Greetings, I am running the ./UnifiedMemoryStreams example given by CUDA 7.0 and receive the following message
GPU Device 0: “GeForce GTX 980” with compute capability 5.2
and later nothing … e.g. it get stuck.
I have writen a small test:
...
int *ret;
cudaStream_t stream = NULL;
cudaStreamCreate(&stream);
if(!stream) {
printf("failed to alloc a stream\n");
return 0;
}
checkCuda(cudaMallocManaged(&ret, 1000000 * sizeof(int), cudaMemAttachHost))
checkCuda(cudaStreamCreate(&stream));
memset((void *)ret, 0xAA, 1000000 * sizeof(int));
printf("after ret setting we have ret[1] == %d\n", ret[1]);
...
And can see that memset get stuck.
The execution environment is a new system 76 running Ubuntu 15.04, and GeForce GTX 980.
P.S. the same test on a different machine (a mac book pro (dual boot)) running Ubuntu 14.04 with CUDA 7.0 and GeForce GT 750M runs this test without any issues.
Please help :-)
Thank you.