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.
661 Commits
32 Branches
325MB
Tree: 607dce96c0
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '607dce96c0'
${ noResults }
FFmpeg/libavcodec/i386/dsputil_mmx_avg.h

373 lines
11KB
Raw Blame History 

  1. /*

  2.  * DSP utils : average functions are compiled twice for 3dnow/mmx2

  3.  * Copyright (c) 2000, 2001 Gerard Lantau.

  4.  * Copyright (c) 2002 Michael Niedermayer

  5.  *

  6.  * This program is free software; you can redistribute it and/or modify

  7.  * it under the terms of the GNU General Public License as published by

  8.  * the Free Software Foundation; either version 2 of the License, or

  9.  * (at your option) any later version.

  10.  *

  11.  * This program 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

  14.  * GNU General Public License for more details.

  15.  *

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

  17.  * along with this program; if not, write to the Free Software

  18.  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

  19.  *

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

  21.  * mostly rewritten by Michael Niedermayer <michaelni@gmx.at>

  22.  */

  23. 

  24. static void DEF(put_pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  25. {

  26.     __asm __volatile(

  27.         "xorl %%eax, %%eax		\n\t"

  28.         ".balign 16			\n\t"

  29.         "1:				\n\t"

  30. 	"movq (%1, %%eax), %%mm0	\n\t"

  31. 	"movq 1(%1, %%eax), %%mm1	\n\t"

  32. 	"movq (%2, %%eax), %%mm2	\n\t"

  33. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  34. 	PAVGB" %%mm1, %%mm0		\n\t"

  35. 	PAVGB" %%mm3, %%mm2		\n\t"

  36. 	"movq %%mm0, (%3, %%eax)	\n\t"

  37. 	"movq %%mm2, (%4, %%eax)	\n\t"

  38.         "addl %5, %%eax			\n\t"

  39. 	"movq (%1, %%eax), %%mm0	\n\t"

  40. 	"movq 1(%1, %%eax), %%mm1	\n\t"

  41. 	"movq (%2, %%eax), %%mm2	\n\t"

  42. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  43. 	PAVGB" %%mm1, %%mm0		\n\t"

  44. 	PAVGB" %%mm3, %%mm2		\n\t"

  45. 	"movq %%mm0, (%3, %%eax)	\n\t"

  46. 	"movq %%mm2, (%4, %%eax)	\n\t"

  47.         "addl %5, %%eax			\n\t"

  48.         "subl $4, %0			\n\t"

  49.         " jnz 1b			\n\t"

  50. 	:"+g"(h)

  51.         :"r"(pixels), "r"(pixels+line_size), "r" (block), "r" (block+line_size),

  52.          "r"(line_size<<1)

  53. 	:"%eax", "memory");

  54. }

  55.  

  56. static void DEF(put_no_rnd_pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  57. {

  58.     __asm __volatile(

  59.         "xorl %%eax, %%eax		\n\t"

  60.         "movq "MANGLE(mm_bone)", %%mm7	\n\t"

  61.         ".balign 16			\n\t"

  62.         "1:				\n\t"

  63. 	"movq (%1, %%eax), %%mm0	\n\t"

  64. 	"movq 1(%1, %%eax), %%mm1	\n\t"

  65. 	"movq (%2, %%eax), %%mm2	\n\t"

  66. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  67.         "psubusb %%mm7, %%mm0		\n\t"

  68.         "psubusb %%mm7, %%mm2		\n\t"

  69. 	PAVGB" %%mm1, %%mm0		\n\t"

  70. 	PAVGB" %%mm3, %%mm2		\n\t"

  71. 	"movq %%mm0, (%3, %%eax)	\n\t"

  72. 	"movq %%mm2, (%4, %%eax)	\n\t"

  73.         "addl %5, %%eax			\n\t"

  74. 	"movq (%1, %%eax), %%mm0	\n\t"

  75. 	"movq 1(%1, %%eax), %%mm1	\n\t"

  76. 	"movq (%2, %%eax), %%mm2	\n\t"

  77. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  78.         "psubusb %%mm7, %%mm0		\n\t"

  79.         "psubusb %%mm7, %%mm2		\n\t"

  80. 	PAVGB" %%mm1, %%mm0		\n\t"

  81. 	PAVGB" %%mm3, %%mm2		\n\t"

  82. 	"movq %%mm0, (%3, %%eax)	\n\t"

  83. 	"movq %%mm2, (%4, %%eax)	\n\t"

  84.         "addl %5, %%eax			\n\t"

  85.         "subl $4, %0			\n\t"

  86.         " jnz 1b			\n\t"

  87. 	:"+g"(h)

  88.         :"r"(pixels), "r"(pixels+line_size), "r" (block), "r" (block+line_size),

  89.          "r"(line_size<<1)

  90. 	:"%eax", "memory");

  91. }

  92. 

  93. static void DEF(put_pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  94. {

  95.     __asm __volatile(

  96.         "xorl %%eax, %%eax		\n\t"

  97. 	"movq (%1), %%mm0		\n\t"

  98.         ".balign 16			\n\t"

  99.         "1:				\n\t"

  100. 	"movq (%2, %%eax), %%mm1	\n\t"

  101. 	"movq (%3, %%eax), %%mm2	\n\t"

  102. 	PAVGB" %%mm1, %%mm0		\n\t"

  103. 	PAVGB" %%mm2, %%mm1		\n\t"

  104. 	"movq %%mm0, (%4, %%eax)	\n\t"

  105. 	"movq %%mm1, (%5, %%eax)	\n\t"

  106.         "addl %6, %%eax			\n\t"

  107. 	"movq (%2, %%eax), %%mm1	\n\t"

  108. 	"movq (%3, %%eax), %%mm0	\n\t"

  109. 	PAVGB" %%mm1, %%mm2		\n\t"

  110. 	PAVGB" %%mm0, %%mm1		\n\t"

  111. 	"movq %%mm2, (%4, %%eax)	\n\t"

  112. 	"movq %%mm1, (%5, %%eax)	\n\t"

  113.         "addl %6, %%eax			\n\t"

  114.         "subl $4, %0			\n\t"

  115.         " jnz 1b			\n\t"

  116. 	:"+g"(h)

  117. 	:"r"(pixels), "r"(pixels+line_size), "r"(pixels+line_size*2), "r" (block), 

  118.          "r" (block+line_size), "g"(line_size<<1)

  119. 	:"%eax",  "memory");

  120. }

  121. 

  122. static void DEF(put_no_rnd_pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  123. {

  124.     __asm __volatile(

  125.         "movq "MANGLE(mm_bone)", %%mm7	\n\t"

  126.         "xorl %%eax, %%eax		\n\t"

  127. 	"movq (%1), %%mm0		\n\t"

  128.         ".balign 16			\n\t"

  129.         "1:				\n\t"

  130. 	"movq (%2, %%eax), %%mm1	\n\t"

  131. 	"movq (%3, %%eax), %%mm2	\n\t"

  132.         "psubusb %%mm7, %%mm1		\n\t"

  133. 	PAVGB" %%mm1, %%mm0		\n\t"

  134. 	PAVGB" %%mm2, %%mm1		\n\t"

  135. 	"movq %%mm0, (%4, %%eax)	\n\t"

  136. 	"movq %%mm1, (%5, %%eax)	\n\t"

  137.         "addl %6, %%eax			\n\t"

  138. 	"movq (%2, %%eax), %%mm1	\n\t"

  139. 	"movq (%3, %%eax), %%mm0	\n\t"

  140.         "psubusb %%mm7, %%mm1		\n\t"

  141. 	PAVGB" %%mm1, %%mm2		\n\t"

  142. 	PAVGB" %%mm0, %%mm1		\n\t"

  143. 	"movq %%mm2, (%4, %%eax)	\n\t"

  144. 	"movq %%mm1, (%5, %%eax)	\n\t"

  145.         "addl %6, %%eax			\n\t"

  146.         "subl $4, %0			\n\t"

  147.         " jnz 1b			\n\t"

  148. 	:"+g"(h)

  149. 	:"r"(pixels), "r"(pixels+line_size), "r"(pixels+line_size*2), "r" (block), 

  150.          "r" (block+line_size), "g"(line_size<<1)

  151. 	:"%eax",  "memory");

  152. }

  153. 

  154. static void DEF(avg_pixels)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  155. {

  156.     __asm __volatile(

  157.         "xorl %%eax, %%eax		\n\t"

  158.         ".balign 16			\n\t"

  159.         "1:				\n\t"

  160. 	"movq (%1, %%eax), %%mm0	\n\t"

  161. 	"movq (%2, %%eax), %%mm2	\n\t"

  162. 	"movq (%3, %%eax), %%mm3	\n\t"

  163. 	"movq (%4, %%eax), %%mm4	\n\t"

  164. 	PAVGB" %%mm3, %%mm0		\n\t"

  165. 	PAVGB" %%mm4, %%mm2		\n\t"

  166. 	"movq %%mm0, (%3, %%eax)	\n\t"

  167. 	"movq %%mm2, (%4, %%eax)	\n\t"

  168.         "addl %5, %%eax			\n\t"

  169. 	"movq (%1, %%eax), %%mm0	\n\t"

  170. 	"movq (%2, %%eax), %%mm2	\n\t"

  171. 	"movq (%3, %%eax), %%mm3	\n\t"

  172. 	"movq (%4, %%eax), %%mm4	\n\t"

  173. 	PAVGB" %%mm3, %%mm0		\n\t"

  174. 	PAVGB" %%mm4, %%mm2		\n\t"

  175. 	"movq %%mm0, (%3, %%eax)	\n\t"

  176. 	"movq %%mm2, (%4, %%eax)	\n\t"

  177.         "addl %5, %%eax			\n\t"

  178.         "subl $4, %0			\n\t"

  179.         " jnz 1b			\n\t"

  180. 	:"+g"(h)

  181.         :"r"(pixels), "r"(pixels+line_size), "r" (block), "r" (block+line_size),

  182.          "r"(line_size<<1)

  183. 	:"%eax", "memory");

  184. }

  185. 

  186. static void DEF(avg_pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  187. {

  188.     __asm __volatile(

  189.         "xorl %%eax, %%eax		\n\t"

  190.         ".balign 16			\n\t"

  191.         "1:				\n\t"

  192. 	"movq (%1, %%eax), %%mm0	\n\t"

  193. 	"movq 1(%1, %%eax), %%mm1	\n\t"

  194. 	"movq (%2, %%eax), %%mm2	\n\t"

  195. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  196. 	PAVGB" %%mm1, %%mm0		\n\t"

  197. 	PAVGB" %%mm3, %%mm2		\n\t"

  198. 	"movq (%3, %%eax), %%mm3	\n\t"

  199. 	"movq (%4, %%eax), %%mm4	\n\t"

  200. 	PAVGB" %%mm3, %%mm0		\n\t"

  201. 	PAVGB" %%mm4, %%mm2		\n\t"

  202. 	"movq %%mm0, (%3, %%eax)	\n\t"

  203. 	"movq %%mm2, (%4, %%eax)	\n\t"

  204.         "addl %5, %%eax			\n\t"

  205. 	"movq (%1, %%eax), %%mm0	\n\t"

  206. 	"movq 1(%1, %%eax), %%mm1	\n\t"

  207. 	"movq (%2, %%eax), %%mm2	\n\t"

  208. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  209. 	PAVGB" %%mm1, %%mm0		\n\t"

  210. 	PAVGB" %%mm3, %%mm2		\n\t"

  211. 	"movq (%3, %%eax), %%mm3	\n\t"

  212. 	"movq (%4, %%eax), %%mm4	\n\t"

  213. 	PAVGB" %%mm3, %%mm0		\n\t"

  214. 	PAVGB" %%mm4, %%mm2		\n\t"

  215. 	"movq %%mm0, (%3, %%eax)	\n\t"

  216. 	"movq %%mm2, (%4, %%eax)	\n\t"

  217.         "addl %5, %%eax			\n\t"

  218.         "subl $4, %0			\n\t"

  219.         " jnz 1b			\n\t"

  220. 	:"+g"(h)

  221.         :"r"(pixels), "r"(pixels+line_size), "r" (block), "r" (block+line_size),

  222.          "r"(line_size<<1)

  223. 	:"%eax", "memory");

  224. }

  225. 

  226. static void DEF(avg_pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  227. {

  228.     __asm __volatile(

  229.         "xorl %%eax, %%eax		\n\t"

  230. 	"movq (%1), %%mm0		\n\t"

  231.         ".balign 16			\n\t"

  232.         "1:				\n\t"

  233. 	"movq (%2, %%eax), %%mm1	\n\t"

  234. 	"movq (%3, %%eax), %%mm2	\n\t"

  235. 	PAVGB" %%mm1, %%mm0		\n\t"

  236. 	PAVGB" %%mm2, %%mm1		\n\t"

  237. 	"movq (%4, %%eax), %%mm3	\n\t"

  238. 	"movq (%5, %%eax), %%mm4	\n\t"

  239. 	PAVGB" %%mm3, %%mm0		\n\t"

  240. 	PAVGB" %%mm4, %%mm1		\n\t"

  241. 	"movq %%mm0, (%4, %%eax)	\n\t"

  242. 	"movq %%mm1, (%5, %%eax)	\n\t"

  243.         "addl %6, %%eax			\n\t"

  244. 	"movq (%2, %%eax), %%mm1	\n\t"

  245. 	"movq (%3, %%eax), %%mm0	\n\t"

  246. 	PAVGB" %%mm1, %%mm2		\n\t"

  247. 	PAVGB" %%mm0, %%mm1		\n\t"

  248. 	"movq (%4, %%eax), %%mm3	\n\t"

  249. 	"movq (%5, %%eax), %%mm4	\n\t"

  250. 	PAVGB" %%mm3, %%mm2		\n\t"

  251. 	PAVGB" %%mm4, %%mm1		\n\t"

  252. 	"movq %%mm2, (%4, %%eax)	\n\t"

  253. 	"movq %%mm1, (%5, %%eax)	\n\t"

  254.         "addl %6, %%eax			\n\t"

  255.         "subl $4, %0			\n\t"

  256.         " jnz 1b			\n\t"

  257. 	:"+g"(h)

  258. 	:"r"(pixels), "r"(pixels+line_size), "r"(pixels+line_size*2), "r" (block), 

  259.          "r" (block+line_size), "g"(line_size<<1)

  260. 	:"%eax",  "memory");

  261. }

  262. 

  263. // Note this is not correctly rounded, but this function is only used for b frames so it doesnt matter 

  264. static void DEF(avg_pixels_xy2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  265. {

  266.     __asm __volatile(

  267.         "movq "MANGLE(mm_bone)", %%mm7	\n\t"

  268.         "xorl %%eax, %%eax		\n\t"

  269. 	"movq (%1), %%mm0		\n\t"

  270. 	"movq 1(%1), %%mm1		\n\t"

  271. 	PAVGB" %%mm1, %%mm0		\n\t"

  272.         ".balign 16			\n\t"

  273.         "1:				\n\t"

  274. 	"movq (%2, %%eax), %%mm1	\n\t"

  275. 	"movq (%3, %%eax), %%mm2	\n\t"

  276. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  277. 	"movq 1(%3, %%eax), %%mm4	\n\t"

  278.         "psubusb %%mm7, %%mm2		\n\t"

  279. 	PAVGB" %%mm3, %%mm1		\n\t"

  280. 	PAVGB" %%mm4, %%mm2		\n\t"

  281. 	PAVGB" %%mm1, %%mm0		\n\t"

  282. 	PAVGB" %%mm2, %%mm1		\n\t"

  283. 	"movq (%4, %%eax), %%mm3	\n\t"

  284. 	"movq (%5, %%eax), %%mm4	\n\t"

  285. 	PAVGB" %%mm3, %%mm0		\n\t"

  286. 	PAVGB" %%mm4, %%mm1		\n\t"

  287. 	"movq %%mm0, (%4, %%eax)	\n\t"

  288. 	"movq %%mm1, (%5, %%eax)	\n\t"

  289.         "addl %6, %%eax			\n\t"

  290. 	"movq (%2, %%eax), %%mm1	\n\t"

  291. 	"movq (%3, %%eax), %%mm0	\n\t"

  292. 	"movq 1(%2, %%eax), %%mm3	\n\t"

  293. 	"movq 1(%3, %%eax), %%mm4	\n\t"

  294. 	PAVGB" %%mm3, %%mm1		\n\t"

  295. 	PAVGB" %%mm4, %%mm0		\n\t"

  296. 	PAVGB" %%mm1, %%mm2		\n\t"

  297. 	PAVGB" %%mm0, %%mm1		\n\t"

  298. 	"movq (%4, %%eax), %%mm3	\n\t"

  299. 	"movq (%5, %%eax), %%mm4	\n\t"

  300. 	PAVGB" %%mm3, %%mm2		\n\t"

  301. 	PAVGB" %%mm4, %%mm1		\n\t"

  302. 	"movq %%mm2, (%4, %%eax)	\n\t"

  303. 	"movq %%mm1, (%5, %%eax)	\n\t"

  304.         "addl %6, %%eax			\n\t"

  305.         "subl $4, %0			\n\t"

  306.         " jnz 1b			\n\t"

  307. 	:"+g"(h)

  308. 	:"r"(pixels), "r"(pixels+line_size), "r"(pixels+line_size*2), "r" (block), 

  309.          "r" (block+line_size), "g"(line_size<<1)

  310. 	:"%eax",  "memory");

  311. }

  312. 

  313. //Note: the sub* functions are no used 

  314. 

  315. static void DEF(sub_pixels_x2)( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)

  316. {

  317.   DCTELEM  *p;

  318.   const UINT8 *pix;

  319.   p = block;

  320.   pix = pixels;

  321.   __asm __volatile(

  322.       "pxor	%%mm7, %%mm7":);

  323.   do {

  324.     __asm __volatile(

  325. 	"movq	1%1, %%mm2\n\t"

  326. 	"movq	%0, %%mm0\n\t"

  327. 	PAVGB"	%1, %%mm2\n\t"

  328. 	"movq	8%0, %%mm1\n\t"

  329. 	"movq	%%mm2, %%mm3\n\t"

  330. 	"punpcklbw %%mm7, %%mm2\n\t"

  331. 	"punpckhbw %%mm7, %%mm3\n\t"

  332. 	"psubsw %%mm2, %%mm0\n\t"

  333. 	"psubsw %%mm3, %%mm1\n\t"

  334. 	"movq	%%mm0, %0\n\t"

  335. 	"movq	%%mm1, 8%0\n\t"

  336. 	:"+m"(*p)

  337. 	:"m"(*pix)

  338. 	:"memory");

  339.    pix += line_size;

  340.    p +=   8;

  341.  } while (--h);

  342. }

  343. 

  344. static void DEF(sub_pixels_y2)( DCTELEM  *block, const UINT8 *pixels, int line_size, int h)

  345. {

  346.   DCTELEM  *p;

  347.   const UINT8 *pix;

  348.   p = block;

  349.   pix = pixels;

  350.   __asm __volatile(

  351.       "pxor	%%mm7, %%mm7":);

  352.   do {

  353.     __asm __volatile(

  354. 	"movq	%2, %%mm2\n\t"

  355. 	"movq	%0, %%mm0\n\t"

  356. 	PAVGB"	%1, %%mm2\n\t"

  357. 	"movq	8%0, %%mm1\n\t"

  358. 	"movq	%%mm2, %%mm3\n\t"

  359. 	"punpcklbw %%mm7, %%mm2\n\t"

  360. 	"punpckhbw %%mm7, %%mm3\n\t"

  361. 	"psubsw %%mm2, %%mm0\n\t"

  362. 	"psubsw %%mm3, %%mm1\n\t"

  363. 	"movq	%%mm0, %0\n\t"

  364. 	"movq	%%mm1, 8%0\n\t"

  365. 	:"+m"(*p)

  366. 	:"m"(*pix), "m"(*(pix+line_size))

  367. 	:"memory");

  368.    pix += line_size;

  369.    p +=   8;

  370.  } while (--h);

  371. }