Mercurial > mplayer.hg

comparison postproc/swscale_template.c @ 18104:7b408d60de9e

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
add support for intel mac. mp3lib is not fixed yet.
author nplourde
date Sat, 15 Apr 2006 20:46:54 +0000
parents fbf94ea858f1
children b10d4b3cb9ec
comparison
equal deleted inserted replaced
18103:26ea12332e67 18104:7b408d60de9e
13 13
14 You should have received a copy of the GNU General Public License 14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software 15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */ 17 */
18
19 #include "asmalign.h"
18 20
19 #undef REAL_MOVNTQ 21 #undef REAL_MOVNTQ
20 #undef MOVNTQ 22 #undef MOVNTQ
21 #undef PAVGB 23 #undef PAVGB
22 #undef PREFETCH 24 #undef PREFETCH
69 "xor %%"REG_a", %%"REG_a" \n\t"\ 71 "xor %%"REG_a", %%"REG_a" \n\t"\
70 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ 72 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\
71 "movq %%mm3, %%mm4 \n\t"\ 73 "movq %%mm3, %%mm4 \n\t"\
72 "lea " offset "(%0), %%"REG_d" \n\t"\ 74 "lea " offset "(%0), %%"REG_d" \n\t"\
73 "mov (%%"REG_d"), %%"REG_S" \n\t"\ 75 "mov (%%"REG_d"), %%"REG_S" \n\t"\
74 ".balign 16 \n\t" /* FIXME Unroll? */\ 76 ASMALIGN16 /* FIXME Unroll? */\
75 "1: \n\t"\ 77 "1: \n\t"\
76 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ 78 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
77 "movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\ 79 "movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\
78 "movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm5\n\t" /* srcData */\ 80 "movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm5\n\t" /* srcData */\
79 "add $16, %%"REG_d" \n\t"\ 81 "add $16, %%"REG_d" \n\t"\
96 "mov (%%"REG_d"), %%"REG_S" \n\t"\ 98 "mov (%%"REG_d"), %%"REG_S" \n\t"\
97 "jb 1b \n\t" 99 "jb 1b \n\t"
98 100
99 #define YSCALEYUV2YV121 \ 101 #define YSCALEYUV2YV121 \
100 "mov %2, %%"REG_a" \n\t"\ 102 "mov %2, %%"REG_a" \n\t"\
101 ".balign 16 \n\t" /* FIXME Unroll? */\ 103 ASMALIGN16 /* FIXME Unroll? */\
102 "1: \n\t"\ 104 "1: \n\t"\
103 "movq (%0, %%"REG_a", 2), %%mm0 \n\t"\ 105 "movq (%0, %%"REG_a", 2), %%mm0 \n\t"\
104 "movq 8(%0, %%"REG_a", 2), %%mm1\n\t"\ 106 "movq 8(%0, %%"REG_a", 2), %%mm1\n\t"\
105 "psraw $7, %%mm0 \n\t"\ 107 "psraw $7, %%mm0 \n\t"\
106 "psraw $7, %%mm1 \n\t"\ 108 "psraw $7, %%mm1 \n\t"\
116 "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize) 118 "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
117 : "%eax", "%ebx", "%ecx", "%edx", "%esi" 119 : "%eax", "%ebx", "%ecx", "%edx", "%esi"
118 */ 120 */
119 #define YSCALEYUV2PACKEDX \ 121 #define YSCALEYUV2PACKEDX \
120 "xor %%"REG_a", %%"REG_a" \n\t"\ 122 "xor %%"REG_a", %%"REG_a" \n\t"\
121 ".balign 16 \n\t"\ 123 ASMALIGN16\
122 "nop \n\t"\ 124 "nop \n\t"\
123 "1: \n\t"\ 125 "1: \n\t"\
124 "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ 126 "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\
125 "mov (%%"REG_d"), %%"REG_S" \n\t"\ 127 "mov (%%"REG_d"), %%"REG_S" \n\t"\
126 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ 128 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\
127 "movq %%mm3, %%mm4 \n\t"\ 129 "movq %%mm3, %%mm4 \n\t"\
128 ".balign 16 \n\t"\ 130 ASMALIGN16\
129 "2: \n\t"\ 131 "2: \n\t"\
130 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ 132 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
131 "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\ 133 "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
132 "movq 4096(%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\ 134 "movq 4096(%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\
133 "add $16, %%"REG_d" \n\t"\ 135 "add $16, %%"REG_d" \n\t"\
141 \ 143 \
142 "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ 144 "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\
143 "mov (%%"REG_d"), %%"REG_S" \n\t"\ 145 "mov (%%"REG_d"), %%"REG_S" \n\t"\
144 "movq "VROUNDER_OFFSET"(%0), %%mm1\n\t"\ 146 "movq "VROUNDER_OFFSET"(%0), %%mm1\n\t"\
145 "movq %%mm1, %%mm7 \n\t"\ 147 "movq %%mm1, %%mm7 \n\t"\
146 ".balign 16 \n\t"\ 148 ASMALIGN16\
147 "2: \n\t"\ 149 "2: \n\t"\
148 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ 150 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
149 "movq (%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y1srcData */\ 151 "movq (%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y1srcData */\
150 "movq 8(%%"REG_S", %%"REG_a", 2), %%mm5 \n\t" /* Y2srcData */\ 152 "movq 8(%%"REG_S", %%"REG_a", 2), %%mm5 \n\t" /* Y2srcData */\
151 "add $16, %%"REG_d" \n\t"\ 153 "add $16, %%"REG_d" \n\t"\
203 "punpcklwd %%mm6, %%mm6 \n\t"\ 205 "punpcklwd %%mm6, %%mm6 \n\t"\
204 "movd %7, %%mm5 \n\t" /*uvalpha1*/\ 206 "movd %7, %%mm5 \n\t" /*uvalpha1*/\
205 "punpcklwd %%mm5, %%mm5 \n\t"\ 207 "punpcklwd %%mm5, %%mm5 \n\t"\
206 "punpcklwd %%mm5, %%mm5 \n\t"\ 208 "punpcklwd %%mm5, %%mm5 \n\t"\
207 "xor %%"REG_a", %%"REG_a" \n\t"\ 209 "xor %%"REG_a", %%"REG_a" \n\t"\
208 ".balign 16 \n\t"\ 210 ASMALIGN16\
209 "1: \n\t"\ 211 "1: \n\t"\
210 "movq (%0, %%"REG_a", 2), %%mm0 \n\t" /*buf0[eax]*/\ 212 "movq (%0, %%"REG_a", 2), %%mm0 \n\t" /*buf0[eax]*/\
211 "movq (%1, %%"REG_a", 2), %%mm1 \n\t" /*buf1[eax]*/\ 213 "movq (%1, %%"REG_a", 2), %%mm1 \n\t" /*buf1[eax]*/\
212 "movq (%2, %%"REG_a",2), %%mm2 \n\t" /* uvbuf0[eax]*/\ 214 "movq (%2, %%"REG_a",2), %%mm2 \n\t" /* uvbuf0[eax]*/\
213 "movq (%3, %%"REG_a",2), %%mm3 \n\t" /* uvbuf1[eax]*/\ 215 "movq (%3, %%"REG_a",2), %%mm3 \n\t" /* uvbuf1[eax]*/\
256 "psraw $3, %%mm0 \n\t"\ 258 "psraw $3, %%mm0 \n\t"\
257 "psraw $3, %%mm1 \n\t"\ 259 "psraw $3, %%mm1 \n\t"\
258 "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c")\n\t"\ 260 "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c")\n\t"\
259 "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c")\n\t"\ 261 "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c")\n\t"\
260 "xor "#index", "#index" \n\t"\ 262 "xor "#index", "#index" \n\t"\
261 ".balign 16 \n\t"\ 263 ASMALIGN16\
262 "1: \n\t"\ 264 "1: \n\t"\
263 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ 265 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
264 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ 266 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
265 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ 267 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
266 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ 268 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
288 290
289 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c) 291 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c)
290 292
291 #define REAL_YSCALEYUV2RGB(index, c) \ 293 #define REAL_YSCALEYUV2RGB(index, c) \
292 "xor "#index", "#index" \n\t"\ 294 "xor "#index", "#index" \n\t"\
293 ".balign 16 \n\t"\ 295 ASMALIGN16\
294 "1: \n\t"\ 296 "1: \n\t"\
295 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ 297 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
296 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ 298 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
297 "movq 4096(%2, "#index"), %%mm5\n\t" /* uvbuf0[eax+2048]*/\ 299 "movq 4096(%2, "#index"), %%mm5\n\t" /* uvbuf0[eax+2048]*/\
298 "movq 4096(%3, "#index"), %%mm4\n\t" /* uvbuf1[eax+2048]*/\ 300 "movq 4096(%3, "#index"), %%mm4\n\t" /* uvbuf1[eax+2048]*/\
354 "pxor %%mm7, %%mm7 \n\t" 356 "pxor %%mm7, %%mm7 \n\t"
355 #define YSCALEYUV2RGB(index, c) REAL_YSCALEYUV2RGB(index, c) 357 #define YSCALEYUV2RGB(index, c) REAL_YSCALEYUV2RGB(index, c)
356 358
357 #define REAL_YSCALEYUV2PACKED1(index, c) \ 359 #define REAL_YSCALEYUV2PACKED1(index, c) \
358 "xor "#index", "#index" \n\t"\ 360 "xor "#index", "#index" \n\t"\
359 ".balign 16 \n\t"\ 361 ASMALIGN16\
360 "1: \n\t"\ 362 "1: \n\t"\
361 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\ 363 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
362 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ 364 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
363 "psraw $7, %%mm3 \n\t" \ 365 "psraw $7, %%mm3 \n\t" \
364 "psraw $7, %%mm4 \n\t" \ 366 "psraw $7, %%mm4 \n\t" \
369 371
370 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c) 372 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c)
371 373
372 #define REAL_YSCALEYUV2RGB1(index, c) \ 374 #define REAL_YSCALEYUV2RGB1(index, c) \
373 "xor "#index", "#index" \n\t"\ 375 "xor "#index", "#index" \n\t"\
374 ".balign 16 \n\t"\ 376 ASMALIGN16\
375 "1: \n\t"\ 377 "1: \n\t"\
376 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\ 378 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
377 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ 379 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
378 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ 380 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
379 "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ 381 "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
418 "pxor %%mm7, %%mm7 \n\t" 420 "pxor %%mm7, %%mm7 \n\t"
419 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c) 421 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c)
420 422
421 #define REAL_YSCALEYUV2PACKED1b(index, c) \ 423 #define REAL_YSCALEYUV2PACKED1b(index, c) \
422 "xor "#index", "#index" \n\t"\ 424 "xor "#index", "#index" \n\t"\
423 ".balign 16 \n\t"\ 425 ASMALIGN16\
424 "1: \n\t"\ 426 "1: \n\t"\
425 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ 427 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
426 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ 428 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
427 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ 429 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
428 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ 430 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
437 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c) 439 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c)
438 440
439 // do vertical chrominance interpolation 441 // do vertical chrominance interpolation
440 #define REAL_YSCALEYUV2RGB1b(index, c) \ 442 #define REAL_YSCALEYUV2RGB1b(index, c) \
441 "xor "#index", "#index" \n\t"\ 443 "xor "#index", "#index" \n\t"\
442 ".balign 16 \n\t"\ 444 ASMALIGN16\
443 "1: \n\t"\ 445 "1: \n\t"\
444 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ 446 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
445 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ 447 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
446 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ 448 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
447 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ 449 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1660 "mov %2, %%"REG_a" \n\t" 1662 "mov %2, %%"REG_a" \n\t"
1661 "movq "MANGLE(bgr2YCoeff)", %%mm6 \n\t" 1663 "movq "MANGLE(bgr2YCoeff)", %%mm6 \n\t"
1662 "movq "MANGLE(w1111)", %%mm5 \n\t" 1664 "movq "MANGLE(w1111)", %%mm5 \n\t"
1663 "pxor %%mm7, %%mm7 \n\t" 1665 "pxor %%mm7, %%mm7 \n\t"
1664 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t" 1666 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t"
1665 ".balign 16 \n\t" 1667 ASMALIGN16
1666 "1: \n\t" 1668 "1: \n\t"
1667 PREFETCH" 64(%0, %%"REG_b") \n\t" 1669 PREFETCH" 64(%0, %%"REG_b") \n\t"
1668 "movd (%0, %%"REG_b"), %%mm0 \n\t" 1670 "movd (%0, %%"REG_b"), %%mm0 \n\t"
1669 "movd 3(%0, %%"REG_b"), %%mm1 \n\t" 1671 "movd 3(%0, %%"REG_b"), %%mm1 \n\t"
1670 "punpcklbw %%mm7, %%mm0 \n\t" 1672 "punpcklbw %%mm7, %%mm0 \n\t"
1746 "movq "MANGLE(w1111)", %%mm5 \n\t" 1748 "movq "MANGLE(w1111)", %%mm5 \n\t"
1747 "movq "MANGLE(bgr2UCoeff)", %%mm6 \n\t" 1749 "movq "MANGLE(bgr2UCoeff)", %%mm6 \n\t"
1748 "pxor %%mm7, %%mm7 \n\t" 1750 "pxor %%mm7, %%mm7 \n\t"
1749 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b" \n\t" 1751 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b" \n\t"
1750 "add %%"REG_b", %%"REG_b" \n\t" 1752 "add %%"REG_b", %%"REG_b" \n\t"
1751 ".balign 16 \n\t" 1753 ASMALIGN16
1752 "1: \n\t" 1754 "1: \n\t"
1753 PREFETCH" 64(%0, %%"REG_b") \n\t" 1755 PREFETCH" 64(%0, %%"REG_b") \n\t"
1754 PREFETCH" 64(%1, %%"REG_b") \n\t" 1756 PREFETCH" 64(%1, %%"REG_b") \n\t"
1755 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) 1757 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
1756 "movq (%0, %%"REG_b"), %%mm0 \n\t" 1758 "movq (%0, %%"REG_b"), %%mm0 \n\t"
2053 asm volatile( 2055 asm volatile(
2054 "pxor %%mm7, %%mm7 \n\t" 2056 "pxor %%mm7, %%mm7 \n\t"
2055 "movq "MANGLE(w02)", %%mm6 \n\t" 2057 "movq "MANGLE(w02)", %%mm6 \n\t"
2056 "push %%"REG_BP" \n\t" // we use 7 regs here ... 2058 "push %%"REG_BP" \n\t" // we use 7 regs here ...
2057 "mov %%"REG_a", %%"REG_BP" \n\t" 2059 "mov %%"REG_a", %%"REG_BP" \n\t"
2058 ".balign 16 \n\t" 2060 ASMALIGN16
2059 "1: \n\t" 2061 "1: \n\t"
2060 "movzwl (%2, %%"REG_BP"), %%eax \n\t" 2062 "movzwl (%2, %%"REG_BP"), %%eax \n\t"
2061 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t" 2063 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t"
2062 "movq (%1, %%"REG_BP", 4), %%mm1\n\t" 2064 "movq (%1, %%"REG_BP", 4), %%mm1\n\t"
2063 "movq 8(%1, %%"REG_BP", 4), %%mm3\n\t" 2065 "movq 8(%1, %%"REG_BP", 4), %%mm3\n\t"
2091 asm volatile( 2093 asm volatile(
2092 "pxor %%mm7, %%mm7 \n\t" 2094 "pxor %%mm7, %%mm7 \n\t"
2093 "movq "MANGLE(w02)", %%mm6 \n\t" 2095 "movq "MANGLE(w02)", %%mm6 \n\t"
2094 "push %%"REG_BP" \n\t" // we use 7 regs here ... 2096 "push %%"REG_BP" \n\t" // we use 7 regs here ...
2095 "mov %%"REG_a", %%"REG_BP" \n\t" 2097 "mov %%"REG_a", %%"REG_BP" \n\t"
2096 ".balign 16 \n\t" 2098 ASMALIGN16
2097 "1: \n\t" 2099 "1: \n\t"
2098 "movzwl (%2, %%"REG_BP"), %%eax \n\t" 2100 "movzwl (%2, %%"REG_BP"), %%eax \n\t"
2099 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t" 2101 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t"
2100 "movq (%1, %%"REG_BP", 8), %%mm1\n\t" 2102 "movq (%1, %%"REG_BP", 8), %%mm1\n\t"
2101 "movq 16(%1, %%"REG_BP", 8), %%mm3\n\t" 2103 "movq 16(%1, %%"REG_BP", 8), %%mm3\n\t"
2140 filterPos-= counter/2; 2142 filterPos-= counter/2;
2141 dst-= counter/2; 2143 dst-= counter/2;
2142 asm volatile( 2144 asm volatile(
2143 "pxor %%mm7, %%mm7 \n\t" 2145 "pxor %%mm7, %%mm7 \n\t"
2144 "movq "MANGLE(w02)", %%mm6 \n\t" 2146 "movq "MANGLE(w02)", %%mm6 \n\t"
2145 ".balign 16 \n\t" 2147 ASMALIGN16
2146 "1: \n\t" 2148 "1: \n\t"
2147 "mov %2, %%"REG_c" \n\t" 2149 "mov %2, %%"REG_c" \n\t"
2148 "movzwl (%%"REG_c", %0), %%eax \n\t" 2150 "movzwl (%%"REG_c", %0), %%eax \n\t"
2149 "movzwl 2(%%"REG_c", %0), %%ebx \n\t" 2151 "movzwl 2(%%"REG_c", %0), %%ebx \n\t"
2150 "mov %5, %%"REG_c" \n\t" 2152 "mov %5, %%"REG_c" \n\t"
2324 //NO MMX just normal asm ... 2326 //NO MMX just normal asm ...
2325 asm volatile( 2327 asm volatile(
2326 "xor %%"REG_a", %%"REG_a" \n\t" // i 2328 "xor %%"REG_a", %%"REG_a" \n\t" // i
2327 "xor %%"REG_b", %%"REG_b" \n\t" // xx 2329 "xor %%"REG_b", %%"REG_b" \n\t" // xx
2328 "xorl %%ecx, %%ecx \n\t" // 2*xalpha 2330 "xorl %%ecx, %%ecx \n\t" // 2*xalpha
2329 ".balign 16 \n\t" 2331 ASMALIGN16
2330 "1: \n\t" 2332 "1: \n\t"
2331 "movzbl (%0, %%"REG_b"), %%edi \n\t" //src[xx] 2333 "movzbl (%0, %%"REG_b"), %%edi \n\t" //src[xx]
2332 "movzbl 1(%0, %%"REG_b"), %%esi \n\t" //src[xx+1] 2334 "movzbl 1(%0, %%"REG_b"), %%esi \n\t" //src[xx+1]
2333 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] 2335 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx]
2334 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha 2336 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha
2521 int xInc_mask = xInc & 0xffff; 2523 int xInc_mask = xInc & 0xffff;
2522 asm volatile( 2524 asm volatile(
2523 "xor %%"REG_a", %%"REG_a" \n\t" // i 2525 "xor %%"REG_a", %%"REG_a" \n\t" // i
2524 "xor %%"REG_b", %%"REG_b" \n\t" // xx 2526 "xor %%"REG_b", %%"REG_b" \n\t" // xx
2525 "xorl %%ecx, %%ecx \n\t" // 2*xalpha 2527 "xorl %%ecx, %%ecx \n\t" // 2*xalpha
2526 ".balign 16 \n\t" 2528 ASMALIGN16
2527 "1: \n\t" 2529 "1: \n\t"
2528 "mov %0, %%"REG_S" \n\t" 2530 "mov %0, %%"REG_S" \n\t"
2529 "movzbl (%%"REG_S", %%"REG_b"), %%edi \n\t" //src[xx] 2531 "movzbl (%%"REG_S", %%"REG_b"), %%edi \n\t" //src[xx]
2530 "movzbl 1(%%"REG_S", %%"REG_b"), %%esi \n\t" //src[xx+1] 2532 "movzbl 1(%%"REG_S", %%"REG_b"), %%esi \n\t" //src[xx+1]
2531 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] 2533 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx]