D3D11 Debug Layer的bug

在开发D3D应用程序时,我们会使用Debug Layer来调试应用程序,以确保我们的程序在最终发布时没有warnings和errors。不过最近在开发应用程序时遇到了这样的问题,就是我把多个网格模型的顶点数据放在一个vertex buffer中,与此同时也把它们的索引数据放在一个index buffer中,然后在绘制不同的网格模型时就需要设置相应的偏移量,代码看上去是这样的:

首先我创建了一个box和一个grid:

CGeometryGenerator::CreateBox( 1.0f, 1.0f, 1.0f, box );
CGeometryGenerator::CreateGrid( 20.0f, 30.0f, , , grid );

接着我记录下每个网格模型的顶点偏移量和索引偏移量:

g_uBoxVertexOffset = ;
g_uGridVertexOffset = g_uBoxVertexOffset + box.vertices.size(); g_uBoxIndexCount = box.indices.size();
g_uGridIndexCount = grid.indices.size(); g_uBoxIndexOffset = ;
g_uGridIndexOffset = g_uBoxIndexOffset + g_uBoxIndexCount;

最后我将box和grid的顶点数据与索引数据存放在一个vertex buffer和index buffer中:

// Copy vertices
UINT uVertexCount = box.vertices.size() + grid.vertices.size() + sphere.vertices.size() + cylinder.vertices.size();
std::vector<VertexPositionNormal> vVertices( uVertexCount );
UINT uVertexIndex = ; for ( UINT vertexIndex = ; vertexIndex < box.vertices.size(); ++vertexIndex, ++uVertexIndex )
{
VertexPositionNormal &vertex = vVertices[uVertexIndex];
VertexPositionNormalTangentColorTexture &boxVertex = box.vertices[vertexIndex]; vertex.position = boxVertex.position;
vertex.normal = boxVertex.normal;
} for ( UINT vertexIndex = ; vertexIndex < grid.vertices.size(); ++vertexIndex, ++uVertexIndex )
{
VertexPositionNormal &vertex = vVertices[uVertexIndex];
VertexPositionNormalTangentColorTexture &gridVertex = grid.vertices[vertexIndex]; vertex.position = gridVertex.position;
vertex.normal = gridVertex.normal;
} // Create vertex buffer
{
D3D11_BUFFER_DESC vertexBufferDesc;
vertexBufferDesc.Usage = D3D11_USAGE_IMMUTABLE;
vertexBufferDesc.ByteWidth = sizeof(VertexPositionNormal) * uVertexCount;
vertexBufferDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER;
vertexBufferDesc.CPUAccessFlags = ;
vertexBufferDesc.MiscFlags = ;
vertexBufferDesc.StructureByteStride = ; D3D11_SUBRESOURCE_DATA subresourceData;
subresourceData.pSysMem = &vVertices[];
subresourceData.SysMemPitch = ;
subresourceData.SysMemSlicePitch = ; HRESULT hr;
V( pD3DDevice->CreateBuffer(&vertexBufferDesc, &subresourceData, g_pShapeVertexBuffer.ReleaseAndGetAddressOf()) );
} // Create indices
UINT uIndexCount = g_uBoxIndexCount + g_uGridIndexCount; std::vector<UINT> vIndices;
vIndices.insert( vIndices.end(), box.indices.begin(), box.indices.end() );
vIndices.insert( vIndices.end(), grid.indices.begin(), grid.indices.end() ); // Create index buffer
{
D3D11_BUFFER_DESC indexBufferDesc;
indexBufferDesc.Usage = D3D11_USAGE_IMMUTABLE;
indexBufferDesc.ByteWidth = sizeof(UINT) * uIndexCount;
indexBufferDesc.BindFlags = D3D11_BIND_INDEX_BUFFER;
indexBufferDesc.CPUAccessFlags = ;
indexBufferDesc.MiscFlags = ;
indexBufferDesc.StructureByteStride = ; D3D11_SUBRESOURCE_DATA subresourceData;
subresourceData.pSysMem = &vIndices[];
subresourceData.SysMemPitch = ;
subresourceData.SysMemSlicePitch = ; HRESULT hr;
V( pD3DDevice->CreateBuffer(&indexBufferDesc, &subresourceData, g_pShapeIndexBuffer.ReleaseAndGetAddressOf()) );
}

因此,在绘制grid时,我就需要设置相应的偏移量来保证读取的顶点和索引数据是正确的,代码如下:

pD3DImmediateContext->DrawIndexed( g_uGridIndexCount, g_uGridIndexOffset, g_uGridVertexOffset );

这个时候,D3D Debug Layer就会出现错误:

D3D11 Debug Layer的bug

起初以为是偏移量设置出错,但是仔细检查了代码,发现没有任何问题,同时渲染结果也是正确的。于是,尝试采用另一种方式,就是在设置vertex buffer时设置偏移量:

UINT stride = sizeof( VertexPositionNormal );
UINT offset = g_uGridVertexOffset * sizeof(VertexPositionNormal);
pD3DImmediateContext->IASetVertexBuffers( , , g_pShapeVertexBuffer.GetAddressOf(), &stride, &offset ); pD3DImmediateContext->DrawIndexed( g_uGridIndexCount, g_uGridIndexOffset, );

此时错误就消失了,渲染也是正确的,也就是说,代码中的偏移量计算是正确的。

于是上网翻阅了一些资料,发现有不少朋友遇到过这个问题,同时确认这个是D3D Debug Layer的bug,也就是当DrawIndexed函数中的StartIndexLocation和BaseVertexLocation参数同时不为0时,就会出现上图的warning。
既然程序是正确的,那么有什么办法可以不让它打印这个warning呢?一个方法就是上面我提到的,在设置vertex buffer时设置偏移量。还有个办法就是通知Debug Layer隐藏这个消息:

// Hide warning: DEVICE_DRAW_INDEX_BUFFER_TOO_SMALL which is a bug of D3D11 Debug Layer
#if defined(DEBUG) | defined(_DEBUG)
ID3D11Debug *pD3DDebug = NULL;
if( SUCCEEDED(pD3DDevice->QueryInterface(__uuidof(ID3D11Debug), reinterpret_cast<VOID**>(&pD3DDebug))) )
{
ID3D11InfoQueue *pD3DInfoQueue = NULL;
if( SUCCEEDED(pD3DDebug->QueryInterface(__uuidof(ID3D11InfoQueue), reinterpret_cast<VOID**>(&pD3DInfoQueue))) )
{
D3D11_MESSAGE_ID arrHiddenMessageID[] =
{
D3D11_MESSAGE_ID_DEVICE_DRAW_INDEX_BUFFER_TOO_SMALL,
}; D3D11_INFO_QUEUE_FILTER filter;
memset( &filter, , sizeof(filter) );
filter.DenyList.NumIDs = ARRAYSIZE( arrHiddenMessageID );
filter.DenyList.pIDList = arrHiddenMessageID; pD3DInfoQueue->AddStorageFilterEntries( &filter );
SAFE_RELEASE( pD3DInfoQueue );
}
SAFE_RELEASE( pD3DDebug );
}
#endif
上一篇:利用Anaconda安装python后,如何安装opencv-python


下一篇:C++学习7-面向对象编程基础(多态性与虚函数、 IO文件流操作)