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.
37273 Commits
32 Branches
297MB
Tree: 110796739a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '110796739a'
${ noResults }
FFmpeg/libavcodec/x86/hpeldsp_init.c

259 lines
10.0KB
Raw Blame History 

  1. /*

  2.  * MMX optimized DSP utils

  3.  * Copyright (c) 2000, 2001 Fabrice Bellard

  4.  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>

  5.  *

  6.  * This file is part of Libav.

  7.  *

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

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

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

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

  12.  *

  13.  * Libav is distributed in the hope that it will be useful,

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

  15.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  16.  * Lesser General Public License for more details.

  17.  *

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

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

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

  21.  *

  22.  * MMX optimization by Nick Kurshev <nickols_k@mail.ru>

  23.  */

  24. 

  25. #include "libavutil/cpu.h"

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

  27. #include "libavcodec/hpeldsp.h"

  28. #include "dsputil_mmx.h"

  29. 

  30. void ff_put_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,

  31.                               ptrdiff_t line_size, int h);

  32. void ff_put_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels,

  33.                              ptrdiff_t line_size, int h);

  34. void ff_put_pixels16_x2_mmxext(uint8_t *block, const uint8_t *pixels,

  35.                                ptrdiff_t line_size, int h);

  36. void ff_put_pixels16_x2_3dnow(uint8_t *block, const uint8_t *pixels,

  37.                               ptrdiff_t line_size, int h);

  38. void ff_put_no_rnd_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,

  39.                                      ptrdiff_t line_size, int h);

  40. void ff_put_no_rnd_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels,

  41.                                     ptrdiff_t line_size, int h);

  42. void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block,

  43.                                            const uint8_t *pixels,

  44.                                            ptrdiff_t line_size, int h);

  45. void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block,

  46.                                           const uint8_t *pixels,

  47.                                           ptrdiff_t line_size, int h);

  48. void ff_put_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels,

  49.                               ptrdiff_t line_size, int h);

  50. void ff_put_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels,

  51.                              ptrdiff_t line_size, int h);

  52. void ff_put_no_rnd_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels,

  53.                                      ptrdiff_t line_size, int h);

  54. void ff_put_no_rnd_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels,

  55.                                     ptrdiff_t line_size, int h);

  56. void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block,

  57.                                            const uint8_t *pixels,

  58.                                            ptrdiff_t line_size, int h);

  59. void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block,

  60.                                           const uint8_t *pixels,

  61.                                           ptrdiff_t line_size, int h);

  62. void ff_avg_pixels8_3dnow(uint8_t *block, const uint8_t *pixels,

  63.                           ptrdiff_t line_size, int h);

  64. void ff_avg_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,

  65.                               ptrdiff_t line_size, int h);

  66. void ff_avg_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels,

  67.                              ptrdiff_t line_size, int h);

  68. void ff_avg_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels,

  69.                               ptrdiff_t line_size, int h);

  70. void ff_avg_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels,

  71.                              ptrdiff_t line_size, int h);

  72. void ff_avg_pixels8_xy2_mmxext(uint8_t *block, const uint8_t *pixels,

  73.                                ptrdiff_t line_size, int h);

  74. void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels,

  75.                               ptrdiff_t line_size, int h);

  76. 

  77. #define avg_pixels8_mmx         ff_avg_pixels8_mmx

  78. #define avg_pixels8_x2_mmx      ff_avg_pixels8_x2_mmx

  79. #define avg_pixels16_mmx        ff_avg_pixels16_mmx

  80. #define put_pixels8_mmx         ff_put_pixels8_mmx

  81. #define put_pixels16_mmx        ff_put_pixels16_mmx

  82. #define avg_no_rnd_pixels16_mmx ff_avg_pixels16_mmx

  83. #define put_no_rnd_pixels8_mmx  ff_put_pixels8_mmx

  84. #define put_no_rnd_pixels16_mmx ff_put_pixels16_mmx

  85. 

  86. #if HAVE_INLINE_ASM

  87. 

  88. /***********************************/

  89. /* MMX no rounding */

  90. #define DEF(x, y) x ## _no_rnd_ ## y ## _mmx

  91. #define SET_RND  MOVQ_WONE

  92. #define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX_NO_RND(a, b, c, d, e, f)

  93. #define PAVGB(a, b, c, e)               PAVGB_MMX_NO_RND(a, b, c, e)

  94. #define OP_AVG(a, b, c, e)              PAVGB_MMX(a, b, c, e)

  95. 

  96. #include "hpeldsp_rnd_template.c"

  97. 

  98. #undef DEF

  99. #undef SET_RND

  100. #undef PAVGBP

  101. #undef PAVGB

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

  103. /* MMX rounding */

  104. 

  105. #define DEF(x, y) x ## _ ## y ## _mmx

  106. #define SET_RND  MOVQ_WTWO

  107. #define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX(a, b, c, d, e, f)

  108. #define PAVGB(a, b, c, e)               PAVGB_MMX(a, b, c, e)

  109. 

  110. #include "hpeldsp_rnd_template.c"

  111. 

  112. #undef DEF

  113. #undef SET_RND

  114. #undef PAVGBP

  115. #undef PAVGB

  116. #undef OP_AVG

  117. 

  118. #endif /* HAVE_INLINE_ASM */

  119. 

  120. 

  121. #if HAVE_YASM

  122. /***********************************/

  123. /* 3Dnow specific */

  124. 

  125. #define DEF(x) x ## _3dnow

  126. 

  127. #include "hpeldsp_avg_template.c"

  128. 

  129. #undef DEF

  130. 

  131. /***********************************/

  132. /* MMXEXT specific */

  133. 

  134. #define DEF(x) x ## _mmxext

  135. 

  136. #include "hpeldsp_avg_template.c"

  137. 

  138. #undef DEF

  139. 

  140. #endif /* HAVE_YASM */

  141. 

  142. #define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU)                                     \

  143.     do {                                                                        \

  144.         c->PFX ## _pixels_tab IDX [0] = PFX ## _pixels ## SIZE ## _     ## CPU; \

  145.         c->PFX ## _pixels_tab IDX [1] = PFX ## _pixels ## SIZE ## _x2_  ## CPU; \

  146.         c->PFX ## _pixels_tab IDX [2] = PFX ## _pixels ## SIZE ## _y2_  ## CPU; \

  147.         c->PFX ## _pixels_tab IDX [3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \

  148.     } while (0)

  149. 

  150. static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int mm_flags)

  151. {

  152. #if HAVE_MMX_INLINE

  153.     SET_HPEL_FUNCS(put,        [0], 16, mmx);

  154.     SET_HPEL_FUNCS(put_no_rnd, [0], 16, mmx);

  155.     SET_HPEL_FUNCS(avg,        [0], 16, mmx);

  156.     SET_HPEL_FUNCS(avg_no_rnd,    , 16, mmx);

  157.     SET_HPEL_FUNCS(put,        [1],  8, mmx);

  158.     SET_HPEL_FUNCS(put_no_rnd, [1],  8, mmx);

  159.     SET_HPEL_FUNCS(avg,        [1],  8, mmx);

  160. #endif /* HAVE_MMX_INLINE */

  161. }

  162. 

  163. static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int mm_flags)

  164. {

  165. #if HAVE_MMXEXT_EXTERNAL

  166.     c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext;

  167.     c->put_pixels_tab[0][2] = put_pixels16_y2_mmxext;

  168. 

  169.     c->avg_pixels_tab[0][0] = avg_pixels16_mmxext;

  170.     c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmxext;

  171.     c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmxext;

  172. 

  173.     c->put_pixels_tab[1][1] = ff_put_pixels8_x2_mmxext;

  174.     c->put_pixels_tab[1][2] = ff_put_pixels8_y2_mmxext;

  175. 

  176.     c->avg_pixels_tab[1][0] = ff_avg_pixels8_mmxext;

  177.     c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_mmxext;

  178.     c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_mmxext;

  179. 

  180.     if (!(flags & CODEC_FLAG_BITEXACT)) {

  181.         c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmxext;

  182.         c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmxext;

  183.         c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_mmxext;

  184.         c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_mmxext;

  185. 

  186.         c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmxext;

  187.         c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_mmxext;

  188.     }

  189. 

  190.     if (flags & CODEC_FLAG_BITEXACT && CONFIG_VP3_DECODER) {

  191.         c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext;

  192.         c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext;

  193.     }

  194. #endif /* HAVE_MMXEXT_EXTERNAL */

  195. }

  196. 

  197. static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int mm_flags)

  198. {

  199. #if HAVE_AMD3DNOW_EXTERNAL

  200.     c->put_pixels_tab[0][1] = ff_put_pixels16_x2_3dnow;

  201.     c->put_pixels_tab[0][2] = put_pixels16_y2_3dnow;

  202. 

  203.     c->avg_pixels_tab[0][0] = avg_pixels16_3dnow;

  204.     c->avg_pixels_tab[0][1] = avg_pixels16_x2_3dnow;

  205.     c->avg_pixels_tab[0][2] = avg_pixels16_y2_3dnow;

  206. 

  207.     c->put_pixels_tab[1][1] = ff_put_pixels8_x2_3dnow;

  208.     c->put_pixels_tab[1][2] = ff_put_pixels8_y2_3dnow;

  209. 

  210.     c->avg_pixels_tab[1][0] = ff_avg_pixels8_3dnow;

  211.     c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_3dnow;

  212.     c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_3dnow;

  213. 

  214.     if (!(flags & CODEC_FLAG_BITEXACT)){

  215.         c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_3dnow;

  216.         c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_3dnow;

  217.         c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_3dnow;

  218.         c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_3dnow;

  219. 

  220.         c->avg_pixels_tab[0][3] = avg_pixels16_xy2_3dnow;

  221.         c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_3dnow;

  222.     }

  223. 

  224.     if (flags & CODEC_FLAG_BITEXACT && CONFIG_VP3_DECODER) {

  225.         c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow;

  226.         c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow;

  227.     }

  228. #endif /* HAVE_AMD3DNOW_EXTERNAL */

  229. }

  230. 

  231. static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int mm_flags)

  232. {

  233. #if HAVE_SSE2_EXTERNAL

  234.     if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {

  235.         // these functions are slower than mmx on AMD, but faster on Intel

  236.         c->put_pixels_tab[0][0]        = ff_put_pixels16_sse2;

  237.         c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2;

  238.         c->avg_pixels_tab[0][0]        = ff_avg_pixels16_sse2;

  239.     }

  240. #endif /* HAVE_SSE2_EXTERNAL */

  241. }

  242. 

  243. void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags)

  244. {

  245.     int mm_flags = av_get_cpu_flags();

  246. 

  247.     if (mm_flags & AV_CPU_FLAG_MMX)

  248.         hpeldsp_init_mmx(c, flags, mm_flags);

  249. 

  250.     if (mm_flags & AV_CPU_FLAG_MMXEXT)

  251.         hpeldsp_init_mmxext(c, flags, mm_flags);

  252. 

  253.     if (mm_flags & AV_CPU_FLAG_3DNOW)

  254.         hpeldsp_init_3dnow(c, flags, mm_flags);

  255. 

  256.     if (mm_flags & AV_CPU_FLAG_SSE2)

  257.         hpeldsp_init_sse2(c, flags, mm_flags);

  258. }