IndexDataManager.cpp

Branch :

Show log
Commit
Author : Jamie Madill
Date : 2015-04-15 14:26:37
Hash : b48e8b07
Message : D3D11: Only rewrite for primitive restart when needed. We would rewrite our index data every draw call. Change the index check to see if we're writing to the same sized / typed static buffer and only rewrite the data if the user re-uploaded. Also add a performance test for the primitive restart workaround. As a future improvement we could avoid creating new D3D objects every time we reinitialize static data, since BufferSubData calls don't change the size of the buffer if the index type remains the same. BUG=476658 Change-Id: I9d2540ad8b1b34fa0142ba0bf794cf572da8c61d Reviewed-on: https://chromium-review.googlesource.com/265838 Tested-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Brandon Jones <bajones@chromium.org>

src/libANGLE/renderer/d3d/IndexDataManager.cpp

//
// Copyright (c) 2002-2014 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.
//

// IndexDataManager.cpp: Defines the IndexDataManager, a class that
// runs the Buffer translation process for index buffers.

#include "libANGLE/renderer/d3d/IndexDataManager.h"
#include "libANGLE/renderer/d3d/BufferD3D.h"
#include "libANGLE/renderer/d3d/IndexBuffer.h"
#include "libANGLE/Buffer.h"
#include "libANGLE/formatutils.h"

namespace rx
{

static void ConvertIndices(GLenum sourceType, GLenum destinationType, const void *input, GLsizei count, void *output)
{
    if (sourceType == GL_UNSIGNED_BYTE)
    {
        ASSERT(destinationType == GL_UNSIGNED_SHORT);
        const GLubyte *in = static_cast<const GLubyte*>(input);
        GLushort *out = static_cast<GLushort*>(output);

        for (GLsizei i = 0; i < count; i++)
        {
            out[i] = in[i];
        }
    }
    else if (sourceType == GL_UNSIGNED_INT)
    {
        ASSERT(destinationType == GL_UNSIGNED_INT);
        memcpy(output, input, count * sizeof(GLuint));
    }
    else if (sourceType == GL_UNSIGNED_SHORT)
    {
        if (destinationType == GL_UNSIGNED_SHORT)
        {
            memcpy(output, input, count * sizeof(GLushort));
        }
        else if (destinationType == GL_UNSIGNED_INT)
        {
            const GLushort *in = static_cast<const GLushort*>(input);
            GLuint *out = static_cast<GLuint*>(output);

            for (GLsizei i = 0; i < count; i++)
            {
                out[i] = in[i];
            }
        }
        else UNREACHABLE();
    }
    else UNREACHABLE();
}

IndexDataManager::IndexDataManager(BufferFactoryD3D *factory, RendererClass rendererClass)
    : mFactory(factory),
      mRendererClass(rendererClass),
      mStreamingBufferShort(nullptr),
      mStreamingBufferInt(nullptr)
{
}

IndexDataManager::~IndexDataManager()
{
    SafeDelete(mStreamingBufferShort);
    SafeDelete(mStreamingBufferInt);
}

gl::Error IndexDataManager::prepareIndexData(GLenum type, GLsizei count, gl::Buffer *buffer, const GLvoid *indices, TranslatedIndexData *translated)
{
    const gl::Type &typeInfo = gl::GetTypeInfo(type);

    GLenum destinationIndexType = (type == GL_UNSIGNED_INT) ? GL_UNSIGNED_INT : GL_UNSIGNED_SHORT;

    unsigned int offset = 0;
    bool alignedOffset = false;

    BufferD3D *storage = NULL;

    if (buffer != NULL)
    {
        offset = static_cast<unsigned int>(reinterpret_cast<uintptr_t>(indices));

        storage = GetImplAs<BufferD3D>(buffer);

        // We'll trust that the compiler will optimize the % below:
        // the operands are unsigned and the divisor is a constant.
        switch (type)
        {
          case GL_UNSIGNED_BYTE:  alignedOffset = (offset % sizeof(GLubyte) == 0);  break;
          case GL_UNSIGNED_SHORT: alignedOffset = (offset % sizeof(GLushort) == 0); break;
          case GL_UNSIGNED_INT:   alignedOffset = (offset % sizeof(GLuint) == 0);   break;
          default: UNREACHABLE(); alignedOffset = false;
        }

        ASSERT(typeInfo.bytes * static_cast<unsigned int>(count) + offset <= storage->getSize());

        const uint8_t *bufferData = NULL;
        gl::Error error = storage->getData(&bufferData);
        if (error.isError())
        {
            return error;
        }

        indices = bufferData + offset;
    }

    StaticIndexBufferInterface *staticBuffer = storage ? storage->getStaticIndexBuffer() : NULL;
    IndexBufferInterface *indexBuffer = NULL;
    bool directStorage = alignedOffset && storage && storage->supportsDirectBinding() &&
                         destinationIndexType == type;
    unsigned int streamOffset = 0;

    // Avoid D3D11's primitive restart index value
    // see http://msdn.microsoft.com/en-us/library/windows/desktop/bb205124(v=vs.85).aspx
    bool primitiveRestartWorkaround = mRendererClass == RENDERER_D3D11 &&
                                      translated->indexRange.end == 0xFFFF &&
                                      type == GL_UNSIGNED_SHORT;

    if (primitiveRestartWorkaround)
    {
        destinationIndexType = GL_UNSIGNED_INT;
        directStorage = false;
    }

    const gl::Type &destTypeInfo = gl::GetTypeInfo(destinationIndexType);

    if (directStorage)
    {
        streamOffset = offset;
    }
    else if (staticBuffer &&
             staticBuffer->getBufferSize() != 0 &&
             alignedOffset &&
             staticBuffer->getIndexType() == destinationIndexType)
    {
        indexBuffer = staticBuffer;

        // Using bit-shift here is faster than using division.
        streamOffset = (offset >> typeInfo.bytesShift) << destTypeInfo.bytesShift;
    }

    if (!directStorage && !indexBuffer)
    {
        gl::Error error = getStreamingIndexBuffer(destinationIndexType, &indexBuffer);
        if (error.isError())
        {
            return error;
        }

        unsigned int convertCount = count;

        if (staticBuffer)
        {
            if (staticBuffer->getBufferSize() == 0 && alignedOffset)
            {
                indexBuffer = staticBuffer;
                // Using bit-shift here is faster than using division.
                convertCount = storage->getSize() >> typeInfo.bytesShift;
            }
            else
            {
                storage->invalidateStaticData();
                staticBuffer = NULL;
            }
        }

        ASSERT(indexBuffer);

        // Using bit-shift here is faster than using division.
        if (convertCount > (std::numeric_limits<unsigned int>::max() >> destTypeInfo.bytesShift))
        {
            return gl::Error(GL_OUT_OF_MEMORY, "Reserving %u indices of %u bytes each exceeds the maximum buffer size.",
                             convertCount, destTypeInfo.bytes);
        }

        unsigned int bufferSizeRequired = convertCount << destTypeInfo.bytesShift;
        error = indexBuffer->reserveBufferSpace(bufferSizeRequired, destinationIndexType);
        if (error.isError())
        {
            return error;
        }

        void* output = NULL;
        error = indexBuffer->mapBuffer(bufferSizeRequired, &output, &streamOffset);
        if (error.isError())
        {
            return error;
        }

        const uint8_t *dataPointer = reinterpret_cast<const uint8_t*>(indices);
        if (staticBuffer)
        {
            error = storage->getData(&dataPointer);
            if (error.isError())
            {
                return error;
            }
        }
        ConvertIndices(type, destinationIndexType, dataPointer, convertCount, output);

        error = indexBuffer->unmapBuffer();
        if (error.isError())
        {
            return error;
        }

        if (staticBuffer)
        {
            // Using bit-shift here is faster than using division.
            streamOffset = (offset >> typeInfo.bytesShift) << destTypeInfo.bytesShift;
        }
    }

    translated->storage = directStorage ? storage : NULL;
    translated->indexBuffer = indexBuffer ? indexBuffer->getIndexBuffer() : NULL;
    translated->serial = directStorage ? storage->getSerial() : indexBuffer->getSerial();
    // Using bit-shift here is faster than using division.
    translated->startIndex = (streamOffset >> destTypeInfo.bytesShift);
    translated->startOffset = streamOffset;
    translated->indexType = destinationIndexType;

    if (storage)
    {
        storage->promoteStaticUsage(count << typeInfo.bytesShift);
    }

    return gl::Error(GL_NO_ERROR);
}

gl::Error IndexDataManager::getStreamingIndexBuffer(GLenum destinationIndexType, IndexBufferInterface **outBuffer)
{
    ASSERT(outBuffer);
    if (destinationIndexType == GL_UNSIGNED_INT)
    {
        if (!mStreamingBufferInt)
        {
            mStreamingBufferInt = new StreamingIndexBufferInterface(mFactory);
            gl::Error error = mStreamingBufferInt->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT);
            if (error.isError())
            {
                SafeDelete(mStreamingBufferInt);
                return error;
            }
        }

        *outBuffer = mStreamingBufferInt;
        return gl::Error(GL_NO_ERROR);
    }
    else
    {
        ASSERT(destinationIndexType == GL_UNSIGNED_SHORT);

        if (!mStreamingBufferShort)
        {
            mStreamingBufferShort = new StreamingIndexBufferInterface(mFactory);
            gl::Error error = mStreamingBufferShort->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_SHORT);
            if (error.isError())
            {
                SafeDelete(mStreamingBufferShort);
                return error;
            }
        }

        *outBuffer = mStreamingBufferShort;
        return gl::Error(GL_NO_ERROR);
    }
}

}

kc3-lang/angle/src/libANGLE/renderer/d3d/IndexDataManager.cpp

Commit

kc3-lang/angle /src/libANGLE/renderer/d3d/IndexDataManager.cpp