I run a mpi+cuda program , but always when run Seventh (foor loop) failed.
I get some message…
cudaSafeCall() Runtime API error in file <c:/Documents and Settings/user/My Documents/Visual Studio 2008/Projects/mpisar/sar_cuda3_modified.cu>, line 368 : no CUDA-capable device is available.
My code 368 line : cudaSafeCall(cudaMalloc( (void**) &cuda_t, sz_t) );
cudaSafeCall() Runtime API error in file <c:/Documents and Settings/user/My Documents/Visual Studio 2008/Projects/mpisar/sar_cuda3_modified.cu>, line 437 : the launch timed out and was terminated.
My code 437 line : cudaSafeCall(cudaMemcpy(host_ss0, cuda_ss0, sz_ss0, cudaMemcpyDeviceToHost));
I has been do cudaFree() , delete , to release memory.
So what problem let the program always run failed ?
Have do release Device ?