API tools faq
paste
matheus__serpa

CUDA

matheus__serpa
Oct 6th, 2014
393
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
C 2.01 KB | None | 0 0
raw download clone embed print report
  1. #include <cuda_runtime.h>
  2.  
  3. static void GPUHandleError( cudaError_t err, const char *file, const int line ) {
  4.     if (err != cudaSuccess) {
  5.         printf( "%s in %s at line %d\n", cudaGetErrorString( err ), file, line );
  6.         exit( EXIT_FAILURE );
  7.     }
  8. }
  9.  
  10. #define GPU_HANDLE_ERROR( err ) (GPUHandleError( err, __FILE__, __LINE__ ))
  11.  
  12. // 1D to 2D to 1D
  13. unsigned int idx = blockIdx.x * blockDim.x + threadIdx.x; // idx < lx * ly
  14. unsigned int x = idx / sizeY;
  15. unsigned int y = idx % sizeY;
  16.  
  17. // 2D to 1D with Blocks
  18. unsigned int x = blockIdx.x;
  19. unsigned int y = blockIdx.y;
  20. unsigned int offset = x + y * gridDim.x;
  21.  
  22. // 2D to 1D with Blocks and Threads
  23. unsigned int x = threadIdx.x + blockIdx.x * blockDim.x;
  24. unsigned int y = threadIdx.y + blockIdx.y * BlockDim.y;
  25. unsigned int offset = x + y * gridDim.x * gridDim.x;
  26.  
  27. // Shared memory
  28. __shared__ float cache[threadsPerBlock];
  29. unsigned int tid = threadIdx.x + blockIdx.x * blockDim.x;
  30. unsigned int cacheIndex = threadIdx.x;
  31. unsigned int stride = blockDim.x * gridDim.x;
  32. __syncthreads();
  33.  
  34.  
  35. // Other hand
  36. cudaSetDevice(0);
  37.  
  38. unsigned int minGridSize, BLOCK_KERNEL;
  39. cudaOccupancyMaxPotentialBlockSize(&minGridSize, &BLOCK_KERNEL, kernel, 0, size);
  40. unsigned int GRID_KERNEL = (size + BLOCK_KERNEL - 1) / BLOCK_KERNEL;
  41.  
  42.  
  43. cudaEvent_t start, stop;
  44. cudaEventCreate(&start); cudaEventCreate(&stop);
  45.  
  46. double *d_var, *h_var;
  47. size_t size = lx * ly * sizeof(double);
  48.  
  49. cudaMalloc((void **) &d_var, size);
  50.  
  51. cudaMemcpy(d_var, h_var, size, cudaMemcpyHostToDevice);
  52. cudaDeviceSynchronize();
  53.  
  54. cudaEventRecord(start, 0);
  55. cudaDeviceSynchronize();
  56.  
  57. kernel<<<GRID_KERNEL, BLOCK_KERNEL>>> (...);
  58. cudaDeviceSynchronize();
  59. cudaGetLastError();
  60.  
  61. cudaEventRecord(&stop, 0);
  62. cudaDeviceSynchronize();
  63.  
  64. float timer; // result in ms. /1000 to convert to s.
  65. cudaEventElapsedTime(&timer, start, stop);
  66.  
  67. cudaEventDestroy(start); cudaEventDestroy(stop);
  68.  
  69. cudaMemcpy(h_var, d_var, size, cudaMemcpyDeviceToHost);
  70. cudaDeviceSynchronize();
  71.  
  72. cudaFree(d_var);
  73. cudaDeviceReset();
Advertisement
Add Comment
Please, Sign In to add comment
Public Pastes
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!