Anyone else experience this or know how to resolve?
Reboot Windows 10 (to establish no other processes running).
Create a new Visual Studio project using the CUDA 9.1 Runtime.
Replace kernel.cu with the below:
–
#include “cuda_runtime.h”
#include “device_launch_parameters.h”
#include <stdio.h>
#define N (1 << 20)
int main()
{
float *x;
cudaError_t cudaStatus;
printf("Allocating unified memory...\n");
cudaStatus = cudaMallocManaged(&x, N * sizeof(float));
printf("Status: %d\n", cudaStatus);
printf("Deallocating unified memory...\n");
cudaFree(x);
printf("Status: %d\n", cudaStatus);
printf("END\n");
fgetc(stdin);
}
Run once with the Nsight next-gen debugger (with or without breakpoints).
Sleep the machine.
Wake the machine.
Run again with the Nsight next-gen debugger.
Observe program does not return status of attempt at cudaFree().
Observe popup window “Operation taking longer than expected” with Terminate button.
Click on Terminate button so Visual Studio becomes responsive again.
Cannot kill host process in Windows Task Manager.
nvidia-smi shows process still running on GPU.
Only way to kill process and get through program again is to reboot.
±----------------------------------------------------------------------------+
| NVIDIA-SMI 391.35 Driver Version: 391.35 |
|-------------------------------±---------------------±---------------------+
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 TITAN V TCC | 00000000:17:00.0 Off | N/A |
| 28% 42C P8 28W / 250W | 561MiB / 12186MiB | 0% Default |
±------------------------------±---------------------±---------------------+