gpgpusim has been built but unable to run a cuda application about gpgpu-sim_distribution HOT 6 CLOSED

Try commenting out cudaGetDevice, cudaSetDevice, cudaStreamCreate calls from the app?
If you try the BFS from Rodinia 3.1, it works.

from gpgpu-sim_distribution.

Actually, it is more than simple device checking. Looking at part of the init process (here), such device information are used to determine the streams and other thing to run on gpu.

Isn't possible for gpgpusim to return a device for cudaGetDevice()?

from gpgpu-sim_distribution.

Hello Mahmood,

Could you run the deviceQuery for the GPGPU-sim 3.0 or 4.0 to get the properties of the simulator? If you could (I hope so), can you share with me? I cannot go beyond the below log. Best wishes.

./deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 100
-> no CUDA-capable device is detected
Result = FAIL

from gpgpu-sim_distribution.

Hi topcuburak

I am also facing same problem. All application compiling but while running it is not finding any CUDA enabled device.

from gpgpu-sim_distribution.

The topic is old, really. I suggest you to test the latest version.

from gpgpu-sim_distribution.

Hi I'm getting the output-

Iterate 10 times with GPU 0 or CPU 1: 0
test 4
ERROR: no CUDA-capable device is detected

I'm a absolute beginner in CUDA and GPGPU-SIM (started yesterday)
Can someone please help.

I'm trying the following code -
#include <cuda_runtime.h>
#include <cuda.h>
#include <stdio.h>
#include <sys/time.h>
#include <math.h>

#define N 32

global void Kernel_double(int niters, int* d_inputs,double* d_outputs)
{
int tid = blockIdx.x * blockDim.x + threadIdx.x;

if (tid<N) {
double val =(double) d_inputs[tid];
/*for (int iter=0; iter < niters; iter++){
val = (sqrt(pow(val,2.0)) + 5.0) - 101.0;
val = (val / 3.0) + 102.0;
val = (val + 1.07) - 103.0;
val = (val / 1.037) + 104.0;
val = (val + 3.00) - 105.0;
val = (val / 0.22) + 106.0;
}*/
val = val + 1.0;
//printf("This is %f\n",val);
d_outputs[tid] = val;

}
}

int main(int argc, char **argv)
{

int niters = 10;
int cpu = 0;
printf("Iterate %d times with GPU 0 or CPU 1: %d\n", niters, cpu);

int inputs[N];
for (int i = 0; i<N; i++){
inputs[i] = i+1;
}

int *d_inputs;
double *d_outputs;
double outputs[N];

cudaMalloc( (void**)&d_inputs, N*sizeof(int));
cudaMalloc( (void**)&d_outputs, N*sizeof(double));
printf("test %d \n", inputs[3]);
cudaMemcpy(d_inputs, inputs, N*sizeof(int), cudaMemcpyHostToDevice);

// printf("test %d \n", d_inputs[1]);
Kernel_double<<<16,2>>>(niters, d_inputs,d_outputs);
cudaDeviceSynchronize();
cudaError_t error = cudaGetLastError();
if(error!=cudaSuccess)
{
fprintf(stderr,"ERROR: %s\n", cudaGetErrorString(error) );
exit(-1);
}
//cudaDeviceSynchronize();
cudaMemcpy(outputs, d_outputs, N*sizeof(double), cudaMemcpyDeviceToHost);
for(int j =0;j<10; j++){
printf("Outputs[%d] is: %f\n",j, outputs[j]);
}
cudaFree(d_inputs);
cudaFree(d_outputs);

return EXIT_SUCCESS;

}

I have nvcc 11.2
gcc 10.2
I'm using the latest GPGPU Sim

from gpgpu-sim_distribution.