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.
28806 Commits
32 Branches
283MB
Tree: 836f47d34b
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '836f47d34b'
${ noResults }
FFmpeg/libavcodec/x86/h264dsp_mmx.c

505 lines
22KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt

  3.  *

  4.  * This file is part of Libav.

  5.  *

  6.  * Libav is free software; you can redistribute it and/or

  7.  * modify it under the terms of the GNU Lesser General Public

  8.  * License as published by the Free Software Foundation; either

  9.  * version 2.1 of the License, or (at your option) any later version.

  10.  *

  11.  * Libav 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 GNU

  14.  * Lesser General Public License for more details.

  15.  *

  16.  * You should have received a copy of the GNU Lesser General Public

  17.  * License along with Libav; if not, write to the Free Software

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

  19.  */

  20. 

  21. #include "libavutil/cpu.h"

  22. #include "libavutil/x86_cpu.h"

  23. #include "libavcodec/h264dsp.h"

  24. #include "dsputil_mmx.h"

  25. 

  26. DECLARE_ALIGNED(8, static const uint64_t, ff_pb_3_1  ) = 0x0103010301030103ULL;

  27. 

  28. /***********************************/

  29. /* IDCT */

  30. #define IDCT_ADD_FUNC(NUM, DEPTH, OPT) \

  31. void ff_h264_idct ## NUM ## _add_ ## DEPTH ## _ ## OPT (uint8_t *dst, int16_t *block, int stride);

  32. 

  33. IDCT_ADD_FUNC(, 10, sse2)

  34. IDCT_ADD_FUNC(_dc, 10, mmx2)

  35. IDCT_ADD_FUNC(8_dc, 10, sse2)

  36. IDCT_ADD_FUNC(8, 10, sse2)

  37. #if HAVE_AVX

  38. IDCT_ADD_FUNC(, 10, avx)

  39. IDCT_ADD_FUNC(8_dc, 10, avx)

  40. IDCT_ADD_FUNC(8, 10, avx)

  41. #endif

  42. 

  43. 

  44. #define IDCT_ADD_REP_FUNC(NUM, REP, DEPTH, OPT) \

  45. void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \

  46.                               (uint8_t *dst, const int *block_offset, \

  47.                               DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  48. 

  49. IDCT_ADD_REP_FUNC(8, 4, 10, sse2)

  50. IDCT_ADD_REP_FUNC(8, 4, 10, avx)

  51. IDCT_ADD_REP_FUNC(, 16, 10, sse2)

  52. IDCT_ADD_REP_FUNC(, 16intra, 10, sse2)

  53. #if HAVE_AVX

  54. IDCT_ADD_REP_FUNC(, 16, 10, avx)

  55. IDCT_ADD_REP_FUNC(, 16intra, 10, avx)

  56. #endif

  57. 

  58. 

  59. #define IDCT_ADD_REP_FUNC2(NUM, REP, DEPTH, OPT) \

  60. void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \

  61.                               (uint8_t **dst, const int *block_offset, \

  62.                               DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  63. IDCT_ADD_REP_FUNC2(, 8, 10, sse2)

  64. #if HAVE_AVX

  65. IDCT_ADD_REP_FUNC2(, 8, 10, avx)

  66. #endif

  67. 

  68. void ff_h264_idct_add_mmx     (uint8_t *dst, int16_t *block, int stride);

  69. void ff_h264_idct8_add_mmx    (uint8_t *dst, int16_t *block, int stride);

  70. void ff_h264_idct8_add_sse2   (uint8_t *dst, int16_t *block, int stride);

  71. void ff_h264_idct_dc_add_mmx2 (uint8_t *dst, int16_t *block, int stride);

  72. void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride);

  73. 

  74. void ff_h264_idct_add16_mmx      (uint8_t *dst, const int *block_offset,

  75.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  76. void ff_h264_idct8_add4_mmx      (uint8_t *dst, const int *block_offset,

  77.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  78. void ff_h264_idct_add16_mmx2     (uint8_t *dst, const int *block_offset,

  79.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  80. void ff_h264_idct_add16intra_mmx (uint8_t *dst, const int *block_offset,

  81.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  82. void ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset,

  83.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  84. void ff_h264_idct8_add4_mmx2     (uint8_t *dst, const int *block_offset,

  85.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  86. void ff_h264_idct8_add4_sse2     (uint8_t *dst, const int *block_offset,

  87.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  88. void ff_h264_idct_add8_mmx       (uint8_t **dest, const int *block_offset,

  89.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  90. void ff_h264_idct_add8_mmx2      (uint8_t **dest, const int *block_offset,

  91.                                   DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

  92. 

  93. void ff_h264_idct_add16_sse2     (uint8_t *dst, const int *block_offset, DCTELEM *block,

  94.                                   int stride, const uint8_t nnzc[6*8]);

  95. void ff_h264_idct_add16intra_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block,

  96.                                   int stride, const uint8_t nnzc[6*8]);

  97. void ff_h264_idct_add8_sse2      (uint8_t **dest, const int *block_offset, DCTELEM *block,

  98.                                   int stride, const uint8_t nnzc[6*8]);

  99. void ff_h264_luma_dc_dequant_idct_mmx (DCTELEM *output, DCTELEM *input, int qmul);

  100. void ff_h264_luma_dc_dequant_idct_sse2(DCTELEM *output, DCTELEM *input, int qmul);

  101. 

  102. /***********************************/

  103. /* deblocking */

  104. 

  105. #define h264_loop_filter_strength_iteration_mmx2(bS, nz, ref, mv, bidir, edges, step, mask_mv, dir, d_idx, mask_dir) \

  106.     do { \

  107.         x86_reg b_idx; \

  108.         mask_mv <<= 3; \

  109.         for( b_idx=0; b_idx<edges; b_idx+=step ) { \

  110.             if (!mask_dir) \

  111.             __asm__ volatile( \

  112.                     "pxor %%mm0, %%mm0 \n\t" \

  113.                     :: \

  114.             ); \

  115.             if(!(mask_mv & b_idx)) { \

  116.                 if(bidir) { \

  117.                     __asm__ volatile( \

  118.                         "movd         %a3(%0,%2), %%mm2 \n" \

  119.                         "punpckldq    %a4(%0,%2), %%mm2 \n" /* { ref0[bn], ref1[bn] } */ \

  120.                         "pshufw $0x44, 12(%0,%2), %%mm0 \n" /* { ref0[b], ref0[b] } */ \

  121.                         "pshufw $0x44, 52(%0,%2), %%mm1 \n" /* { ref1[b], ref1[b] } */ \

  122.                         "pshufw $0x4E, %%mm2, %%mm3 \n" \

  123.                         "psubb         %%mm2, %%mm0 \n" /* { ref0[b]!=ref0[bn], ref0[b]!=ref1[bn] } */ \

  124.                         "psubb         %%mm3, %%mm1 \n" /* { ref1[b]!=ref1[bn], ref1[b]!=ref0[bn] } */ \

  125.  \

  126.                         "por           %%mm1, %%mm0 \n" \

  127.                         "movq   %a5(%1,%2,4), %%mm1 \n" \

  128.                         "movq   %a6(%1,%2,4), %%mm2 \n" \

  129.                         "movq          %%mm1, %%mm3 \n" \

  130.                         "movq          %%mm2, %%mm4 \n" \

  131.                         "psubw   48(%1,%2,4), %%mm1 \n" \

  132.                         "psubw   56(%1,%2,4), %%mm2 \n" \

  133.                         "psubw  208(%1,%2,4), %%mm3 \n" \

  134.                         "psubw  216(%1,%2,4), %%mm4 \n" \

  135.                         "packsswb      %%mm2, %%mm1 \n" \

  136.                         "packsswb      %%mm4, %%mm3 \n" \

  137.                         "paddb         %%mm6, %%mm1 \n" \

  138.                         "paddb         %%mm6, %%mm3 \n" \

  139.                         "psubusb       %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \

  140.                         "psubusb       %%mm5, %%mm3 \n" \

  141.                         "packsswb      %%mm3, %%mm1 \n" \

  142.  \

  143.                         "por           %%mm1, %%mm0 \n" \

  144.                         "movq   %a7(%1,%2,4), %%mm1 \n" \

  145.                         "movq   %a8(%1,%2,4), %%mm2 \n" \

  146.                         "movq          %%mm1, %%mm3 \n" \

  147.                         "movq          %%mm2, %%mm4 \n" \

  148.                         "psubw   48(%1,%2,4), %%mm1 \n" \

  149.                         "psubw   56(%1,%2,4), %%mm2 \n" \

  150.                         "psubw  208(%1,%2,4), %%mm3 \n" \

  151.                         "psubw  216(%1,%2,4), %%mm4 \n" \

  152.                         "packsswb      %%mm2, %%mm1 \n" \

  153.                         "packsswb      %%mm4, %%mm3 \n" \

  154.                         "paddb         %%mm6, %%mm1 \n" \

  155.                         "paddb         %%mm6, %%mm3 \n" \

  156.                         "psubusb       %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \

  157.                         "psubusb       %%mm5, %%mm3 \n" \

  158.                         "packsswb      %%mm3, %%mm1 \n" \

  159.  \

  160.                         "pshufw $0x4E, %%mm1, %%mm1 \n" \

  161.                         "por           %%mm1, %%mm0 \n" \

  162.                         "pshufw $0x4E, %%mm0, %%mm1 \n" \

  163.                         "pminub        %%mm1, %%mm0 \n" \

  164.                         ::"r"(ref), \

  165.                           "r"(mv), \

  166.                           "r"(b_idx), \

  167.                           "i"(d_idx+12), \

  168.                           "i"(d_idx+52), \

  169.                           "i"(d_idx*4+48), \

  170.                           "i"(d_idx*4+56), \

  171.                           "i"(d_idx*4+208), \

  172.                           "i"(d_idx*4+216) \

  173.                     ); \

  174.                 } else { \

  175.                     __asm__ volatile( \

  176.                         "movd   12(%0,%2), %%mm0 \n" \

  177.                         "psubb %a3(%0,%2), %%mm0 \n" /* ref[b] != ref[bn] */ \

  178.                         "movq   48(%1,%2,4), %%mm1 \n" \

  179.                         "movq   56(%1,%2,4), %%mm2 \n" \

  180.                         "psubw %a4(%1,%2,4), %%mm1 \n" \

  181.                         "psubw %a5(%1,%2,4), %%mm2 \n" \

  182.                         "packsswb   %%mm2, %%mm1 \n" \

  183.                         "paddb      %%mm6, %%mm1 \n" \

  184.                         "psubusb    %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \

  185.                         "packsswb   %%mm1, %%mm1 \n" \

  186.                         "por        %%mm1, %%mm0 \n" \

  187.                         ::"r"(ref), \

  188.                           "r"(mv), \

  189.                           "r"(b_idx), \

  190.                           "i"(d_idx+12), \

  191.                           "i"(d_idx*4+48), \

  192.                           "i"(d_idx*4+56) \

  193.                     ); \

  194.                 } \

  195.             } \

  196.             __asm__ volatile( \

  197.                 "movd 12(%0,%1), %%mm1 \n" \

  198.                 "por %a2(%0,%1), %%mm1 \n" /* nnz[b] || nnz[bn] */ \

  199.                 ::"r"(nnz), \

  200.                   "r"(b_idx), \

  201.                   "i"(d_idx+12) \

  202.             ); \

  203.             __asm__ volatile( \

  204.                 "pminub    %%mm7, %%mm1 \n" \

  205.                 "pminub    %%mm7, %%mm0 \n" \

  206.                 "psllw        $1, %%mm1 \n" \

  207.                 "pxor      %%mm2, %%mm2 \n" \

  208.                 "pmaxub    %%mm0, %%mm1 \n" \

  209.                 "punpcklbw %%mm2, %%mm1 \n" \

  210.                 "movq      %%mm1, %a1(%0,%2) \n" \

  211.                 ::"r"(bS), \

  212.                   "i"(32*dir), \

  213.                   "r"(b_idx) \

  214.                 :"memory" \

  215.             ); \

  216.         } \

  217.     } while (0)

  218. 

  219. static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],

  220.                                             int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field ) {

  221.     __asm__ volatile(

  222.         "movq %0, %%mm7 \n"

  223.         "movq %1, %%mm6 \n"

  224.         ::"m"(ff_pb_1), "m"(ff_pb_3)

  225.     );

  226.     if(field)

  227.         __asm__ volatile(

  228.             "movq %0, %%mm6 \n"

  229.             ::"m"(ff_pb_3_1)

  230.         );

  231.     __asm__ volatile(

  232.         "movq  %%mm6, %%mm5 \n"

  233.         "paddb %%mm5, %%mm5 \n"

  234.     :);

  235. 

  236.     // could do a special case for dir==0 && edges==1, but it only reduces the

  237.     // average filter time by 1.2%

  238.     step  <<= 3;

  239.     edges <<= 3;

  240.     h264_loop_filter_strength_iteration_mmx2(bS, nnz, ref, mv, bidir, edges, step, mask_mv1, 1, -8,  0);

  241.     h264_loop_filter_strength_iteration_mmx2(bS, nnz, ref, mv, bidir,    32,    8, mask_mv0, 0, -1, -1);

  242. 

  243.     __asm__ volatile(

  244.         "movq   (%0), %%mm0 \n\t"

  245.         "movq  8(%0), %%mm1 \n\t"

  246.         "movq 16(%0), %%mm2 \n\t"

  247.         "movq 24(%0), %%mm3 \n\t"

  248.         TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4)

  249.         "movq %%mm0,   (%0) \n\t"

  250.         "movq %%mm3,  8(%0) \n\t"

  251.         "movq %%mm4, 16(%0) \n\t"

  252.         "movq %%mm2, 24(%0) \n\t"

  253.         ::"r"(bS[0])

  254.         :"memory"

  255.     );

  256. }

  257. 

  258. #define LF_FUNC(DIR, TYPE, DEPTH, OPT) \

  259. void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *pix, int stride, \

  260.                                                                 int alpha, int beta, int8_t *tc0);

  261. #define LF_IFUNC(DIR, TYPE, DEPTH, OPT) \

  262. void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *pix, int stride, \

  263.                                                                 int alpha, int beta);

  264. 

  265. #define LF_FUNCS(type, depth)\

  266. LF_FUNC (h,  chroma,       depth, mmxext)\

  267. LF_IFUNC(h,  chroma_intra, depth, mmxext)\

  268. LF_FUNC (v,  chroma,       depth, mmxext)\

  269. LF_IFUNC(v,  chroma_intra, depth, mmxext)\

  270. LF_FUNC (h,  luma,         depth, mmxext)\

  271. LF_IFUNC(h,  luma_intra,   depth, mmxext)\

  272. LF_FUNC (h,  luma,         depth, sse2)\

  273. LF_IFUNC(h,  luma_intra,   depth, sse2)\

  274. LF_FUNC (v,  luma,         depth, sse2)\

  275. LF_IFUNC(v,  luma_intra,   depth, sse2)\

  276. LF_FUNC (h,  chroma,       depth, sse2)\

  277. LF_IFUNC(h,  chroma_intra, depth, sse2)\

  278. LF_FUNC (v,  chroma,       depth, sse2)\

  279. LF_IFUNC(v,  chroma_intra, depth, sse2)\

  280. LF_FUNC (h,  luma,         depth,  avx)\

  281. LF_IFUNC(h,  luma_intra,   depth,  avx)\

  282. LF_FUNC (v,  luma,         depth,  avx)\

  283. LF_IFUNC(v,  luma_intra,   depth,  avx)\

  284. LF_FUNC (h,  chroma,       depth,  avx)\

  285. LF_IFUNC(h,  chroma_intra, depth,  avx)\

  286. LF_FUNC (v,  chroma,       depth,  avx)\

  287. LF_IFUNC(v,  chroma_intra, depth,  avx)

  288. 

  289. LF_FUNCS( uint8_t,  8)

  290. LF_FUNCS(uint16_t, 10)

  291. 

  292. #if ARCH_X86_32

  293. LF_FUNC (v8, luma,             8, mmxext)

  294. static void ff_deblock_v_luma_8_mmxext(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)

  295. {

  296.     if((tc0[0] & tc0[1]) >= 0)

  297.         ff_deblock_v8_luma_8_mmxext(pix+0, stride, alpha, beta, tc0);

  298.     if((tc0[2] & tc0[3]) >= 0)

  299.         ff_deblock_v8_luma_8_mmxext(pix+8, stride, alpha, beta, tc0+2);

  300. }

  301. LF_IFUNC(v8, luma_intra,        8, mmxext)

  302. static void ff_deblock_v_luma_intra_8_mmxext(uint8_t *pix, int stride, int alpha, int beta)

  303. {

  304.     ff_deblock_v8_luma_intra_8_mmxext(pix+0, stride, alpha, beta);

  305.     ff_deblock_v8_luma_intra_8_mmxext(pix+8, stride, alpha, beta);

  306. }

  307. #endif /* ARCH_X86_32 */

  308. 

  309. LF_FUNC (v,  luma,            10, mmxext)

  310. LF_IFUNC(v,  luma_intra,      10, mmxext)

  311. 

  312. /***********************************/

  313. /* weighted prediction */

  314. 

  315. #define H264_WEIGHT(W, H, OPT) \

  316. void ff_h264_weight_ ## W ## x ## H ## _ ## OPT(uint8_t *dst, \

  317.     int stride, int log2_denom, int weight, int offset);

  318. 

  319. #define H264_BIWEIGHT(W, H, OPT) \

  320. void ff_h264_biweight_ ## W ## x ## H ## _ ## OPT(uint8_t *dst, \

  321.     uint8_t *src, int stride, int log2_denom, int weightd, \

  322.     int weights, int offset);

  323. 

  324. #define H264_BIWEIGHT_MMX(W,H) \

  325. H264_WEIGHT  (W, H, mmx2) \

  326. H264_BIWEIGHT(W, H, mmx2)

  327. 

  328. #define H264_BIWEIGHT_MMX_SSE(W,H) \

  329. H264_BIWEIGHT_MMX(W, H) \

  330. H264_WEIGHT      (W, H, sse2) \

  331. H264_BIWEIGHT    (W, H, sse2) \

  332. H264_BIWEIGHT    (W, H, ssse3)

  333. 

  334. H264_BIWEIGHT_MMX_SSE(16, 16)

  335. H264_BIWEIGHT_MMX_SSE(16,  8)

  336. H264_BIWEIGHT_MMX_SSE( 8, 16)

  337. H264_BIWEIGHT_MMX_SSE( 8,  8)

  338. H264_BIWEIGHT_MMX_SSE( 8,  4)

  339. H264_BIWEIGHT_MMX    ( 4,  8)

  340. H264_BIWEIGHT_MMX    ( 4,  4)

  341. H264_BIWEIGHT_MMX    ( 4,  2)

  342. 

  343. void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)

  344. {

  345.     int mm_flags = av_get_cpu_flags();

  346. 

  347.     if (bit_depth == 8) {

  348.     if (mm_flags & AV_CPU_FLAG_MMX2) {

  349.         c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2;

  350.     }

  351. #if HAVE_YASM

  352.     if (mm_flags & AV_CPU_FLAG_MMX) {

  353.         c->h264_idct_dc_add=

  354.         c->h264_idct_add= ff_h264_idct_add_mmx;

  355.         c->h264_idct8_dc_add=

  356.         c->h264_idct8_add= ff_h264_idct8_add_mmx;

  357. 

  358.         c->h264_idct_add16     = ff_h264_idct_add16_mmx;

  359.         c->h264_idct8_add4     = ff_h264_idct8_add4_mmx;

  360.         c->h264_idct_add8      = ff_h264_idct_add8_mmx;

  361.         c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx;

  362.         c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_mmx;

  363. 

  364.         if (mm_flags & AV_CPU_FLAG_MMX2) {

  365.             c->h264_idct_dc_add= ff_h264_idct_dc_add_mmx2;

  366.             c->h264_idct8_dc_add= ff_h264_idct8_dc_add_mmx2;

  367.             c->h264_idct_add16     = ff_h264_idct_add16_mmx2;

  368.             c->h264_idct8_add4     = ff_h264_idct8_add4_mmx2;

  369.             c->h264_idct_add8      = ff_h264_idct_add8_mmx2;

  370.             c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;

  371. 

  372.             c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_8_mmxext;

  373.             c->h264_h_loop_filter_chroma= ff_deblock_h_chroma_8_mmxext;

  374.             c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_8_mmxext;

  375.             c->h264_h_loop_filter_chroma_intra= ff_deblock_h_chroma_intra_8_mmxext;

  376. #if ARCH_X86_32

  377.             c->h264_v_loop_filter_luma= ff_deblock_v_luma_8_mmxext;

  378.             c->h264_h_loop_filter_luma= ff_deblock_h_luma_8_mmxext;

  379.             c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmxext;

  380.             c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmxext;

  381. #endif

  382.             c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_mmx2;

  383.             c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_mmx2;

  384.             c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_mmx2;

  385.             c->weight_h264_pixels_tab[3]= ff_h264_weight_8x8_mmx2;

  386.             c->weight_h264_pixels_tab[4]= ff_h264_weight_8x4_mmx2;

  387.             c->weight_h264_pixels_tab[5]= ff_h264_weight_4x8_mmx2;

  388.             c->weight_h264_pixels_tab[6]= ff_h264_weight_4x4_mmx2;

  389.             c->weight_h264_pixels_tab[7]= ff_h264_weight_4x2_mmx2;

  390. 

  391.             c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_mmx2;

  392.             c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_mmx2;

  393.             c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_mmx2;

  394.             c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_mmx2;

  395.             c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_mmx2;

  396.             c->biweight_h264_pixels_tab[5]= ff_h264_biweight_4x8_mmx2;

  397.             c->biweight_h264_pixels_tab[6]= ff_h264_biweight_4x4_mmx2;

  398.             c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;

  399. 

  400.             if (mm_flags&AV_CPU_FLAG_SSE2) {

  401.                 c->h264_idct8_add = ff_h264_idct8_add_sse2;

  402.                 c->h264_idct8_add4= ff_h264_idct8_add4_sse2;

  403.                 c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_sse2;

  404. 

  405.                 c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_sse2;

  406.                 c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_sse2;

  407.                 c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_sse2;

  408.                 c->weight_h264_pixels_tab[3]= ff_h264_weight_8x8_sse2;

  409.                 c->weight_h264_pixels_tab[4]= ff_h264_weight_8x4_sse2;

  410. 

  411.                 c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_sse2;

  412.                 c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_sse2;

  413.                 c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_sse2;

  414.                 c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_sse2;

  415.                 c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_sse2;

  416. 

  417. #if HAVE_ALIGNED_STACK

  418.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_sse2;

  419.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_sse2;

  420.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_sse2;

  421.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_sse2;

  422. #endif

  423. 

  424.                 c->h264_idct_add16 = ff_h264_idct_add16_sse2;

  425.                 c->h264_idct_add8  = ff_h264_idct_add8_sse2;

  426.                 c->h264_idct_add16intra = ff_h264_idct_add16intra_sse2;

  427.             }

  428.             if (mm_flags&AV_CPU_FLAG_SSSE3) {

  429.                 c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_ssse3;

  430.                 c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_ssse3;

  431.                 c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_ssse3;

  432.                 c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_ssse3;

  433.                 c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_ssse3;

  434.             }

  435.             if (mm_flags&AV_CPU_FLAG_AVX) {

  436. #if HAVE_ALIGNED_STACK

  437.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_avx;

  438.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_avx;

  439.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_avx;

  440.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_avx;

  441. #endif

  442.             }

  443.         }

  444.     }

  445. #endif

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

  447. #if HAVE_YASM

  448.     if (mm_flags & AV_CPU_FLAG_MMX) {

  449.         if (mm_flags & AV_CPU_FLAG_MMX2) {

  450. #if ARCH_X86_32

  451.             c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_mmxext;

  452.             c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_mmxext;

  453.             c->h264_v_loop_filter_luma= ff_deblock_v_luma_10_mmxext;

  454.             c->h264_h_loop_filter_luma= ff_deblock_h_luma_10_mmxext;

  455.             c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_mmxext;

  456.             c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_mmxext;

  457. #endif

  458.             c->h264_idct_dc_add= ff_h264_idct_dc_add_10_mmx2;

  459.             if (mm_flags&AV_CPU_FLAG_SSE2) {

  460.                 c->h264_idct_add       = ff_h264_idct_add_10_sse2;

  461.                 c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_10_sse2;

  462.                 c->h264_idct8_add      = ff_h264_idct8_add_10_sse2;

  463. 

  464.                 c->h264_idct_add16     = ff_h264_idct_add16_10_sse2;

  465.                 c->h264_idct8_add4     = ff_h264_idct8_add4_10_sse2;

  466.                 c->h264_idct_add8      = ff_h264_idct_add8_10_sse2;

  467.                 c->h264_idct_add16intra= ff_h264_idct_add16intra_10_sse2;

  468. 

  469.                 c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_sse2;

  470.                 c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_sse2;

  471. #if HAVE_ALIGNED_STACK

  472.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_sse2;

  473.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_sse2;

  474.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_sse2;

  475.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_sse2;

  476. #endif

  477.             }

  478. #if HAVE_AVX

  479.             if (mm_flags&AV_CPU_FLAG_AVX) {

  480.                 c->h264_idct_dc_add    =

  481.                 c->h264_idct_add       = ff_h264_idct_add_10_avx;

  482.                 c->h264_idct8_add      = ff_h264_idct8_add_10_avx;

  483.                 c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_10_avx;

  484. 

  485.                 c->h264_idct_add16     = ff_h264_idct_add16_10_avx;

  486.                 c->h264_idct8_add4     = ff_h264_idct8_add4_10_avx;

  487.                 c->h264_idct_add8      = ff_h264_idct_add8_10_avx;

  488.                 c->h264_idct_add16intra= ff_h264_idct_add16intra_10_avx;

  489. 

  490.                 c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_avx;

  491.                 c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_avx;

  492. #if HAVE_ALIGNED_STACK

  493.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_avx;

  494.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_avx;

  495.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_avx;

  496.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_avx;

  497. #endif

  498.             }

  499. #endif /* HAVE_AVX */

  500.         }

  501.     }

  502. #endif

  503.     }

  504. }