multiformat_conversion_codelets_opencl.c 2.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2011 Institut National de Recherche en Informatique et Automatique
  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. #include <starpu.h>
  17. extern struct starpu_opencl_program opencl_conversion_program;
  18. void cpu_to_opencl_opencl_func(void *buffers[], void *args)
  19. {
  20. (void) args;
  21. int id, devid;
  22. cl_int err;
  23. cl_kernel kernel;
  24. cl_command_queue queue;
  25. cl_event event;
  26. unsigned n = STARPU_MULTIFORMAT_GET_NX(buffers[0]);
  27. cl_mem src = (cl_mem) STARPU_MULTIFORMAT_GET_CPU_PTR(buffers[0]);
  28. cl_mem dst = (cl_mem) STARPU_MULTIFORMAT_GET_OPENCL_PTR(buffers[0]);
  29. id = starpu_worker_get_id();
  30. devid = starpu_worker_get_devid(id);
  31. err = starpu_opencl_load_kernel(&kernel,
  32. &queue,
  33. &opencl_conversion_program,
  34. "cpu_to_opencl_opencl",
  35. devid);
  36. if (err != CL_SUCCESS)
  37. STARPU_OPENCL_REPORT_ERROR(err);
  38. err = clSetKernelArg(kernel, 0, sizeof(src), &src);
  39. if (err != CL_SUCCESS)
  40. STARPU_OPENCL_REPORT_ERROR(err);
  41. err = clSetKernelArg(kernel, 1, sizeof(dst), &dst);
  42. if (err != CL_SUCCESS)
  43. STARPU_OPENCL_REPORT_ERROR(err);
  44. err = clSetKernelArg(kernel, 2, sizeof(n), &n);
  45. if (err != CL_SUCCESS)
  46. STARPU_OPENCL_REPORT_ERROR(err);
  47. {
  48. size_t global=n;
  49. size_t local;
  50. size_t s;
  51. cl_device_id device;
  52. starpu_opencl_get_device(devid, &device);
  53. err = clGetKernelWorkGroupInfo (kernel,
  54. device,
  55. CL_KERNEL_WORK_GROUP_SIZE,
  56. sizeof(local),
  57. &local,
  58. &s);
  59. if (err != CL_SUCCESS)
  60. STARPU_OPENCL_REPORT_ERROR(err);
  61. if (local > global)
  62. local = global;
  63. err = clEnqueueNDRangeKernel(queue,
  64. kernel,
  65. 1,
  66. NULL,
  67. &global,
  68. &local,
  69. 0,
  70. NULL,
  71. &event);
  72. if (err != CL_SUCCESS)
  73. STARPU_OPENCL_REPORT_ERROR(err);
  74. }
  75. clFinish(queue);
  76. starpu_opencl_collect_stats(event);
  77. clReleaseEvent(event);
  78. starpu_opencl_release_kernel(kernel);
  79. }