multiformat_conversion_codelets_opencl.c 3.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112
  1. /* StarPU --- Runtime system for heterogeneous multicore architectures.
  2. *
  3. * Copyright (C) 2011-2020 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. #include <starpu.h>
  17. #include "../../../helper.h"
  18. #define KERNEL_LOCATION "tests/datawizard/interfaces/multiformat/multiformat_conversion_codelets_kernel.cl"
  19. static struct starpu_opencl_program opencl_conversion_program;
  20. void cpu_to_opencl_opencl_func(void *buffers[], void *args)
  21. {
  22. STARPU_SKIP_IF_VALGRIND;
  23. (void) args;
  24. int id, devid, ret;
  25. cl_int err;
  26. cl_kernel kernel;
  27. cl_command_queue queue;
  28. cl_event event;
  29. unsigned n = STARPU_MULTIFORMAT_GET_NX(buffers[0]);
  30. cl_mem src = (cl_mem) STARPU_MULTIFORMAT_GET_CPU_PTR(buffers[0]);
  31. cl_mem dst = (cl_mem) STARPU_MULTIFORMAT_GET_OPENCL_PTR(buffers[0]);
  32. id = starpu_worker_get_id_check();
  33. devid = starpu_worker_get_devid(id);
  34. ret = starpu_opencl_load_opencl_from_file(KERNEL_LOCATION,
  35. &opencl_conversion_program,
  36. NULL);
  37. STARPU_CHECK_RETURN_VALUE(ret, "starpu_opencl_load_opencl_from_file");
  38. err = starpu_opencl_load_kernel(&kernel,
  39. &queue,
  40. &opencl_conversion_program,
  41. "cpu_to_opencl_opencl",
  42. devid);
  43. if (err != CL_SUCCESS)
  44. STARPU_OPENCL_REPORT_ERROR(err);
  45. err = clSetKernelArg(kernel, 0, sizeof(src), &src);
  46. if (err != CL_SUCCESS)
  47. STARPU_OPENCL_REPORT_ERROR(err);
  48. err = clSetKernelArg(kernel, 1, sizeof(dst), &dst);
  49. if (err != CL_SUCCESS)
  50. STARPU_OPENCL_REPORT_ERROR(err);
  51. err = clSetKernelArg(kernel, 2, sizeof(n), &n);
  52. if (err != CL_SUCCESS)
  53. STARPU_OPENCL_REPORT_ERROR(err);
  54. {
  55. size_t global=n;
  56. size_t local;
  57. size_t s;
  58. cl_device_id device;
  59. starpu_opencl_get_device(devid, &device);
  60. err = clGetKernelWorkGroupInfo (kernel,
  61. device,
  62. CL_KERNEL_WORK_GROUP_SIZE,
  63. sizeof(local),
  64. &local,
  65. &s);
  66. if (err != CL_SUCCESS)
  67. STARPU_OPENCL_REPORT_ERROR(err);
  68. if (local > global)
  69. local = global;
  70. else
  71. global = (global + local-1) / local * local;
  72. err = clEnqueueNDRangeKernel(queue,
  73. kernel,
  74. 1,
  75. NULL,
  76. &global,
  77. &local,
  78. 0,
  79. NULL,
  80. &event);
  81. if (err != CL_SUCCESS)
  82. STARPU_OPENCL_REPORT_ERROR(err);
  83. }
  84. clFinish(queue);
  85. starpu_opencl_collect_stats(event);
  86. clReleaseEvent(event);
  87. starpu_opencl_release_kernel(kernel);
  88. ret = starpu_opencl_unload_opencl(&opencl_conversion_program);
  89. STARPU_CHECK_RETURN_VALUE(ret, "starpu_opencl_unload_opencl");
  90. }