matrix_cuda.cu 2.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2011 Institut National de Recherche en Informatique et Automatique
  4. *
  5. * StarPU is free software; you can redistribute it and/or modify
  6. * it under the terms of the GNU Lesser General Public License as published by
  7. * the Free Software Foundation; either version 2.1 of the License, or (at
  8. * your option) any later version.
  9. *
  10. * StarPU is distributed in the hope that it will be useful, but
  11. * WITHOUT ANY WARRANTY; without even the implied warranty of
  12. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  13. *
  14. * See the GNU Lesser General Public License in COPYING.LGPL for more details.
  15. */
  16. #include <starpu.h>
  17. #include <starpu_cuda.h>
  18. #include "../test_interfaces.h"
  19. extern struct test_config matrix_config;
  20. __global__ void matrix_cuda(int *val, unsigned n, int *err, int factor)
  21. {
  22. unsigned i = blockIdx.x*blockDim.x + threadIdx.x;
  23. if (i >= n)
  24. return;
  25. if (val[i] != i*factor)
  26. *err = 1;
  27. else
  28. val[i] = -val[i];
  29. }
  30. extern "C" void test_matrix_cuda_func(void *buffers[], void *args)
  31. {
  32. int factor;
  33. int *ret;
  34. int *val;
  35. cudaError_t error;
  36. unsigned int nx, ny, n;
  37. nx = STARPU_MATRIX_GET_NX(buffers[0]);
  38. ny = STARPU_MATRIX_GET_NY(buffers[0]);
  39. n = nx * ny;
  40. unsigned threads_per_block = 64;
  41. unsigned nblocks = (n + threads_per_block-1) / threads_per_block;
  42. factor = *(int *) args;
  43. val = (int *) STARPU_MATRIX_GET_PTR(buffers[0]);
  44. error = cudaMalloc(&ret, sizeof(int));
  45. if (error != cudaSuccess)
  46. STARPU_CUDA_REPORT_ERROR(error);
  47. error = cudaMemcpy(ret,
  48. &matrix_config.copy_failed,
  49. sizeof(int),
  50. cudaMemcpyHostToDevice);
  51. if (error != cudaSuccess)
  52. STARPU_CUDA_REPORT_ERROR(error);
  53. matrix_cuda<<<nblocks,threads_per_block,2,starpu_cuda_get_local_stream()>>>(val, n, ret, factor);
  54. error = cudaMemcpy(&matrix_config.copy_failed,
  55. ret,
  56. sizeof(int),
  57. cudaMemcpyDeviceToHost);
  58. cudaFree(ret);
  59. cudaStreamSynchronize(starpu_cuda_get_local_stream());
  60. }