Hi,
I’ve been looking for an example of how to sample a 3D texture created in OpenGL (without PBO) with the CUDA Driver API.
all without deprecated function.
Got it working to write to 2D textures via surface objects but sampling textures won’t work.
Anyone?
Cheers,
Florian
This is what I do:
// init GL texture
glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA32UI, width, height, depth, 0, format, type, NULL);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE);
// upload data to texture ...
// ...
err = cuGraphicsGLRegisterImage(&m_cuGpuMocTree, (GLuint)(*m_gpuMocTree.get()), GL_TEXTURE_3D, CU_GRAPHICS_REGISTER_FLAGS_NONE);
assert(err == CUDA_SUCCESS);
// ...
CUarray gpuMoctree;
err = cuGraphicsSubResourceGetMappedArray(&gpuMoctree, m_cuGpuMocTree, 0, 0);
assert(err == CUDA_SUCCESS);
CUtexref texrefMoctree;
err = cuModuleGetTexRef(&texrefMoctree, m_raycastProgram, "moctree");
assert(err == CUDA_SUCCESS);
err = cuTexRefSetArray(texrefMoctree, gpuMoctree, CU_TRSA_OVERRIDE_FORMAT);
assert(err == CUDA_SUCCESS);
err = cuTexRefSetAddressMode(texrefMoctree, 0, CU_TR_ADDRESS_MODE_CLAMP);
assert(err == CUDA_SUCCESS);
err = cuTexRefSetAddressMode(texrefMoctree, 1, CU_TR_ADDRESS_MODE_CLAMP);
assert(err == CUDA_SUCCESS);
err = cuTexRefSetAddressMode(texrefMoctree, 2, CU_TR_ADDRESS_MODE_CLAMP);
assert(err == CUDA_SUCCESS);
err = cuTexRefSetFilterMode(texrefMoctree, CU_TR_FILTER_MODE_POINT);
assert(err == CUDA_SUCCESS);
err = cuTexRefSetFlags(texrefMoctree, CU_TRSF_READ_AS_INTEGER);
assert(err == CUDA_SUCCESS);
err = cuTexRefSetFormat(texrefMoctree, CU_AD_FORMAT_UNSIGNED_INT32, 4);
assert(err == CUDA_SUCCESS);
// ...
cuLaunchKernel(m_raycastIntersect,....);
// ...
// tear down stuff
And the CUDA C program:
// ...
texture<uint4, cudaTextureType3D, cudaReadModeElementType> moctree;
// ...
extern "C" __global__ void raycastIntersect(...)
{
// ...
uint4 at = tex3D(moctree, x, y, z);
// ...
}
All of the components are always zero and the “moctree” texref shows “???” when putting a breakpoint with CUDA Debugger. All the asserts do pass. This is on a GTX Titan on Win8 x64.
Anything clearly wrong here?
Cheers!