Re: [Starpu-devel] StarPU 1.2.0 not working with CUDA 8 and Ubuntu

[Gabriel Gazolla <gabrielgazolla@gmail.com>]

Hello Samuel,

Sorry for the huge delay, I was on vacation and after that forgot to answer the email for you, was using CUDA 7.5 so I didn`t need to worry about this issue, but now I need to use StarPU on CUDA 8...

Attached on this email are the files you requested:

- configure output with verbose enabled

- make check output

- test.c software you sent me to test and its output

- starpumachinedisplay output is also attached.

Tell me if you need me to do any other tests...

I did this test on a machine using Ubuntu 14.04 LTS and CUDA 8. Cuda 7.5 everything works fine, but CUDA 8 dont, I am using starpu 1.2.0 release of 25.08.2016.

Hope to hear back from you soon,

Thanks,

Joao Gazolla

On Wed, Nov 23, 2016 at 4:50 PM, Samuel Thibault <samuel.thibault@inria.fr> wrote:

Gabriel Gazolla, on Wed 23 Nov 2016 16:38:24 -0200, wrote:
> If you try to compile StarPU 1.2 with CUDA 8 probably you will receive a lot of
> warnings about sm_20. Did you notice that?

Yes. It's "funny" in that we don't request that architecture at all...

It seems like nvcc just always emit that warning unless we explicitly
request for a more recent architecture... But then if we request for an
explicit more recent architecture, later on it'll become obsolete and
that'll turn into an actual error... I guess I'll just disable the
warning for releases.

Samuel

Attachment: configure, verbose enabled
Description: Binary data

Attachment: makeCheck
Description: Binary data

driver version 8000
runtime version 8000

CUDA0
GeForce GTX 670MX
2.939 GB
5 MP

#include <stdio.h>
#include <cuda.h>
#include <cuda_runtime.h>

int main(void) {
	int n, i, version;
	cudaError_t err;

	err = cudaGetDeviceCount(&n);
	if (err)
	{
		fprintf(stderr,"cuda error %d\n", err);
		exit(1);
	}
	cudaDriverGetVersion(&version);
	printf("driver version %d\n", version);
	cudaRuntimeGetVersion(&version);
	printf("runtime version %d\n", version);
	printf("\n");

	for (i = 0; i < n; i++) {
		struct cudaDeviceProp props;
		printf("CUDA%d\n", i);
		err = cudaGetDeviceProperties(&props, i);
		if (err)
		{
			fprintf(stderr,"cuda error %d\n", err);
			continue;
		}
		printf("%s\n", props.name);
		printf("%0.3f GB\n", (float) props.totalGlobalMem / (1<<30));
		printf("%u MP\n", props.multiProcessorCount);
		printf("\n");
	}
	return 0;
}

Attachment: starPUMachineDisplayOUTPUT
Description: Binary data