You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

247 lines
15KB

  1. /*
  2. * HEVC video decoder
  3. *
  4. * Copyright (C) 2012 - 2013 Guillaume Martres
  5. *
  6. * This file is part of Libav.
  7. *
  8. * Libav is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU Lesser General Public
  10. * License as published by the Free Software Foundation; either
  11. * version 2.1 of the License, or (at your option) any later version.
  12. *
  13. * Libav is distributed in the hope that it will be useful,
  14. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  16. * Lesser General Public License for more details.
  17. *
  18. * You should have received a copy of the GNU Lesser General Public
  19. * License along with Libav; if not, write to the Free Software
  20. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21. */
  22. #include "hevcdsp.h"
  23. static const int8_t transform[32][32] = {
  24. { 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
  25. 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 },
  26. { 90, 90, 88, 85, 82, 78, 73, 67, 61, 54, 46, 38, 31, 22, 13, 4,
  27. -4, -13, -22, -31, -38, -46, -54, -61, -67, -73, -78, -82, -85, -88, -90, -90 },
  28. { 90, 87, 80, 70, 57, 43, 25, 9, -9, -25, -43, -57, -70, -80, -87, -90,
  29. -90, -87, -80, -70, -57, -43, -25, -9, 9, 25, 43, 57, 70, 80, 87, 90 },
  30. { 90, 82, 67, 46, 22, -4, -31, -54, -73, -85, -90, -88, -78, -61, -38, -13,
  31. 13, 38, 61, 78, 88, 90, 85, 73, 54, 31, 4, -22, -46, -67, -82, -90 },
  32. { 89, 75, 50, 18, -18, -50, -75, -89, -89, -75, -50, -18, 18, 50, 75, 89,
  33. 89, 75, 50, 18, -18, -50, -75, -89, -89, -75, -50, -18, 18, 50, 75, 89 },
  34. { 88, 67, 31, -13, -54, -82, -90, -78, -46, -4, 38, 73, 90, 85, 61, 22,
  35. -22, -61, -85, -90, -73, -38, 4, 46, 78, 90, 82, 54, 13, -31, -67, -88 },
  36. { 87, 57, 9, -43, -80, -90, -70, -25, 25, 70, 90, 80, 43, -9, -57, -87,
  37. -87, -57, -9, 43, 80, 90, 70, 25, -25, -70, -90, -80, -43, 9, 57, 87 },
  38. { 85, 46, -13, -67, -90, -73, -22, 38, 82, 88, 54, -4, -61, -90, -78, -31,
  39. 31, 78, 90, 61, 4, -54, -88, -82, -38, 22, 73, 90, 67, 13, -46, -85 },
  40. { 83, 36, -36, -83, -83, -36, 36, 83, 83, 36, -36, -83, -83, -36, 36, 83,
  41. 83, 36, -36, -83, -83, -36, 36, 83, 83, 36, -36, -83, -83, -36, 36, 83 },
  42. { 82, 22, -54, -90, -61, 13, 78, 85, 31, -46, -90, -67, 4, 73, 88, 38,
  43. -38, -88, -73, -4, 67, 90, 46, -31, -85, -78, -13, 61, 90, 54, -22, -82 },
  44. { 80, 9, -70, -87, -25, 57, 90, 43, -43, -90, -57, 25, 87, 70, -9, -80,
  45. -80, -9, 70, 87, 25, -57, -90, -43, 43, 90, 57, -25, -87, -70, 9, 80 },
  46. { 78, -4, -82, -73, 13, 85, 67, -22, -88, -61, 31, 90, 54, -38, -90, -46,
  47. 46, 90, 38, -54, -90, -31, 61, 88, 22, -67, -85, -13, 73, 82, 4, -78 },
  48. { 75, -18, -89, -50, 50, 89, 18, -75, -75, 18, 89, 50, -50, -89, -18, 75,
  49. 75, -18, -89, -50, 50, 89, 18, -75, -75, 18, 89, 50, -50, -89, -18, 75 },
  50. { 73, -31, -90, -22, 78, 67, -38, -90, -13, 82, 61, -46, -88, -4, 85, 54,
  51. -54, -85, 4, 88, 46, -61, -82, 13, 90, 38, -67, -78, 22, 90, 31, -73 },
  52. { 70, -43, -87, 9, 90, 25, -80, -57, 57, 80, -25, -90, -9, 87, 43, -70,
  53. -70, 43, 87, -9, -90, -25, 80, 57, -57, -80, 25, 90, 9, -87, -43, 70 },
  54. { 67, -54, -78, 38, 85, -22, -90, 4, 90, 13, -88, -31, 82, 46, -73, -61,
  55. 61, 73, -46, -82, 31, 88, -13, -90, -4, 90, 22, -85, -38, 78, 54, -67 },
  56. { 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64,
  57. 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64 },
  58. { 61, -73, -46, 82, 31, -88, -13, 90, -4, -90, 22, 85, -38, -78, 54, 67,
  59. -67, -54, 78, 38, -85, -22, 90, 4, -90, 13, 88, -31, -82, 46, 73, -61 },
  60. { 57, -80, -25, 90, -9, -87, 43, 70, -70, -43, 87, 9, -90, 25, 80, -57,
  61. -57, 80, 25, -90, 9, 87, -43, -70, 70, 43, -87, -9, 90, -25, -80, 57 },
  62. { 54, -85, -4, 88, -46, -61, 82, 13, -90, 38, 67, -78, -22, 90, -31, -73,
  63. 73, 31, -90, 22, 78, -67, -38, 90, -13, -82, 61, 46, -88, 4, 85, -54 },
  64. { 50, -89, 18, 75, -75, -18, 89, -50, -50, 89, -18, -75, 75, 18, -89, 50,
  65. 50, -89, 18, 75, -75, -18, 89, -50, -50, 89, -18, -75, 75, 18, -89, 50 },
  66. { 46, -90, 38, 54, -90, 31, 61, -88, 22, 67, -85, 13, 73, -82, 4, 78,
  67. -78, -4, 82, -73, -13, 85, -67, -22, 88, -61, -31, 90, -54, -38, 90, -46 },
  68. { 43, -90, 57, 25, -87, 70, 9, -80, 80, -9, -70, 87, -25, -57, 90, -43,
  69. -43, 90, -57, -25, 87, -70, -9, 80, -80, 9, 70, -87, 25, 57, -90, 43 },
  70. { 38, -88, 73, -4, -67, 90, -46, -31, 85, -78, 13, 61, -90, 54, 22, -82,
  71. 82, -22, -54, 90, -61, -13, 78, -85, 31, 46, -90, 67, 4, -73, 88, -38 },
  72. { 36, -83, 83, -36, -36, 83, -83, 36, 36, -83, 83, -36, -36, 83, -83, 36,
  73. 36, -83, 83, -36, -36, 83, -83, 36, 36, -83, 83, -36, -36, 83, -83, 36 },
  74. { 31, -78, 90, -61, 4, 54, -88, 82, -38, -22, 73, -90, 67, -13, -46, 85,
  75. -85, 46, 13, -67, 90, -73, 22, 38, -82, 88, -54, -4, 61, -90, 78, -31 },
  76. { 25, -70, 90, -80, 43, 9, -57, 87, -87, 57, -9, -43, 80, -90, 70, -25,
  77. -25, 70, -90, 80, -43, -9, 57, -87, 87, -57, 9, 43, -80, 90, -70, 25 },
  78. { 22, -61, 85, -90, 73, -38, -4, 46, -78, 90, -82, 54, -13, -31, 67, -88,
  79. 88, -67, 31, 13, -54, 82, -90, 78, -46, 4, 38, -73, 90, -85, 61, -22 },
  80. { 18, -50, 75, -89, 89, -75, 50, -18, -18, 50, -75, 89, -89, 75, -50, 18,
  81. 18, -50, 75, -89, 89, -75, 50, -18, -18, 50, -75, 89, -89, 75, -50, 18 },
  82. { 13, -38, 61, -78, 88, -90, 85, -73, 54, -31, 4, 22, -46, 67, -82, 90,
  83. -90, 82, -67, 46, -22, -4, 31, -54, 73, -85, 90, -88, 78, -61, 38, -13 },
  84. { 9, -25, 43, -57, 70, -80, 87, -90, 90, -87, 80, -70, 57, -43, 25, -9,
  85. -9, 25, -43, 57, -70, 80, -87, 90, -90, 87, -80, 70, -57, 43, -25, 9 },
  86. { 4, -13, 22, -31, 38, -46, 54, -61, 67, -73, 78, -82, 85, -88, 90, -90,
  87. 90, -90, 88, -85, 82, -78, 73, -67, 61, -54, 46, -38, 31, -22, 13, -4 },
  88. };
  89. DECLARE_ALIGNED(16, const int16_t, ff_hevc_epel_coeffs[7][16]) = {
  90. { -2, 58, 10, -2, -2, 58, 10, -2, -2, 58, 10, -2, -2, 58, 10, -2 },
  91. { -4, 54, 16, -2, -4, 54, 16, -2, -4, 54, 16, -2, -4, 54, 16, -2 },
  92. { -6, 46, 28, -4, -6, 46, 28, -4, -6, 46, 28, -4, -6, 46, 28, -4 },
  93. { -4, 36, 36, -4, -4, 36, 36, -4, -4, 36, 36, -4, -4, 36, 36, -4 },
  94. { -4, 28, 46, -6, -4, 28, 46, -6, -4, 28, 46, -6, -4, 28, 46, -6 },
  95. { -2, 16, 54, -4, -2, 16, 54, -4, -2, 16, 54, -4, -2, 16, 54, -4 },
  96. { -2, 10, 58, -2, -2, 10, 58, -2, -2, 10, 58, -2, -2, 10, 58, -2 },
  97. };
  98. DECLARE_ALIGNED(16, const int8_t, ff_hevc_epel_coeffs8[7][16]) = {
  99. { -2, 58, 10, -2, -2, 58, 10, -2, -2, 58, 10, -2, -2, 58, 10, -2 },
  100. { -4, 54, 16, -2, -4, 54, 16, -2, -4, 54, 16, -2, -4, 54, 16, -2 },
  101. { -6, 46, 28, -4, -6, 46, 28, -4, -6, 46, 28, -4, -6, 46, 28, -4 },
  102. { -4, 36, 36, -4, -4, 36, 36, -4, -4, 36, 36, -4, -4, 36, 36, -4 },
  103. { -4, 28, 46, -6, -4, 28, 46, -6, -4, 28, 46, -6, -4, 28, 46, -6 },
  104. { -2, 16, 54, -4, -2, 16, 54, -4, -2, 16, 54, -4, -2, 16, 54, -4 },
  105. { -2, 10, 58, -2, -2, 10, 58, -2, -2, 10, 58, -2, -2, 10, 58, -2 },
  106. };
  107. DECLARE_ALIGNED(16, const int16_t, ff_hevc_qpel_coeffs[3][8]) = {
  108. { -1, 4, -10, 58, 17, -5, 1, 0 },
  109. { -1, 4, -11, 40, 40, -11, 4, -1 },
  110. { 0, 1, -5, 17, 58, -10, 4, -1 },
  111. };
  112. DECLARE_ALIGNED(16, const int8_t, ff_hevc_qpel_coeffs8[3][16]) = {
  113. { -1, 4, -10, 58, 17, -5, 1, 0, -1, 4, -10, 58, 17, -5, 1, 0 },
  114. { -1, 4, -11, 40, 40, -11, 4, -1, -1, 4, -11, 40, 40, -11, 4, -1 },
  115. { 0, 1, -5, 17, 58, -10, 4, -1, 0, 1, -5, 17, 58, -10, 4, -1 },
  116. };
  117. #define BIT_DEPTH 8
  118. #include "hevcdsp_template.c"
  119. #undef BIT_DEPTH
  120. #define BIT_DEPTH 9
  121. #include "hevcdsp_template.c"
  122. #undef BIT_DEPTH
  123. #define BIT_DEPTH 10
  124. #include "hevcdsp_template.c"
  125. #undef BIT_DEPTH
  126. void ff_hevc_dsp_init(HEVCDSPContext *hevcdsp, int bit_depth)
  127. {
  128. #undef FUNC
  129. #define FUNC(a, depth) a ## _ ## depth
  130. #define QPEL_FUNC(i, width, depth) \
  131. hevcdsp->put_hevc_qpel[0][0][i] = FUNC(put_hevc_qpel_pixels_ ## width, depth); \
  132. hevcdsp->put_hevc_qpel[0][1][i] = FUNC(put_hevc_qpel_h_ ## width, depth); \
  133. hevcdsp->put_hevc_qpel[1][0][i] = FUNC(put_hevc_qpel_v_ ## width, depth); \
  134. hevcdsp->put_hevc_qpel[1][1][i] = FUNC(put_hevc_qpel_hv_ ## width, depth); \
  135. #define EPEL_FUNC(i, width, depth) \
  136. hevcdsp->put_hevc_epel[0][0][i] = FUNC(put_hevc_epel_pixels_ ## width, depth); \
  137. hevcdsp->put_hevc_epel[0][1][i] = FUNC(put_hevc_epel_h_ ## width, depth); \
  138. hevcdsp->put_hevc_epel[1][0][i] = FUNC(put_hevc_epel_v_ ## width, depth); \
  139. hevcdsp->put_hevc_epel[1][1][i] = FUNC(put_hevc_epel_hv_ ## width, depth); \
  140. #define PRED_FUNC(i, width, depth) \
  141. hevcdsp->put_unweighted_pred[i] = FUNC(put_unweighted_pred_ ## width, depth); \
  142. hevcdsp->put_unweighted_pred_avg[i] = FUNC(put_unweighted_pred_avg_ ## width, depth); \
  143. hevcdsp->weighted_pred[i] = FUNC(put_weighted_pred_ ## width, depth); \
  144. hevcdsp->weighted_pred_avg[i] = FUNC(put_weighted_pred_avg_ ## width, depth); \
  145. #define PRED_FUNC_CHROMA(i, width, depth) \
  146. hevcdsp->put_unweighted_pred_chroma[i] = FUNC(put_unweighted_pred_ ## width, depth); \
  147. hevcdsp->put_unweighted_pred_avg_chroma[i] = FUNC(put_unweighted_pred_avg_ ## width, depth); \
  148. hevcdsp->weighted_pred_chroma[i] = FUNC(put_weighted_pred_ ## width, depth); \
  149. hevcdsp->weighted_pred_avg_chroma[i] = FUNC(put_weighted_pred_avg_ ## width, depth); \
  150. #define HEVC_DSP(depth) \
  151. hevcdsp->put_pcm = FUNC(put_pcm, depth); \
  152. hevcdsp->transquant_bypass[0] = FUNC(transquant_bypass4x4, depth); \
  153. hevcdsp->transquant_bypass[1] = FUNC(transquant_bypass8x8, depth); \
  154. hevcdsp->transquant_bypass[2] = FUNC(transquant_bypass16x16, depth); \
  155. hevcdsp->transquant_bypass[3] = FUNC(transquant_bypass32x32, depth); \
  156. hevcdsp->transform_skip = FUNC(transform_skip, depth); \
  157. hevcdsp->transform_4x4_luma_add = FUNC(transform_4x4_luma_add, depth); \
  158. hevcdsp->transform_add[0] = FUNC(transform_4x4_add, depth); \
  159. hevcdsp->transform_add[1] = FUNC(transform_8x8_add, depth); \
  160. hevcdsp->transform_add[2] = FUNC(transform_16x16_add, depth); \
  161. hevcdsp->transform_add[3] = FUNC(transform_32x32_add, depth); \
  162. \
  163. hevcdsp->sao_band_filter[0] = FUNC(sao_band_filter_0, depth); \
  164. hevcdsp->sao_band_filter[1] = FUNC(sao_band_filter_1, depth); \
  165. hevcdsp->sao_band_filter[2] = FUNC(sao_band_filter_2, depth); \
  166. hevcdsp->sao_band_filter[3] = FUNC(sao_band_filter_3, depth); \
  167. \
  168. hevcdsp->sao_edge_filter[0] = FUNC(sao_edge_filter_0, depth); \
  169. hevcdsp->sao_edge_filter[1] = FUNC(sao_edge_filter_1, depth); \
  170. hevcdsp->sao_edge_filter[2] = FUNC(sao_edge_filter_2, depth); \
  171. hevcdsp->sao_edge_filter[3] = FUNC(sao_edge_filter_3, depth); \
  172. \
  173. QPEL_FUNC(0, 4, depth); \
  174. QPEL_FUNC(1, 8, depth); \
  175. QPEL_FUNC(2, 12, depth); \
  176. QPEL_FUNC(3, 16, depth); \
  177. QPEL_FUNC(4, 24, depth); \
  178. QPEL_FUNC(5, 32, depth); \
  179. QPEL_FUNC(6, 48, depth); \
  180. QPEL_FUNC(7, 64, depth); \
  181. \
  182. EPEL_FUNC(0, 2, depth); \
  183. EPEL_FUNC(1, 4, depth); \
  184. EPEL_FUNC(2, 6, depth); \
  185. EPEL_FUNC(3, 8, depth); \
  186. EPEL_FUNC(4, 12, depth); \
  187. EPEL_FUNC(5, 16, depth); \
  188. EPEL_FUNC(6, 24, depth); \
  189. EPEL_FUNC(7, 32, depth); \
  190. \
  191. PRED_FUNC(0, 4, depth); \
  192. PRED_FUNC(1, 8, depth); \
  193. PRED_FUNC(2, 12, depth); \
  194. PRED_FUNC(3, 16, depth); \
  195. PRED_FUNC(4, 24, depth); \
  196. PRED_FUNC(5, 32, depth); \
  197. PRED_FUNC(6, 48, depth); \
  198. PRED_FUNC(7, 64, depth); \
  199. PRED_FUNC_CHROMA(0, 2, depth); \
  200. PRED_FUNC_CHROMA(1, 4, depth); \
  201. PRED_FUNC_CHROMA(2, 6, depth); \
  202. PRED_FUNC_CHROMA(3, 8, depth); \
  203. PRED_FUNC_CHROMA(4, 12, depth); \
  204. PRED_FUNC_CHROMA(5, 16, depth); \
  205. PRED_FUNC_CHROMA(6, 24, depth); \
  206. PRED_FUNC_CHROMA(7, 32, depth); \
  207. \
  208. hevcdsp->hevc_h_loop_filter_luma = FUNC(hevc_h_loop_filter_luma, depth); \
  209. hevcdsp->hevc_v_loop_filter_luma = FUNC(hevc_v_loop_filter_luma, depth); \
  210. hevcdsp->hevc_h_loop_filter_chroma = FUNC(hevc_h_loop_filter_chroma, depth); \
  211. hevcdsp->hevc_v_loop_filter_chroma = FUNC(hevc_v_loop_filter_chroma, depth); \
  212. hevcdsp->hevc_h_loop_filter_luma_c = FUNC(hevc_h_loop_filter_luma, depth); \
  213. hevcdsp->hevc_v_loop_filter_luma_c = FUNC(hevc_v_loop_filter_luma, depth); \
  214. hevcdsp->hevc_h_loop_filter_chroma_c = FUNC(hevc_h_loop_filter_chroma, depth); \
  215. hevcdsp->hevc_v_loop_filter_chroma_c = FUNC(hevc_v_loop_filter_chroma, depth);
  216. switch (bit_depth) {
  217. case 9:
  218. HEVC_DSP(9);
  219. break;
  220. case 10:
  221. HEVC_DSP(10);
  222. break;
  223. default:
  224. HEVC_DSP(8);
  225. break;
  226. }
  227. if (ARCH_X86)
  228. ff_hevc_dsp_init_x86(hevcdsp, bit_depth);
  229. }