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.
43135 Commits
32 Branches
303MB
Tree: ec7ecb8811
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'ec7ecb8811'
${ noResults }
FFmpeg/libavcodec/x86/h264dsp_mmx.c

364 lines
15KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

  6.  * FFmpeg 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.  * FFmpeg is distributed in the hope that it will be useful,

  12.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  13.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the 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 FFmpeg; 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. /***********************************/

  27. /* IDCT */

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

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

  30. 

  31. IDCT_ADD_FUNC(, 8, mmx)

  32. IDCT_ADD_FUNC(, 10, sse2)

  33. IDCT_ADD_FUNC(_dc, 8, mmx2)

  34. IDCT_ADD_FUNC(_dc, 10, mmx2)

  35. IDCT_ADD_FUNC(8_dc, 8, mmx2)

  36. IDCT_ADD_FUNC(8_dc, 10, sse2)

  37. IDCT_ADD_FUNC(8, 8, mmx)

  38. IDCT_ADD_FUNC(8, 8, sse2)

  39. IDCT_ADD_FUNC(8, 10, sse2)

  40. #if HAVE_AVX

  41. IDCT_ADD_FUNC(, 10, avx)

  42. IDCT_ADD_FUNC(8_dc, 10, avx)

  43. IDCT_ADD_FUNC(8, 10, avx)

  44. #endif

  45. 

  46. 

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

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

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

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

  51. 

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

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

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

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

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

  57. IDCT_ADD_REP_FUNC(, 16, 8, mmx)

  58. IDCT_ADD_REP_FUNC(, 16, 8, mmx2)

  59. IDCT_ADD_REP_FUNC(, 16, 8, sse2)

  60. IDCT_ADD_REP_FUNC(, 16, 10, sse2)

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

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

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

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

  65. #if HAVE_AVX

  66. IDCT_ADD_REP_FUNC(, 16, 10, avx)

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

  68. #endif

  69. 

  70. 

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

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

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

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

  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. #if HAVE_AVX

  80. IDCT_ADD_REP_FUNC2(, 8, 10, avx)

  81. #endif

  82. 

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

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

  85. 

  86. /***********************************/

  87. /* deblocking */

  88. 

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

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

  91.                                        int bidir, int edges, int step,

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

  93. 

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

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

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

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

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

  99.                                                                 int alpha, int beta);

  100. 

  101. #define LF_FUNCS(type, depth)\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  124. 

  125. LF_FUNCS( uint8_t,  8)

  126. LF_FUNCS(uint16_t, 10)

  127. 

  128. #if ARCH_X86_32 && HAVE_YASM

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

  130. static void ff_deblock_v_luma_8_mmx2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)

  131. {

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

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

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

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

  136. }

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

  138. static void ff_deblock_v_luma_intra_8_mmx2(uint8_t *pix, int stride, int alpha, int beta)

  139. {

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

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

  142. }

  143. #endif /* ARCH_X86_32 */

  144. 

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

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

  147. 

  148. /***********************************/

  149. /* weighted prediction */

  150. 

  151. #define H264_WEIGHT(W, OPT) \

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

  153.     int stride, int height, int log2_denom, int weight, int offset);

  154. 

  155. #define H264_BIWEIGHT(W, OPT) \

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

  157.     uint8_t *src, int stride, int height, int log2_denom, int weightd, \

  158.     int weights, int offset);

  159. 

  160. #define H264_BIWEIGHT_MMX(W) \

  161. H264_WEIGHT  (W, mmx2) \

  162. H264_BIWEIGHT(W, mmx2)

  163. 

  164. #define H264_BIWEIGHT_MMX_SSE(W) \

  165. H264_BIWEIGHT_MMX(W) \

  166. H264_WEIGHT      (W, sse2) \

  167. H264_BIWEIGHT    (W, sse2) \

  168. H264_BIWEIGHT    (W, ssse3)

  169. 

  170. H264_BIWEIGHT_MMX_SSE(16)

  171. H264_BIWEIGHT_MMX_SSE( 8)

  172. H264_BIWEIGHT_MMX    ( 4)

  173. 

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

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

  176.     int stride, int height, int log2_denom, int weight, int offset);

  177. 

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

  179. void ff_h264_biweight_ ## W ## _ ## DEPTH ## _ ## OPT \

  180.     (uint8_t *dst, uint8_t *src, int stride, int height, int log2_denom, \

  181.      int weightd, int weights, int offset);

  182. 

  183. #define H264_BIWEIGHT_10_SSE(W, DEPTH) \

  184. H264_WEIGHT_10  (W, DEPTH, sse2) \

  185. H264_WEIGHT_10  (W, DEPTH, sse4) \

  186. H264_BIWEIGHT_10(W, DEPTH, sse2) \

  187. H264_BIWEIGHT_10(W, DEPTH, sse4)

  188. 

  189. H264_BIWEIGHT_10_SSE(16, 10)

  190. H264_BIWEIGHT_10_SSE( 8, 10)

  191. H264_BIWEIGHT_10_SSE( 4, 10)

  192. 

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

  194. {

  195. #if HAVE_YASM

  196.     int mm_flags = av_get_cpu_flags();

  197. 

  198.     if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMX2) {

  199.         c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2;

  200.     }

  201. 

  202.     if (bit_depth == 8) {

  203.     if (mm_flags & AV_CPU_FLAG_MMX) {

  204.         c->h264_idct_dc_add         =

  205.         c->h264_idct_add            = ff_h264_idct_add_8_mmx;

  206.         c->h264_idct8_dc_add        =

  207.         c->h264_idct8_add           = ff_h264_idct8_add_8_mmx;

  208. 

  209.         c->h264_idct_add16          = ff_h264_idct_add16_8_mmx;

  210.         c->h264_idct8_add4          = ff_h264_idct8_add4_8_mmx;

  211.         if (chroma_format_idc == 1)

  212.             c->h264_idct_add8       = ff_h264_idct_add8_8_mmx;

  213.         c->h264_idct_add16intra     = ff_h264_idct_add16intra_8_mmx;

  214.         if (mm_flags & AV_CPU_FLAG_CMOV)

  215.             c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx;

  216. 

  217.         if (mm_flags & AV_CPU_FLAG_MMX2) {

  218.             c->h264_idct_dc_add    = ff_h264_idct_dc_add_8_mmx2;

  219.             c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_8_mmx2;

  220.             c->h264_idct_add16     = ff_h264_idct_add16_8_mmx2;

  221.             c->h264_idct8_add4     = ff_h264_idct8_add4_8_mmx2;

  222.             if (chroma_format_idc == 1)

  223.                 c->h264_idct_add8  = ff_h264_idct_add8_8_mmx2;

  224.             c->h264_idct_add16intra= ff_h264_idct_add16intra_8_mmx2;

  225. 

  226.             c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_8_mmx2;

  227.             c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_8_mmx2;

  228.             if (chroma_format_idc == 1) {

  229.                 c->h264_h_loop_filter_chroma= ff_deblock_h_chroma_8_mmx2;

  230.                 c->h264_h_loop_filter_chroma_intra= ff_deblock_h_chroma_intra_8_mmx2;

  231.             }

  232. #if ARCH_X86_32

  233.             c->h264_v_loop_filter_luma= ff_deblock_v_luma_8_mmx2;

  234.             c->h264_h_loop_filter_luma= ff_deblock_h_luma_8_mmx2;

  235.             c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmx2;

  236.             c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmx2;

  237. #endif

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

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

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

  241. 

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

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

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

  245. 

  246.             if (mm_flags&AV_CPU_FLAG_SSE2) {

  247.                 c->h264_idct8_add           = ff_h264_idct8_add_8_sse2;

  248. 

  249.                 c->h264_idct_add16          = ff_h264_idct_add16_8_sse2;

  250.                 c->h264_idct8_add4          = ff_h264_idct8_add4_8_sse2;

  251.                 if (chroma_format_idc == 1)

  252.                     c->h264_idct_add8       = ff_h264_idct_add8_8_sse2;

  253.                 c->h264_idct_add16intra     = ff_h264_idct_add16intra_8_sse2;

  254.                 c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_sse2;

  255. 

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

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

  258. 

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

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

  261. 

  262. #if HAVE_ALIGNED_STACK

  263.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_sse2;

  264.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_sse2;

  265.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_sse2;

  266.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_sse2;

  267. #endif

  268.             }

  269.             if (mm_flags&AV_CPU_FLAG_SSSE3) {

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

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

  272.             }

  273.             if (HAVE_AVX && mm_flags&AV_CPU_FLAG_AVX) {

  274. #if HAVE_ALIGNED_STACK

  275.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_avx;

  276.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_avx;

  277.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_avx;

  278.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_avx;

  279. #endif

  280.             }

  281.         }

  282.     }

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

  284.     if (mm_flags & AV_CPU_FLAG_MMX) {

  285.         if (mm_flags & AV_CPU_FLAG_MMX2) {

  286. #if ARCH_X86_32

  287.             c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_mmx2;

  288.             c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_mmx2;

  289.             c->h264_v_loop_filter_luma= ff_deblock_v_luma_10_mmx2;

  290.             c->h264_h_loop_filter_luma= ff_deblock_h_luma_10_mmx2;

  291.             c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_mmx2;

  292.             c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_mmx2;

  293. #endif

  294.             c->h264_idct_dc_add= ff_h264_idct_dc_add_10_mmx2;

  295.             if (mm_flags&AV_CPU_FLAG_SSE2) {

  296.                 c->h264_idct_add       = ff_h264_idct_add_10_sse2;

  297.                 c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_10_sse2;

  298. 

  299.                 c->h264_idct_add16     = ff_h264_idct_add16_10_sse2;

  300.                 if (chroma_format_idc == 1)

  301.                     c->h264_idct_add8  = ff_h264_idct_add8_10_sse2;

  302.                 c->h264_idct_add16intra= ff_h264_idct_add16intra_10_sse2;

  303. #if HAVE_ALIGNED_STACK

  304.                 c->h264_idct8_add      = ff_h264_idct8_add_10_sse2;

  305.                 c->h264_idct8_add4     = ff_h264_idct8_add4_10_sse2;

  306. #endif

  307. 

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

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

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

  311. 

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

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

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

  315. 

  316.                 c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_sse2;

  317.                 c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_sse2;

  318. #if HAVE_ALIGNED_STACK

  319.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_sse2;

  320.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_sse2;

  321.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_sse2;

  322.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_sse2;

  323. #endif

  324.             }

  325.             if (mm_flags&AV_CPU_FLAG_SSE4) {

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

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

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

  329. 

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

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

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

  333.             }

  334. #if HAVE_AVX

  335.             if (mm_flags&AV_CPU_FLAG_AVX) {

  336.                 c->h264_idct_dc_add    =

  337.                 c->h264_idct_add       = ff_h264_idct_add_10_avx;

  338.                 c->h264_idct8_dc_add   = ff_h264_idct8_dc_add_10_avx;

  339. 

  340.                 c->h264_idct_add16     = ff_h264_idct_add16_10_avx;

  341.                 if (chroma_format_idc == 1)

  342.                     c->h264_idct_add8  = ff_h264_idct_add8_10_avx;

  343.                 c->h264_idct_add16intra= ff_h264_idct_add16intra_10_avx;

  344. #if HAVE_ALIGNED_STACK

  345.                 c->h264_idct8_add      = ff_h264_idct8_add_10_avx;

  346.                 c->h264_idct8_add4     = ff_h264_idct8_add4_10_avx;

  347. #endif

  348. 

  349.                 c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_10_avx;

  350.                 c->h264_v_loop_filter_chroma_intra= ff_deblock_v_chroma_intra_10_avx;

  351. #if HAVE_ALIGNED_STACK

  352.                 c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_avx;

  353.                 c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_avx;

  354.                 c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_avx;

  355.                 c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_avx;

  356. #endif

  357.             }

  358. #endif /* HAVE_AVX */

  359.         }

  360.     }

  361.     }

  362. #endif

  363. }