vector_scal_cuda.cu 1.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748
  1. /*
  2. * StarPU
  3. * Copyright (C) Université Bordeaux 1, CNRS 2008-2009 (see AUTHORS file)
  4. *
  5. * This program 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. * This program 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. /*
  17. * This example complements vector_scale.c: here we implement a CUDA version.
  18. */
  19. #include <starpu.h>
  20. #include <starpu_cuda.h>
  21. static __global__ void vector_mult_cuda(float *val, unsigned n,
  22. float factor)
  23. {
  24. unsigned i = threadIdx.x;
  25. if (i < n)
  26. val[i] *= factor;
  27. }
  28. extern "C" void scal_cuda_func(void *buffers[], void *_args)
  29. {
  30. float *factor = (float *)_args;
  31. /* length of the vector */
  32. unsigned n = STARPU_VECTOR_GET_NX(buffers[0]);
  33. /* local copy of the vector pointer */
  34. float *val = (float *)STARPU_VECTOR_GET_PTR(buffers[0]);
  35. unsigned threads_per_block = 64;
  36. unsigned nblocks = (n + threads_per_block-1) / threads_per_block;
  37. vector_mult_cuda<<<nblocks,threads_per_block,0,starpu_cuda_get_local_stream()>>>(val, n, *factor);
  38. cudaStreamSynchronize(starpu_cuda_get_local_stream());
  39. }