects/src/libGLESv2.vcxproj.filters

@@ -303,6 +303,9 @@
     <ClInclude Include="..\..\src\libGLESv2\renderer\VertexArrayImpl.h">
       <Filter>src\libGLESv2\renderer</Filter>
     </ClInclude>
+    <None Include="..\..\src\libGLESv2\renderer\copyvertex.inl">
+      <Filter>src\libGLESv2\renderer</Filter>
+    </None>
     <ClInclude Include="..\..\src\libGLESv2\renderer\SwapChain.h">
       <Filter>src\libGLESv2\renderer</Filter>
     </ClInclude>

src/libGLESv2.gypi

@@ -118,6 +118,7 @@
             'libGLESv2/renderer/copyimage.cpp',
             'libGLESv2/renderer/copyimage.h',
             'libGLESv2/renderer/copyvertex.h',
+            'libGLESv2/renderer/copyvertex.inl',
             'libGLESv2/renderer/generatemip.h',
             'libGLESv2/renderer/generatemip.inl',
             'libGLESv2/renderer/imageformats.h',

src/libGLESv2/formatutils.h

@@ -31,7 +31,7 @@ typedef void (*ColorReadFunction)(const void *source, void *dest);
 typedef void (*ColorWriteFunction)(const void *source, void *dest);
 typedef void (*ColorCopyFunction)(const void *source, void *dest);
 
-typedef void (*VertexCopyFunction)(const void *input, size_t stride, size_t count, void *output);
+typedef void (*VertexCopyFunction)(const uint8_t *input, size_t stride, size_t count, uint8_t *output);
 
 namespace gl
 {

src/libGLESv2/renderer/copyvertex.inl

+//
+// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+//
+
+namespace rx
+{
+
+template <typename T, size_t componentCount, uint32_t widenDefaultValueBits>
+inline void CopyNativeVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output)
+{
+    const size_t attribSize = sizeof(T)* componentCount;
+    const T defaultValue = gl::bitCast<T>(widenDefaultValueBits);
+    const bool widen = (widenDefaultValueBits != 0);
+
+    if (attribSize == stride && !widen)
+    {
+        memcpy(output, input, count * attribSize);
+    }
+    else
+    {
+        size_t outputStride = widen ? 4 : componentCount;
+
+        for (size_t i = 0; i < count; i++)
+        {
+            const T *offsetInput = reinterpret_cast<const T*>(input + (i * stride));
+            T *offsetOutput = reinterpret_cast<T*>(output) + i * outputStride;
+
+            for (size_t j = 0; j < componentCount; j++)
+            {
+                offsetOutput[j] = offsetInput[j];
+            }
+
+            if (widen)
+            {
+                offsetOutput[3] = defaultValue;
+            }
+        }
+    }
+}
+
+template <size_t componentCount>
+inline void Copy32FixedTo32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output)
+{
+    static const float divisor = 1.0f / (1 << 16);
+
+    for (size_t i = 0; i < count; i++)
+    {
+        const GLfixed* offsetInput = reinterpret_cast<const GLfixed*>(input + (stride * i));
+        float* offsetOutput = reinterpret_cast<float*>(output) + i * componentCount;
+
+        for (size_t j = 0; j < componentCount; j++)
+        {
+            offsetOutput[j] = static_cast<float>(offsetInput[j]) * divisor;
+        }
+    }
+}
+
+template <typename T, size_t componentCount, bool normalized>
+inline void CopyTo32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output)
+{
+    typedef std::numeric_limits<T> NL;
+
+    for (size_t i = 0; i < count; i++)
+    {
+        const T *offsetInput = reinterpret_cast<const T*>(input + (stride * i));
+        float *offsetOutput = reinterpret_cast<float*>(output) + i * componentCount;
+
+        for (size_t j = 0; j < componentCount; j++)
+        {
+            if (normalized)
+            {
+                if (NL::is_signed)
+                {
+                    const float divisor = 1.0f / (2 * static_cast<float>(NL::max()) + 1);
+                    offsetOutput[j] = (2 * static_cast<float>(offsetInput[j]) + 1) * divisor;
+                }
+                else
+                {
+                    offsetOutput[j] =  static_cast<float>(offsetInput[j]) / NL::max();
+                }
+            }
+            else
+            {
+                offsetOutput[j] =  static_cast<float>(offsetInput[j]);
+            }
+        }
+    }
+}
+
+namespace priv
+{
+
+template <bool isSigned, bool normalized, bool toFloat>
+static inline void CopyPackedRGB(uint32_t data, uint8_t *output)
+{
+    const uint32_t rgbSignMask = 0x200;       // 1 set at the 9 bit
+    const uint32_t negativeMask = 0xFFFFFC00; // All bits from 10 to 31 set to 1
+
+    if (toFloat)
+    {
+        GLfloat *floatOutput = reinterpret_cast<GLfloat*>(output);
+        if (isSigned)
+        {
+            GLfloat finalValue = 0;
+            if (data & rgbSignMask)
+            {
+                int negativeNumber = data | negativeMask;
+                finalValue = static_cast<GLfloat>(negativeNumber);
+            }
+            else
+            {
+                finalValue = static_cast<GLfloat>(data);
+            }
+
+            if (normalized)
+            {
+                const int32_t maxValue = 0x1FF;      // 1 set in bits 0 through 8
+                const int32_t minValue = 0xFFFFFE01; // Inverse of maxValue
+
+                // A 10-bit two's complement number has the possibility of being minValue - 1 but
+                // OpenGL's normalization rules dictate that it should be clamped to minValue in this
+                // case.
+                if (finalValue < minValue)
+                {
+                    finalValue = minValue;
+                }
+
+                const int32_t halfRange = (maxValue - minValue) >> 1;
+                *floatOutput = ((finalValue - minValue) / halfRange) - 1.0f;
+            }
+            else
+            {
+                *floatOutput = finalValue;
+            }
+        }
+        else
+        {
+            if (normalized)
+            {
+                const uint32_t maxValue = 0x3FF; // 1 set in bits 0 through 9
+                *floatOutput = static_cast<GLfloat>(data) / static_cast<GLfloat>(maxValue);
+            }
+            else
+            {
+                *floatOutput = static_cast<GLfloat>(data);
+            }
+        }
+    }
+    else
+    {
+        if (isSigned)
+        {
+            GLshort *intOutput = reinterpret_cast<GLshort*>(output);
+
+            if (data & rgbSignMask)
+            {
+                *intOutput = data | negativeMask;
+            }
+            else
+            {
+                *intOutput = data;
+            }
+        }
+        else
+        {
+            GLushort *uintOutput = reinterpret_cast<GLushort*>(output);
+            *uintOutput = data;
+        }
+    }
+}
+
+template <bool isSigned, bool normalized, bool toFloat>
+inline void CopyPackedAlpha(uint32_t data, uint8_t *output)
+{
+    if (toFloat)
+    {
+        GLfloat *floatOutput = reinterpret_cast<GLfloat*>(output);
+        if (isSigned)
+        {
+            if (normalized)
+            {
+                switch (data)
+                {
+                  case 0x0: *floatOutput =  0.0f; break;
+                  case 0x1: *floatOutput =  1.0f; break;
+                  case 0x2: *floatOutput = -1.0f; break;
+                  case 0x3: *floatOutput = -1.0f; break;
+                  default: UNREACHABLE();
+                }
+            }
+            else
+            {
+                switch (data)
+                {
+                  case 0x0: *floatOutput =  0.0f; break;
+                  case 0x1: *floatOutput =  1.0f; break;
+                  case 0x2: *floatOutput = -2.0f; break;
+                  case 0x3: *floatOutput = -1.0f; break;
+                  default: UNREACHABLE();
+                }
+            }
+        }
+        else
+        {
+            if (normalized)
+            {
+                switch (data)
+                {
+                  case 0x0: *floatOutput = 0.0f / 3.0f; break;
+                  case 0x1: *floatOutput = 1.0f / 3.0f; break;
+                  case 0x2: *floatOutput = 2.0f / 3.0f; break;
+                  case 0x3: *floatOutput = 3.0f / 3.0f; break;
+                  default: UNREACHABLE();
+                }
+            }
+            else
+            {
+                switch (data)
+                {
+                  case 0x0: *floatOutput = 0.0f; break;
+                  case 0x1: *floatOutput = 1.0f; break;
+                  case 0x2: *floatOutput = 2.0f; break;
+                  case 0x3: *floatOutput = 3.0f; break;
+                  default: UNREACHABLE();
+                }
+            }
+        }
+    }
+    else
+    {
+        if (isSigned)
+        {
+            GLshort *intOutput = reinterpret_cast<GLshort*>(output);
+            switch (data)
+            {
+              case 0x0: *intOutput =  0; break;
+              case 0x1: *intOutput =  1; break;
+              case 0x2: *intOutput = -2; break;
+              case 0x3: *intOutput = -1; break;
+              default: UNREACHABLE();
+            }
+        }
+        else
+        {
+            GLushort *uintOutput = reinterpret_cast<GLushort*>(output);
+            switch (data)
+            {
+              case 0x0: *uintOutput = 0; break;
+              case 0x1: *uintOutput = 1; break;
+              case 0x2: *uintOutput = 2; break;
+              case 0x3: *uintOutput = 3; break;
+              default: UNREACHABLE();
+            }
+        }
+    }
+}
+
+}
+
+template <bool isSigned, bool normalized, bool toFloat>
+inline void CopyXYZ10W2ToXYZW32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output)
+{
+    const size_t outputComponentSize = toFloat ? 4 : 2;
+    const size_t componentCount = 4;
+
+    const uint32_t rgbMask = 0x3FF; // 1 set in bits 0 through 9
+    const size_t redShift = 0;    // red is bits 0 through 9
+    const size_t greenShift = 10; // green is bits 10 through 19
+    const size_t blueShift = 20;  // blue is bits 20 through 29
+
+    const uint32_t alphaMask = 0x3; // 1 set in bits 0 and 1
+    const size_t alphaShift = 30; // Alpha is the 30 and 31 bits
+
+    for (size_t i = 0; i < count; i++)
+    {
+        GLuint packedValue = *reinterpret_cast<const GLuint*>(input + (i * stride));
+        uint8_t *offsetOutput = output + (i * outputComponentSize * componentCount);
+
+        priv::CopyPackedRGB<isSigned, normalized, toFloat>(  (packedValue >> redShift)   & rgbMask,   offsetOutput + (0 * outputComponentSize));
+        priv::CopyPackedRGB<isSigned, normalized, toFloat>(  (packedValue >> greenShift) & rgbMask,   offsetOutput + (1 * outputComponentSize));
+        priv::CopyPackedRGB<isSigned, normalized, toFloat>(  (packedValue >> blueShift)  & rgbMask,   offsetOutput + (2 * outputComponentSize));
+        priv::CopyPackedAlpha<isSigned, normalized, toFloat>((packedValue >> alphaShift) & alphaMask, offsetOutput + (3 * outputComponentSize));
+    }
+}
+
+}

src/libGLESv2/renderer/d3d/d3d11/VertexBuffer11.cpp

@@ -83,24 +83,24 @@ bool VertexBuffer11::storeVertexAttributes(const gl::VertexAttribute &attrib, co
             return false;
         }
 
-        char* output = reinterpret_cast<char*>(mappedResource.pData) + offset;
+        uint8_t* output = reinterpret_cast<uint8_t*>(mappedResource.pData) + offset;
 
-        const char *input = NULL;
+        const uint8_t *input = NULL;
         if (attrib.enabled)
         {
             if (buffer)
             {
                 Buffer11 *storage = Buffer11::makeBuffer11(buffer->getImplementation());
-                input = static_cast<const char*>(storage->getData()) + static_cast<int>(attrib.offset);
+                input = static_cast<const uint8_t*>(storage->getData()) + static_cast<int>(attrib.offset);
             }
             else
             {
-                input = static_cast<const char*>(attrib.pointer);
+                input = static_cast<const uint8_t*>(attrib.pointer);
             }
         }
         else
         {
-            input = reinterpret_cast<const char*>(currentValue.FloatValues);
+            input = reinterpret_cast<const uint8_t*>(currentValue.FloatValues);
         }
 
         if (instances == 0 || attrib.divisor == 0)

src/libGLESv2/renderer/d3d/d3d9/VertexBuffer9.cpp

@@ -77,7 +77,7 @@ bool VertexBuffer9::storeVertexAttributes(const gl::VertexAttribute &attrib, con
 
         DWORD lockFlags = mDynamicUsage ? D3DLOCK_NOOVERWRITE : 0;
 
-        void *mapPtr = NULL;
+        uint8_t *mapPtr = NULL;
 
         unsigned int mapSize;
         if (!spaceRequired(attrib, count, instances, &mapSize))
@@ -85,7 +85,7 @@ bool VertexBuffer9::storeVertexAttributes(const gl::VertexAttribute &attrib, con
             return false;
         }
 
-        HRESULT result = mVertexBuffer->Lock(offset, mapSize, &mapPtr, lockFlags);
+        HRESULT result = mVertexBuffer->Lock(offset, mapSize, reinterpret_cast<void**>(&mapPtr), lockFlags);
 
         if (FAILED(result))
         {
@@ -93,22 +93,22 @@ bool VertexBuffer9::storeVertexAttributes(const gl::VertexAttribute &attrib, con
             return false;
         }
 
-        const char *input = NULL;
+        const uint8_t *input = NULL;
         if (attrib.enabled)
         {
             if (buffer)
             {
                 BufferImpl *storage = buffer->getImplementation();
-                input = static_cast<const char*>(storage->getData()) + static_cast<int>(attrib.offset);
+                input = static_cast<const uint8_t*>(storage->getData()) + static_cast<int>(attrib.offset);
             }
             else
             {
-                input = static_cast<const char*>(attrib.pointer);
+                input = static_cast<const uint8_t*>(attrib.pointer);
             }
         }
         else
         {
-            input = reinterpret_cast<const char*>(currentValue.FloatValues);
+            input = reinterpret_cast<const uint8_t*>(currentValue.FloatValues);
         }
 
         if (instances == 0 || attrib.divisor == 0)

src/libGLESv2/renderer/vertexconversion.h

@@ -154,11 +154,13 @@ struct VertexDataConverter
     static const bool identity = (WidenRule::initialWidth == WidenRule::finalWidth) && Converter::identity;
     static const std::size_t finalSize = WidenRule::finalWidth * sizeof(OutputType);
 
-    static void convertArray(const InputType *in, std::size_t stride, std::size_t n, OutputType *out)
+    static void convertArray(const uint8_t *input, size_t stride, size_t n, uint8_t *output)
     {
+        OutputType *out = reinterpret_cast<OutputType*>(output);
+
         for (std::size_t i = 0; i < n; i++)
         {
-            const InputType *ein = pointerAddBytes(in, i * stride);
+            const InputType *ein = reinterpret_cast<const InputType*>(input + i * stride);
 
             copyComponent(out, ein, 0, static_cast<OutputType>(DefaultValueRule::zero()));
             copyComponent(out, ein, 1, static_cast<OutputType>(DefaultValueRule::zero()));
@@ -169,19 +171,7 @@ struct VertexDataConverter
         }
     }
 
-    static void convertArray(const void *in, std::size_t stride, std::size_t n, void *out)
-    {
-        return convertArray(static_cast<const InputType*>(in), stride, n, static_cast<OutputType*>(out));
-    }
-
   private:
-    // Advance the given pointer by a number of bytes (not pointed-to elements).
-    template <class T>
-    static T *pointerAddBytes(T *basePtr, std::size_t numBytes)
-    {
-        return reinterpret_cast<T *>(reinterpret_cast<uintptr_t>(basePtr) + numBytes);
-    }
-
     static void copyComponent(OutputType *out, const InputType *in, std::size_t elementindex, OutputType defaultvalue)
     {
         if (WidenRule::finalWidth > elementindex)