csr_cuda.cu 2.1 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2011,2012 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 csr_config;
  20. __global__ void csr_cuda(int *nzval, uint32_t nnz, int *err, int factor)
  21. {
  22. unsigned i = blockIdx.x*blockDim.x + threadIdx.x;
  23. if (i >= nnz)
  24. return;
  25. if (nzval[i] != (i+1)*factor)
  26. *err = 1;
  27. else
  28. nzval[i] = -nzval[i];
  29. }
  30. extern "C" void test_csr_cuda_func(void *buffers[], void *args)
  31. {
  32. int factor;
  33. int *ret;
  34. int *val;
  35. cudaError_t error;
  36. uint32_t nnz = STARPU_CSR_GET_NNZ(buffers[0]);
  37. unsigned threads_per_block = 64;
  38. unsigned nblocks = (nnz + threads_per_block-1) / threads_per_block;
  39. factor = *(int *) args;
  40. val = (int *) STARPU_CSR_GET_NZVAL(buffers[0]);
  41. error = cudaMalloc(&ret, sizeof(int));
  42. if (error != cudaSuccess)
  43. STARPU_CUDA_REPORT_ERROR(error);
  44. error = cudaMemcpyAsync(ret,
  45. &csr_config.copy_failed,
  46. sizeof(int),
  47. cudaMemcpyHostToDevice, starpu_cuda_get_local_stream());
  48. if (error != cudaSuccess)
  49. STARPU_CUDA_REPORT_ERROR(error);
  50. csr_cuda<<<nblocks,threads_per_block,2,starpu_cuda_get_local_stream()>>> (val, nnz, ret, factor);
  51. error = cudaMemcpyAsync(&csr_config.copy_failed,
  52. ret,
  53. sizeof(int),
  54. cudaMemcpyDeviceToHost, starpu_cuda_get_local_stream());
  55. if (error != cudaSuccess)
  56. STARPU_CUDA_REPORT_ERROR(error);
  57. cudaFree(ret);
  58. cudaStreamSynchronize(starpu_cuda_get_local_stream());
  59. }