axpy_opencl.c 2.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2012 Inria
  4. * Copyright (C) 2012,2015-2017 CNRS
  5. * Copyright (C) 2012,2014,2015 Université de Bordeaux
  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. /* OpenCL codelet for axpy */
  19. #include <starpu.h>
  20. #include "axpy.h"
  21. extern struct starpu_opencl_program opencl_program;
  22. void axpy_opencl(void *buffers[], void *_args)
  23. {
  24. TYPE *alpha = _args;
  25. int id, devid;
  26. cl_int err;
  27. cl_kernel kernel;
  28. cl_command_queue queue;
  29. unsigned n = STARPU_VECTOR_GET_NX(buffers[0]);
  30. cl_mem x = (cl_mem) STARPU_VECTOR_GET_DEV_HANDLE(buffers[0]);
  31. cl_mem y = (cl_mem) STARPU_VECTOR_GET_DEV_HANDLE(buffers[1]);
  32. id = starpu_worker_get_id_check();
  33. devid = starpu_worker_get_devid(id);
  34. err = starpu_opencl_load_kernel(&kernel, &queue, &opencl_program, "_axpy_opencl", devid);
  35. if (err != CL_SUCCESS)
  36. STARPU_OPENCL_REPORT_ERROR(err);
  37. err = clSetKernelArg(kernel, 0, sizeof(x), &x);
  38. err|= clSetKernelArg(kernel, 1, sizeof(y), &y);
  39. err|= clSetKernelArg(kernel, 2, sizeof(n), &n);
  40. err|= clSetKernelArg(kernel, 3, sizeof(*alpha), alpha);
  41. if (err)
  42. STARPU_OPENCL_REPORT_ERROR(err);
  43. {
  44. size_t global=n;
  45. size_t local;
  46. size_t s;
  47. cl_device_id device;
  48. starpu_opencl_get_device(devid, &device);
  49. err = clGetKernelWorkGroupInfo (kernel, device, CL_KERNEL_WORK_GROUP_SIZE, sizeof(local), &local, &s);
  50. if (err != CL_SUCCESS)
  51. STARPU_OPENCL_REPORT_ERROR(err);
  52. if (local > global)
  53. local=global;
  54. err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global, &local, 0, NULL, NULL);
  55. if (err != CL_SUCCESS)
  56. STARPU_OPENCL_REPORT_ERROR(err);
  57. }
  58. starpu_opencl_release_kernel(kernel);
  59. }