hevc_idct.c 3.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105
  1. /*
  2. * Copyright (c) 2016 Alexandra Hájková
  3. *
  4. * This file is part of FFmpeg.
  5. *
  6. * FFmpeg is free software; you can redistribute it and/or modify
  7. * it under the terms of the GNU General Public License as published by
  8. * the Free Software Foundation; either version 2 of the License, or
  9. * (at your option) any later version.
  10. *
  11. * FFmpeg is distributed in the hope that it will be useful,
  12. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  14. * GNU General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU General Public License along
  17. * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
  18. * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  19. */
  20. #include <string.h>
  21. #include "libavutil/intreadwrite.h"
  22. #include "libavcodec/hevcdsp.h"
  23. #include "checkasm.h"
  24. #define randomize_buffers(buf, size) \
  25. do { \
  26. int j; \
  27. for (j = 0; j < size; j++) { \
  28. int16_t r = rnd(); \
  29. AV_WN16A(buf + j, r); \
  30. } \
  31. } while (0)
  32. static void check_idct(HEVCDSPContext h, int bit_depth)
  33. {
  34. int i;
  35. LOCAL_ALIGNED(32, int16_t, coeffs0, [32 * 32]);
  36. LOCAL_ALIGNED(32, int16_t, coeffs1, [32 * 32]);
  37. for (i = 2; i <= 5; i++) {
  38. int block_size = 1 << i;
  39. int size = block_size * block_size;
  40. int col_limit = block_size;
  41. declare_func(void, int16_t *coeffs, int col_limit);
  42. randomize_buffers(coeffs0, size);
  43. memcpy(coeffs1, coeffs0, sizeof(*coeffs0) * size);
  44. if (check_func(h.idct[i - 2], "hevc_idct_%dx%d_%d", block_size, block_size, bit_depth)) {
  45. call_ref(coeffs0, col_limit);
  46. call_new(coeffs1, col_limit);
  47. if (memcmp(coeffs0, coeffs1, sizeof(*coeffs0) * size))
  48. fail();
  49. bench_new(coeffs1, col_limit);
  50. }
  51. }
  52. }
  53. static void check_idct_dc(HEVCDSPContext h, int bit_depth)
  54. {
  55. int i;
  56. LOCAL_ALIGNED(32, int16_t, coeffs0, [32 * 32]);
  57. LOCAL_ALIGNED(32, int16_t, coeffs1, [32 * 32]);
  58. for (i = 2; i <= 5; i++) {
  59. int block_size = 1 << i;
  60. int size = block_size * block_size;
  61. declare_func_emms(AV_CPU_FLAG_MMXEXT, void, int16_t *coeffs);
  62. randomize_buffers(coeffs0, size);
  63. memcpy(coeffs1, coeffs0, sizeof(*coeffs0) * size);
  64. if (check_func(h.idct_dc[i - 2], "hevc_idct_%dx%d_dc_%d", block_size, block_size, bit_depth)) {
  65. call_ref(coeffs0);
  66. call_new(coeffs1);
  67. if (memcmp(coeffs0, coeffs1, sizeof(*coeffs0) * size))
  68. fail();
  69. bench_new(coeffs1);
  70. }
  71. }
  72. }
  73. void checkasm_check_hevc_idct(void)
  74. {
  75. int bit_depth;
  76. for (bit_depth = 8; bit_depth <= 12; bit_depth += 2) {
  77. HEVCDSPContext h;
  78. ff_hevc_dsp_init(&h, bit_depth);
  79. check_idct_dc(h, bit_depth);
  80. }
  81. report("idct_dc");
  82. for (bit_depth = 8; bit_depth <= 12; bit_depth += 2) {
  83. HEVCDSPContext h;
  84. ff_hevc_dsp_init(&h, bit_depth);
  85. check_idct(h, bit_depth);
  86. }
  87. report("idct");
  88. }