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.
1018 Commits
32 Branches
388MB
Tree: a10678da8a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a10678da8a'
${ noResults }
FFmpeg/libavcodec/i386/dsputil_mmx_rnd.h

306 lines
8.4KB
Raw Blame History 

  1. /*

  2.  * DSP utils mmx functions are compiled twice for rnd/no_rnd

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

  4.  *

  5.  * This library is free software; you can redistribute it and/or

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

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

  8.  * version 2 of the License, or (at your option) any later version.

  9.  *

  10.  * This library is distributed in the hope that it will be useful,

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

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

  13.  * Lesser General Public License for more details.

  14.  *

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

  16.  * License along with this library; if not, write to the Free Software

  17.  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

  18.  *

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

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

  21.  * and improved by Zdenek Kabelac <kabi@users.sf.net>

  22.  */

  23. 

  24. // put_pixels

  25. static void DEF(put, pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  26. {

  27.     MOVQ_BFE(mm6);

  28.     __asm __volatile(

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

  30. 	".balign 8			\n\t"

  31. 	"1:				\n\t"

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

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

  34. 	"movq	(%1, %3), %%mm2		\n\t"

  35. 	"movq	1(%1, %3), %%mm3	\n\t"

  36. 	PAVGBP(%%mm0, %%mm1, %%mm4,   %%mm2, %%mm3, %%mm5)

  37. 	"movq	%%mm4, (%2)		\n\t"

  38. 	"movq	%%mm5, (%2, %3)		\n\t"

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

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

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

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

  43. 	"movq	(%1, %3), %%mm2		\n\t"

  44. 	"movq	1(%1, %3), %%mm3	\n\t"

  45. 	PAVGBP(%%mm0, %%mm1, %%mm4,   %%mm2, %%mm3, %%mm5)

  46. 	"movq	%%mm4, (%2)		\n\t"

  47. 	"movq	%%mm5, (%2, %3)		\n\t"

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

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

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

  51. 	"jnz	1b			\n\t"

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

  53. 	:"r"(line_size)

  54. 	:"eax", "memory");

  55. }

  56. 

  57. static void DEF(put, pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  58. {

  59.     MOVQ_BFE(mm6);

  60.     __asm __volatile(

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

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

  63. 	".balign 8			\n\t"

  64. 	"1:				\n\t"

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

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

  67. 	PAVGBP(%%mm1, %%mm0, %%mm4,   %%mm2, %%mm1, %%mm5)

  68. 	"movq	%%mm4, (%2)		\n\t"

  69. 	"movq	%%mm5, (%2, %3)		\n\t"

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

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

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

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

  74. 	PAVGBP(%%mm1, %%mm2, %%mm4,   %%mm0, %%mm1, %%mm5)

  75. 	"movq	%%mm4, (%2)		\n\t"

  76. 	"movq	%%mm5, (%2, %3)		\n\t"

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

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

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

  80. 	"jnz	1b			\n\t"

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

  82. 	:"r"(line_size)

  83. 	:"eax", "memory");

  84. }

  85. 

  86. static void DEF(put, pixels_xy2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  87. {

  88.     MOVQ_ZERO(mm7);

  89.     SET_RND(mm6); // =2 for rnd  and  =1 for no_rnd version

  90.     __asm __volatile(

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

  92. 	"movq	1(%1), %%mm4		\n\t"

  93. 	"movq	%%mm0, %%mm1		\n\t"

  94. 	"movq	%%mm4, %%mm5		\n\t"

  95. 	"punpcklbw %%mm7, %%mm0		\n\t"

  96. 	"punpcklbw %%mm7, %%mm4		\n\t"

  97. 	"punpckhbw %%mm7, %%mm1		\n\t"

  98. 	"punpckhbw %%mm7, %%mm5		\n\t"

  99. 	"paddusw %%mm0, %%mm4		\n\t"

  100. 	"paddusw %%mm1, %%mm5		\n\t"

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

  102. 	"addl	%3, %1			\n\t"

  103. 	".balign 8      		\n\t"

  104. 	"1:				\n\t"

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

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

  107. 	"movq	%%mm0, %%mm1		\n\t"

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

  109. 	"punpcklbw %%mm7, %%mm0		\n\t"

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

  111. 	"punpckhbw %%mm7, %%mm1		\n\t"

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

  113. 	"paddusw %%mm2, %%mm0	 	\n\t"

  114. 	"paddusw %%mm3, %%mm1		\n\t"

  115. 	"paddusw %%mm6, %%mm4		\n\t"

  116. 	"paddusw %%mm6, %%mm5		\n\t"

  117. 	"paddusw %%mm0, %%mm4		\n\t"

  118. 	"paddusw %%mm1, %%mm5		\n\t"

  119. 	"psrlw	$2, %%mm4		\n\t"

  120. 	"psrlw	$2, %%mm5		\n\t"

  121. 	"packuswb  %%mm5, %%mm4		\n\t"

  122. 	"movq	%%mm4, (%2, %%eax)	\n\t"

  123. 	"addl	%3, %%eax		\n\t"

  124. 

  125. 	"movq	(%1, %%eax), %%mm2	\n\t" // 0 <-> 2   1 <-> 3

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

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

  128. 	"movq	%%mm4, %%mm5		\n\t"

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

  130. 	"punpcklbw %%mm7, %%mm4		\n\t"

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

  132. 	"punpckhbw %%mm7, %%mm5		\n\t"

  133. 	"paddusw %%mm2, %%mm4	 	\n\t"

  134. 	"paddusw %%mm3, %%mm5		\n\t"

  135. 	"paddusw %%mm6, %%mm0		\n\t"

  136. 	"paddusw %%mm6, %%mm1		\n\t"

  137. 	"paddusw %%mm4, %%mm0		\n\t"

  138. 	"paddusw %%mm5, %%mm1		\n\t"

  139. 	"psrlw	$2, %%mm0		\n\t"

  140. 	"psrlw	$2, %%mm1		\n\t"

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

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

  143. 	"addl	%3, %%eax		\n\t"

  144. 

  145. 	"subl	$2, %0			\n\t"

  146. 	"jnz	1b			\n\t"

  147. 	:"+g"(h), "+S"(pixels)

  148. 	:"D"(block), "r"(line_size)

  149. 	:"eax", "memory");

  150. }

  151. 

  152. // avg_pixels

  153. // in case more speed is needed - unroling would certainly help

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

  155. {

  156.     MOVQ_BFE(mm6);

  157.     JUMPALIGN();

  158.     do {

  159. 	__asm __volatile(

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

  161. 	     "movq  %1, %%mm1		\n\t"

  162. 	     PAVGB(%%mm0, %%mm1, %%mm2, %%mm6)

  163. 	     "movq  %%mm2, %0		\n\t"

  164. 	     :"+m"(*block)

  165. 	     :"m"(*pixels)

  166. 	     :"memory");

  167. 	pixels += line_size;

  168. 	block += line_size;

  169.     }

  170.     while (--h);

  171. }

  172. 

  173. static void DEF(avg, pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  174. {

  175.     MOVQ_BFE(mm6);

  176.     JUMPALIGN();

  177.     do {

  178. 	__asm __volatile(

  179. 	    "movq  %1, %%mm0		\n\t"

  180. 	    "movq  1%1, %%mm1		\n\t"

  181. 	    "movq  %0, %%mm3		\n\t"

  182. 	    PAVGB(%%mm0, %%mm1, %%mm2, %%mm6)

  183. 	    PAVGB(%%mm3, %%mm2, %%mm0, %%mm6)

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

  185. 	    :"+m"(*block)

  186. 	    :"m"(*pixels)

  187. 	    :"memory");

  188. 	pixels += line_size;

  189. 	block += line_size;

  190.     } while (--h);

  191. }

  192. 

  193. static void DEF(avg, pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  194. {

  195.     MOVQ_BFE(mm6);

  196.     __asm __volatile(

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

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

  199. 	".balign 8			\n\t"

  200. 	"1:				\n\t"

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

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

  203. 	PAVGBP(%%mm1, %%mm0, %%mm4,   %%mm2, %%mm1, %%mm5)

  204. 	"movq	(%2), %%mm3		\n\t"

  205. 	PAVGB(%%mm3, %%mm4, %%mm0, %%mm6)

  206. 	"movq	(%2, %3), %%mm3		\n\t"

  207. 	PAVGB(%%mm3, %%mm5, %%mm1, %%mm6)

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

  209. 	"movq	%%mm1, (%2, %3)		\n\t"

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

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

  212. 

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

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

  215. 	PAVGBP(%%mm1, %%mm2, %%mm4,   %%mm0, %%mm1, %%mm5)

  216. 	"movq	(%2), %%mm3		\n\t"

  217. 	PAVGB(%%mm3, %%mm4, %%mm0, %%mm6)

  218. 	"movq	(%2, %3), %%mm3		\n\t"

  219. 	PAVGB(%%mm3, %%mm5, %%mm1, %%mm6)

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

  221. 	"movq	%%mm1, (%2, %3)		\n\t"

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

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

  224. 

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

  226. 	"jnz	1b			\n\t"

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

  228. 	:"r"(line_size)

  229. 	:"eax", "memory");

  230. }

  231. 

  232. // this routine is 'slightly' suboptimal but mostly unused

  233. static void DEF(avg, pixels_xy2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)

  234. {

  235.     MOVQ_ZERO(mm7);

  236.     SET_RND(mm6); // =2 for rnd  and  =1 for no_rnd version

  237.     __asm __volatile(

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

  239. 	"movq	1(%1), %%mm4		\n\t"

  240. 	"movq	%%mm0, %%mm1		\n\t"

  241. 	"movq	%%mm4, %%mm5		\n\t"

  242. 	"punpcklbw %%mm7, %%mm0		\n\t"

  243. 	"punpcklbw %%mm7, %%mm4		\n\t"

  244. 	"punpckhbw %%mm7, %%mm1		\n\t"

  245. 	"punpckhbw %%mm7, %%mm5		\n\t"

  246. 	"paddusw %%mm0, %%mm4		\n\t"

  247. 	"paddusw %%mm1, %%mm5		\n\t"

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

  249. 	"addl	%3, %1			\n\t"

  250. 	".balign 8			\n\t"

  251. 	"1:				\n\t"

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

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

  254. 	"movq	%%mm0, %%mm1		\n\t"

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

  256. 	"punpcklbw %%mm7, %%mm0		\n\t"

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

  258. 	"punpckhbw %%mm7, %%mm1		\n\t"

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

  260. 	"paddusw %%mm2, %%mm0	 	\n\t"

  261. 	"paddusw %%mm3, %%mm1		\n\t"

  262. 	"paddusw %%mm6, %%mm4		\n\t"

  263. 	"paddusw %%mm6, %%mm5		\n\t"

  264. 	"paddusw %%mm0, %%mm4		\n\t"

  265. 	"paddusw %%mm1, %%mm5		\n\t"

  266. 	"psrlw	$2, %%mm4		\n\t"

  267. 	"psrlw	$2, %%mm5		\n\t"

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

  269. 	"packuswb  %%mm5, %%mm4		\n\t"

  270. 		"pcmpeqd %%mm2, %%mm2	\n\t"

  271. 		"paddb %%mm2, %%mm2	\n\t"

  272. 		PAVGB(%%mm3, %%mm4, %%mm5, %%mm2)

  273. 		"movq	%%mm5, (%2, %%eax)	\n\t"

  274. 	"addl	%3, %%eax		\n\t"

  275. 

  276. 	"movq	(%1, %%eax), %%mm2	\n\t" // 0 <-> 2   1 <-> 3

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

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

  279. 	"movq	%%mm4, %%mm5		\n\t"

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

  281. 	"punpcklbw %%mm7, %%mm4		\n\t"

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

  283. 	"punpckhbw %%mm7, %%mm5		\n\t"

  284. 	"paddusw %%mm2, %%mm4	 	\n\t"

  285. 	"paddusw %%mm3, %%mm5		\n\t"

  286. 	"paddusw %%mm6, %%mm0		\n\t"

  287. 	"paddusw %%mm6, %%mm1		\n\t"

  288. 	"paddusw %%mm4, %%mm0		\n\t"

  289. 	"paddusw %%mm5, %%mm1		\n\t"

  290. 	"psrlw	$2, %%mm0		\n\t"

  291. 	"psrlw	$2, %%mm1		\n\t"

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

  293. 	"packuswb  %%mm1, %%mm0		\n\t"

  294. 		"pcmpeqd %%mm2, %%mm2	\n\t"

  295. 		"paddb %%mm2, %%mm2	\n\t"

  296. 		PAVGB(%%mm3, %%mm0, %%mm1, %%mm2)

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

  298. 	"addl	%3, %%eax		\n\t"

  299. 

  300. 	"subl	$2, %0			\n\t"

  301. 	"jnz	1b			\n\t"

  302. 	:"+g"(h), "+S"(pixels)

  303. 	:"D"(block), "r"(line_size)

  304. 	:"eax", "memory");

  305. }