I have (RW)StructuredBuffer initialized to eg. 50MB. In each frame I need to fill it with different data in interval <10, 50> MB
If i use
deviceContext->UpdateSubresource( gpuBuffer->buffer, 0, NULL, data, 0, 0 );
works fine untill data has size 50MB
But if data are smaller, application crash on this line, no DX debug output nothing.
So i used for smaller data D3D11_BOX and fill it this way:
D3D11_BOX destRegion;
destRegion.left = 0;
destRegion.right = dataSize;
destRegion.top = 0;
destRegion.bottom = 1;
destRegion.front = 0;
destRegion.back = 1;
I hoped that data will be filled from index 0 - dataSize... BUT ...application doesn´t crash this time, however buffer has incorrect content... I got "blank" results if i vizualize content of buffer.
MSDN says nothing about D3D11_BOX and StructuredBuffers, so i really don´t know how to solve this.
That sounds like a pointer error to me. Are you sure you aren't exceeding the size of the buffer???
Also, you could try to use mapping instead of update subresource and see if you meet a similar issue. That would indicate that there is a general problem with the overall setup, or if it works then it would mean that you have an issue with update subresource.
// Create the constant buffer pointer so we can access the shader constant buffer from within this class.
HRESULT hr = this->device->CreateBuffer(&bufferDesc, NULL, &gpuBuffer.buffer);
if (FAILED(hr))
{
MyUtils::Logger::LogError("Failed to create buffer %s", bufferName.GetConstString());
return;
}
In that case you will need to use two buffers - one with staging usage and the other one as you have already created. Then you write to the staging buffer, and then use copy resource to put the info into the default usage buffer.
It isn't ideal to do this copying, but from experience it is pretty fast since both resources are already on the GPU. In any case, you are only trying to isolate the problem, so this might still be helpful.
I changed data representation from uint8 array to int array (inside shader i work with int buffer, uin8 representation was in byte form) and now it works fine..
