coo_cuda.cu 2.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2012 Inria
  4. * Copyright (C) 2012,2015,2017 CNRS
  5. *
  6. * StarPU is free software; you can redistribute it and/or modify
  7. * it under the terms of the GNU Lesser General Public License as published by
  8. * the Free Software Foundation; either version 2.1 of the License, or (at
  9. * your option) any later version.
  10. *
  11. * StarPU is distributed in the hope that it will be useful, but
  12. * WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  14. *
  15. * See the GNU Lesser General Public License in COPYING.LGPL for more details.
  16. */
  17. #include <starpu.h>
  18. #include "../test_interfaces.h"
  19. extern struct test_config coo_config;
  20. __global__ void coo_cuda(int *val, uint32_t 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] *= -1;
  29. }
  30. extern "C" void test_coo_cuda_func(void *buffers[], void *args)
  31. {
  32. int factor;
  33. int *ret;
  34. int *val;
  35. cudaError_t error;
  36. uint32_t nvalues = STARPU_COO_GET_NVALUES(buffers[0]);
  37. unsigned threads_per_block = 64;
  38. unsigned nblocks = (nvalues + threads_per_block-1) / threads_per_block;
  39. factor = *(int *) args;
  40. val = (int *) STARPU_COO_GET_VALUES(buffers[0]);
  41. error = cudaMalloc(&ret, sizeof(int));
  42. if (error != cudaSuccess)
  43. STARPU_CUDA_REPORT_ERROR(error);
  44. error = cudaMemcpy(ret,
  45. &coo_config.copy_failed,
  46. sizeof(int),
  47. cudaMemcpyHostToDevice);
  48. if (error != cudaSuccess)
  49. STARPU_CUDA_REPORT_ERROR(error);
  50. coo_cuda<<<nblocks,threads_per_block,2,starpu_cuda_get_local_stream()>>>
  51. (val, nvalues, ret, factor);
  52. error = cudaMemcpy(&coo_config.copy_failed,
  53. ret,
  54. sizeof(int),
  55. cudaMemcpyDeviceToHost);
  56. if (error != cudaSuccess)
  57. STARPU_CUDA_REPORT_ERROR(error);
  58. cudaFree(ret);
  59. cudaStreamSynchronize(starpu_cuda_get_local_stream());
  60. }