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.
735 Commits
32 Branches
850MB
Tree: 789bee1264
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '789bee1264'
${ noResults }
FFmpeg/libavcodec/i386/dsputil_mmx_avg.h

352 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. /* XXX: we use explicit registers to avoid a gcc 2.95.2 register asm

  25.    clobber bug - now it will work with 2.95.2 and also with -fPIC

  26.  */

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

  28. {

  29.     __asm __volatile(

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

  31.         ".balign 16			\n\t"

  32.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  52.         " jnz 1b			\n\t"

  53. 	:"+g"(h)

  54.         :"D"(pixels), "S"(pixels+line_size), "r" (block), "r" (block+line_size),

  55.         "g"(line_size<<1)

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

  57. }

  58.  

  59. /* GL: this function does incorrect rounding if overflow */

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

  61. {

  62.     __asm __volatile(

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

  64. 	MOVQ_BONE(%%mm7)

  65.         ".balign 16			\n\t"

  66.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  90.         " jnz 1b			\n\t"

  91. 	:"+g"(h)

  92.         :"D"(pixels), "S"(pixels+line_size), "r" (block), "r" (block+line_size),

  93.         "r"(line_size<<1)

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

  95. }

  96. 

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

  98. {

  99. #if 1

  100.     // Michael - measure me

  101.     __asm __volatile(

  102.         "lea (%3, %3), %%eax            \n\t"

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

  104.         "subl %3, %2			\n\t"

  105.         ".balign 16			\n\t"

  106.         "1:				\n\t"

  107. 	"movq (%1, %3), %%mm1		\n\t"

  108. 	"movq (%1, %%eax), %%mm2	\n\t"

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

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

  111.         "addl %%eax, %1			\n\t"

  112. 	"movq %%mm0, (%2, %3)		\n\t"

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

  114. 	"movq (%1, %3), %%mm1		\n\t"

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

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

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

  118.         "addl %%eax, %2			\n\t"

  119.         "addl %%eax, %1			\n\t"

  120. 	"movq %%mm2, (%2, %3)		\n\t"

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

  122.         "addl %%eax, %2			\n\t"

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

  124.         "jnz 1b				\n\t"

  125.         :"+g"(h), "+D"(pixels), "+S" (block)

  126. 	:"c"(line_size)

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

  128. #else

  129.    // kabi measure me

  130.     __asm __volatile(

  131. 	"movq (%2), %%mm0		\n\t"

  132.         "addl %1, %2			\n\t"

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

  134.         "leal (%1, %2), %%edi		\n\t"

  135.         "leal (%1, %3), %%esi		\n\t"

  136.         "addl %1, %1			\n\t"

  137.         ".balign 16			\n\t"

  138.         "1:				\n\t"

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

  140. 	"movq (%%edi, %%eax), %%mm2	\n\t"

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

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

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

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

  145.         "addl %1, %%eax			\n\t"

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

  147. 	"movq (%%edi, %%eax), %%mm0	\n\t"

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

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

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

  151. 	"movq %%mm1, (%%esi, %%eax)	\n\t"

  152.         "addl %1, %%eax			\n\t"

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

  154.         " jnz 1b			\n\t"

  155. 	:"+g"(h), "+r"(line_size), "+r"(pixels)

  156. 	: "r" (block)

  157. 	: "%eax", "%esi", "%edi", "memory");

  158. #endif

  159. }

  160. 

  161. /* GL: this function does incorrect rounding if overflow */

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

  163. {

  164.     __asm __volatile(

  165. 	MOVQ_BONE(%%mm7)

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

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

  168.         ".balign 16			\n\t"

  169.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  187.         " jnz 1b			\n\t"

  188. 	:"+g"(h)

  189. 	:"D"(pixels), "S"(pixels+line_size), "r"(pixels+line_size*2), "r" (block),

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

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

  192. }

  193. 

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

  195. {

  196.     __asm __volatile(

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

  198.         ".balign 16			\n\t"

  199.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

  210. 	"movq (%2, %%eax), %%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.         :"D"(pixels), "S"(pixels+line_size), "r" (block), "r" (block+line_size),

  222.          "g"(line_size<<1)

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

  224. }

  225. 

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

  227. {

  228.     __asm __volatile(

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

  230.         ".balign 16			\n\t"

  231.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  259.         " jnz 1b			\n\t"

  260. 	:"+g"(h)

  261.         :"D"(pixels), "S"(pixels+line_size), "r" (block), "r" (block+line_size),

  262.          "g"(line_size<<1)

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

  264. }

  265. 

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

  267. {

  268.     __asm __volatile(

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

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

  271.         ".balign 16			\n\t"

  272.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  296.         " jnz 1b			\n\t"

  297. 	:"+g"(h)

  298. 	:"D"(pixels), "S"(pixels+line_size), "r"(pixels+line_size*2), "r" (block),

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

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

  301. }

  302. 

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

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

  305. {

  306.     __asm __volatile(

  307. 	MOVQ_BONE(%%mm7)

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

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

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

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

  312.         ".balign 16			\n\t"

  313.         "1:				\n\t"

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  346.         " jnz 1b			\n\t"

  347. 	:"+g"(h)

  348. 	:"D"(pixels), "S"(pixels+line_size), "r"(pixels+line_size*2), "r" (block),

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

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

  351. }