vector_scal_opencl.c 2.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2010-2020 Université de Bordeaux, CNRS (LaBRI UMR 5800), Inria
  4. *
  5. * StarPU 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. * StarPU 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_scal.c: here we implement a OpenCL version.
  18. */
  19. #include <starpu.h>
  20. extern struct starpu_opencl_program opencl_program;
  21. void scal_opencl_func(void *buffers[], void *_args)
  22. {
  23. float *factor = _args;
  24. int id, devid;
  25. cl_int err;
  26. cl_kernel kernel;
  27. cl_command_queue queue;
  28. /* length of the vector */
  29. unsigned int 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_check();
  33. devid = starpu_worker_get_devid(id);
  34. err = starpu_opencl_load_kernel(&kernel, &queue, &opencl_program, "vector_mult_opencl", devid);
  35. if (err != CL_SUCCESS) STARPU_OPENCL_REPORT_ERROR(err);
  36. err = clSetKernelArg(kernel, 0, sizeof(n), &n);
  37. err |= clSetKernelArg(kernel, 1, sizeof(val), &val);
  38. err |= clSetKernelArg(kernel, 2, sizeof(*factor), factor);
  39. if (err) STARPU_OPENCL_REPORT_ERROR(err);
  40. {
  41. size_t global=n;
  42. size_t local;
  43. size_t s;
  44. cl_device_id device;
  45. starpu_opencl_get_device(devid, &device);
  46. err = clGetKernelWorkGroupInfo (kernel, device, CL_KERNEL_WORK_GROUP_SIZE, sizeof(local), &local, &s);
  47. if (err != CL_SUCCESS) STARPU_OPENCL_REPORT_ERROR(err);
  48. if (local > global) local=global;
  49. err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global, &local, 0, NULL, NULL);
  50. if (err != CL_SUCCESS) STARPU_OPENCL_REPORT_ERROR(err);
  51. }
  52. starpu_opencl_release_kernel(kernel);
  53. }