Hi.
I have a Win Form 2012 application that uses CUDA 5.5. I have a 3D graphic (triangle mesh) made of about 1.7 Million vertices. I am using VBOs for rendering the graphic. Vertex, normal and color are interleaved in the following pattern Vx,Vy,Vz;Nx,Ny,Nz;R,G,B. Color is assigned to each vertex after doing some computation on the GPU. If I compile the program with compute_10,sm_11 options, everything works perfect. However, if I change the the option to compute_20,sm_20 I do not see any colors (as if GPU did not do it’s job). I was wondering if anybody else has seen the same problem or knows how to fix the issue. Attached are the screenshots of the two versions.
Here is how I perform the calculations:
size_t starta;
err = cudaGraphicsMapResources( 1, &vertexVBO.cudaResource, NULL );
if (err != cudaSuccess) { printf(“Error cudaGrpahicMapResources vertexVBO: %s \n)”, cudaGetErrorString(err));}
err = cudaGraphicsResourceGetMappedPointer( (void **)&vertexPtr, &starta, vertexVBO.cudaResource);
if (err != cudaSuccess) { printf(“Error cudaGraphicsResourceGetMappedPointer vertexVBO: %s \n)”, cudaGetErrorString(err));}
Kernel<<<numBlocks, numThreads>>>(params);
cudaDeviceSynchronize();
I would appreciate any help in this regard.
Thanks
Vj
