falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
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.
65817 Commits
32 Branches
698MB
Tree: dad7f15567
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'dad7f15567'
${ noResults }
FFmpeg/libavcodec/x86/hevcdsp_init.c

620 lines
30KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2013 Seppo Tomperi

  3.  * Copyright (c) 2013 - 2014 Pierre-Edouard Lepere

  4.  *

  5.  *

  6.  * This file is part of FFmpeg.

  7.  *

  8.  * FFmpeg 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.  * FFmpeg 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 FFmpeg; if not, write to the Free Software

  20.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  21.  */

  22. 

  23. #include "config.h"

  24. #include "libavutil/cpu.h"

  25. #include "libavutil/x86/asm.h"

  26. #include "libavutil/x86/cpu.h"

  27. #include "libavcodec/get_bits.h" /* required for hevcdsp.h GetBitContext */

  28. #include "libavcodec/hevcdsp.h"

  29. #include "libavcodec/x86/hevcdsp.h"

  30. 

  31. #define LFC_FUNC(DIR, DEPTH, OPT) \

  32. void ff_hevc_ ## DIR ## _loop_filter_chroma_ ## DEPTH ## _ ## OPT(uint8_t *pix, ptrdiff_t stride, int *tc, uint8_t *no_p, uint8_t *no_q);

  33. 

  34. #define LFL_FUNC(DIR, DEPTH, OPT) \

  35. void ff_hevc_ ## DIR ## _loop_filter_luma_ ## DEPTH ## _ ## OPT(uint8_t *pix, ptrdiff_t stride, int beta, int *tc, uint8_t *no_p, uint8_t *no_q);

  36. 

  37. #define LFC_FUNCS(type, depth, opt) \

  38.     LFC_FUNC(h, depth, opt)  \

  39.     LFC_FUNC(v, depth, opt)

  40. 

  41. #define LFL_FUNCS(type, depth, opt) \

  42.     LFL_FUNC(h, depth, opt)  \

  43.     LFL_FUNC(v, depth, opt)

  44. 

  45. LFC_FUNCS(uint8_t,   8, sse2)

  46. LFC_FUNCS(uint8_t,  10, sse2)

  47. LFC_FUNCS(uint8_t,  12, sse2)

  48. LFC_FUNCS(uint8_t,   8, avx)

  49. LFC_FUNCS(uint8_t,  10, avx)

  50. LFC_FUNCS(uint8_t,  12, avx)

  51. LFL_FUNCS(uint8_t,   8, sse2)

  52. LFL_FUNCS(uint8_t,  10, sse2)

  53. LFL_FUNCS(uint8_t,  12, sse2)

  54. LFL_FUNCS(uint8_t,   8, ssse3)

  55. LFL_FUNCS(uint8_t,  10, ssse3)

  56. LFL_FUNCS(uint8_t,  12, ssse3)

  57. LFL_FUNCS(uint8_t,   8, avx)

  58. LFL_FUNCS(uint8_t,  10, avx)

  59. LFL_FUNCS(uint8_t,  12, avx)

  60. 

  61. #define IDCT_FUNCS(W, opt) \

  62. void ff_hevc_idct##W##_dc_8_##opt(int16_t *coeffs); \

  63. void ff_hevc_idct##W##_dc_10_##opt(int16_t *coeffs); \

  64. void ff_hevc_idct##W##_dc_12_##opt(int16_t *coeffs)

  65. 

  66. IDCT_FUNCS(4x4,   mmxext);

  67. IDCT_FUNCS(8x8,   mmxext);

  68. IDCT_FUNCS(8x8,   sse2);

  69. IDCT_FUNCS(16x16, sse2);

  70. IDCT_FUNCS(32x32, sse2);

  71. IDCT_FUNCS(16x16, avx2);

  72. IDCT_FUNCS(32x32, avx2);

  73. 

  74. #define mc_rep_func(name, bitd, step, W, opt) \

  75. void ff_hevc_put_hevc_##name##W##_##bitd##_##opt(int16_t *_dst,                                                 \

  76.                                                 uint8_t *_src, ptrdiff_t _srcstride, int height,                \

  77.                                                 intptr_t mx, intptr_t my, int width)                            \

  78. {                                                                                                               \

  79.     int i;                                                                                                      \

  80.     uint8_t *src;                                                                                               \

  81.     int16_t *dst;                                                                                               \

  82.     for (i = 0; i < W; i += step) {                                                                             \

  83.         src  = _src + (i * ((bitd + 7) / 8));                                                                   \

  84.         dst = _dst + i;                                                                                         \

  85.         ff_hevc_put_hevc_##name##step##_##bitd##_##opt(dst, src, _srcstride, height, mx, my, width);            \

  86.     }                                                                                                           \

  87. }

  88. #define mc_rep_uni_func(name, bitd, step, W, opt) \

  89. void ff_hevc_put_hevc_uni_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride,                        \

  90.                                                     uint8_t *_src, ptrdiff_t _srcstride, int height,            \

  91.                                                     intptr_t mx, intptr_t my, int width)                        \

  92. {                                                                                                               \

  93.     int i;                                                                                                      \

  94.     uint8_t *src;                                                                                               \

  95.     uint8_t *dst;                                                                                               \

  96.     for (i = 0; i < W; i += step) {                                                                             \

  97.         src = _src + (i * ((bitd + 7) / 8));                                                                    \

  98.         dst = _dst + (i * ((bitd + 7) / 8));                                                                    \

  99.         ff_hevc_put_hevc_uni_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride,                     \

  100.                                                           height, mx, my, width);                               \

  101.     }                                                                                                           \

  102. }

  103. #define mc_rep_bi_func(name, bitd, step, W, opt) \

  104. void ff_hevc_put_hevc_bi_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, uint8_t *_src,          \

  105.                                                    ptrdiff_t _srcstride, int16_t* _src2,                        \

  106.                                                    int height, intptr_t mx, intptr_t my, int width)             \

  107. {                                                                                                               \

  108.     int i;                                                                                                      \

  109.     uint8_t  *src;                                                                                              \

  110.     uint8_t  *dst;                                                                                              \

  111.     int16_t  *src2;                                                                                             \

  112.     for (i = 0; i < W ; i += step) {                                                                            \

  113.         src  = _src + (i * ((bitd + 7) / 8));                                                                   \

  114.         dst  = _dst + (i * ((bitd + 7) / 8));                                                                   \

  115.         src2 = _src2 + i;                                                                                       \

  116.         ff_hevc_put_hevc_bi_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride, src2,                \

  117.                                                           height, mx, my, width);                               \

  118.     }                                                                                                           \

  119. }

  120. 

  121. #define mc_rep_funcs(name, bitd, step, W, opt)        \

  122.     mc_rep_func(name, bitd, step, W, opt);            \

  123.     mc_rep_uni_func(name, bitd, step, W, opt);        \

  124.     mc_rep_bi_func(name, bitd, step, W, opt)

  125. 

  126. 

  127. #if ARCH_X86_64 && HAVE_SSE4_EXTERNAL

  128. 

  129. mc_rep_funcs(pel_pixels, 8, 16, 64, sse4);

  130. mc_rep_funcs(pel_pixels, 8, 16, 48, sse4);

  131. mc_rep_funcs(pel_pixels, 8, 16, 32, sse4);

  132. mc_rep_funcs(pel_pixels, 8,  8, 24, sse4);

  133. mc_rep_funcs(pel_pixels,10,  8, 64, sse4);

  134. mc_rep_funcs(pel_pixels,10,  8, 48, sse4);

  135. mc_rep_funcs(pel_pixels,10,  8, 32, sse4);

  136. mc_rep_funcs(pel_pixels,10,  8, 24, sse4);

  137. mc_rep_funcs(pel_pixels,10,  8, 16, sse4);

  138. mc_rep_funcs(pel_pixels,10,  4, 12, sse4);

  139. mc_rep_funcs(pel_pixels,12,  8, 64, sse4);

  140. mc_rep_funcs(pel_pixels,12,  8, 48, sse4);

  141. mc_rep_funcs(pel_pixels,12,  8, 32, sse4);

  142. mc_rep_funcs(pel_pixels,12,  8, 24, sse4);

  143. mc_rep_funcs(pel_pixels,12,  8, 16, sse4);

  144. mc_rep_funcs(pel_pixels,12,  4, 12, sse4);

  145. 

  146. mc_rep_funcs(epel_h, 8, 16, 64, sse4);

  147. mc_rep_funcs(epel_h, 8, 16, 48, sse4);

  148. mc_rep_funcs(epel_h, 8, 16, 32, sse4);

  149. mc_rep_funcs(epel_h, 8,  8, 24, sse4);

  150. mc_rep_funcs(epel_h,10,  8, 64, sse4);

  151. mc_rep_funcs(epel_h,10,  8, 48, sse4);

  152. mc_rep_funcs(epel_h,10,  8, 32, sse4);

  153. mc_rep_funcs(epel_h,10,  8, 24, sse4);

  154. mc_rep_funcs(epel_h,10,  8, 16, sse4);

  155. mc_rep_funcs(epel_h,10,  4, 12, sse4);

  156. mc_rep_funcs(epel_h,12,  8, 64, sse4);

  157. mc_rep_funcs(epel_h,12,  8, 48, sse4);

  158. mc_rep_funcs(epel_h,12,  8, 32, sse4);

  159. mc_rep_funcs(epel_h,12,  8, 24, sse4);

  160. mc_rep_funcs(epel_h,12,  8, 16, sse4);

  161. mc_rep_funcs(epel_h,12,  4, 12, sse4);

  162. mc_rep_funcs(epel_v, 8, 16, 64, sse4);

  163. mc_rep_funcs(epel_v, 8, 16, 48, sse4);

  164. mc_rep_funcs(epel_v, 8, 16, 32, sse4);

  165. mc_rep_funcs(epel_v, 8,  8, 24, sse4);

  166. mc_rep_funcs(epel_v,10,  8, 64, sse4);

  167. mc_rep_funcs(epel_v,10,  8, 48, sse4);

  168. mc_rep_funcs(epel_v,10,  8, 32, sse4);

  169. mc_rep_funcs(epel_v,10,  8, 24, sse4);

  170. mc_rep_funcs(epel_v,10,  8, 16, sse4);

  171. mc_rep_funcs(epel_v,10,  4, 12, sse4);

  172. mc_rep_funcs(epel_v,12,  8, 64, sse4);

  173. mc_rep_funcs(epel_v,12,  8, 48, sse4);

  174. mc_rep_funcs(epel_v,12,  8, 32, sse4);

  175. mc_rep_funcs(epel_v,12,  8, 24, sse4);

  176. mc_rep_funcs(epel_v,12,  8, 16, sse4);

  177. mc_rep_funcs(epel_v,12,  4, 12, sse4);

  178. mc_rep_funcs(epel_hv, 8,  8, 64, sse4);

  179. mc_rep_funcs(epel_hv, 8,  8, 48, sse4);

  180. mc_rep_funcs(epel_hv, 8,  8, 32, sse4);

  181. mc_rep_funcs(epel_hv, 8,  8, 24, sse4);

  182. mc_rep_funcs(epel_hv, 8,  8, 16, sse4);

  183. mc_rep_funcs(epel_hv, 8,  4, 12, sse4);

  184. mc_rep_funcs(epel_hv,10,  8, 64, sse4);

  185. mc_rep_funcs(epel_hv,10,  8, 48, sse4);

  186. mc_rep_funcs(epel_hv,10,  8, 32, sse4);

  187. mc_rep_funcs(epel_hv,10,  8, 24, sse4);

  188. mc_rep_funcs(epel_hv,10,  8, 16, sse4);

  189. mc_rep_funcs(epel_hv,10,  4, 12, sse4);

  190. mc_rep_funcs(epel_hv,12,  8, 64, sse4);

  191. mc_rep_funcs(epel_hv,12,  8, 48, sse4);

  192. mc_rep_funcs(epel_hv,12,  8, 32, sse4);

  193. mc_rep_funcs(epel_hv,12,  8, 24, sse4);

  194. mc_rep_funcs(epel_hv,12,  8, 16, sse4);

  195. mc_rep_funcs(epel_hv,12,  4, 12, sse4);

  196. 

  197. mc_rep_funcs(qpel_h, 8, 16, 64, sse4);

  198. mc_rep_funcs(qpel_h, 8, 16, 48, sse4);

  199. mc_rep_funcs(qpel_h, 8, 16, 32, sse4);

  200. mc_rep_funcs(qpel_h, 8,  8, 24, sse4);

  201. mc_rep_funcs(qpel_h,10,  8, 64, sse4);

  202. mc_rep_funcs(qpel_h,10,  8, 48, sse4);

  203. mc_rep_funcs(qpel_h,10,  8, 32, sse4);

  204. mc_rep_funcs(qpel_h,10,  8, 24, sse4);

  205. mc_rep_funcs(qpel_h,10,  8, 16, sse4);

  206. mc_rep_funcs(qpel_h,10,  4, 12, sse4);

  207. mc_rep_funcs(qpel_h,12,  8, 64, sse4);

  208. mc_rep_funcs(qpel_h,12,  8, 48, sse4);

  209. mc_rep_funcs(qpel_h,12,  8, 32, sse4);

  210. mc_rep_funcs(qpel_h,12,  8, 24, sse4);

  211. mc_rep_funcs(qpel_h,12,  8, 16, sse4);

  212. mc_rep_funcs(qpel_h,12,  4, 12, sse4);

  213. mc_rep_funcs(qpel_v, 8, 16, 64, sse4);

  214. mc_rep_funcs(qpel_v, 8, 16, 48, sse4);

  215. mc_rep_funcs(qpel_v, 8, 16, 32, sse4);

  216. mc_rep_funcs(qpel_v, 8,  8, 24, sse4);

  217. mc_rep_funcs(qpel_v,10,  8, 64, sse4);

  218. mc_rep_funcs(qpel_v,10,  8, 48, sse4);

  219. mc_rep_funcs(qpel_v,10,  8, 32, sse4);

  220. mc_rep_funcs(qpel_v,10,  8, 24, sse4);

  221. mc_rep_funcs(qpel_v,10,  8, 16, sse4);

  222. mc_rep_funcs(qpel_v,10,  4, 12, sse4);

  223. mc_rep_funcs(qpel_v,12,  8, 64, sse4);

  224. mc_rep_funcs(qpel_v,12,  8, 48, sse4);

  225. mc_rep_funcs(qpel_v,12,  8, 32, sse4);

  226. mc_rep_funcs(qpel_v,12,  8, 24, sse4);

  227. mc_rep_funcs(qpel_v,12,  8, 16, sse4);

  228. mc_rep_funcs(qpel_v,12,  4, 12, sse4);

  229. mc_rep_funcs(qpel_hv, 8,  8, 64, sse4);

  230. mc_rep_funcs(qpel_hv, 8,  8, 48, sse4);

  231. mc_rep_funcs(qpel_hv, 8,  8, 32, sse4);

  232. mc_rep_funcs(qpel_hv, 8,  8, 24, sse4);

  233. mc_rep_funcs(qpel_hv, 8,  8, 16, sse4);

  234. mc_rep_funcs(qpel_hv, 8,  4, 12, sse4);

  235. mc_rep_funcs(qpel_hv,10,  8, 64, sse4);

  236. mc_rep_funcs(qpel_hv,10,  8, 48, sse4);

  237. mc_rep_funcs(qpel_hv,10,  8, 32, sse4);

  238. mc_rep_funcs(qpel_hv,10,  8, 24, sse4);

  239. mc_rep_funcs(qpel_hv,10,  8, 16, sse4);

  240. mc_rep_funcs(qpel_hv,10,  4, 12, sse4);

  241. mc_rep_funcs(qpel_hv,12,  8, 64, sse4);

  242. mc_rep_funcs(qpel_hv,12,  8, 48, sse4);

  243. mc_rep_funcs(qpel_hv,12,  8, 32, sse4);

  244. mc_rep_funcs(qpel_hv,12,  8, 24, sse4);

  245. mc_rep_funcs(qpel_hv,12,  8, 16, sse4);

  246. mc_rep_funcs(qpel_hv,12,  4, 12, sse4);

  247. 

  248. #define mc_rep_uni_w(bitd, step, W, opt) \

  249. void ff_hevc_put_hevc_uni_w##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, int16_t *_src, ptrdiff_t _srcstride,\

  250.                                                int height, int denom,  int _wx, int _ox)                                \

  251. {                                                                                                                       \

  252.     int i;                                                                                                              \

  253.     int16_t *src;                                                                                                       \

  254.     uint8_t *dst;                                                                                                       \

  255.     for (i = 0; i < W; i += step) {                                                                                     \

  256.         src= _src + i;                                                                                                  \

  257.         dst= _dst + (i * ((bitd + 7) / 8));                                                                             \

  258.         ff_hevc_put_hevc_uni_w##step##_##bitd##_##opt(dst, dststride, src, _srcstride,                                  \

  259.                                                      height, denom, _wx, _ox);                                          \

  260.     }                                                                                                                   \

  261. }

  262. 

  263. mc_rep_uni_w(8, 6, 12, sse4);

  264. mc_rep_uni_w(8, 8, 16, sse4);

  265. mc_rep_uni_w(8, 8, 24, sse4);

  266. mc_rep_uni_w(8, 8, 32, sse4);

  267. mc_rep_uni_w(8, 8, 48, sse4);

  268. mc_rep_uni_w(8, 8, 64, sse4);

  269. 

  270. mc_rep_uni_w(10, 6, 12, sse4);

  271. mc_rep_uni_w(10, 8, 16, sse4);

  272. mc_rep_uni_w(10, 8, 24, sse4);

  273. mc_rep_uni_w(10, 8, 32, sse4);

  274. mc_rep_uni_w(10, 8, 48, sse4);

  275. mc_rep_uni_w(10, 8, 64, sse4);

  276. 

  277. mc_rep_uni_w(12, 6, 12, sse4);

  278. mc_rep_uni_w(12, 8, 16, sse4);

  279. mc_rep_uni_w(12, 8, 24, sse4);

  280. mc_rep_uni_w(12, 8, 32, sse4);

  281. mc_rep_uni_w(12, 8, 48, sse4);

  282. mc_rep_uni_w(12, 8, 64, sse4);

  283. 

  284. #define mc_rep_bi_w(bitd, step, W, opt) \

  285. void ff_hevc_put_hevc_bi_w##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, int16_t *_src, ptrdiff_t _srcstride, \

  286.                                               int16_t *_src2, int height,                                               \

  287.                                               int denom,  int _wx0,  int _wx1, int _ox0, int _ox1)                      \

  288. {                                                                                                                       \

  289.     int i;                                                                                                              \

  290.     int16_t *src;                                                                                                       \

  291.     int16_t *src2;                                                                                                      \

  292.     uint8_t *dst;                                                                                                       \

  293.     for (i = 0; i < W; i += step) {                                                                                     \

  294.         src  = _src  + i;                                                                                               \

  295.         src2 = _src2 + i;                                                                                               \

  296.         dst  = _dst  + (i * ((bitd + 7) / 8));                                                                          \

  297.         ff_hevc_put_hevc_bi_w##step##_##bitd##_##opt(dst, dststride, src, _srcstride, src2,                             \

  298.                                                     height, denom, _wx0, _wx1, _ox0, _ox1);                             \

  299.     }                                                                                                                   \

  300. }

  301. 

  302. mc_rep_bi_w(8, 6, 12, sse4);

  303. mc_rep_bi_w(8, 8, 16, sse4);

  304. mc_rep_bi_w(8, 8, 24, sse4);

  305. mc_rep_bi_w(8, 8, 32, sse4);

  306. mc_rep_bi_w(8, 8, 48, sse4);

  307. mc_rep_bi_w(8, 8, 64, sse4);

  308. 

  309. mc_rep_bi_w(10, 6, 12, sse4);

  310. mc_rep_bi_w(10, 8, 16, sse4);

  311. mc_rep_bi_w(10, 8, 24, sse4);

  312. mc_rep_bi_w(10, 8, 32, sse4);

  313. mc_rep_bi_w(10, 8, 48, sse4);

  314. mc_rep_bi_w(10, 8, 64, sse4);

  315. 

  316. mc_rep_bi_w(12, 6, 12, sse4);

  317. mc_rep_bi_w(12, 8, 16, sse4);

  318. mc_rep_bi_w(12, 8, 24, sse4);

  319. mc_rep_bi_w(12, 8, 32, sse4);

  320. mc_rep_bi_w(12, 8, 48, sse4);

  321. mc_rep_bi_w(12, 8, 64, sse4);

  322. 

  323. #define mc_uni_w_func(name, bitd, W, opt) \

  324. void ff_hevc_put_hevc_uni_w_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t _dststride,         \

  325.                                                       uint8_t *_src, ptrdiff_t _srcstride,          \

  326.                                                       int height, int denom,                        \

  327.                                                       int _wx, int _ox,                             \

  328.                                                       intptr_t mx, intptr_t my, int width)          \

  329. {                                                                                                   \

  330.     LOCAL_ALIGNED_16(int16_t, temp, [71 * MAX_PB_SIZE]);                                            \

  331.     ff_hevc_put_hevc_##name##W##_##bitd##_##opt(temp, _src, _srcstride, height, mx, my, width);     \

  332.     ff_hevc_put_hevc_uni_w##W##_##bitd##_##opt(_dst, _dststride, temp, MAX_PB_SIZE, height, denom, _wx, _ox);\

  333. }

  334. 

  335. #define mc_uni_w_funcs(name, bitd, opt)       \

  336.         mc_uni_w_func(name, bitd, 4, opt);    \

  337.         mc_uni_w_func(name, bitd, 8, opt);    \

  338.         mc_uni_w_func(name, bitd, 12, opt);   \

  339.         mc_uni_w_func(name, bitd, 16, opt);   \

  340.         mc_uni_w_func(name, bitd, 24, opt);   \

  341.         mc_uni_w_func(name, bitd, 32, opt);   \

  342.         mc_uni_w_func(name, bitd, 48, opt);   \

  343.         mc_uni_w_func(name, bitd, 64, opt)

  344. 

  345. mc_uni_w_funcs(pel_pixels, 8, sse4);

  346. mc_uni_w_func(pel_pixels, 8, 6, sse4);

  347. mc_uni_w_funcs(epel_h, 8, sse4);

  348. mc_uni_w_func(epel_h, 8, 6, sse4);

  349. mc_uni_w_funcs(epel_v, 8, sse4);

  350. mc_uni_w_func(epel_v, 8, 6, sse4);

  351. mc_uni_w_funcs(epel_hv, 8, sse4);

  352. mc_uni_w_func(epel_hv, 8, 6, sse4);

  353. mc_uni_w_funcs(qpel_h, 8, sse4);

  354. mc_uni_w_funcs(qpel_v, 8, sse4);

  355. mc_uni_w_funcs(qpel_hv, 8, sse4);

  356. 

  357. mc_uni_w_funcs(pel_pixels, 10, sse4);

  358. mc_uni_w_func(pel_pixels, 10, 6, sse4);

  359. mc_uni_w_funcs(epel_h, 10, sse4);

  360. mc_uni_w_func(epel_h, 10, 6, sse4);

  361. mc_uni_w_funcs(epel_v, 10, sse4);

  362. mc_uni_w_func(epel_v, 10, 6, sse4);

  363. mc_uni_w_funcs(epel_hv, 10, sse4);

  364. mc_uni_w_func(epel_hv, 10, 6, sse4);

  365. mc_uni_w_funcs(qpel_h, 10, sse4);

  366. mc_uni_w_funcs(qpel_v, 10, sse4);

  367. mc_uni_w_funcs(qpel_hv, 10, sse4);

  368. 

  369. mc_uni_w_funcs(pel_pixels, 12, sse4);

  370. mc_uni_w_func(pel_pixels, 12, 6, sse4);

  371. mc_uni_w_funcs(epel_h, 12, sse4);

  372. mc_uni_w_func(epel_h, 12, 6, sse4);

  373. mc_uni_w_funcs(epel_v, 12, sse4);

  374. mc_uni_w_func(epel_v, 12, 6, sse4);

  375. mc_uni_w_funcs(epel_hv, 12, sse4);

  376. mc_uni_w_func(epel_hv, 12, 6, sse4);

  377. mc_uni_w_funcs(qpel_h, 12, sse4);

  378. mc_uni_w_funcs(qpel_v, 12, sse4);

  379. mc_uni_w_funcs(qpel_hv, 12, sse4);

  380. 

  381. #define mc_bi_w_func(name, bitd, W, opt) \

  382. void ff_hevc_put_hevc_bi_w_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t _dststride,           \

  383.                                                      uint8_t *_src, ptrdiff_t _srcstride,            \

  384.                                                      int16_t *_src2,                                 \

  385.                                                      int height, int denom,                          \

  386.                                                      int _wx0, int _wx1, int _ox0, int _ox1,         \

  387.                                                      intptr_t mx, intptr_t my, int width)            \

  388. {                                                                                                    \

  389.     LOCAL_ALIGNED_16(int16_t, temp, [71 * MAX_PB_SIZE]);                                             \

  390.     ff_hevc_put_hevc_##name##W##_##bitd##_##opt(temp, _src, _srcstride, height, mx, my, width);      \

  391.     ff_hevc_put_hevc_bi_w##W##_##bitd##_##opt(_dst, _dststride, temp, MAX_PB_SIZE, _src2,            \

  392.                                              height, denom, _wx0, _wx1, _ox0, _ox1);                 \

  393. }

  394. 

  395. #define mc_bi_w_funcs(name, bitd, opt)       \

  396.         mc_bi_w_func(name, bitd, 4, opt);    \

  397.         mc_bi_w_func(name, bitd, 8, opt);    \

  398.         mc_bi_w_func(name, bitd, 12, opt);   \

  399.         mc_bi_w_func(name, bitd, 16, opt);   \

  400.         mc_bi_w_func(name, bitd, 24, opt);   \

  401.         mc_bi_w_func(name, bitd, 32, opt);   \

  402.         mc_bi_w_func(name, bitd, 48, opt);   \

  403.         mc_bi_w_func(name, bitd, 64, opt)

  404. 

  405. mc_bi_w_funcs(pel_pixels, 8, sse4);

  406. mc_bi_w_func(pel_pixels, 8, 6, sse4);

  407. mc_bi_w_funcs(epel_h, 8, sse4);

  408. mc_bi_w_func(epel_h, 8, 6, sse4);

  409. mc_bi_w_funcs(epel_v, 8, sse4);

  410. mc_bi_w_func(epel_v, 8, 6, sse4);

  411. mc_bi_w_funcs(epel_hv, 8, sse4);

  412. mc_bi_w_func(epel_hv, 8, 6, sse4);

  413. mc_bi_w_funcs(qpel_h, 8, sse4);

  414. mc_bi_w_funcs(qpel_v, 8, sse4);

  415. mc_bi_w_funcs(qpel_hv, 8, sse4);

  416. 

  417. mc_bi_w_funcs(pel_pixels, 10, sse4);

  418. mc_bi_w_func(pel_pixels, 10, 6, sse4);

  419. mc_bi_w_funcs(epel_h, 10, sse4);

  420. mc_bi_w_func(epel_h, 10, 6, sse4);

  421. mc_bi_w_funcs(epel_v, 10, sse4);

  422. mc_bi_w_func(epel_v, 10, 6, sse4);

  423. mc_bi_w_funcs(epel_hv, 10, sse4);

  424. mc_bi_w_func(epel_hv, 10, 6, sse4);

  425. mc_bi_w_funcs(qpel_h, 10, sse4);

  426. mc_bi_w_funcs(qpel_v, 10, sse4);

  427. mc_bi_w_funcs(qpel_hv, 10, sse4);

  428. 

  429. mc_bi_w_funcs(pel_pixels, 12, sse4);

  430. mc_bi_w_func(pel_pixels, 12, 6, sse4);

  431. mc_bi_w_funcs(epel_h, 12, sse4);

  432. mc_bi_w_func(epel_h, 12, 6, sse4);

  433. mc_bi_w_funcs(epel_v, 12, sse4);

  434. mc_bi_w_func(epel_v, 12, 6, sse4);

  435. mc_bi_w_funcs(epel_hv, 12, sse4);

  436. mc_bi_w_func(epel_hv, 12, 6, sse4);

  437. mc_bi_w_funcs(qpel_h, 12, sse4);

  438. mc_bi_w_funcs(qpel_v, 12, sse4);

  439. mc_bi_w_funcs(qpel_hv, 12, sse4);

  440. #endif //ARCH_X86_64 && HAVE_SSE4_EXTERNAL

  441. 

  442. 

  443. #define EPEL_LINKS(pointer, my, mx, fname, bitd, opt )           \

  444.         PEL_LINK(pointer, 1, my , mx , fname##4 ,  bitd, opt ); \

  445.         PEL_LINK(pointer, 2, my , mx , fname##6 ,  bitd, opt ); \

  446.         PEL_LINK(pointer, 3, my , mx , fname##8 ,  bitd, opt ); \

  447.         PEL_LINK(pointer, 4, my , mx , fname##12,  bitd, opt ); \

  448.         PEL_LINK(pointer, 5, my , mx , fname##16,  bitd, opt ); \

  449.         PEL_LINK(pointer, 6, my , mx , fname##24,  bitd, opt ); \

  450.         PEL_LINK(pointer, 7, my , mx , fname##32,  bitd, opt ); \

  451.         PEL_LINK(pointer, 8, my , mx , fname##48,  bitd, opt ); \

  452.         PEL_LINK(pointer, 9, my , mx , fname##64,  bitd, opt )

  453. #define QPEL_LINKS(pointer, my, mx, fname, bitd, opt)           \

  454.         PEL_LINK(pointer, 1, my , mx , fname##4 ,  bitd, opt ); \

  455.         PEL_LINK(pointer, 3, my , mx , fname##8 ,  bitd, opt ); \

  456.         PEL_LINK(pointer, 4, my , mx , fname##12,  bitd, opt ); \

  457.         PEL_LINK(pointer, 5, my , mx , fname##16,  bitd, opt ); \

  458.         PEL_LINK(pointer, 6, my , mx , fname##24,  bitd, opt ); \

  459.         PEL_LINK(pointer, 7, my , mx , fname##32,  bitd, opt ); \

  460.         PEL_LINK(pointer, 8, my , mx , fname##48,  bitd, opt ); \

  461.         PEL_LINK(pointer, 9, my , mx , fname##64,  bitd, opt )

  462. 

  463. 

  464. void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)

  465. {

  466.     int cpu_flags = av_get_cpu_flags();

  467. 

  468.     if (bit_depth == 8) {

  469.         if (EXTERNAL_MMXEXT(cpu_flags)) {

  470.             c->idct_dc[0] = ff_hevc_idct4x4_dc_8_mmxext;

  471.             c->idct_dc[1] = ff_hevc_idct8x8_dc_8_mmxext;

  472.             c->transform_add[0]    =  ff_hevc_transform_add4_8_mmxext;

  473.         }

  474.         if (EXTERNAL_SSE2(cpu_flags)) {

  475.             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2;

  476.             c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_8_sse2;

  477.             if (ARCH_X86_64) {

  478.                 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_sse2;

  479.                 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_sse2;

  480.             }

  481.             c->idct_dc[1] = ff_hevc_idct8x8_dc_8_sse2;

  482.             c->idct_dc[2] = ff_hevc_idct16x16_dc_8_sse2;

  483.             c->idct_dc[3] = ff_hevc_idct32x32_dc_8_sse2;

  484. 

  485.             c->transform_add[1]    = ff_hevc_transform_add8_8_sse2;

  486.             c->transform_add[2]    = ff_hevc_transform_add16_8_sse2;

  487.             c->transform_add[3]    = ff_hevc_transform_add32_8_sse2;

  488.         }

  489.         if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) {

  490.             c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_ssse3;

  491.             c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_ssse3;

  492.         }

  493.         if (EXTERNAL_SSE4(cpu_flags) && ARCH_X86_64) {

  494. 

  495.             EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels,  8, sse4);

  496.             EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h,      8, sse4);

  497.             EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v,      8, sse4);

  498.             EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv,     8, sse4);

  499. 

  500.             QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 8, sse4);

  501.             QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h,     8, sse4);

  502.             QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v,     8, sse4);

  503.             QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv,    8, sse4);

  504.         }

  505.         if (EXTERNAL_AVX(cpu_flags)) {

  506.             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_avx;

  507.             c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_8_avx;

  508.             if (ARCH_X86_64) {

  509.                 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_avx;

  510.                 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_avx;

  511.             }

  512.             c->transform_add[1]    = ff_hevc_transform_add8_8_avx;

  513.             c->transform_add[2]    = ff_hevc_transform_add16_8_avx;

  514.             c->transform_add[3]    = ff_hevc_transform_add32_8_avx;

  515.         }

  516.         if (EXTERNAL_AVX2(cpu_flags)) {

  517.             c->idct_dc[2] = ff_hevc_idct16x16_dc_8_avx2;

  518.             c->idct_dc[3] = ff_hevc_idct32x32_dc_8_avx2;

  519.         }

  520.     } else if (bit_depth == 10) {

  521.         if (EXTERNAL_MMXEXT(cpu_flags)) {

  522.             c->transform_add[0] = ff_hevc_transform_add4_10_mmxext;

  523.             c->idct_dc[0] = ff_hevc_idct4x4_dc_10_mmxext;

  524.             c->idct_dc[1] = ff_hevc_idct8x8_dc_10_mmxext;

  525.         }

  526.         if (EXTERNAL_SSE2(cpu_flags)) {

  527.             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_sse2;

  528.             c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_10_sse2;

  529.             if (ARCH_X86_64) {

  530.                 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_sse2;

  531.                 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_sse2;

  532.             }

  533. 

  534.             c->idct_dc[1] = ff_hevc_idct8x8_dc_10_sse2;

  535.             c->idct_dc[2] = ff_hevc_idct16x16_dc_10_sse2;

  536.             c->idct_dc[3] = ff_hevc_idct32x32_dc_10_sse2;

  537. 

  538.             c->transform_add[1]    = ff_hevc_transform_add8_10_sse2;

  539.             c->transform_add[2]    = ff_hevc_transform_add16_10_sse2;

  540.             c->transform_add[3]    = ff_hevc_transform_add32_10_sse2;

  541.         }

  542.         if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) {

  543.             c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3;

  544.             c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_ssse3;

  545.         }

  546.         if (EXTERNAL_SSE4(cpu_flags) && ARCH_X86_64) {

  547.             EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 10, sse4);

  548.             EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h,     10, sse4);

  549.             EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v,     10, sse4);

  550.             EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv,    10, sse4);

  551. 

  552.             QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 10, sse4);

  553.             QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h,     10, sse4);

  554.             QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v,     10, sse4);

  555.             QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv,    10, sse4);

  556.         }

  557.         if (EXTERNAL_AVX(cpu_flags)) {

  558.             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_avx;

  559.             c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_10_avx;

  560.             if (ARCH_X86_64) {

  561.                 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_avx;

  562.                 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_avx;

  563.             }

  564.         }

  565.         if (EXTERNAL_AVX2(cpu_flags)) {

  566. 

  567.             c->idct_dc[2] = ff_hevc_idct16x16_dc_10_avx2;

  568.             c->idct_dc[3] = ff_hevc_idct32x32_dc_10_avx2;

  569. 

  570.             c->transform_add[2] = ff_hevc_transform_add16_10_avx2;

  571.             c->transform_add[3] = ff_hevc_transform_add32_10_avx2;

  572. 

  573.         }

  574.     } else if (bit_depth == 12) {

  575.         if (EXTERNAL_MMXEXT(cpu_flags)) {

  576.             c->idct_dc[0] = ff_hevc_idct4x4_dc_12_mmxext;

  577.             c->idct_dc[1] = ff_hevc_idct8x8_dc_12_mmxext;

  578.         }

  579.         if (EXTERNAL_SSE2(cpu_flags)) {

  580.             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_12_sse2;

  581.             c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_12_sse2;

  582.             if (ARCH_X86_64) {

  583.                 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_sse2;

  584.                 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_sse2;

  585.             }

  586. 

  587.             c->idct_dc[1] = ff_hevc_idct8x8_dc_12_sse2;

  588.             c->idct_dc[2] = ff_hevc_idct16x16_dc_12_sse2;

  589.             c->idct_dc[3] = ff_hevc_idct32x32_dc_12_sse2;

  590.         }

  591.         if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) {

  592.             c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_ssse3;

  593.             c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_ssse3;

  594.         }

  595.         if (EXTERNAL_SSE4(cpu_flags) && ARCH_X86_64) {

  596.             EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 12, sse4);

  597.             EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h,     12, sse4);

  598.             EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v,     12, sse4);

  599.             EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv,    12, sse4);

  600. 

  601.             QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 12, sse4);

  602.             QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h,     12, sse4);

  603.             QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v,     12, sse4);

  604.             QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv,    12, sse4);

  605.         }

  606.         if (EXTERNAL_AVX(cpu_flags)) {

  607.             c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_12_avx;

  608.             c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_12_avx;

  609.             if (ARCH_X86_64) {

  610.                 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_avx;

  611.                 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_avx;

  612.             }

  613.         }

  614.         if (EXTERNAL_AVX2(cpu_flags)) {

  615.             c->idct_dc[2] = ff_hevc_idct16x16_dc_12_avx2;

  616.             c->idct_dc[3] = ff_hevc_idct32x32_dc_12_avx2;

  617.         }

  618.     }

  619. }