vector_scal_opencl.c 2.5 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2010, 2012, 2013 Centre National de la Recherche Scientifique
  4. * Copyright (C) 2010 Institut National de Recherche en Informatique et Automatique
  5. * Copyright (C) 2011 Université de Bordeaux 1
  6. *
  7. * StarPU is free software; you can redistribute it and/or modify
  8. * it under the terms of the GNU Lesser General Public License as published by
  9. * the Free Software Foundation; either version 2.1 of the License, or (at
  10. * your option) any later version.
  11. *
  12. * StarPU is distributed in the hope that it will be useful, but
  13. * WITHOUT ANY WARRANTY; without even the implied warranty of
  14. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  15. *
  16. * See the GNU Lesser General Public License in COPYING.LGPL for more details.
  17. */
  18. //! [To be included]
  19. #include <starpu.h>
  20. extern struct starpu_opencl_program programs;
  21. void scal_opencl_func(void *buffers[], void *_args)
  22. {
  23. float *factor = _args;
  24. int id, devid, err;
  25. cl_kernel kernel;
  26. cl_command_queue queue;
  27. cl_event event;
  28. /* length of the vector */
  29. unsigned n = STARPU_VECTOR_GET_NX(buffers[0]);
  30. /* OpenCL copy of the vector pointer */
  31. cl_mem val = (cl_mem)STARPU_VECTOR_GET_DEV_HANDLE(buffers[0]);
  32. id = starpu_worker_get_id();
  33. devid = starpu_worker_get_devid(id);
  34. err = starpu_opencl_load_kernel(&kernel, &queue, &programs, "vector_mult_opencl",
  35. devid);
  36. if (err != CL_SUCCESS) STARPU_OPENCL_REPORT_ERROR(err);
  37. err = clSetKernelArg(kernel, 0, sizeof(n), &n);
  38. err |= clSetKernelArg(kernel, 1, sizeof(val), &val);
  39. err |= clSetKernelArg(kernel, 2, sizeof(*factor), factor);
  40. if (err) STARPU_OPENCL_REPORT_ERROR(err);
  41. {
  42. size_t global=n;
  43. size_t local;
  44. size_t s;
  45. cl_device_id device;
  46. starpu_opencl_get_device(devid, &device);
  47. err = clGetKernelWorkGroupInfo (kernel, device, CL_KERNEL_WORK_GROUP_SIZE,
  48. sizeof(local), &local, &s);
  49. if (err != CL_SUCCESS) STARPU_OPENCL_REPORT_ERROR(err);
  50. if (local > global) local=global;
  51. err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global, &local, 0,
  52. NULL, &event);
  53. if (err != CL_SUCCESS) STARPU_OPENCL_REPORT_ERROR(err);
  54. }
  55. clFinish(queue);
  56. starpu_opencl_collect_stats(event);
  57. clReleaseEvent(event);
  58. starpu_opencl_release_kernel(kernel);
  59. }
  60. //! [To be included]