Dynamic Parallelism on GTX 980 ti: Unknown error
I am trying to dynamically parallelism on a GTX 980 ti card. All attempts to run the code return an "unknown error". Simple code is shown below with compilation options.
I can execute kernels in depth=0
no problem. The first time the child is called, an error is thrown.
cudaDeviceSynchronize()
were included after looking at other questions here, but didn't fix the issue.
Any ideas? Could this be a driver issue?
Edit 1:
OS: Linux-x86_64
Nvidia driver version: 384.59
nvcc version 7.5.17
There are two 980 ti connected to PCIe x16 Gen3. The system also has windows installed on another SSD with a RAID array.
#include <cuda.h>
#include <fstream>
#include <stdio.h>
#include <stdlib.h>
__global__ void ker_two(){
int two=0;
two++;
}
__global__ void ker_one(){
int one=0;
one++;
ker_two<<<1,1>>>();
cudaDeviceSynchronize();
};
int main( ){
ker_one<<<1,1>>>();
cudaDeviceSynchronize();
cudaError_t err = cudaGetLastError();
if (err != cudaSuccess)
printf("Cuda Error: %s\n", cudaGetErrorString(err));//*/
return 0;
}
compiled with
nvcc -arch=compute_52 -rdc=true -lcudadevrt test.cu
source to share
I can (?) Reproduce the error on a machine with a Maxwell Titan card. This is a Fedora 24 distribution with CUDA 8.0.61 installed manually. The driver version is 375.51.
However - it seems the problem only occurs on my system when I call cudaDeviceSynchronize()
inside ker_one()
, whether I call the second core or not. Perhaps the problem you are seeing, not dynamic parallelism per se.
Considering @talonmies comment, it might even be a driver issue.
source to share