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.
35406 Commits
32 Branches
776MB
Tree: c3e15f7b39
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c3e15f7b39'
${ noResults }
FFmpeg/libavcodec/x86/h264dsp_init.c

375 lines
17KB
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/asm.h"

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

  24. #include "libavcodec/h264dsp.h"

  25. #include "dsputil_mmx.h"

  26. 

  27. /***********************************/

  28. /* IDCT */

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

  30. void ff_h264_idct ## NUM ## _add_ ## DEPTH ## _ ## OPT(uint8_t *dst,    \

  31.                                                        int16_t *block,  \

  32.                                                        int stride);

  33. 

  34. IDCT_ADD_FUNC(, 8, mmx)

  35. IDCT_ADD_FUNC(, 10, sse2)

  36. IDCT_ADD_FUNC(_dc, 8, mmx2)

  37. IDCT_ADD_FUNC(_dc, 10, mmx2)

  38. IDCT_ADD_FUNC(8_dc, 8, mmx2)

  39. IDCT_ADD_FUNC(8_dc, 10, sse2)

  40. IDCT_ADD_FUNC(8, 8, mmx)

  41. IDCT_ADD_FUNC(8, 8, sse2)

  42. IDCT_ADD_FUNC(8, 10, sse2)

  43. IDCT_ADD_FUNC(, 10, avx)

  44. IDCT_ADD_FUNC(8_dc, 10, avx)

  45. IDCT_ADD_FUNC(8, 10, avx)

  46. 

  47. 

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

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

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

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

  52. 

  53. IDCT_ADD_REP_FUNC(8, 4, 8, mmx)

  54. IDCT_ADD_REP_FUNC(8, 4, 8, mmx2)

  55. IDCT_ADD_REP_FUNC(8, 4, 8, sse2)

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

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

  58. IDCT_ADD_REP_FUNC(, 16, 8, mmx)

  59. IDCT_ADD_REP_FUNC(, 16, 8, mmx2)

  60. IDCT_ADD_REP_FUNC(, 16, 8, sse2)

  61. IDCT_ADD_REP_FUNC(, 16, 10, sse2)

  62. IDCT_ADD_REP_FUNC(, 16intra, 8, mmx)

  63. IDCT_ADD_REP_FUNC(, 16intra, 8, mmx2)

  64. IDCT_ADD_REP_FUNC(, 16intra, 8, sse2)

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

  66. IDCT_ADD_REP_FUNC(, 16, 10, avx)

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

  68. 

  69. 

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

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

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

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

  74. 

  75. IDCT_ADD_REP_FUNC2(, 8, 8, mmx)

  76. IDCT_ADD_REP_FUNC2(, 8, 8, mmx2)

  77. IDCT_ADD_REP_FUNC2(, 8, 8, sse2)

  78. IDCT_ADD_REP_FUNC2(, 8, 10, sse2)

  79. IDCT_ADD_REP_FUNC2(, 8, 10, avx)

  80. 

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

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

  83. 

  84. /***********************************/

  85. /* deblocking */

  86. 

  87. void ff_h264_loop_filter_strength_mmx2(int16_t bS[2][4][4], uint8_t nnz[40],

  88.                                        int8_t ref[2][40], int16_t mv[2][40][2],

  89.                                        int bidir, int edges, int step,

  90.                                        int mask_mv0, int mask_mv1, int field);

  91. 

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

  93. void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix,  \

  94.                                                                int stride,    \

  95.                                                                int alpha,     \

  96.                                                                int beta,      \

  97.                                                                int8_t *tc0);

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

  99. void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix,  \

  100.                                                                int stride,    \

  101.                                                                int alpha,     \

  102.                                                                int beta);

  103. 

  104. #define LF_FUNCS(type, depth)                   \

  105. LF_FUNC(h,  chroma,       depth, mmx2)          \

  106. LF_IFUNC(h, chroma_intra, depth, mmx2)          \

  107. LF_FUNC(v,  chroma,       depth, mmx2)          \

  108. LF_IFUNC(v, chroma_intra, depth, mmx2)          \

  109. LF_FUNC(h,  luma,         depth, mmx2)          \

  110. LF_IFUNC(h, luma_intra,   depth, mmx2)          \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  127. 

  128. LF_FUNCS(uint8_t,   8)

  129. LF_FUNCS(uint16_t, 10)

  130. 

  131. #if ARCH_X86_32

  132. LF_FUNC(v8, luma, 8, mmx2)

  133. static void ff_deblock_v_luma_8_mmx2(uint8_t *pix, int stride, int alpha,

  134.                                      int beta, int8_t *tc0)

  135. {

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

  137.         ff_deblock_v8_luma_8_mmx2(pix + 0, stride, alpha, beta, tc0);

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

  139.         ff_deblock_v8_luma_8_mmx2(pix + 8, stride, alpha, beta, tc0 + 2);

  140. }

  141. 

  142. LF_IFUNC(v8, luma_intra, 8, mmx2)

  143. static void ff_deblock_v_luma_intra_8_mmx2(uint8_t *pix, int stride,

  144.                                            int alpha, int beta)

  145. {

  146.     ff_deblock_v8_luma_intra_8_mmx2(pix + 0, stride, alpha, beta);

  147.     ff_deblock_v8_luma_intra_8_mmx2(pix + 8, stride, alpha, beta);

  148. }

  149. #endif /* ARCH_X86_32 */

  150. 

  151. LF_FUNC(v,  luma,       10, mmx2)

  152. LF_IFUNC(v, luma_intra, 10, mmx2)

  153. 

  154. /***********************************/

  155. /* weighted prediction */

  156. 

  157. #define H264_WEIGHT(W, OPT)                                             \

  158. void ff_h264_weight_ ## W ## _ ## OPT(uint8_t *dst, int stride,         \

  159.                                       int height, int log2_denom,       \

  160.                                       int weight, int offset);

  161. 

  162. #define H264_BIWEIGHT(W, OPT)                                           \

  163. void ff_h264_biweight_ ## W ## _ ## OPT(uint8_t *dst, uint8_t *src,     \

  164.                                         int stride, int height,         \

  165.                                         int log2_denom, int weightd,    \

  166.                                         int weights, int offset);

  167. 

  168. #define H264_BIWEIGHT_MMX(W)                    \

  169.     H264_WEIGHT(W, mmx2)                        \

  170.     H264_BIWEIGHT(W, mmx2)

  171. 

  172. #define H264_BIWEIGHT_MMX_SSE(W)                \

  173.     H264_BIWEIGHT_MMX(W)                        \

  174.     H264_WEIGHT(W, sse2)                        \

  175.     H264_BIWEIGHT(W, sse2)                      \

  176.     H264_BIWEIGHT(W, ssse3)

  177. 

  178. H264_BIWEIGHT_MMX_SSE(16)

  179. H264_BIWEIGHT_MMX_SSE(8)

  180. H264_BIWEIGHT_MMX(4)

  181. 

  182. #define H264_WEIGHT_10(W, DEPTH, OPT)                                   \

  183. void ff_h264_weight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst,       \

  184.                                                     int stride,         \

  185.                                                     int height,         \

  186.                                                     int log2_denom,     \

  187.                                                     int weight,         \

  188.                                                     int offset);

  189. 

  190. #define H264_BIWEIGHT_10(W, DEPTH, OPT)                                 \

  191. void ff_h264_biweight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst,     \

  192.                                                       uint8_t *src,     \

  193.                                                       int stride,       \

  194.                                                       int height,       \

  195.                                                       int log2_denom,   \

  196.                                                       int weightd,      \

  197.                                                       int weights,      \

  198.                                                       int offset);

  199. 

  200. #define H264_BIWEIGHT_10_SSE(W, DEPTH)          \

  201.     H264_WEIGHT_10(W, DEPTH, sse2)              \

  202.     H264_WEIGHT_10(W, DEPTH, sse4)              \

  203.     H264_BIWEIGHT_10(W, DEPTH, sse2)            \

  204.     H264_BIWEIGHT_10(W, DEPTH, sse4)

  205. 

  206. H264_BIWEIGHT_10_SSE(16, 10)

  207. H264_BIWEIGHT_10_SSE(8,  10)

  208. H264_BIWEIGHT_10_SSE(4,  10)

  209. 

  210. void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,

  211.                          const int chroma_format_idc)

  212. {

  213.     int mm_flags = av_get_cpu_flags();

  214. 

  215.     if (chroma_format_idc == 1 && EXTERNAL_MMXEXT(mm_flags))

  216.         c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2;

  217. 

  218.     if (bit_depth == 8) {

  219.         if (EXTERNAL_MMX(mm_flags)) {

  220.             c->h264_idct_dc_add   =

  221.             c->h264_idct_add      = ff_h264_idct_add_8_mmx;

  222.             c->h264_idct8_dc_add  =

  223.             c->h264_idct8_add     = ff_h264_idct8_add_8_mmx;

  224. 

  225.             c->h264_idct_add16 = ff_h264_idct_add16_8_mmx;

  226.             c->h264_idct8_add4 = ff_h264_idct8_add4_8_mmx;

  227.             if (chroma_format_idc == 1)

  228.                 c->h264_idct_add8 = ff_h264_idct_add8_8_mmx;

  229.             c->h264_idct_add16intra = ff_h264_idct_add16intra_8_mmx;

  230.             if (mm_flags & AV_CPU_FLAG_CMOV)

  231.                 c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx;

  232. 

  233.             if (EXTERNAL_MMXEXT(mm_flags)) {

  234.                 c->h264_idct_dc_add  = ff_h264_idct_dc_add_8_mmx2;

  235.                 c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmx2;

  236.                 c->h264_idct_add16   = ff_h264_idct_add16_8_mmx2;

  237.                 c->h264_idct8_add4   = ff_h264_idct8_add4_8_mmx2;

  238.                 if (chroma_format_idc == 1)

  239.                     c->h264_idct_add8 = ff_h264_idct_add8_8_mmx2;

  240.                 c->h264_idct_add16intra = ff_h264_idct_add16intra_8_mmx2;

  241. 

  242.                 c->h264_v_loop_filter_chroma       = ff_deblock_v_chroma_8_mmx2;

  243.                 c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_8_mmx2;

  244.                 if (chroma_format_idc == 1) {

  245.                     c->h264_h_loop_filter_chroma       = ff_deblock_h_chroma_8_mmx2;

  246.                     c->h264_h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_mmx2;

  247.                 }

  248. #if ARCH_X86_32

  249.                 c->h264_v_loop_filter_luma       = ff_deblock_v_luma_8_mmx2;

  250.                 c->h264_h_loop_filter_luma       = ff_deblock_h_luma_8_mmx2;

  251.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmx2;

  252.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmx2;

  253. #endif /* ARCH_X86_32 */

  254.                 c->weight_h264_pixels_tab[0] = ff_h264_weight_16_mmx2;

  255.                 c->weight_h264_pixels_tab[1] = ff_h264_weight_8_mmx2;

  256.                 c->weight_h264_pixels_tab[2] = ff_h264_weight_4_mmx2;

  257. 

  258.                 c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_mmx2;

  259.                 c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_mmx2;

  260.                 c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_mmx2;

  261. 

  262.                 if (EXTERNAL_SSE2(mm_flags)) {

  263.                     c->h264_idct8_add  = ff_h264_idct8_add_8_sse2;

  264. 

  265.                     c->h264_idct_add16 = ff_h264_idct_add16_8_sse2;

  266.                     c->h264_idct8_add4 = ff_h264_idct8_add4_8_sse2;

  267.                     if (chroma_format_idc == 1)

  268.                         c->h264_idct_add8 = ff_h264_idct_add8_8_sse2;

  269.                     c->h264_idct_add16intra      = ff_h264_idct_add16intra_8_sse2;

  270.                     c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_sse2;

  271. 

  272.                     c->weight_h264_pixels_tab[0] = ff_h264_weight_16_sse2;

  273.                     c->weight_h264_pixels_tab[1] = ff_h264_weight_8_sse2;

  274. 

  275.                     c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_sse2;

  276.                     c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_sse2;

  277. 

  278. #if HAVE_ALIGNED_STACK

  279.                     c->h264_v_loop_filter_luma       = ff_deblock_v_luma_8_sse2;

  280.                     c->h264_h_loop_filter_luma       = ff_deblock_h_luma_8_sse2;

  281.                     c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_sse2;

  282.                     c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_sse2;

  283. #endif /* HAVE_ALIGNED_STACK */

  284.                 }

  285.                 if (EXTERNAL_SSSE3(mm_flags)) {

  286.                     c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_ssse3;

  287.                     c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_ssse3;

  288.                 }

  289.                 if (EXTERNAL_AVX(mm_flags) && HAVE_ALIGNED_STACK) {

  290.                     c->h264_v_loop_filter_luma       = ff_deblock_v_luma_8_avx;

  291.                     c->h264_h_loop_filter_luma       = ff_deblock_h_luma_8_avx;

  292.                     c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_avx;

  293.                     c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_avx;

  294.                 }

  295.             }

  296.         }

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

  298.         if (EXTERNAL_MMX(mm_flags)) {

  299.             if (EXTERNAL_MMXEXT(mm_flags)) {

  300. #if ARCH_X86_32

  301.                 c->h264_v_loop_filter_chroma       = ff_deblock_v_chroma_10_mmx2;

  302.                 c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmx2;

  303.                 c->h264_v_loop_filter_luma         = ff_deblock_v_luma_10_mmx2;

  304.                 c->h264_h_loop_filter_luma         = ff_deblock_h_luma_10_mmx2;

  305.                 c->h264_v_loop_filter_luma_intra   = ff_deblock_v_luma_intra_10_mmx2;

  306.                 c->h264_h_loop_filter_luma_intra   = ff_deblock_h_luma_intra_10_mmx2;

  307. #endif /* ARCH_X86_32 */

  308.                 c->h264_idct_dc_add = ff_h264_idct_dc_add_10_mmx2;

  309.                 if (EXTERNAL_SSE2(mm_flags)) {

  310.                     c->h264_idct_add     = ff_h264_idct_add_10_sse2;

  311.                     c->h264_idct8_dc_add = ff_h264_idct8_dc_add_10_sse2;

  312. 

  313.                     c->h264_idct_add16 = ff_h264_idct_add16_10_sse2;

  314.                     if (chroma_format_idc == 1)

  315.                         c->h264_idct_add8 = ff_h264_idct_add8_10_sse2;

  316.                     c->h264_idct_add16intra = ff_h264_idct_add16intra_10_sse2;

  317. #if HAVE_ALIGNED_STACK

  318.                     c->h264_idct8_add  = ff_h264_idct8_add_10_sse2;

  319.                     c->h264_idct8_add4 = ff_h264_idct8_add4_10_sse2;

  320. #endif /* HAVE_ALIGNED_STACK */

  321. 

  322.                     c->weight_h264_pixels_tab[0] = ff_h264_weight_16_10_sse2;

  323.                     c->weight_h264_pixels_tab[1] = ff_h264_weight_8_10_sse2;

  324.                     c->weight_h264_pixels_tab[2] = ff_h264_weight_4_10_sse2;

  325. 

  326.                     c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_10_sse2;

  327.                     c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_10_sse2;

  328.                     c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_10_sse2;

  329. 

  330.                     c->h264_v_loop_filter_chroma       = ff_deblock_v_chroma_10_sse2;

  331.                     c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_sse2;

  332. #if HAVE_ALIGNED_STACK

  333.                     c->h264_v_loop_filter_luma       = ff_deblock_v_luma_10_sse2;

  334.                     c->h264_h_loop_filter_luma       = ff_deblock_h_luma_10_sse2;

  335.                     c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_sse2;

  336.                     c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_sse2;

  337. #endif /* HAVE_ALIGNED_STACK */

  338.                 }

  339.                 if (EXTERNAL_SSE4(mm_flags)) {

  340.                     c->weight_h264_pixels_tab[0] = ff_h264_weight_16_10_sse4;

  341.                     c->weight_h264_pixels_tab[1] = ff_h264_weight_8_10_sse4;

  342.                     c->weight_h264_pixels_tab[2] = ff_h264_weight_4_10_sse4;

  343. 

  344.                     c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_10_sse4;

  345.                     c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_10_sse4;

  346.                     c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_10_sse4;

  347.                 }

  348.                 if (EXTERNAL_AVX(mm_flags)) {

  349.                     c->h264_idct_dc_add  =

  350.                     c->h264_idct_add     = ff_h264_idct_add_10_avx;

  351.                     c->h264_idct8_dc_add = ff_h264_idct8_dc_add_10_avx;

  352. 

  353.                     c->h264_idct_add16 = ff_h264_idct_add16_10_avx;

  354.                     if (chroma_format_idc == 1)

  355.                         c->h264_idct_add8 = ff_h264_idct_add8_10_avx;

  356.                     c->h264_idct_add16intra = ff_h264_idct_add16intra_10_avx;

  357. #if HAVE_ALIGNED_STACK

  358.                     c->h264_idct8_add  = ff_h264_idct8_add_10_avx;

  359.                     c->h264_idct8_add4 = ff_h264_idct8_add4_10_avx;

  360. #endif /* HAVE_ALIGNED_STACK */

  361. 

  362.                     c->h264_v_loop_filter_chroma       = ff_deblock_v_chroma_10_avx;

  363.                     c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_avx;

  364. #if HAVE_ALIGNED_STACK

  365.                     c->h264_v_loop_filter_luma         = ff_deblock_v_luma_10_avx;

  366.                     c->h264_h_loop_filter_luma         = ff_deblock_h_luma_10_avx;

  367.                     c->h264_v_loop_filter_luma_intra   = ff_deblock_v_luma_intra_10_avx;

  368.                     c->h264_h_loop_filter_luma_intra   = ff_deblock_h_luma_intra_10_avx;

  369. #endif /* HAVE_ALIGNED_STACK */

  370.                 }

  371.             }

  372.         }

  373.     }

  374. }