test nvidia-smi

test.cu

// compile me by nvcc copy.cu -o test
#include <iostream>

int main(int argc, char const *argv[])
{
  const size_t MB = 10;
  const size_t len = MB*256;


  float *d_ptr, *h_ptr;

  h_ptr = new float[len];

  cudaMalloc(&d_ptr, len*sizeof(float)); 

  for (int i = 0; i < 150000000; ++i)
  {
    cudaMemcpy(d_ptr, h_ptr, len*sizeof(float), cudaMemcpyHostToDevice);

  }
  return 0;
}