Hi, for gpgpu, I want to use ps to read int, and process, and write int to rendertarget. read int is ok, since texture of int is successfully created. but create(rendertarget of R32G32B32_UINT) failed, why? Thanks. My card is 8800GTX. ps. Now I tested DXGI_FORMAT_R32G32B32A32_UINT and it works! why 3 channels can't work and 4 channels work? isn't it that dx10 is "all caps supported"?

	descTex.Width = texW;
	descTex.Height = texH;
	descTex.ArraySize = 1;
	descTex.SampleDesc.Count = 1;
	descTex.SampleDesc.Quality = 0;
	descTex.Format = DXGI_FORMAT_R32G32B32_UINT; //failed! can only DXGI_FORMAT_R32G32B32A32_FLOAT;
	descTex.Usage = D3D10_USAGE_DEFAULT;
	descTex.BindFlags = D3D10_BIND_RENDER_TARGET; //rendertarget
	descTex.CPUAccessFlags = 0;
	descTex.MiscFlags = 0;
	descTex.MipLevels = 1;
	g_pd3dDevice->CreateTexture2D(&descTex, NULL, &pTex);

ps. I have set a fake format during the earliest creation of device:

	UINT createDeviceFlags = 0;
#ifdef _DEBUG
	createDeviceFlags |= D3D10_CREATE_DEVICE_DEBUG;
#endif
	DXGI_SWAP_CHAIN_DESC sd;
	ZeroMemory( &sd, sizeof(sd) );
	sd.BufferCount = 1;
	sd.BufferDesc.Width = 256;
	sd.BufferDesc.Height = 256;
	sd.BufferDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;	//
	sd.BufferDesc.RefreshRate.Numerator = 60;
	sd.BufferDesc.RefreshRate.Denominator = 1;
	sd.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT;
	sd.OutputWindow = g_hWnd;
	sd.SampleDesc.Count = 1;
	sd.SampleDesc.Quality = 0;
	sd.Windowed = TRUE;
	V_RETURN(D3D10CreateDeviceAndSwapChain(NULL, D3D10_DRIVER_TYPE_HARDWARE, NULL, createDeviceFlags, D3D10_SDK_VERSION, &sd, &g_pSwapChain, &g_pd3dDevice));
#endif

by the sd.BufferDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM; I never meant to use this format, since I actually render to a much bigger texture (4k*4k), which has nothing to do with the swap chain. Actually I only render *once* and don't enter the ProcMsg loop. Does this have problem? Thanks a lot!!1 [Edited by - yk_cadcg on May 13, 2007 1:00:42 AM]

Well there isn’t a caps structure no more but there is still a CheckFormatSupport method. GPUs need only support a set of usages for every format (the list is not in the SDK) anything beyond is optional and need to checked with CheckFormatSupport.

If you don’t need the back buffer you can create it with any valid format and size. But you can although create a device without a swapchain and backbuffers at all.