Mercurial > mplayer.hg
annotate libswscale/swscale_template.c @ 19172:bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
the +-1 issue is limited to >2tap vertical filters, so bilinear upscale was unaffected
the new code is sometime faster sometimes slower but the difference is significant (~20%) so its optional and enabled with arnd=1
| author | michael |
|---|---|
| date | Mon, 24 Jul 2006 10:36:06 +0000 |
| parents | 8579acff875e |
| children | dbdc58b6e9bb |
| rev | line source |
|---|---|
| 18861 | 1 /* |
| 2 Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at> | |
| 3 | |
| 4 This program is free software; you can redistribute it and/or modify | |
| 5 it under the terms of the GNU General Public License as published by | |
| 6 the Free Software Foundation; either version 2 of the License, or | |
| 7 (at your option) any later version. | |
| 8 | |
| 9 This program is distributed in the hope that it will be useful, | |
| 10 but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
| 12 GNU General Public License for more details. | |
| 13 | |
| 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 | |
| 16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
| 17 */ | |
| 18 | |
| 19 #include "asmalign.h" | |
| 20 | |
| 21 #undef REAL_MOVNTQ | |
| 22 #undef MOVNTQ | |
| 23 #undef PAVGB | |
| 24 #undef PREFETCH | |
| 25 #undef PREFETCHW | |
| 26 #undef EMMS | |
| 27 #undef SFENCE | |
| 28 | |
| 29 #ifdef HAVE_3DNOW | |
| 30 /* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */ | |
| 31 #define EMMS "femms" | |
| 32 #else | |
| 33 #define EMMS "emms" | |
| 34 #endif | |
| 35 | |
| 36 #ifdef HAVE_3DNOW | |
| 37 #define PREFETCH "prefetch" | |
| 38 #define PREFETCHW "prefetchw" | |
| 39 #elif defined ( HAVE_MMX2 ) | |
| 40 #define PREFETCH "prefetchnta" | |
| 41 #define PREFETCHW "prefetcht0" | |
| 42 #else | |
| 43 #define PREFETCH "/nop" | |
| 44 #define PREFETCHW "/nop" | |
| 45 #endif | |
| 46 | |
| 47 #ifdef HAVE_MMX2 | |
| 48 #define SFENCE "sfence" | |
| 49 #else | |
| 50 #define SFENCE "/nop" | |
| 51 #endif | |
| 52 | |
| 53 #ifdef HAVE_MMX2 | |
| 54 #define PAVGB(a,b) "pavgb " #a ", " #b " \n\t" | |
| 55 #elif defined (HAVE_3DNOW) | |
| 56 #define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t" | |
| 57 #endif | |
| 58 | |
| 59 #ifdef HAVE_MMX2 | |
| 60 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t" | |
| 61 #else | |
| 62 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t" | |
| 63 #endif | |
| 64 #define MOVNTQ(a,b) REAL_MOVNTQ(a,b) | |
| 65 | |
| 66 #ifdef HAVE_ALTIVEC | |
| 67 #include "swscale_altivec_template.c" | |
| 68 #endif | |
| 69 | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
70 #define YSCALEYUV2YV12X(x, offset, dest, width) \ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
71 asm volatile(\ |
| 18861 | 72 "xor %%"REG_a", %%"REG_a" \n\t"\ |
| 73 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ | |
| 74 "movq %%mm3, %%mm4 \n\t"\ | |
| 75 "lea " offset "(%0), %%"REG_d" \n\t"\ | |
| 76 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 77 ASMALIGN16 /* FIXME Unroll? */\ | |
| 78 "1: \n\t"\ | |
| 79 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ | |
| 80 "movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\ | |
| 81 "movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm5\n\t" /* srcData */\ | |
| 82 "add $16, %%"REG_d" \n\t"\ | |
| 83 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 84 "test %%"REG_S", %%"REG_S" \n\t"\ | |
| 85 "pmulhw %%mm0, %%mm2 \n\t"\ | |
| 86 "pmulhw %%mm0, %%mm5 \n\t"\ | |
| 87 "paddw %%mm2, %%mm3 \n\t"\ | |
| 88 "paddw %%mm5, %%mm4 \n\t"\ | |
| 89 " jnz 1b \n\t"\ | |
| 90 "psraw $3, %%mm3 \n\t"\ | |
| 91 "psraw $3, %%mm4 \n\t"\ | |
| 92 "packuswb %%mm4, %%mm3 \n\t"\ | |
| 93 MOVNTQ(%%mm3, (%1, %%REGa))\ | |
| 94 "add $8, %%"REG_a" \n\t"\ | |
| 95 "cmp %2, %%"REG_a" \n\t"\ | |
| 96 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ | |
| 97 "movq %%mm3, %%mm4 \n\t"\ | |
| 98 "lea " offset "(%0), %%"REG_d" \n\t"\ | |
| 99 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
100 "jb 1b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
101 :: "r" (&c->redDither),\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
102 "r" (dest), "p" (width)\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
103 : "%"REG_a, "%"REG_d, "%"REG_S\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
104 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
105 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
106 #define YSCALEYUV2YV12X_ACCURATE(x, offset, dest, width) \ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
107 asm volatile(\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
108 "lea " offset "(%0), %%"REG_d" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
109 "xor %%"REG_a", %%"REG_a" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
110 "pxor %%mm4, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
111 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
112 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
113 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
114 "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
115 ASMALIGN16 \ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
116 "1: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
117 "movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm0\n\t" /* srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
118 "movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
119 "mov 4(%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
120 "movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm1\n\t" /* srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
121 "movq %%mm0, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
122 "punpcklwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
123 "punpckhwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
124 "movq 8(%%"REG_d"), %%mm1 \n\t" /* filterCoeff */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
125 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
126 "pmaddwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
127 "paddd %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
128 "paddd %%mm3, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
129 "movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm3\n\t" /* srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
130 "mov 16(%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
131 "add $16, %%"REG_d" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
132 "test %%"REG_S", %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
133 "movq %%mm2, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
134 "punpcklwd %%mm3, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
135 "punpckhwd %%mm3, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
136 "pmaddwd %%mm1, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
137 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
138 "paddd %%mm2, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
139 "paddd %%mm0, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
140 " jnz 1b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
141 "psrad $16, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
142 "psrad $16, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
143 "psrad $16, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
144 "psrad $16, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
145 "movq "VROUNDER_OFFSET"(%0), %%mm0\n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
146 "packssdw %%mm5, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
147 "packssdw %%mm7, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
148 "paddw %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
149 "paddw %%mm0, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
150 "psraw $3, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
151 "psraw $3, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
152 "packuswb %%mm6, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
153 MOVNTQ(%%mm4, (%1, %%REGa))\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
154 "add $8, %%"REG_a" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
155 "cmp %2, %%"REG_a" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
156 "lea " offset "(%0), %%"REG_d" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
157 "pxor %%mm4, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
158 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
159 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
160 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
161 "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
162 "jb 1b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
163 :: "r" (&c->redDither),\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
164 "r" (dest), "p" (width)\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
165 : "%"REG_a, "%"REG_d, "%"REG_S\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
166 ); |
| 18861 | 167 |
| 168 #define YSCALEYUV2YV121 \ | |
| 169 "mov %2, %%"REG_a" \n\t"\ | |
| 170 ASMALIGN16 /* FIXME Unroll? */\ | |
| 171 "1: \n\t"\ | |
| 172 "movq (%0, %%"REG_a", 2), %%mm0 \n\t"\ | |
| 173 "movq 8(%0, %%"REG_a", 2), %%mm1\n\t"\ | |
| 174 "psraw $7, %%mm0 \n\t"\ | |
| 175 "psraw $7, %%mm1 \n\t"\ | |
| 176 "packuswb %%mm1, %%mm0 \n\t"\ | |
| 177 MOVNTQ(%%mm0, (%1, %%REGa))\ | |
| 178 "add $8, %%"REG_a" \n\t"\ | |
| 179 "jnc 1b \n\t" | |
| 180 | |
| 181 /* | |
| 182 :: "m" (-lumFilterSize), "m" (-chrFilterSize), | |
| 183 "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4), | |
| 184 "r" (dest), "m" (dstW), | |
| 185 "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize) | |
| 186 : "%eax", "%ebx", "%ecx", "%edx", "%esi" | |
| 187 */ | |
| 188 #define YSCALEYUV2PACKEDX \ | |
| 189 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 190 ASMALIGN16\ | |
| 191 "nop \n\t"\ | |
| 192 "1: \n\t"\ | |
| 193 "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ | |
| 194 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 195 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ | |
| 196 "movq %%mm3, %%mm4 \n\t"\ | |
| 197 ASMALIGN16\ | |
| 198 "2: \n\t"\ | |
| 199 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ | |
| 200 "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\ | |
| 201 "movq 4096(%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\ | |
| 202 "add $16, %%"REG_d" \n\t"\ | |
| 203 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 204 "pmulhw %%mm0, %%mm2 \n\t"\ | |
| 205 "pmulhw %%mm0, %%mm5 \n\t"\ | |
| 206 "paddw %%mm2, %%mm3 \n\t"\ | |
| 207 "paddw %%mm5, %%mm4 \n\t"\ | |
| 208 "test %%"REG_S", %%"REG_S" \n\t"\ | |
| 209 " jnz 2b \n\t"\ | |
| 210 \ | |
| 211 "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ | |
| 212 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 213 "movq "VROUNDER_OFFSET"(%0), %%mm1\n\t"\ | |
| 214 "movq %%mm1, %%mm7 \n\t"\ | |
| 215 ASMALIGN16\ | |
| 216 "2: \n\t"\ | |
| 217 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ | |
| 218 "movq (%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y1srcData */\ | |
| 219 "movq 8(%%"REG_S", %%"REG_a", 2), %%mm5 \n\t" /* Y2srcData */\ | |
| 220 "add $16, %%"REG_d" \n\t"\ | |
| 221 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 222 "pmulhw %%mm0, %%mm2 \n\t"\ | |
| 223 "pmulhw %%mm0, %%mm5 \n\t"\ | |
| 224 "paddw %%mm2, %%mm1 \n\t"\ | |
| 225 "paddw %%mm5, %%mm7 \n\t"\ | |
| 226 "test %%"REG_S", %%"REG_S" \n\t"\ | |
| 227 " jnz 2b \n\t"\ | |
| 228 | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
229 #define YSCALEYUV2PACKEDX_ACCURATE \ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
230 "xor %%"REG_a", %%"REG_a" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
231 ASMALIGN16\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
232 "nop \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
233 "1: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
234 "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
235 "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
236 "pxor %%mm4, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
237 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
238 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
239 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
240 ASMALIGN16\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
241 "2: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
242 "movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
243 "movq 4096(%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
244 "mov 4(%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
245 "movq (%%"REG_S", %%"REG_a"), %%mm1 \n\t" /* UsrcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
246 "movq %%mm0, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
247 "punpcklwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
248 "punpckhwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
249 "movq 8(%%"REG_d"), %%mm1 \n\t" /* filterCoeff */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
250 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
251 "pmaddwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
252 "paddd %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
253 "paddd %%mm3, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
254 "movq 4096(%%"REG_S", %%"REG_a"), %%mm3 \n\t" /* VsrcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
255 "mov 16(%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
256 "add $16, %%"REG_d" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
257 "test %%"REG_S", %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
258 "movq %%mm2, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
259 "punpcklwd %%mm3, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
260 "punpckhwd %%mm3, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
261 "pmaddwd %%mm1, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
262 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
263 "paddd %%mm2, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
264 "paddd %%mm0, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
265 " jnz 2b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
266 "psrad $16, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
267 "psrad $16, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
268 "psrad $16, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
269 "psrad $16, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
270 "movq "VROUNDER_OFFSET"(%0), %%mm0\n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
271 "packssdw %%mm5, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
272 "packssdw %%mm7, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
273 "paddw %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
274 "paddw %%mm0, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
275 "movq %%mm4, "U_TEMP"(%0) \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
276 "movq %%mm6, "V_TEMP"(%0) \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
277 \ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
278 "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
279 "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
280 "pxor %%mm1, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
281 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
282 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
283 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
284 ASMALIGN16\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
285 "2: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
286 "movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
287 "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
288 "mov 4(%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
289 "movq (%%"REG_S", %%"REG_a", 2), %%mm4 \n\t" /* Y1srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
290 "movq %%mm0, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
291 "punpcklwd %%mm4, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
292 "punpckhwd %%mm4, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
293 "movq 8(%%"REG_d"), %%mm4 \n\t" /* filterCoeff */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
294 "pmaddwd %%mm4, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
295 "pmaddwd %%mm4, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
296 "paddd %%mm0, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
297 "paddd %%mm3, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
298 "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3 \n\t" /* Y2srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
299 "mov 16(%%"REG_d"), %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
300 "add $16, %%"REG_d" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
301 "test %%"REG_S", %%"REG_S" \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
302 "movq %%mm2, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
303 "punpcklwd %%mm3, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
304 "punpckhwd %%mm3, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
305 "pmaddwd %%mm4, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
306 "pmaddwd %%mm4, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
307 "paddd %%mm2, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
308 "paddd %%mm0, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
309 " jnz 2b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
310 "psrad $16, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
311 "psrad $16, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
312 "psrad $16, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
313 "psrad $16, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
314 "movq "VROUNDER_OFFSET"(%0), %%mm0\n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
315 "packssdw %%mm5, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
316 "packssdw %%mm6, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
317 "paddw %%mm0, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
318 "paddw %%mm0, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
319 "movq "U_TEMP"(%0), %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
320 "movq "V_TEMP"(%0), %%mm4 \n\t"\ |
| 18861 | 321 |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
322 #define YSCALEYUV2RGBX(YSCALEYUV2PACKEDX) \ |
| 18861 | 323 YSCALEYUV2PACKEDX\ |
| 324 "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\ | |
| 325 "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\ | |
| 326 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 327 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 328 "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\ | |
| 329 "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\ | |
| 330 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 331 "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\ | |
| 332 "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\ | |
| 333 "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 334 "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 335 "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\ | |
| 336 "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\ | |
| 337 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 338 "paddw %%mm3, %%mm4 \n\t"\ | |
| 339 "movq %%mm2, %%mm0 \n\t"\ | |
| 340 "movq %%mm5, %%mm6 \n\t"\ | |
| 341 "movq %%mm4, %%mm3 \n\t"\ | |
| 342 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 343 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 344 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 345 "paddw %%mm1, %%mm2 \n\t"\ | |
| 346 "paddw %%mm1, %%mm5 \n\t"\ | |
| 347 "paddw %%mm1, %%mm4 \n\t"\ | |
| 348 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 349 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 350 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 351 "paddw %%mm7, %%mm0 \n\t"\ | |
| 352 "paddw %%mm7, %%mm6 \n\t"\ | |
| 353 "paddw %%mm7, %%mm3 \n\t"\ | |
| 354 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 355 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 356 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 357 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 358 "pxor %%mm7, %%mm7 \n\t" | |
| 359 #if 0 | |
| 360 #define FULL_YSCALEYUV2RGB \ | |
| 361 "pxor %%mm7, %%mm7 \n\t"\ | |
| 362 "movd %6, %%mm6 \n\t" /*yalpha1*/\ | |
| 363 "punpcklwd %%mm6, %%mm6 \n\t"\ | |
| 364 "punpcklwd %%mm6, %%mm6 \n\t"\ | |
| 365 "movd %7, %%mm5 \n\t" /*uvalpha1*/\ | |
| 366 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 367 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 368 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 369 ASMALIGN16\ | |
| 370 "1: \n\t"\ | |
| 371 "movq (%0, %%"REG_a", 2), %%mm0 \n\t" /*buf0[eax]*/\ | |
| 372 "movq (%1, %%"REG_a", 2), %%mm1 \n\t" /*buf1[eax]*/\ | |
| 373 "movq (%2, %%"REG_a",2), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 374 "movq (%3, %%"REG_a",2), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 375 "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 376 "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\ | |
| 377 "pmulhw %%mm6, %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 378 "pmulhw %%mm5, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\ | |
| 379 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 380 "movq 4096(%2, %%"REG_a",2), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 381 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 382 "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 383 "movq 4096(%3, %%"REG_a",2), %%mm0 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 384 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\ | |
| 385 "psubw %%mm0, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\ | |
| 386 "psubw "MANGLE(w80)", %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 387 "psubw "MANGLE(w400)", %%mm3 \n\t" /* 8(U-128)*/\ | |
| 388 "pmulhw "MANGLE(yCoeff)", %%mm1 \n\t"\ | |
| 389 \ | |
| 390 \ | |
| 391 "pmulhw %%mm5, %%mm4 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\ | |
| 392 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 393 "pmulhw "MANGLE(ubCoeff)", %%mm3\n\t"\ | |
| 394 "psraw $4, %%mm0 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 395 "pmulhw "MANGLE(ugCoeff)", %%mm2\n\t"\ | |
| 396 "paddw %%mm4, %%mm0 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\ | |
| 397 "psubw "MANGLE(w400)", %%mm0 \n\t" /* (V-128)8*/\ | |
| 398 \ | |
| 399 \ | |
| 400 "movq %%mm0, %%mm4 \n\t" /* (V-128)8*/\ | |
| 401 "pmulhw "MANGLE(vrCoeff)", %%mm0\n\t"\ | |
| 402 "pmulhw "MANGLE(vgCoeff)", %%mm4\n\t"\ | |
| 403 "paddw %%mm1, %%mm3 \n\t" /* B*/\ | |
| 404 "paddw %%mm1, %%mm0 \n\t" /* R*/\ | |
| 405 "packuswb %%mm3, %%mm3 \n\t"\ | |
| 406 \ | |
| 407 "packuswb %%mm0, %%mm0 \n\t"\ | |
| 408 "paddw %%mm4, %%mm2 \n\t"\ | |
| 409 "paddw %%mm2, %%mm1 \n\t" /* G*/\ | |
| 410 \ | |
| 411 "packuswb %%mm1, %%mm1 \n\t" | |
| 412 #endif | |
| 413 | |
| 414 #define REAL_YSCALEYUV2PACKED(index, c) \ | |
| 415 "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\ | |
| 416 "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1\n\t"\ | |
| 417 "psraw $3, %%mm0 \n\t"\ | |
| 418 "psraw $3, %%mm1 \n\t"\ | |
| 419 "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c")\n\t"\ | |
| 420 "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c")\n\t"\ | |
| 421 "xor "#index", "#index" \n\t"\ | |
| 422 ASMALIGN16\ | |
| 423 "1: \n\t"\ | |
| 424 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 425 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 426 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 427 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 428 "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\ | |
| 429 "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\ | |
| 430 "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\ | |
| 431 "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\ | |
| 432 "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\ | |
| 433 "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 434 "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 435 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\ | |
| 436 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\ | |
| 437 "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\ | |
| 438 "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\ | |
| 439 "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\ | |
| 440 "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\ | |
| 441 "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 442 "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 443 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 444 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 445 "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 446 "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 447 "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 448 "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 449 | |
| 450 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c) | |
| 451 | |
| 452 #define REAL_YSCALEYUV2RGB(index, c) \ | |
| 453 "xor "#index", "#index" \n\t"\ | |
| 454 ASMALIGN16\ | |
| 455 "1: \n\t"\ | |
| 456 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 457 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 458 "movq 4096(%2, "#index"), %%mm5\n\t" /* uvbuf0[eax+2048]*/\ | |
| 459 "movq 4096(%3, "#index"), %%mm4\n\t" /* uvbuf1[eax+2048]*/\ | |
| 460 "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\ | |
| 461 "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\ | |
| 462 "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\ | |
| 463 "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\ | |
| 464 "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\ | |
| 465 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 466 "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 467 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\ | |
| 468 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\ | |
| 469 "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\ | |
| 470 "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\ | |
| 471 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 472 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 473 "pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\ | |
| 474 "pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\ | |
| 475 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 476 "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\ | |
| 477 "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\ | |
| 478 "movq 8(%0, "#index", 2), %%mm6\n\t" /*buf0[eax]*/\ | |
| 479 "movq 8(%1, "#index", 2), %%mm7\n\t" /*buf1[eax]*/\ | |
| 480 "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 481 "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 482 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 483 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 484 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 485 "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 486 "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 487 "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 488 "pmulhw "UB_COEFF"("#c"), %%mm2\n\t"\ | |
| 489 "pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\ | |
| 490 "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 491 "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 492 "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\ | |
| 493 "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\ | |
| 494 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 495 "paddw %%mm3, %%mm4 \n\t"\ | |
| 496 "movq %%mm2, %%mm0 \n\t"\ | |
| 497 "movq %%mm5, %%mm6 \n\t"\ | |
| 498 "movq %%mm4, %%mm3 \n\t"\ | |
| 499 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 500 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 501 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 502 "paddw %%mm1, %%mm2 \n\t"\ | |
| 503 "paddw %%mm1, %%mm5 \n\t"\ | |
| 504 "paddw %%mm1, %%mm4 \n\t"\ | |
| 505 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 506 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 507 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 508 "paddw %%mm7, %%mm0 \n\t"\ | |
| 509 "paddw %%mm7, %%mm6 \n\t"\ | |
| 510 "paddw %%mm7, %%mm3 \n\t"\ | |
| 511 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 512 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 513 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 514 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 515 "pxor %%mm7, %%mm7 \n\t" | |
| 516 #define YSCALEYUV2RGB(index, c) REAL_YSCALEYUV2RGB(index, c) | |
| 517 | |
| 518 #define REAL_YSCALEYUV2PACKED1(index, c) \ | |
| 519 "xor "#index", "#index" \n\t"\ | |
| 520 ASMALIGN16\ | |
| 521 "1: \n\t"\ | |
| 522 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\ | |
| 523 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 524 "psraw $7, %%mm3 \n\t" \ | |
| 525 "psraw $7, %%mm4 \n\t" \ | |
| 526 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 527 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 528 "psraw $7, %%mm1 \n\t" \ | |
| 529 "psraw $7, %%mm7 \n\t" \ | |
| 530 | |
| 531 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c) | |
| 532 | |
| 533 #define REAL_YSCALEYUV2RGB1(index, c) \ | |
| 534 "xor "#index", "#index" \n\t"\ | |
| 535 ASMALIGN16\ | |
| 536 "1: \n\t"\ | |
| 537 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\ | |
| 538 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 539 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 540 "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 541 "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\ | |
| 542 "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\ | |
| 543 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 544 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 545 "pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\ | |
| 546 "pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\ | |
| 547 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 548 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 549 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 550 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 551 "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 552 "pmulhw "UB_COEFF"("#c"), %%mm2\n\t"\ | |
| 553 "pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\ | |
| 554 "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 555 "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 556 "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\ | |
| 557 "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\ | |
| 558 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 559 "paddw %%mm3, %%mm4 \n\t"\ | |
| 560 "movq %%mm2, %%mm0 \n\t"\ | |
| 561 "movq %%mm5, %%mm6 \n\t"\ | |
| 562 "movq %%mm4, %%mm3 \n\t"\ | |
| 563 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 564 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 565 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 566 "paddw %%mm1, %%mm2 \n\t"\ | |
| 567 "paddw %%mm1, %%mm5 \n\t"\ | |
| 568 "paddw %%mm1, %%mm4 \n\t"\ | |
| 569 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 570 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 571 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 572 "paddw %%mm7, %%mm0 \n\t"\ | |
| 573 "paddw %%mm7, %%mm6 \n\t"\ | |
| 574 "paddw %%mm7, %%mm3 \n\t"\ | |
| 575 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 576 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 577 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 578 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 579 "pxor %%mm7, %%mm7 \n\t" | |
| 580 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c) | |
| 581 | |
| 582 #define REAL_YSCALEYUV2PACKED1b(index, c) \ | |
| 583 "xor "#index", "#index" \n\t"\ | |
| 584 ASMALIGN16\ | |
| 585 "1: \n\t"\ | |
| 586 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 587 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 588 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 589 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 590 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\ | |
| 591 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\ | |
| 592 "psrlw $8, %%mm3 \n\t" \ | |
| 593 "psrlw $8, %%mm4 \n\t" \ | |
| 594 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 595 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 596 "psraw $7, %%mm1 \n\t" \ | |
| 597 "psraw $7, %%mm7 \n\t" | |
| 598 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c) | |
| 599 | |
| 600 // do vertical chrominance interpolation | |
| 601 #define REAL_YSCALEYUV2RGB1b(index, c) \ | |
| 602 "xor "#index", "#index" \n\t"\ | |
| 603 ASMALIGN16\ | |
| 604 "1: \n\t"\ | |
| 605 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 606 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 607 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 608 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 609 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\ | |
| 610 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\ | |
| 611 "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\ | |
| 612 "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\ | |
| 613 "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\ | |
| 614 "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\ | |
| 615 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 616 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 617 "pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\ | |
| 618 "pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\ | |
| 619 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 620 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 621 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 622 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 623 "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 624 "pmulhw "UB_COEFF"("#c"), %%mm2\n\t"\ | |
| 625 "pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\ | |
| 626 "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 627 "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 628 "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\ | |
| 629 "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\ | |
| 630 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 631 "paddw %%mm3, %%mm4 \n\t"\ | |
| 632 "movq %%mm2, %%mm0 \n\t"\ | |
| 633 "movq %%mm5, %%mm6 \n\t"\ | |
| 634 "movq %%mm4, %%mm3 \n\t"\ | |
| 635 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 636 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 637 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 638 "paddw %%mm1, %%mm2 \n\t"\ | |
| 639 "paddw %%mm1, %%mm5 \n\t"\ | |
| 640 "paddw %%mm1, %%mm4 \n\t"\ | |
| 641 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 642 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 643 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 644 "paddw %%mm7, %%mm0 \n\t"\ | |
| 645 "paddw %%mm7, %%mm6 \n\t"\ | |
| 646 "paddw %%mm7, %%mm3 \n\t"\ | |
| 647 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 648 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 649 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 650 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 651 "pxor %%mm7, %%mm7 \n\t" | |
| 652 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c) | |
| 653 | |
| 654 #define REAL_WRITEBGR32(dst, dstw, index) \ | |
| 655 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 656 "movq %%mm2, %%mm1 \n\t" /* B */\ | |
| 657 "movq %%mm5, %%mm6 \n\t" /* R */\ | |
| 658 "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\ | |
| 659 "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\ | |
| 660 "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\ | |
| 661 "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\ | |
| 662 "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\ | |
| 663 "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\ | |
| 664 "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\ | |
| 665 "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 666 "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\ | |
| 667 "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\ | |
| 668 \ | |
| 669 MOVNTQ(%%mm0, (dst, index, 4))\ | |
| 670 MOVNTQ(%%mm2, 8(dst, index, 4))\ | |
| 671 MOVNTQ(%%mm1, 16(dst, index, 4))\ | |
| 672 MOVNTQ(%%mm3, 24(dst, index, 4))\ | |
| 673 \ | |
| 674 "add $8, "#index" \n\t"\ | |
| 675 "cmp "#dstw", "#index" \n\t"\ | |
| 676 " jb 1b \n\t" | |
| 677 #define WRITEBGR32(dst, dstw, index) REAL_WRITEBGR32(dst, dstw, index) | |
| 678 | |
| 679 #define REAL_WRITEBGR16(dst, dstw, index) \ | |
| 680 "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\ | |
| 681 "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\ | |
| 682 "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\ | |
| 683 "psrlq $3, %%mm2 \n\t"\ | |
| 684 \ | |
| 685 "movq %%mm2, %%mm1 \n\t"\ | |
| 686 "movq %%mm4, %%mm3 \n\t"\ | |
| 687 \ | |
| 688 "punpcklbw %%mm7, %%mm3 \n\t"\ | |
| 689 "punpcklbw %%mm5, %%mm2 \n\t"\ | |
| 690 "punpckhbw %%mm7, %%mm4 \n\t"\ | |
| 691 "punpckhbw %%mm5, %%mm1 \n\t"\ | |
| 692 \ | |
| 693 "psllq $3, %%mm3 \n\t"\ | |
| 694 "psllq $3, %%mm4 \n\t"\ | |
| 695 \ | |
| 696 "por %%mm3, %%mm2 \n\t"\ | |
| 697 "por %%mm4, %%mm1 \n\t"\ | |
| 698 \ | |
| 699 MOVNTQ(%%mm2, (dst, index, 2))\ | |
| 700 MOVNTQ(%%mm1, 8(dst, index, 2))\ | |
| 701 \ | |
| 702 "add $8, "#index" \n\t"\ | |
| 703 "cmp "#dstw", "#index" \n\t"\ | |
| 704 " jb 1b \n\t" | |
| 705 #define WRITEBGR16(dst, dstw, index) REAL_WRITEBGR16(dst, dstw, index) | |
| 706 | |
| 707 #define REAL_WRITEBGR15(dst, dstw, index) \ | |
| 708 "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\ | |
| 709 "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\ | |
| 710 "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\ | |
| 711 "psrlq $3, %%mm2 \n\t"\ | |
| 712 "psrlq $1, %%mm5 \n\t"\ | |
| 713 \ | |
| 714 "movq %%mm2, %%mm1 \n\t"\ | |
| 715 "movq %%mm4, %%mm3 \n\t"\ | |
| 716 \ | |
| 717 "punpcklbw %%mm7, %%mm3 \n\t"\ | |
| 718 "punpcklbw %%mm5, %%mm2 \n\t"\ | |
| 719 "punpckhbw %%mm7, %%mm4 \n\t"\ | |
| 720 "punpckhbw %%mm5, %%mm1 \n\t"\ | |
| 721 \ | |
| 722 "psllq $2, %%mm3 \n\t"\ | |
| 723 "psllq $2, %%mm4 \n\t"\ | |
| 724 \ | |
| 725 "por %%mm3, %%mm2 \n\t"\ | |
| 726 "por %%mm4, %%mm1 \n\t"\ | |
| 727 \ | |
| 728 MOVNTQ(%%mm2, (dst, index, 2))\ | |
| 729 MOVNTQ(%%mm1, 8(dst, index, 2))\ | |
| 730 \ | |
| 731 "add $8, "#index" \n\t"\ | |
| 732 "cmp "#dstw", "#index" \n\t"\ | |
| 733 " jb 1b \n\t" | |
| 734 #define WRITEBGR15(dst, dstw, index) REAL_WRITEBGR15(dst, dstw, index) | |
| 735 | |
| 736 #define WRITEBGR24OLD(dst, dstw, index) \ | |
| 737 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 738 "movq %%mm2, %%mm1 \n\t" /* B */\ | |
| 739 "movq %%mm5, %%mm6 \n\t" /* R */\ | |
| 740 "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\ | |
| 741 "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\ | |
| 742 "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\ | |
| 743 "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\ | |
| 744 "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\ | |
| 745 "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\ | |
| 746 "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\ | |
| 747 "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 748 "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\ | |
| 749 "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\ | |
| 750 \ | |
| 751 "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\ | |
| 752 "psrlq $8, %%mm0 \n\t" /* 00RGB0RG 0 */\ | |
| 753 "pand "MANGLE(bm00000111)", %%mm4\n\t" /* 00000RGB 0 */\ | |
| 754 "pand "MANGLE(bm11111000)", %%mm0\n\t" /* 00RGB000 0.5 */\ | |
| 755 "por %%mm4, %%mm0 \n\t" /* 00RGBRGB 0 */\ | |
| 756 "movq %%mm2, %%mm4 \n\t" /* 0RGB0RGB 1 */\ | |
| 757 "psllq $48, %%mm2 \n\t" /* GB000000 1 */\ | |
| 758 "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\ | |
| 759 \ | |
| 760 "movq %%mm4, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 761 "psrld $16, %%mm4 \n\t" /* 000R000R 1 */\ | |
| 762 "psrlq $24, %%mm2 \n\t" /* 0000RGB0 1.5 */\ | |
| 763 "por %%mm4, %%mm2 \n\t" /* 000RRGBR 1 */\ | |
| 764 "pand "MANGLE(bm00001111)", %%mm2\n\t" /* 0000RGBR 1 */\ | |
| 765 "movq %%mm1, %%mm4 \n\t" /* 0RGB0RGB 2 */\ | |
| 766 "psrlq $8, %%mm1 \n\t" /* 00RGB0RG 2 */\ | |
| 767 "pand "MANGLE(bm00000111)", %%mm4\n\t" /* 00000RGB 2 */\ | |
| 768 "pand "MANGLE(bm11111000)", %%mm1\n\t" /* 00RGB000 2.5 */\ | |
| 769 "por %%mm4, %%mm1 \n\t" /* 00RGBRGB 2 */\ | |
| 770 "movq %%mm1, %%mm4 \n\t" /* 00RGBRGB 2 */\ | |
| 771 "psllq $32, %%mm1 \n\t" /* BRGB0000 2 */\ | |
| 772 "por %%mm1, %%mm2 \n\t" /* BRGBRGBR 1 */\ | |
| 773 \ | |
| 774 "psrlq $32, %%mm4 \n\t" /* 000000RG 2.5 */\ | |
| 775 "movq %%mm3, %%mm5 \n\t" /* 0RGB0RGB 3 */\ | |
| 776 "psrlq $8, %%mm3 \n\t" /* 00RGB0RG 3 */\ | |
| 777 "pand "MANGLE(bm00000111)", %%mm5\n\t" /* 00000RGB 3 */\ | |
| 778 "pand "MANGLE(bm11111000)", %%mm3\n\t" /* 00RGB000 3.5 */\ | |
| 779 "por %%mm5, %%mm3 \n\t" /* 00RGBRGB 3 */\ | |
| 780 "psllq $16, %%mm3 \n\t" /* RGBRGB00 3 */\ | |
| 781 "por %%mm4, %%mm3 \n\t" /* RGBRGBRG 2.5 */\ | |
| 782 \ | |
| 783 MOVNTQ(%%mm0, (dst))\ | |
| 784 MOVNTQ(%%mm2, 8(dst))\ | |
| 785 MOVNTQ(%%mm3, 16(dst))\ | |
| 786 "add $24, "#dst" \n\t"\ | |
| 787 \ | |
| 788 "add $8, "#index" \n\t"\ | |
| 789 "cmp "#dstw", "#index" \n\t"\ | |
| 790 " jb 1b \n\t" | |
| 791 | |
| 792 #define WRITEBGR24MMX(dst, dstw, index) \ | |
| 793 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 794 "movq %%mm2, %%mm1 \n\t" /* B */\ | |
| 795 "movq %%mm5, %%mm6 \n\t" /* R */\ | |
| 796 "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\ | |
| 797 "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\ | |
| 798 "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\ | |
| 799 "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\ | |
| 800 "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\ | |
| 801 "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\ | |
| 802 "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\ | |
| 803 "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 804 "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\ | |
| 805 "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\ | |
| 806 \ | |
| 807 "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\ | |
| 808 "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\ | |
| 809 "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\ | |
| 810 "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\ | |
| 811 \ | |
| 812 "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\ | |
| 813 "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\ | |
| 814 "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\ | |
| 815 "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\ | |
| 816 \ | |
| 817 "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\ | |
| 818 "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\ | |
| 819 "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\ | |
| 820 "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\ | |
| 821 \ | |
| 822 "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\ | |
| 823 "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\ | |
| 824 "psllq $40, %%mm2 \n\t" /* GB000000 1 */\ | |
| 825 "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\ | |
| 826 MOVNTQ(%%mm0, (dst))\ | |
| 827 \ | |
| 828 "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\ | |
| 829 "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\ | |
| 830 "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\ | |
| 831 "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\ | |
| 832 MOVNTQ(%%mm6, 8(dst))\ | |
| 833 \ | |
| 834 "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\ | |
| 835 "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\ | |
| 836 "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\ | |
| 837 MOVNTQ(%%mm5, 16(dst))\ | |
| 838 \ | |
| 839 "add $24, "#dst" \n\t"\ | |
| 840 \ | |
| 841 "add $8, "#index" \n\t"\ | |
| 842 "cmp "#dstw", "#index" \n\t"\ | |
| 843 " jb 1b \n\t" | |
| 844 | |
| 845 #define WRITEBGR24MMX2(dst, dstw, index) \ | |
| 846 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 847 "movq "MANGLE(M24A)", %%mm0 \n\t"\ | |
| 848 "movq "MANGLE(M24C)", %%mm7 \n\t"\ | |
| 849 "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\ | |
| 850 "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\ | |
| 851 "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\ | |
| 852 \ | |
| 853 "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\ | |
| 854 "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\ | |
| 855 "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\ | |
| 856 \ | |
| 857 "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\ | |
| 858 "por %%mm1, %%mm6 \n\t"\ | |
| 859 "por %%mm3, %%mm6 \n\t"\ | |
| 860 MOVNTQ(%%mm6, (dst))\ | |
| 861 \ | |
| 862 "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\ | |
| 863 "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\ | |
| 864 "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\ | |
| 865 "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\ | |
| 866 \ | |
| 867 "pand "MANGLE(M24B)", %%mm1 \n\t" /* B5 B4 B3 */\ | |
| 868 "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\ | |
| 869 "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\ | |
| 870 \ | |
| 871 "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\ | |
| 872 "por %%mm3, %%mm6 \n\t"\ | |
| 873 MOVNTQ(%%mm6, 8(dst))\ | |
| 874 \ | |
| 875 "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\ | |
| 876 "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\ | |
| 877 "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\ | |
| 878 \ | |
| 879 "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\ | |
| 880 "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\ | |
| 881 "pand "MANGLE(M24B)", %%mm6 \n\t" /* R7 R6 R5 */\ | |
| 882 \ | |
| 883 "por %%mm1, %%mm3 \n\t"\ | |
| 884 "por %%mm3, %%mm6 \n\t"\ | |
| 885 MOVNTQ(%%mm6, 16(dst))\ | |
| 886 \ | |
| 887 "add $24, "#dst" \n\t"\ | |
| 888 \ | |
| 889 "add $8, "#index" \n\t"\ | |
| 890 "cmp "#dstw", "#index" \n\t"\ | |
| 891 " jb 1b \n\t" | |
| 892 | |
| 893 #ifdef HAVE_MMX2 | |
| 894 #undef WRITEBGR24 | |
| 895 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX2(dst, dstw, index) | |
| 896 #else | |
| 897 #undef WRITEBGR24 | |
| 898 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index) | |
| 899 #endif | |
| 900 | |
| 901 #define REAL_WRITEYUY2(dst, dstw, index) \ | |
| 902 "packuswb %%mm3, %%mm3 \n\t"\ | |
| 903 "packuswb %%mm4, %%mm4 \n\t"\ | |
| 904 "packuswb %%mm7, %%mm1 \n\t"\ | |
| 905 "punpcklbw %%mm4, %%mm3 \n\t"\ | |
| 906 "movq %%mm1, %%mm7 \n\t"\ | |
| 907 "punpcklbw %%mm3, %%mm1 \n\t"\ | |
| 908 "punpckhbw %%mm3, %%mm7 \n\t"\ | |
| 909 \ | |
| 910 MOVNTQ(%%mm1, (dst, index, 2))\ | |
| 911 MOVNTQ(%%mm7, 8(dst, index, 2))\ | |
| 912 \ | |
| 913 "add $8, "#index" \n\t"\ | |
| 914 "cmp "#dstw", "#index" \n\t"\ | |
| 915 " jb 1b \n\t" | |
| 916 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index) | |
| 917 | |
| 918 | |
| 919 static inline void RENAME(yuv2yuvX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
| 920 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
| 921 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, long dstW, long chrDstW) | |
| 922 { | |
| 923 #ifdef HAVE_MMX | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
924 if(c->flags & SWS_ACCURATE_RND){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
925 if(uDest){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
926 YSCALEYUV2YV12X_ACCURATE( 0, CHR_MMX_FILTER_OFFSET, uDest, chrDstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
927 YSCALEYUV2YV12X_ACCURATE(4096, CHR_MMX_FILTER_OFFSET, vDest, chrDstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
928 } |
| 18861 | 929 |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
930 YSCALEYUV2YV12X_ACCURATE(0, LUM_MMX_FILTER_OFFSET, dest, dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
931 }else{ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
932 if(uDest){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
933 YSCALEYUV2YV12X( 0, CHR_MMX_FILTER_OFFSET, uDest, chrDstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
934 YSCALEYUV2YV12X(4096, CHR_MMX_FILTER_OFFSET, vDest, chrDstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
935 } |
| 18861 | 936 |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
937 YSCALEYUV2YV12X(0, LUM_MMX_FILTER_OFFSET, dest, dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
938 } |
| 18861 | 939 #else |
| 940 #ifdef HAVE_ALTIVEC | |
| 941 yuv2yuvX_altivec_real(lumFilter, lumSrc, lumFilterSize, | |
| 942 chrFilter, chrSrc, chrFilterSize, | |
| 943 dest, uDest, vDest, dstW, chrDstW); | |
| 944 #else //HAVE_ALTIVEC | |
| 945 yuv2yuvXinC(lumFilter, lumSrc, lumFilterSize, | |
| 946 chrFilter, chrSrc, chrFilterSize, | |
| 947 dest, uDest, vDest, dstW, chrDstW); | |
| 948 #endif //!HAVE_ALTIVEC | |
| 949 #endif | |
| 950 } | |
| 951 | |
| 952 static inline void RENAME(yuv2nv12X)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
| 953 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
| 954 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat) | |
| 955 { | |
| 956 yuv2nv12XinC(lumFilter, lumSrc, lumFilterSize, | |
| 957 chrFilter, chrSrc, chrFilterSize, | |
| 958 dest, uDest, dstW, chrDstW, dstFormat); | |
| 959 } | |
| 960 | |
| 961 static inline void RENAME(yuv2yuv1)(int16_t *lumSrc, int16_t *chrSrc, | |
| 962 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, long dstW, long chrDstW) | |
| 963 { | |
| 964 #ifdef HAVE_MMX | |
| 965 if(uDest != NULL) | |
| 966 { | |
| 967 asm volatile( | |
| 968 YSCALEYUV2YV121 | |
| 969 :: "r" (chrSrc + chrDstW), "r" (uDest + chrDstW), | |
| 970 "g" (-chrDstW) | |
| 971 : "%"REG_a | |
| 972 ); | |
| 973 | |
| 974 asm volatile( | |
| 975 YSCALEYUV2YV121 | |
| 976 :: "r" (chrSrc + 2048 + chrDstW), "r" (vDest + chrDstW), | |
| 977 "g" (-chrDstW) | |
| 978 : "%"REG_a | |
| 979 ); | |
| 980 } | |
| 981 | |
| 982 asm volatile( | |
| 983 YSCALEYUV2YV121 | |
| 984 :: "r" (lumSrc + dstW), "r" (dest + dstW), | |
| 985 "g" (-dstW) | |
| 986 : "%"REG_a | |
| 987 ); | |
| 988 #else | |
| 989 int i; | |
| 990 for(i=0; i<dstW; i++) | |
| 991 { | |
| 992 int val= lumSrc[i]>>7; | |
| 993 | |
| 994 if(val&256){ | |
| 995 if(val<0) val=0; | |
| 996 else val=255; | |
| 997 } | |
| 998 | |
| 999 dest[i]= val; | |
| 1000 } | |
| 1001 | |
| 1002 if(uDest != NULL) | |
| 1003 for(i=0; i<chrDstW; i++) | |
| 1004 { | |
| 1005 int u=chrSrc[i]>>7; | |
| 1006 int v=chrSrc[i + 2048]>>7; | |
| 1007 | |
| 1008 if((u|v)&256){ | |
| 1009 if(u<0) u=0; | |
| 1010 else if (u>255) u=255; | |
| 1011 if(v<0) v=0; | |
| 1012 else if (v>255) v=255; | |
| 1013 } | |
| 1014 | |
| 1015 uDest[i]= u; | |
| 1016 vDest[i]= v; | |
| 1017 } | |
| 1018 #endif | |
| 1019 } | |
| 1020 | |
| 1021 | |
| 1022 /** | |
| 1023 * vertical scale YV12 to RGB | |
| 1024 */ | |
| 1025 static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
| 1026 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
| 1027 uint8_t *dest, long dstW, long dstY) | |
| 1028 { | |
| 1029 long dummy=0; | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1030 #ifdef HAVE_MMX |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1031 if(c->flags & SWS_ACCURATE_RND){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1032 switch(c->dstFormat){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1033 case IMGFMT_BGR32: |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1034 asm volatile( |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1035 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX_ACCURATE) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1036 WRITEBGR32(%4, %5, %%REGa) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1037 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1038 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1039 "m" (dummy), "m" (dummy), "m" (dummy), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1040 "r" (dest), "m" (dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1041 : "%"REG_a, "%"REG_d, "%"REG_S |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1042 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1043 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1044 case IMGFMT_BGR24: |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1045 asm volatile( |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1046 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX_ACCURATE) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1047 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t" //FIXME optimize |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1048 "add %4, %%"REG_b" \n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1049 WRITEBGR24(%%REGb, %5, %%REGa) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1050 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1051 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1052 "m" (dummy), "m" (dummy), "m" (dummy), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1053 "r" (dest), "m" (dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1054 : "%"REG_a, "%"REG_b, "%"REG_d, "%"REG_S //FIXME ebx |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1055 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1056 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1057 case IMGFMT_BGR15: |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1058 asm volatile( |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1059 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX_ACCURATE) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1060 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1061 #ifdef DITHER1XBPP |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1062 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1063 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1064 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1065 #endif |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1066 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1067 WRITEBGR15(%4, %5, %%REGa) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1068 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1069 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1070 "m" (dummy), "m" (dummy), "m" (dummy), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1071 "r" (dest), "m" (dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1072 : "%"REG_a, "%"REG_d, "%"REG_S |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1073 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1074 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1075 case IMGFMT_BGR16: |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1076 asm volatile( |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1077 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX_ACCURATE) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1078 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1079 #ifdef DITHER1XBPP |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1080 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1081 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1082 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1083 #endif |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1084 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1085 WRITEBGR16(%4, %5, %%REGa) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1086 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1087 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1088 "m" (dummy), "m" (dummy), "m" (dummy), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1089 "r" (dest), "m" (dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1090 : "%"REG_a, "%"REG_d, "%"REG_S |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1091 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1092 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1093 case IMGFMT_YUY2: |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1094 asm volatile( |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1095 YSCALEYUV2PACKEDX_ACCURATE |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1096 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1097 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1098 "psraw $3, %%mm3 \n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1099 "psraw $3, %%mm4 \n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1100 "psraw $3, %%mm1 \n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1101 "psraw $3, %%mm7 \n\t" |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1102 WRITEYUY2(%4, %5, %%REGa) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1103 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1104 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1105 "m" (dummy), "m" (dummy), "m" (dummy), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1106 "r" (dest), "m" (dstW) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1107 : "%"REG_a, "%"REG_d, "%"REG_S |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1108 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1109 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1110 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1111 }else{ |
| 18861 | 1112 switch(c->dstFormat) |
| 1113 { | |
| 1114 case IMGFMT_BGR32: | |
| 1115 { | |
| 1116 asm volatile( | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1117 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX) |
| 18861 | 1118 WRITEBGR32(%4, %5, %%REGa) |
| 1119 | |
| 1120 :: "r" (&c->redDither), | |
| 1121 "m" (dummy), "m" (dummy), "m" (dummy), | |
| 1122 "r" (dest), "m" (dstW) | |
| 1123 : "%"REG_a, "%"REG_d, "%"REG_S | |
| 1124 ); | |
| 1125 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1126 return; |
| 18861 | 1127 case IMGFMT_BGR24: |
| 1128 { | |
| 1129 asm volatile( | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1130 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX) |
| 18861 | 1131 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t" //FIXME optimize |
| 1132 "add %4, %%"REG_b" \n\t" | |
| 1133 WRITEBGR24(%%REGb, %5, %%REGa) | |
| 1134 | |
| 1135 :: "r" (&c->redDither), | |
| 1136 "m" (dummy), "m" (dummy), "m" (dummy), | |
| 1137 "r" (dest), "m" (dstW) | |
| 1138 : "%"REG_a, "%"REG_b, "%"REG_d, "%"REG_S //FIXME ebx | |
| 1139 ); | |
| 1140 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1141 return; |
| 18861 | 1142 case IMGFMT_BGR15: |
| 1143 { | |
| 1144 asm volatile( | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1145 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX) |
| 18861 | 1146 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ |
| 1147 #ifdef DITHER1XBPP | |
| 1148 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1149 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1150 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1151 #endif | |
| 1152 | |
| 1153 WRITEBGR15(%4, %5, %%REGa) | |
| 1154 | |
| 1155 :: "r" (&c->redDither), | |
| 1156 "m" (dummy), "m" (dummy), "m" (dummy), | |
| 1157 "r" (dest), "m" (dstW) | |
| 1158 : "%"REG_a, "%"REG_d, "%"REG_S | |
| 1159 ); | |
| 1160 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1161 return; |
| 18861 | 1162 case IMGFMT_BGR16: |
| 1163 { | |
| 1164 asm volatile( | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1165 YSCALEYUV2RGBX(YSCALEYUV2PACKEDX) |
| 18861 | 1166 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ |
| 1167 #ifdef DITHER1XBPP | |
| 1168 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1169 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1170 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1171 #endif | |
| 1172 | |
| 1173 WRITEBGR16(%4, %5, %%REGa) | |
| 1174 | |
| 1175 :: "r" (&c->redDither), | |
| 1176 "m" (dummy), "m" (dummy), "m" (dummy), | |
| 1177 "r" (dest), "m" (dstW) | |
| 1178 : "%"REG_a, "%"REG_d, "%"REG_S | |
| 1179 ); | |
| 1180 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1181 return; |
| 18861 | 1182 case IMGFMT_YUY2: |
| 1183 { | |
| 1184 asm volatile( | |
| 1185 YSCALEYUV2PACKEDX | |
| 1186 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1187 | |
| 1188 "psraw $3, %%mm3 \n\t" | |
| 1189 "psraw $3, %%mm4 \n\t" | |
| 1190 "psraw $3, %%mm1 \n\t" | |
| 1191 "psraw $3, %%mm7 \n\t" | |
| 1192 WRITEYUY2(%4, %5, %%REGa) | |
| 1193 | |
| 1194 :: "r" (&c->redDither), | |
| 1195 "m" (dummy), "m" (dummy), "m" (dummy), | |
| 1196 "r" (dest), "m" (dstW) | |
| 1197 : "%"REG_a, "%"REG_d, "%"REG_S | |
| 1198 ); | |
| 1199 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1200 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1201 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1202 } |
| 18861 | 1203 #endif |
| 1204 #ifdef HAVE_ALTIVEC | |
| 1205 /* The following list of supported dstFormat values should | |
| 1206 match what's found in the body of altivec_yuv2packedX() */ | |
| 1207 if(c->dstFormat==IMGFMT_ABGR || c->dstFormat==IMGFMT_BGRA || | |
| 1208 c->dstFormat==IMGFMT_BGR24 || c->dstFormat==IMGFMT_RGB24 || | |
| 1209 c->dstFormat==IMGFMT_RGBA || c->dstFormat==IMGFMT_ARGB) | |
| 1210 altivec_yuv2packedX (c, lumFilter, lumSrc, lumFilterSize, | |
| 1211 chrFilter, chrSrc, chrFilterSize, | |
| 1212 dest, dstW, dstY); | |
| 1213 else | |
| 1214 #endif | |
| 1215 yuv2packedXinC(c, lumFilter, lumSrc, lumFilterSize, | |
| 1216 chrFilter, chrSrc, chrFilterSize, | |
| 1217 dest, dstW, dstY); | |
| 1218 } | |
| 1219 | |
| 1220 /** | |
| 1221 * vertical bilinear scale YV12 to RGB | |
| 1222 */ | |
| 1223 static inline void RENAME(yuv2packed2)(SwsContext *c, uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1, | |
| 1224 uint8_t *dest, int dstW, int yalpha, int uvalpha, int y) | |
| 1225 { | |
| 1226 int yalpha1=yalpha^4095; | |
| 1227 int uvalpha1=uvalpha^4095; | |
| 1228 int i; | |
| 1229 | |
| 1230 #if 0 //isn't used | |
| 1231 if(flags&SWS_FULL_CHR_H_INT) | |
| 1232 { | |
| 1233 switch(dstFormat) | |
| 1234 { | |
| 1235 #ifdef HAVE_MMX | |
| 1236 case IMGFMT_BGR32: | |
| 1237 asm volatile( | |
| 1238 | |
| 1239 | |
| 1240 FULL_YSCALEYUV2RGB | |
| 1241 "punpcklbw %%mm1, %%mm3 \n\t" // BGBGBGBG | |
| 1242 "punpcklbw %%mm7, %%mm0 \n\t" // R0R0R0R0 | |
| 1243 | |
| 1244 "movq %%mm3, %%mm1 \n\t" | |
| 1245 "punpcklwd %%mm0, %%mm3 \n\t" // BGR0BGR0 | |
| 1246 "punpckhwd %%mm0, %%mm1 \n\t" // BGR0BGR0 | |
| 1247 | |
| 1248 MOVNTQ(%%mm3, (%4, %%REGa, 4)) | |
| 1249 MOVNTQ(%%mm1, 8(%4, %%REGa, 4)) | |
| 1250 | |
| 1251 "add $4, %%"REG_a" \n\t" | |
| 1252 "cmp %5, %%"REG_a" \n\t" | |
| 1253 " jb 1b \n\t" | |
| 1254 | |
| 1255 | |
| 1256 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" ((long)dstW), | |
| 1257 "m" (yalpha1), "m" (uvalpha1) | |
| 1258 : "%"REG_a | |
| 1259 ); | |
| 1260 break; | |
| 1261 case IMGFMT_BGR24: | |
| 1262 asm volatile( | |
| 1263 | |
| 1264 FULL_YSCALEYUV2RGB | |
| 1265 | |
| 1266 // lsb ... msb | |
| 1267 "punpcklbw %%mm1, %%mm3 \n\t" // BGBGBGBG | |
| 1268 "punpcklbw %%mm7, %%mm0 \n\t" // R0R0R0R0 | |
| 1269 | |
| 1270 "movq %%mm3, %%mm1 \n\t" | |
| 1271 "punpcklwd %%mm0, %%mm3 \n\t" // BGR0BGR0 | |
| 1272 "punpckhwd %%mm0, %%mm1 \n\t" // BGR0BGR0 | |
| 1273 | |
| 1274 "movq %%mm3, %%mm2 \n\t" // BGR0BGR0 | |
| 1275 "psrlq $8, %%mm3 \n\t" // GR0BGR00 | |
| 1276 "pand "MANGLE(bm00000111)", %%mm2\n\t" // BGR00000 | |
| 1277 "pand "MANGLE(bm11111000)", %%mm3\n\t" // 000BGR00 | |
| 1278 "por %%mm2, %%mm3 \n\t" // BGRBGR00 | |
| 1279 "movq %%mm1, %%mm2 \n\t" | |
| 1280 "psllq $48, %%mm1 \n\t" // 000000BG | |
| 1281 "por %%mm1, %%mm3 \n\t" // BGRBGRBG | |
| 1282 | |
| 1283 "movq %%mm2, %%mm1 \n\t" // BGR0BGR0 | |
| 1284 "psrld $16, %%mm2 \n\t" // R000R000 | |
| 1285 "psrlq $24, %%mm1 \n\t" // 0BGR0000 | |
| 1286 "por %%mm2, %%mm1 \n\t" // RBGRR000 | |
| 1287 | |
| 1288 "mov %4, %%"REG_b" \n\t" | |
| 1289 "add %%"REG_a", %%"REG_b" \n\t" | |
| 1290 | |
| 1291 #ifdef HAVE_MMX2 | |
| 1292 //FIXME Alignment | |
| 1293 "movntq %%mm3, (%%"REG_b", %%"REG_a", 2)\n\t" | |
| 1294 "movntq %%mm1, 8(%%"REG_b", %%"REG_a", 2)\n\t" | |
| 1295 #else | |
| 1296 "movd %%mm3, (%%"REG_b", %%"REG_a", 2) \n\t" | |
| 1297 "psrlq $32, %%mm3 \n\t" | |
| 1298 "movd %%mm3, 4(%%"REG_b", %%"REG_a", 2) \n\t" | |
| 1299 "movd %%mm1, 8(%%"REG_b", %%"REG_a", 2) \n\t" | |
| 1300 #endif | |
| 1301 "add $4, %%"REG_a" \n\t" | |
| 1302 "cmp %5, %%"REG_a" \n\t" | |
| 1303 " jb 1b \n\t" | |
| 1304 | |
| 1305 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "m" (dest), "m" (dstW), | |
| 1306 "m" (yalpha1), "m" (uvalpha1) | |
| 1307 : "%"REG_a, "%"REG_b | |
| 1308 ); | |
| 1309 break; | |
| 1310 case IMGFMT_BGR15: | |
| 1311 asm volatile( | |
| 1312 | |
| 1313 FULL_YSCALEYUV2RGB | |
| 1314 #ifdef DITHER1XBPP | |
| 1315 "paddusb "MANGLE(g5Dither)", %%mm1\n\t" | |
| 1316 "paddusb "MANGLE(r5Dither)", %%mm0\n\t" | |
| 1317 "paddusb "MANGLE(b5Dither)", %%mm3\n\t" | |
| 1318 #endif | |
| 1319 "punpcklbw %%mm7, %%mm1 \n\t" // 0G0G0G0G | |
| 1320 "punpcklbw %%mm7, %%mm3 \n\t" // 0B0B0B0B | |
| 1321 "punpcklbw %%mm7, %%mm0 \n\t" // 0R0R0R0R | |
| 1322 | |
| 1323 "psrlw $3, %%mm3 \n\t" | |
| 1324 "psllw $2, %%mm1 \n\t" | |
| 1325 "psllw $7, %%mm0 \n\t" | |
| 1326 "pand "MANGLE(g15Mask)", %%mm1 \n\t" | |
| 1327 "pand "MANGLE(r15Mask)", %%mm0 \n\t" | |
| 1328 | |
| 1329 "por %%mm3, %%mm1 \n\t" | |
| 1330 "por %%mm1, %%mm0 \n\t" | |
| 1331 | |
| 1332 MOVNTQ(%%mm0, (%4, %%REGa, 2)) | |
| 1333 | |
| 1334 "add $4, %%"REG_a" \n\t" | |
| 1335 "cmp %5, %%"REG_a" \n\t" | |
| 1336 " jb 1b \n\t" | |
| 1337 | |
| 1338 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstW), | |
| 1339 "m" (yalpha1), "m" (uvalpha1) | |
| 1340 : "%"REG_a | |
| 1341 ); | |
| 1342 break; | |
| 1343 case IMGFMT_BGR16: | |
| 1344 asm volatile( | |
| 1345 | |
| 1346 FULL_YSCALEYUV2RGB | |
| 1347 #ifdef DITHER1XBPP | |
| 1348 "paddusb "MANGLE(g6Dither)", %%mm1\n\t" | |
| 1349 "paddusb "MANGLE(r5Dither)", %%mm0\n\t" | |
| 1350 "paddusb "MANGLE(b5Dither)", %%mm3\n\t" | |
| 1351 #endif | |
| 1352 "punpcklbw %%mm7, %%mm1 \n\t" // 0G0G0G0G | |
| 1353 "punpcklbw %%mm7, %%mm3 \n\t" // 0B0B0B0B | |
| 1354 "punpcklbw %%mm7, %%mm0 \n\t" // 0R0R0R0R | |
| 1355 | |
| 1356 "psrlw $3, %%mm3 \n\t" | |
| 1357 "psllw $3, %%mm1 \n\t" | |
| 1358 "psllw $8, %%mm0 \n\t" | |
| 1359 "pand "MANGLE(g16Mask)", %%mm1 \n\t" | |
| 1360 "pand "MANGLE(r16Mask)", %%mm0 \n\t" | |
| 1361 | |
| 1362 "por %%mm3, %%mm1 \n\t" | |
| 1363 "por %%mm1, %%mm0 \n\t" | |
| 1364 | |
| 1365 MOVNTQ(%%mm0, (%4, %%REGa, 2)) | |
| 1366 | |
| 1367 "add $4, %%"REG_a" \n\t" | |
| 1368 "cmp %5, %%"REG_a" \n\t" | |
| 1369 " jb 1b \n\t" | |
| 1370 | |
| 1371 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstW), | |
| 1372 "m" (yalpha1), "m" (uvalpha1) | |
| 1373 : "%"REG_a | |
| 1374 ); | |
| 1375 break; | |
| 1376 #endif | |
| 1377 case IMGFMT_RGB32: | |
| 1378 #ifndef HAVE_MMX | |
| 1379 case IMGFMT_BGR32: | |
| 1380 #endif | |
| 1381 if(dstFormat==IMGFMT_BGR32) | |
| 1382 { | |
| 1383 int i; | |
| 1384 #ifdef WORDS_BIGENDIAN | |
| 1385 dest++; | |
| 1386 #endif | |
| 1387 for(i=0;i<dstW;i++){ | |
| 1388 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1389 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1390 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1391 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1392 dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)]; | |
| 1393 dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)]; | |
| 1394 dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)]; | |
| 1395 dest+= 4; | |
| 1396 } | |
| 1397 } | |
| 1398 else if(dstFormat==IMGFMT_BGR24) | |
| 1399 { | |
| 1400 int i; | |
| 1401 for(i=0;i<dstW;i++){ | |
| 1402 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1403 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1404 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1405 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1406 dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)]; | |
| 1407 dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)]; | |
| 1408 dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)]; | |
| 1409 dest+= 3; | |
| 1410 } | |
| 1411 } | |
| 1412 else if(dstFormat==IMGFMT_BGR16) | |
| 1413 { | |
| 1414 int i; | |
| 1415 for(i=0;i<dstW;i++){ | |
| 1416 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1417 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1418 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1419 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1420 | |
| 1421 ((uint16_t*)dest)[i] = | |
| 1422 clip_table16b[(Y + yuvtab_40cf[U]) >>13] | | |
| 1423 clip_table16g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] | | |
| 1424 clip_table16r[(Y + yuvtab_3343[V]) >>13]; | |
| 1425 } | |
| 1426 } | |
| 1427 else if(dstFormat==IMGFMT_BGR15) | |
| 1428 { | |
| 1429 int i; | |
| 1430 for(i=0;i<dstW;i++){ | |
| 1431 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1432 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1433 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1434 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1435 | |
| 1436 ((uint16_t*)dest)[i] = | |
| 1437 clip_table15b[(Y + yuvtab_40cf[U]) >>13] | | |
| 1438 clip_table15g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] | | |
| 1439 clip_table15r[(Y + yuvtab_3343[V]) >>13]; | |
| 1440 } | |
| 1441 } | |
| 1442 }//FULL_UV_IPOL | |
| 1443 else | |
| 1444 { | |
| 1445 #endif // if 0 | |
| 1446 #ifdef HAVE_MMX | |
| 1447 switch(c->dstFormat) | |
| 1448 { | |
| 1449 //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :( | |
| 1450 case IMGFMT_BGR32: | |
| 1451 asm volatile( | |
| 1452 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1453 "mov %4, %%"REG_b" \n\t" | |
| 1454 "push %%"REG_BP" \n\t" | |
| 1455 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1456 WRITEBGR32(%%REGb, 8280(%5), %%REGBP) | |
| 1457 "pop %%"REG_BP" \n\t" | |
| 1458 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1459 | |
| 1460 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1461 "a" (&c->redDither) | |
| 1462 ); | |
| 1463 return; | |
| 1464 case IMGFMT_BGR24: | |
| 1465 asm volatile( | |
| 1466 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1467 "mov %4, %%"REG_b" \n\t" | |
| 1468 "push %%"REG_BP" \n\t" | |
| 1469 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1470 WRITEBGR24(%%REGb, 8280(%5), %%REGBP) | |
| 1471 "pop %%"REG_BP" \n\t" | |
| 1472 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1473 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1474 "a" (&c->redDither) | |
| 1475 ); | |
| 1476 return; | |
| 1477 case IMGFMT_BGR15: | |
| 1478 asm volatile( | |
| 1479 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1480 "mov %4, %%"REG_b" \n\t" | |
| 1481 "push %%"REG_BP" \n\t" | |
| 1482 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1483 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1484 #ifdef DITHER1XBPP | |
| 1485 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1486 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1487 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1488 #endif | |
| 1489 | |
| 1490 WRITEBGR15(%%REGb, 8280(%5), %%REGBP) | |
| 1491 "pop %%"REG_BP" \n\t" | |
| 1492 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1493 | |
| 1494 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1495 "a" (&c->redDither) | |
| 1496 ); | |
| 1497 return; | |
| 1498 case IMGFMT_BGR16: | |
| 1499 asm volatile( | |
| 1500 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1501 "mov %4, %%"REG_b" \n\t" | |
| 1502 "push %%"REG_BP" \n\t" | |
| 1503 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1504 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1505 #ifdef DITHER1XBPP | |
| 1506 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1507 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1508 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1509 #endif | |
| 1510 | |
| 1511 WRITEBGR16(%%REGb, 8280(%5), %%REGBP) | |
| 1512 "pop %%"REG_BP" \n\t" | |
| 1513 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1514 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1515 "a" (&c->redDither) | |
| 1516 ); | |
| 1517 return; | |
| 1518 case IMGFMT_YUY2: | |
| 1519 asm volatile( | |
| 1520 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1521 "mov %4, %%"REG_b" \n\t" | |
| 1522 "push %%"REG_BP" \n\t" | |
| 1523 YSCALEYUV2PACKED(%%REGBP, %5) | |
| 1524 WRITEYUY2(%%REGb, 8280(%5), %%REGBP) | |
| 1525 "pop %%"REG_BP" \n\t" | |
| 1526 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1527 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1528 "a" (&c->redDither) | |
| 1529 ); | |
| 1530 return; | |
| 1531 default: break; | |
| 1532 } | |
| 1533 #endif //HAVE_MMX | |
| 1534 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C) | |
| 1535 } | |
| 1536 | |
| 1537 /** | |
| 1538 * YV12 to RGB without scaling or interpolating | |
| 1539 */ | |
| 1540 static inline void RENAME(yuv2packed1)(SwsContext *c, uint16_t *buf0, uint16_t *uvbuf0, uint16_t *uvbuf1, | |
| 1541 uint8_t *dest, int dstW, int uvalpha, int dstFormat, int flags, int y) | |
| 1542 { | |
| 1543 const int yalpha1=0; | |
| 1544 int i; | |
| 1545 | |
| 1546 uint16_t *buf1= buf0; //FIXME needed for the rgb1/bgr1 | |
| 1547 const int yalpha= 4096; //FIXME ... | |
| 1548 | |
| 1549 if(flags&SWS_FULL_CHR_H_INT) | |
| 1550 { | |
| 1551 RENAME(yuv2packed2)(c, buf0, buf0, uvbuf0, uvbuf1, dest, dstW, 0, uvalpha, y); | |
| 1552 return; | |
| 1553 } | |
| 1554 | |
| 1555 #ifdef HAVE_MMX | |
| 1556 if( uvalpha < 2048 ) // note this is not correct (shifts chrominance by 0.5 pixels) but its a bit faster | |
| 1557 { | |
| 1558 switch(dstFormat) | |
| 1559 { | |
| 1560 case IMGFMT_BGR32: | |
| 1561 asm volatile( | |
| 1562 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1563 "mov %4, %%"REG_b" \n\t" | |
| 1564 "push %%"REG_BP" \n\t" | |
| 1565 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1566 WRITEBGR32(%%REGb, 8280(%5), %%REGBP) | |
| 1567 "pop %%"REG_BP" \n\t" | |
| 1568 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1569 | |
| 1570 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1571 "a" (&c->redDither) | |
| 1572 ); | |
| 1573 return; | |
| 1574 case IMGFMT_BGR24: | |
| 1575 asm volatile( | |
| 1576 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1577 "mov %4, %%"REG_b" \n\t" | |
| 1578 "push %%"REG_BP" \n\t" | |
| 1579 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1580 WRITEBGR24(%%REGb, 8280(%5), %%REGBP) | |
| 1581 "pop %%"REG_BP" \n\t" | |
| 1582 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1583 | |
| 1584 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1585 "a" (&c->redDither) | |
| 1586 ); | |
| 1587 return; | |
| 1588 case IMGFMT_BGR15: | |
| 1589 asm volatile( | |
| 1590 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1591 "mov %4, %%"REG_b" \n\t" | |
| 1592 "push %%"REG_BP" \n\t" | |
| 1593 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1594 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1595 #ifdef DITHER1XBPP | |
| 1596 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1597 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1598 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1599 #endif | |
| 1600 WRITEBGR15(%%REGb, 8280(%5), %%REGBP) | |
| 1601 "pop %%"REG_BP" \n\t" | |
| 1602 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1603 | |
| 1604 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1605 "a" (&c->redDither) | |
| 1606 ); | |
| 1607 return; | |
| 1608 case IMGFMT_BGR16: | |
| 1609 asm volatile( | |
| 1610 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1611 "mov %4, %%"REG_b" \n\t" | |
| 1612 "push %%"REG_BP" \n\t" | |
| 1613 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1614 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1615 #ifdef DITHER1XBPP | |
| 1616 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1617 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1618 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1619 #endif | |
| 1620 | |
| 1621 WRITEBGR16(%%REGb, 8280(%5), %%REGBP) | |
| 1622 "pop %%"REG_BP" \n\t" | |
| 1623 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1624 | |
| 1625 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1626 "a" (&c->redDither) | |
| 1627 ); | |
| 1628 return; | |
| 1629 case IMGFMT_YUY2: | |
| 1630 asm volatile( | |
| 1631 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1632 "mov %4, %%"REG_b" \n\t" | |
| 1633 "push %%"REG_BP" \n\t" | |
| 1634 YSCALEYUV2PACKED1(%%REGBP, %5) | |
| 1635 WRITEYUY2(%%REGb, 8280(%5), %%REGBP) | |
| 1636 "pop %%"REG_BP" \n\t" | |
| 1637 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1638 | |
| 1639 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1640 "a" (&c->redDither) | |
| 1641 ); | |
| 1642 return; | |
| 1643 } | |
| 1644 } | |
| 1645 else | |
| 1646 { | |
| 1647 switch(dstFormat) | |
| 1648 { | |
| 1649 case IMGFMT_BGR32: | |
| 1650 asm volatile( | |
| 1651 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1652 "mov %4, %%"REG_b" \n\t" | |
| 1653 "push %%"REG_BP" \n\t" | |
| 1654 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1655 WRITEBGR32(%%REGb, 8280(%5), %%REGBP) | |
| 1656 "pop %%"REG_BP" \n\t" | |
| 1657 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1658 | |
| 1659 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1660 "a" (&c->redDither) | |
| 1661 ); | |
| 1662 return; | |
| 1663 case IMGFMT_BGR24: | |
| 1664 asm volatile( | |
| 1665 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1666 "mov %4, %%"REG_b" \n\t" | |
| 1667 "push %%"REG_BP" \n\t" | |
| 1668 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1669 WRITEBGR24(%%REGb, 8280(%5), %%REGBP) | |
| 1670 "pop %%"REG_BP" \n\t" | |
| 1671 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1672 | |
| 1673 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1674 "a" (&c->redDither) | |
| 1675 ); | |
| 1676 return; | |
| 1677 case IMGFMT_BGR15: | |
| 1678 asm volatile( | |
| 1679 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1680 "mov %4, %%"REG_b" \n\t" | |
| 1681 "push %%"REG_BP" \n\t" | |
| 1682 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1683 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1684 #ifdef DITHER1XBPP | |
| 1685 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1686 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1687 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1688 #endif | |
| 1689 WRITEBGR15(%%REGb, 8280(%5), %%REGBP) | |
| 1690 "pop %%"REG_BP" \n\t" | |
| 1691 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1692 | |
| 1693 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1694 "a" (&c->redDither) | |
| 1695 ); | |
| 1696 return; | |
| 1697 case IMGFMT_BGR16: | |
| 1698 asm volatile( | |
| 1699 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1700 "mov %4, %%"REG_b" \n\t" | |
| 1701 "push %%"REG_BP" \n\t" | |
| 1702 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1703 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1704 #ifdef DITHER1XBPP | |
| 1705 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1706 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1707 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1708 #endif | |
| 1709 | |
| 1710 WRITEBGR16(%%REGb, 8280(%5), %%REGBP) | |
| 1711 "pop %%"REG_BP" \n\t" | |
| 1712 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1713 | |
| 1714 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1715 "a" (&c->redDither) | |
| 1716 ); | |
| 1717 return; | |
| 1718 case IMGFMT_YUY2: | |
| 1719 asm volatile( | |
| 1720 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1721 "mov %4, %%"REG_b" \n\t" | |
| 1722 "push %%"REG_BP" \n\t" | |
| 1723 YSCALEYUV2PACKED1b(%%REGBP, %5) | |
| 1724 WRITEYUY2(%%REGb, 8280(%5), %%REGBP) | |
| 1725 "pop %%"REG_BP" \n\t" | |
| 1726 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1727 | |
| 1728 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1729 "a" (&c->redDither) | |
| 1730 ); | |
| 1731 return; | |
| 1732 } | |
| 1733 } | |
| 1734 #endif | |
| 1735 if( uvalpha < 2048 ) | |
| 1736 { | |
| 1737 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1_C, YSCALE_YUV_2_PACKED1_C) | |
| 1738 }else{ | |
| 1739 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1B_C, YSCALE_YUV_2_PACKED1B_C) | |
| 1740 } | |
| 1741 } | |
| 1742 | |
| 1743 //FIXME yuy2* can read upto 7 samples to much | |
| 1744 | |
| 1745 static inline void RENAME(yuy2ToY)(uint8_t *dst, uint8_t *src, long width) | |
| 1746 { | |
| 1747 #ifdef HAVE_MMX | |
| 1748 asm volatile( | |
| 1749 "movq "MANGLE(bm01010101)", %%mm2\n\t" | |
| 1750 "mov %0, %%"REG_a" \n\t" | |
| 1751 "1: \n\t" | |
| 1752 "movq (%1, %%"REG_a",2), %%mm0 \n\t" | |
| 1753 "movq 8(%1, %%"REG_a",2), %%mm1 \n\t" | |
| 1754 "pand %%mm2, %%mm0 \n\t" | |
| 1755 "pand %%mm2, %%mm1 \n\t" | |
| 1756 "packuswb %%mm1, %%mm0 \n\t" | |
| 1757 "movq %%mm0, (%2, %%"REG_a") \n\t" | |
| 1758 "add $8, %%"REG_a" \n\t" | |
| 1759 " js 1b \n\t" | |
| 1760 : : "g" (-width), "r" (src+width*2), "r" (dst+width) | |
| 1761 : "%"REG_a | |
| 1762 ); | |
| 1763 #else | |
| 1764 int i; | |
| 1765 for(i=0; i<width; i++) | |
| 1766 dst[i]= src[2*i]; | |
| 1767 #endif | |
| 1768 } | |
| 1769 | |
| 1770 static inline void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width) | |
| 1771 { | |
| 1772 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) | |
| 1773 asm volatile( | |
| 1774 "movq "MANGLE(bm01010101)", %%mm4\n\t" | |
| 1775 "mov %0, %%"REG_a" \n\t" | |
| 1776 "1: \n\t" | |
| 1777 "movq (%1, %%"REG_a",4), %%mm0 \n\t" | |
| 1778 "movq 8(%1, %%"REG_a",4), %%mm1 \n\t" | |
| 1779 "movq (%2, %%"REG_a",4), %%mm2 \n\t" | |
| 1780 "movq 8(%2, %%"REG_a",4), %%mm3 \n\t" | |
| 1781 PAVGB(%%mm2, %%mm0) | |
| 1782 PAVGB(%%mm3, %%mm1) | |
| 1783 "psrlw $8, %%mm0 \n\t" | |
| 1784 "psrlw $8, %%mm1 \n\t" | |
| 1785 "packuswb %%mm1, %%mm0 \n\t" | |
| 1786 "movq %%mm0, %%mm1 \n\t" | |
| 1787 "psrlw $8, %%mm0 \n\t" | |
| 1788 "pand %%mm4, %%mm1 \n\t" | |
| 1789 "packuswb %%mm0, %%mm0 \n\t" | |
| 1790 "packuswb %%mm1, %%mm1 \n\t" | |
| 1791 "movd %%mm0, (%4, %%"REG_a") \n\t" | |
| 1792 "movd %%mm1, (%3, %%"REG_a") \n\t" | |
| 1793 "add $4, %%"REG_a" \n\t" | |
| 1794 " js 1b \n\t" | |
| 1795 : : "g" (-width), "r" (src1+width*4), "r" (src2+width*4), "r" (dstU+width), "r" (dstV+width) | |
| 1796 : "%"REG_a | |
| 1797 ); | |
| 1798 #else | |
| 1799 int i; | |
| 1800 for(i=0; i<width; i++) | |
| 1801 { | |
| 1802 dstU[i]= (src1[4*i + 1] + src2[4*i + 1])>>1; | |
| 1803 dstV[i]= (src1[4*i + 3] + src2[4*i + 3])>>1; | |
| 1804 } | |
| 1805 #endif | |
| 1806 } | |
| 1807 | |
| 1808 //this is allmost identical to the previous, end exists only cuz yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses | |
| 1809 static inline void RENAME(uyvyToY)(uint8_t *dst, uint8_t *src, long width) | |
| 1810 { | |
| 1811 #ifdef HAVE_MMX | |
| 1812 asm volatile( | |
| 1813 "mov %0, %%"REG_a" \n\t" | |
| 1814 "1: \n\t" | |
| 1815 "movq (%1, %%"REG_a",2), %%mm0 \n\t" | |
| 1816 "movq 8(%1, %%"REG_a",2), %%mm1 \n\t" | |
| 1817 "psrlw $8, %%mm0 \n\t" | |
| 1818 "psrlw $8, %%mm1 \n\t" | |
| 1819 "packuswb %%mm1, %%mm0 \n\t" | |
| 1820 "movq %%mm0, (%2, %%"REG_a") \n\t" | |
| 1821 "add $8, %%"REG_a" \n\t" | |
| 1822 " js 1b \n\t" | |
| 1823 : : "g" (-width), "r" (src+width*2), "r" (dst+width) | |
| 1824 : "%"REG_a | |
| 1825 ); | |
| 1826 #else | |
| 1827 int i; | |
| 1828 for(i=0; i<width; i++) | |
| 1829 dst[i]= src[2*i+1]; | |
| 1830 #endif | |
| 1831 } | |
| 1832 | |
| 1833 static inline void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width) | |
| 1834 { | |
| 1835 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) | |
| 1836 asm volatile( | |
| 1837 "movq "MANGLE(bm01010101)", %%mm4\n\t" | |
| 1838 "mov %0, %%"REG_a" \n\t" | |
| 1839 "1: \n\t" | |
| 1840 "movq (%1, %%"REG_a",4), %%mm0 \n\t" | |
| 1841 "movq 8(%1, %%"REG_a",4), %%mm1 \n\t" | |
| 1842 "movq (%2, %%"REG_a",4), %%mm2 \n\t" | |
| 1843 "movq 8(%2, %%"REG_a",4), %%mm3 \n\t" | |
| 1844 PAVGB(%%mm2, %%mm0) | |
| 1845 PAVGB(%%mm3, %%mm1) | |
| 1846 "pand %%mm4, %%mm0 \n\t" | |
| 1847 "pand %%mm4, %%mm1 \n\t" | |
| 1848 "packuswb %%mm1, %%mm0 \n\t" | |
| 1849 "movq %%mm0, %%mm1 \n\t" | |
| 1850 "psrlw $8, %%mm0 \n\t" | |
| 1851 "pand %%mm4, %%mm1 \n\t" | |
| 1852 "packuswb %%mm0, %%mm0 \n\t" | |
| 1853 "packuswb %%mm1, %%mm1 \n\t" | |
| 1854 "movd %%mm0, (%4, %%"REG_a") \n\t" | |
| 1855 "movd %%mm1, (%3, %%"REG_a") \n\t" | |
| 1856 "add $4, %%"REG_a" \n\t" | |
| 1857 " js 1b \n\t" | |
| 1858 : : "g" (-width), "r" (src1+width*4), "r" (src2+width*4), "r" (dstU+width), "r" (dstV+width) | |
| 1859 : "%"REG_a | |
| 1860 ); | |
| 1861 #else | |
| 1862 int i; | |
| 1863 for(i=0; i<width; i++) | |
| 1864 { | |
| 1865 dstU[i]= (src1[4*i + 0] + src2[4*i + 0])>>1; | |
| 1866 dstV[i]= (src1[4*i + 2] + src2[4*i + 2])>>1; | |
| 1867 } | |
| 1868 #endif | |
| 1869 } | |
| 1870 | |
| 1871 static inline void RENAME(bgr32ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 1872 { | |
| 1873 int i; | |
| 1874 for(i=0; i<width; i++) | |
| 1875 { | |
| 1876 int b= ((uint32_t*)src)[i]&0xFF; | |
| 1877 int g= (((uint32_t*)src)[i]>>8)&0xFF; | |
| 1878 int r= (((uint32_t*)src)[i]>>16)&0xFF; | |
| 1879 | |
| 1880 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 1881 } | |
| 1882 } | |
| 1883 | |
| 1884 static inline void RENAME(bgr32ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 1885 { | |
| 1886 int i; | |
| 1887 for(i=0; i<width; i++) | |
| 1888 { | |
| 1889 const int a= ((uint32_t*)src1)[2*i+0]; | |
| 1890 const int e= ((uint32_t*)src1)[2*i+1]; | |
| 1891 const int c= ((uint32_t*)src2)[2*i+0]; | |
| 1892 const int d= ((uint32_t*)src2)[2*i+1]; | |
| 1893 const int l= (a&0xFF00FF) + (e&0xFF00FF) + (c&0xFF00FF) + (d&0xFF00FF); | |
| 1894 const int h= (a&0x00FF00) + (e&0x00FF00) + (c&0x00FF00) + (d&0x00FF00); | |
| 1895 const int b= l&0x3FF; | |
| 1896 const int g= h>>8; | |
| 1897 const int r= l>>16; | |
| 1898 | |
| 1899 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 1900 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 1901 } | |
| 1902 } | |
| 1903 | |
| 1904 static inline void RENAME(bgr24ToY)(uint8_t *dst, uint8_t *src, long width) | |
| 1905 { | |
| 1906 #ifdef HAVE_MMX | |
| 1907 asm volatile( | |
| 1908 "mov %2, %%"REG_a" \n\t" | |
| 1909 "movq "MANGLE(bgr2YCoeff)", %%mm6 \n\t" | |
| 1910 "movq "MANGLE(w1111)", %%mm5 \n\t" | |
| 1911 "pxor %%mm7, %%mm7 \n\t" | |
| 1912 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t" | |
| 1913 ASMALIGN16 | |
| 1914 "1: \n\t" | |
| 1915 PREFETCH" 64(%0, %%"REG_b") \n\t" | |
| 1916 "movd (%0, %%"REG_b"), %%mm0 \n\t" | |
| 1917 "movd 3(%0, %%"REG_b"), %%mm1 \n\t" | |
| 1918 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 1919 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 1920 "movd 6(%0, %%"REG_b"), %%mm2 \n\t" | |
| 1921 "movd 9(%0, %%"REG_b"), %%mm3 \n\t" | |
| 1922 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 1923 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 1924 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 1925 "pmaddwd %%mm6, %%mm1 \n\t" | |
| 1926 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 1927 "pmaddwd %%mm6, %%mm3 \n\t" | |
| 1928 #ifndef FAST_BGR2YV12 | |
| 1929 "psrad $8, %%mm0 \n\t" | |
| 1930 "psrad $8, %%mm1 \n\t" | |
| 1931 "psrad $8, %%mm2 \n\t" | |
| 1932 "psrad $8, %%mm3 \n\t" | |
| 1933 #endif | |
| 1934 "packssdw %%mm1, %%mm0 \n\t" | |
| 1935 "packssdw %%mm3, %%mm2 \n\t" | |
| 1936 "pmaddwd %%mm5, %%mm0 \n\t" | |
| 1937 "pmaddwd %%mm5, %%mm2 \n\t" | |
| 1938 "packssdw %%mm2, %%mm0 \n\t" | |
| 1939 "psraw $7, %%mm0 \n\t" | |
| 1940 | |
| 1941 "movd 12(%0, %%"REG_b"), %%mm4 \n\t" | |
| 1942 "movd 15(%0, %%"REG_b"), %%mm1 \n\t" | |
| 1943 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 1944 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 1945 "movd 18(%0, %%"REG_b"), %%mm2 \n\t" | |
| 1946 "movd 21(%0, %%"REG_b"), %%mm3 \n\t" | |
| 1947 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 1948 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 1949 "pmaddwd %%mm6, %%mm4 \n\t" | |
| 1950 "pmaddwd %%mm6, %%mm1 \n\t" | |
| 1951 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 1952 "pmaddwd %%mm6, %%mm3 \n\t" | |
| 1953 #ifndef FAST_BGR2YV12 | |
| 1954 "psrad $8, %%mm4 \n\t" | |
| 1955 "psrad $8, %%mm1 \n\t" | |
| 1956 "psrad $8, %%mm2 \n\t" | |
| 1957 "psrad $8, %%mm3 \n\t" | |
| 1958 #endif | |
| 1959 "packssdw %%mm1, %%mm4 \n\t" | |
| 1960 "packssdw %%mm3, %%mm2 \n\t" | |
| 1961 "pmaddwd %%mm5, %%mm4 \n\t" | |
| 1962 "pmaddwd %%mm5, %%mm2 \n\t" | |
| 1963 "add $24, %%"REG_b" \n\t" | |
| 1964 "packssdw %%mm2, %%mm4 \n\t" | |
| 1965 "psraw $7, %%mm4 \n\t" | |
| 1966 | |
| 1967 "packuswb %%mm4, %%mm0 \n\t" | |
| 1968 "paddusb "MANGLE(bgr2YOffset)", %%mm0 \n\t" | |
| 1969 | |
| 1970 "movq %%mm0, (%1, %%"REG_a") \n\t" | |
| 1971 "add $8, %%"REG_a" \n\t" | |
| 1972 " js 1b \n\t" | |
| 1973 : : "r" (src+width*3), "r" (dst+width), "g" (-width) | |
| 1974 : "%"REG_a, "%"REG_b | |
| 1975 ); | |
| 1976 #else | |
| 1977 int i; | |
| 1978 for(i=0; i<width; i++) | |
| 1979 { | |
| 1980 int b= src[i*3+0]; | |
| 1981 int g= src[i*3+1]; | |
| 1982 int r= src[i*3+2]; | |
| 1983 | |
| 1984 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 1985 } | |
| 1986 #endif | |
| 1987 } | |
| 1988 | |
| 1989 static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width) | |
| 1990 { | |
| 1991 #ifdef HAVE_MMX | |
| 1992 asm volatile( | |
| 1993 "mov %4, %%"REG_a" \n\t" | |
| 1994 "movq "MANGLE(w1111)", %%mm5 \n\t" | |
| 1995 "movq "MANGLE(bgr2UCoeff)", %%mm6 \n\t" | |
| 1996 "pxor %%mm7, %%mm7 \n\t" | |
| 1997 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_b" \n\t" | |
| 1998 "add %%"REG_b", %%"REG_b" \n\t" | |
| 1999 ASMALIGN16 | |
| 2000 "1: \n\t" | |
| 2001 PREFETCH" 64(%0, %%"REG_b") \n\t" | |
| 2002 PREFETCH" 64(%1, %%"REG_b") \n\t" | |
| 2003 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) | |
| 2004 "movq (%0, %%"REG_b"), %%mm0 \n\t" | |
| 2005 "movq (%1, %%"REG_b"), %%mm1 \n\t" | |
| 2006 "movq 6(%0, %%"REG_b"), %%mm2 \n\t" | |
| 2007 "movq 6(%1, %%"REG_b"), %%mm3 \n\t" | |
| 2008 PAVGB(%%mm1, %%mm0) | |
| 2009 PAVGB(%%mm3, %%mm2) | |
| 2010 "movq %%mm0, %%mm1 \n\t" | |
| 2011 "movq %%mm2, %%mm3 \n\t" | |
| 2012 "psrlq $24, %%mm0 \n\t" | |
| 2013 "psrlq $24, %%mm2 \n\t" | |
| 2014 PAVGB(%%mm1, %%mm0) | |
| 2015 PAVGB(%%mm3, %%mm2) | |
| 2016 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2017 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2018 #else | |
| 2019 "movd (%0, %%"REG_b"), %%mm0 \n\t" | |
| 2020 "movd (%1, %%"REG_b"), %%mm1 \n\t" | |
| 2021 "movd 3(%0, %%"REG_b"), %%mm2 \n\t" | |
| 2022 "movd 3(%1, %%"REG_b"), %%mm3 \n\t" | |
| 2023 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2024 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 2025 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2026 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 2027 "paddw %%mm1, %%mm0 \n\t" | |
| 2028 "paddw %%mm3, %%mm2 \n\t" | |
| 2029 "paddw %%mm2, %%mm0 \n\t" | |
| 2030 "movd 6(%0, %%"REG_b"), %%mm4 \n\t" | |
| 2031 "movd 6(%1, %%"REG_b"), %%mm1 \n\t" | |
| 2032 "movd 9(%0, %%"REG_b"), %%mm2 \n\t" | |
| 2033 "movd 9(%1, %%"REG_b"), %%mm3 \n\t" | |
| 2034 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 2035 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 2036 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2037 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 2038 "paddw %%mm1, %%mm4 \n\t" | |
| 2039 "paddw %%mm3, %%mm2 \n\t" | |
| 2040 "paddw %%mm4, %%mm2 \n\t" | |
| 2041 "psrlw $2, %%mm0 \n\t" | |
| 2042 "psrlw $2, %%mm2 \n\t" | |
| 2043 #endif | |
| 2044 "movq "MANGLE(bgr2VCoeff)", %%mm1 \n\t" | |
| 2045 "movq "MANGLE(bgr2VCoeff)", %%mm3 \n\t" | |
| 2046 | |
| 2047 "pmaddwd %%mm0, %%mm1 \n\t" | |
| 2048 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2049 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 2050 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 2051 #ifndef FAST_BGR2YV12 | |
| 2052 "psrad $8, %%mm0 \n\t" | |
| 2053 "psrad $8, %%mm1 \n\t" | |
| 2054 "psrad $8, %%mm2 \n\t" | |
| 2055 "psrad $8, %%mm3 \n\t" | |
| 2056 #endif | |
| 2057 "packssdw %%mm2, %%mm0 \n\t" | |
| 2058 "packssdw %%mm3, %%mm1 \n\t" | |
| 2059 "pmaddwd %%mm5, %%mm0 \n\t" | |
| 2060 "pmaddwd %%mm5, %%mm1 \n\t" | |
| 2061 "packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0 | |
| 2062 "psraw $7, %%mm0 \n\t" | |
| 2063 | |
| 2064 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) | |
| 2065 "movq 12(%0, %%"REG_b"), %%mm4 \n\t" | |
| 2066 "movq 12(%1, %%"REG_b"), %%mm1 \n\t" | |
| 2067 "movq 18(%0, %%"REG_b"), %%mm2 \n\t" | |
| 2068 "movq 18(%1, %%"REG_b"), %%mm3 \n\t" | |
| 2069 PAVGB(%%mm1, %%mm4) | |
| 2070 PAVGB(%%mm3, %%mm2) | |
| 2071 "movq %%mm4, %%mm1 \n\t" | |
| 2072 "movq %%mm2, %%mm3 \n\t" | |
| 2073 "psrlq $24, %%mm4 \n\t" | |
| 2074 "psrlq $24, %%mm2 \n\t" | |
| 2075 PAVGB(%%mm1, %%mm4) | |
| 2076 PAVGB(%%mm3, %%mm2) | |
| 2077 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 2078 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2079 #else | |
| 2080 "movd 12(%0, %%"REG_b"), %%mm4 \n\t" | |
| 2081 "movd 12(%1, %%"REG_b"), %%mm1 \n\t" | |
| 2082 "movd 15(%0, %%"REG_b"), %%mm2 \n\t" | |
| 2083 "movd 15(%1, %%"REG_b"), %%mm3 \n\t" | |
| 2084 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 2085 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 2086 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2087 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 2088 "paddw %%mm1, %%mm4 \n\t" | |
| 2089 "paddw %%mm3, %%mm2 \n\t" | |
| 2090 "paddw %%mm2, %%mm4 \n\t" | |
| 2091 "movd 18(%0, %%"REG_b"), %%mm5 \n\t" | |
| 2092 "movd 18(%1, %%"REG_b"), %%mm1 \n\t" | |
| 2093 "movd 21(%0, %%"REG_b"), %%mm2 \n\t" | |
| 2094 "movd 21(%1, %%"REG_b"), %%mm3 \n\t" | |
| 2095 "punpcklbw %%mm7, %%mm5 \n\t" | |
| 2096 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 2097 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2098 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 2099 "paddw %%mm1, %%mm5 \n\t" | |
| 2100 "paddw %%mm3, %%mm2 \n\t" | |
| 2101 "paddw %%mm5, %%mm2 \n\t" | |
| 2102 "movq "MANGLE(w1111)", %%mm5 \n\t" | |
| 2103 "psrlw $2, %%mm4 \n\t" | |
| 2104 "psrlw $2, %%mm2 \n\t" | |
| 2105 #endif | |
| 2106 "movq "MANGLE(bgr2VCoeff)", %%mm1 \n\t" | |
| 2107 "movq "MANGLE(bgr2VCoeff)", %%mm3 \n\t" | |
| 2108 | |
| 2109 "pmaddwd %%mm4, %%mm1 \n\t" | |
| 2110 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2111 "pmaddwd %%mm6, %%mm4 \n\t" | |
| 2112 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 2113 #ifndef FAST_BGR2YV12 | |
| 2114 "psrad $8, %%mm4 \n\t" | |
| 2115 "psrad $8, %%mm1 \n\t" | |
| 2116 "psrad $8, %%mm2 \n\t" | |
| 2117 "psrad $8, %%mm3 \n\t" | |
| 2118 #endif | |
| 2119 "packssdw %%mm2, %%mm4 \n\t" | |
| 2120 "packssdw %%mm3, %%mm1 \n\t" | |
| 2121 "pmaddwd %%mm5, %%mm4 \n\t" | |
| 2122 "pmaddwd %%mm5, %%mm1 \n\t" | |
| 2123 "add $24, %%"REG_b" \n\t" | |
| 2124 "packssdw %%mm1, %%mm4 \n\t" // V3 V2 U3 U2 | |
| 2125 "psraw $7, %%mm4 \n\t" | |
| 2126 | |
| 2127 "movq %%mm0, %%mm1 \n\t" | |
| 2128 "punpckldq %%mm4, %%mm0 \n\t" | |
| 2129 "punpckhdq %%mm4, %%mm1 \n\t" | |
| 2130 "packsswb %%mm1, %%mm0 \n\t" | |
| 2131 "paddb "MANGLE(bgr2UVOffset)", %%mm0 \n\t" | |
| 2132 | |
| 2133 "movd %%mm0, (%2, %%"REG_a") \n\t" | |
| 2134 "punpckhdq %%mm0, %%mm0 \n\t" | |
| 2135 "movd %%mm0, (%3, %%"REG_a") \n\t" | |
| 2136 "add $4, %%"REG_a" \n\t" | |
| 2137 " js 1b \n\t" | |
| 2138 : : "r" (src1+width*6), "r" (src2+width*6), "r" (dstU+width), "r" (dstV+width), "g" (-width) | |
| 2139 : "%"REG_a, "%"REG_b | |
| 2140 ); | |
| 2141 #else | |
| 2142 int i; | |
| 2143 for(i=0; i<width; i++) | |
| 2144 { | |
| 2145 int b= src1[6*i + 0] + src1[6*i + 3] + src2[6*i + 0] + src2[6*i + 3]; | |
| 2146 int g= src1[6*i + 1] + src1[6*i + 4] + src2[6*i + 1] + src2[6*i + 4]; | |
| 2147 int r= src1[6*i + 2] + src1[6*i + 5] + src2[6*i + 2] + src2[6*i + 5]; | |
| 2148 | |
| 2149 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 2150 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 2151 } | |
| 2152 #endif | |
| 2153 } | |
| 2154 | |
| 2155 static inline void RENAME(bgr16ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2156 { | |
| 2157 int i; | |
| 2158 for(i=0; i<width; i++) | |
| 2159 { | |
| 2160 int d= ((uint16_t*)src)[i]; | |
| 2161 int b= d&0x1F; | |
| 2162 int g= (d>>5)&0x3F; | |
| 2163 int r= (d>>11)&0x1F; | |
| 2164 | |
| 2165 dst[i]= ((2*RY*r + GY*g + 2*BY*b)>>(RGB2YUV_SHIFT-2)) + 16; | |
| 2166 } | |
| 2167 } | |
| 2168 | |
| 2169 static inline void RENAME(bgr16ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2170 { | |
| 2171 int i; | |
| 2172 for(i=0; i<width; i++) | |
| 2173 { | |
| 2174 int d0= ((uint32_t*)src1)[i]; | |
| 2175 int d1= ((uint32_t*)src2)[i]; | |
| 2176 | |
| 2177 int dl= (d0&0x07E0F81F) + (d1&0x07E0F81F); | |
| 2178 int dh= ((d0>>5)&0x07C0F83F) + ((d1>>5)&0x07C0F83F); | |
| 2179 | |
| 2180 int dh2= (dh>>11) + (dh<<21); | |
| 2181 int d= dh2 + dl; | |
| 2182 | |
| 2183 int b= d&0x7F; | |
| 2184 int r= (d>>11)&0x7F; | |
| 2185 int g= d>>21; | |
| 2186 dstU[i]= ((2*RU*r + GU*g + 2*BU*b)>>(RGB2YUV_SHIFT+2-2)) + 128; | |
| 2187 dstV[i]= ((2*RV*r + GV*g + 2*BV*b)>>(RGB2YUV_SHIFT+2-2)) + 128; | |
| 2188 } | |
| 2189 } | |
| 2190 | |
| 2191 static inline void RENAME(bgr15ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2192 { | |
| 2193 int i; | |
| 2194 for(i=0; i<width; i++) | |
| 2195 { | |
| 2196 int d= ((uint16_t*)src)[i]; | |
| 2197 int b= d&0x1F; | |
| 2198 int g= (d>>5)&0x1F; | |
| 2199 int r= (d>>10)&0x1F; | |
| 2200 | |
| 2201 dst[i]= ((RY*r + GY*g + BY*b)>>(RGB2YUV_SHIFT-3)) + 16; | |
| 2202 } | |
| 2203 } | |
| 2204 | |
| 2205 static inline void RENAME(bgr15ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2206 { | |
| 2207 int i; | |
| 2208 for(i=0; i<width; i++) | |
| 2209 { | |
| 2210 int d0= ((uint32_t*)src1)[i]; | |
| 2211 int d1= ((uint32_t*)src2)[i]; | |
| 2212 | |
| 2213 int dl= (d0&0x03E07C1F) + (d1&0x03E07C1F); | |
| 2214 int dh= ((d0>>5)&0x03E0F81F) + ((d1>>5)&0x03E0F81F); | |
| 2215 | |
| 2216 int dh2= (dh>>11) + (dh<<21); | |
| 2217 int d= dh2 + dl; | |
| 2218 | |
| 2219 int b= d&0x7F; | |
| 2220 int r= (d>>10)&0x7F; | |
| 2221 int g= d>>21; | |
| 2222 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2-3)) + 128; | |
| 2223 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2-3)) + 128; | |
| 2224 } | |
| 2225 } | |
| 2226 | |
| 2227 | |
| 2228 static inline void RENAME(rgb32ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2229 { | |
| 2230 int i; | |
| 2231 for(i=0; i<width; i++) | |
| 2232 { | |
| 2233 int r= ((uint32_t*)src)[i]&0xFF; | |
| 2234 int g= (((uint32_t*)src)[i]>>8)&0xFF; | |
| 2235 int b= (((uint32_t*)src)[i]>>16)&0xFF; | |
| 2236 | |
| 2237 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 2238 } | |
| 2239 } | |
| 2240 | |
| 2241 static inline void RENAME(rgb32ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2242 { | |
| 2243 int i; | |
| 2244 for(i=0; i<width; i++) | |
| 2245 { | |
| 2246 const int a= ((uint32_t*)src1)[2*i+0]; | |
| 2247 const int e= ((uint32_t*)src1)[2*i+1]; | |
| 2248 const int c= ((uint32_t*)src2)[2*i+0]; | |
| 2249 const int d= ((uint32_t*)src2)[2*i+1]; | |
| 2250 const int l= (a&0xFF00FF) + (e&0xFF00FF) + (c&0xFF00FF) + (d&0xFF00FF); | |
| 2251 const int h= (a&0x00FF00) + (e&0x00FF00) + (c&0x00FF00) + (d&0x00FF00); | |
| 2252 const int r= l&0x3FF; | |
| 2253 const int g= h>>8; | |
| 2254 const int b= l>>16; | |
| 2255 | |
| 2256 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 2257 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 2258 } | |
| 2259 } | |
| 2260 | |
| 2261 static inline void RENAME(rgb24ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2262 { | |
| 2263 int i; | |
| 2264 for(i=0; i<width; i++) | |
| 2265 { | |
| 2266 int r= src[i*3+0]; | |
| 2267 int g= src[i*3+1]; | |
| 2268 int b= src[i*3+2]; | |
| 2269 | |
| 2270 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 2271 } | |
| 2272 } | |
| 2273 | |
| 2274 static inline void RENAME(rgb24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2275 { | |
| 2276 int i; | |
| 2277 for(i=0; i<width; i++) | |
| 2278 { | |
| 2279 int r= src1[6*i + 0] + src1[6*i + 3] + src2[6*i + 0] + src2[6*i + 3]; | |
| 2280 int g= src1[6*i + 1] + src1[6*i + 4] + src2[6*i + 1] + src2[6*i + 4]; | |
| 2281 int b= src1[6*i + 2] + src1[6*i + 5] + src2[6*i + 2] + src2[6*i + 5]; | |
| 2282 | |
| 2283 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 2284 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2)) + 128; | |
| 2285 } | |
| 2286 } | |
| 2287 | |
| 2288 | |
| 2289 // Bilinear / Bicubic scaling | |
| 2290 static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc, | |
| 2291 int16_t *filter, int16_t *filterPos, long filterSize) | |
| 2292 { | |
| 2293 #ifdef HAVE_MMX | |
| 2294 assert(filterSize % 4 == 0 && filterSize>0); | |
| 2295 if(filterSize==4) // allways true for upscaling, sometimes for down too | |
| 2296 { | |
| 2297 long counter= -2*dstW; | |
| 2298 filter-= counter*2; | |
| 2299 filterPos-= counter/2; | |
| 2300 dst-= counter/2; | |
| 2301 asm volatile( | |
| 2302 "pxor %%mm7, %%mm7 \n\t" | |
| 2303 "movq "MANGLE(w02)", %%mm6 \n\t" | |
| 2304 "push %%"REG_BP" \n\t" // we use 7 regs here ... | |
| 2305 "mov %%"REG_a", %%"REG_BP" \n\t" | |
| 2306 ASMALIGN16 | |
| 2307 "1: \n\t" | |
| 2308 "movzwl (%2, %%"REG_BP"), %%eax \n\t" | |
| 2309 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t" | |
| 2310 "movq (%1, %%"REG_BP", 4), %%mm1\n\t" | |
| 2311 "movq 8(%1, %%"REG_BP", 4), %%mm3\n\t" | |
| 2312 "movd (%3, %%"REG_a"), %%mm0 \n\t" | |
| 2313 "movd (%3, %%"REG_b"), %%mm2 \n\t" | |
| 2314 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2315 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2316 "pmaddwd %%mm1, %%mm0 \n\t" | |
| 2317 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2318 "psrad $8, %%mm0 \n\t" | |
| 2319 "psrad $8, %%mm3 \n\t" | |
| 2320 "packssdw %%mm3, %%mm0 \n\t" | |
| 2321 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 2322 "packssdw %%mm0, %%mm0 \n\t" | |
| 2323 "movd %%mm0, (%4, %%"REG_BP") \n\t" | |
| 2324 "add $4, %%"REG_BP" \n\t" | |
| 2325 " jnc 1b \n\t" | |
| 2326 | |
| 2327 "pop %%"REG_BP" \n\t" | |
| 2328 : "+a" (counter) | |
| 2329 : "c" (filter), "d" (filterPos), "S" (src), "D" (dst) | |
| 2330 : "%"REG_b | |
| 2331 ); | |
| 2332 } | |
| 2333 else if(filterSize==8) | |
| 2334 { | |
| 2335 long counter= -2*dstW; | |
| 2336 filter-= counter*4; | |
| 2337 filterPos-= counter/2; | |
| 2338 dst-= counter/2; | |
| 2339 asm volatile( | |
| 2340 "pxor %%mm7, %%mm7 \n\t" | |
| 2341 "movq "MANGLE(w02)", %%mm6 \n\t" | |
| 2342 "push %%"REG_BP" \n\t" // we use 7 regs here ... | |
| 2343 "mov %%"REG_a", %%"REG_BP" \n\t" | |
| 2344 ASMALIGN16 | |
| 2345 "1: \n\t" | |
| 2346 "movzwl (%2, %%"REG_BP"), %%eax \n\t" | |
| 2347 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t" | |
| 2348 "movq (%1, %%"REG_BP", 8), %%mm1\n\t" | |
| 2349 "movq 16(%1, %%"REG_BP", 8), %%mm3\n\t" | |
| 2350 "movd (%3, %%"REG_a"), %%mm0 \n\t" | |
| 2351 "movd (%3, %%"REG_b"), %%mm2 \n\t" | |
| 2352 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2353 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2354 "pmaddwd %%mm1, %%mm0 \n\t" | |
| 2355 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2356 | |
| 2357 "movq 8(%1, %%"REG_BP", 8), %%mm1\n\t" | |
| 2358 "movq 24(%1, %%"REG_BP", 8), %%mm5\n\t" | |
| 2359 "movd 4(%3, %%"REG_a"), %%mm4 \n\t" | |
| 2360 "movd 4(%3, %%"REG_b"), %%mm2 \n\t" | |
| 2361 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 2362 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2363 "pmaddwd %%mm1, %%mm4 \n\t" | |
| 2364 "pmaddwd %%mm2, %%mm5 \n\t" | |
| 2365 "paddd %%mm4, %%mm0 \n\t" | |
| 2366 "paddd %%mm5, %%mm3 \n\t" | |
| 2367 | |
| 2368 "psrad $8, %%mm0 \n\t" | |
| 2369 "psrad $8, %%mm3 \n\t" | |
| 2370 "packssdw %%mm3, %%mm0 \n\t" | |
| 2371 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 2372 "packssdw %%mm0, %%mm0 \n\t" | |
| 2373 "movd %%mm0, (%4, %%"REG_BP") \n\t" | |
| 2374 "add $4, %%"REG_BP" \n\t" | |
| 2375 " jnc 1b \n\t" | |
| 2376 | |
| 2377 "pop %%"REG_BP" \n\t" | |
| 2378 : "+a" (counter) | |
| 2379 : "c" (filter), "d" (filterPos), "S" (src), "D" (dst) | |
| 2380 : "%"REG_b | |
| 2381 ); | |
| 2382 } | |
| 2383 else | |
| 2384 { | |
| 2385 uint8_t *offset = src+filterSize; | |
| 2386 long counter= -2*dstW; | |
| 2387 // filter-= counter*filterSize/2; | |
| 2388 filterPos-= counter/2; | |
| 2389 dst-= counter/2; | |
| 2390 asm volatile( | |
| 2391 "pxor %%mm7, %%mm7 \n\t" | |
| 2392 "movq "MANGLE(w02)", %%mm6 \n\t" | |
| 2393 ASMALIGN16 | |
| 2394 "1: \n\t" | |
| 2395 "mov %2, %%"REG_c" \n\t" | |
| 2396 "movzwl (%%"REG_c", %0), %%eax \n\t" | |
| 2397 "movzwl 2(%%"REG_c", %0), %%ebx \n\t" | |
| 2398 "mov %5, %%"REG_c" \n\t" | |
| 2399 "pxor %%mm4, %%mm4 \n\t" | |
| 2400 "pxor %%mm5, %%mm5 \n\t" | |
| 2401 "2: \n\t" | |
| 2402 "movq (%1), %%mm1 \n\t" | |
| 2403 "movq (%1, %6), %%mm3 \n\t" | |
| 2404 "movd (%%"REG_c", %%"REG_a"), %%mm0\n\t" | |
| 2405 "movd (%%"REG_c", %%"REG_b"), %%mm2\n\t" | |
| 2406 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2407 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2408 "pmaddwd %%mm1, %%mm0 \n\t" | |
| 2409 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2410 "paddd %%mm3, %%mm5 \n\t" | |
| 2411 "paddd %%mm0, %%mm4 \n\t" | |
| 2412 "add $8, %1 \n\t" | |
| 2413 "add $4, %%"REG_c" \n\t" | |
| 2414 "cmp %4, %%"REG_c" \n\t" | |
| 2415 " jb 2b \n\t" | |
| 2416 "add %6, %1 \n\t" | |
| 2417 "psrad $8, %%mm4 \n\t" | |
| 2418 "psrad $8, %%mm5 \n\t" | |
| 2419 "packssdw %%mm5, %%mm4 \n\t" | |
| 2420 "pmaddwd %%mm6, %%mm4 \n\t" | |
| 2421 "packssdw %%mm4, %%mm4 \n\t" | |
| 2422 "mov %3, %%"REG_a" \n\t" | |
| 2423 "movd %%mm4, (%%"REG_a", %0) \n\t" | |
| 2424 "add $4, %0 \n\t" | |
| 2425 " jnc 1b \n\t" | |
| 2426 | |
| 2427 : "+r" (counter), "+r" (filter) | |
| 2428 : "m" (filterPos), "m" (dst), "m"(offset), | |
| 2429 "m" (src), "r" (filterSize*2) | |
| 2430 : "%"REG_b, "%"REG_a, "%"REG_c | |
| 2431 ); | |
| 2432 } | |
| 2433 #else | |
| 2434 #ifdef HAVE_ALTIVEC | |
| 2435 hScale_altivec_real(dst, dstW, src, srcW, xInc, filter, filterPos, filterSize); | |
| 2436 #else | |
| 2437 int i; | |
| 2438 for(i=0; i<dstW; i++) | |
| 2439 { | |
| 2440 int j; | |
| 2441 int srcPos= filterPos[i]; | |
| 2442 int val=0; | |
| 2443 // printf("filterPos: %d\n", filterPos[i]); | |
| 2444 for(j=0; j<filterSize; j++) | |
| 2445 { | |
| 2446 // printf("filter: %d, src: %d\n", filter[i], src[srcPos + j]); | |
| 2447 val += ((int)src[srcPos + j])*filter[filterSize*i + j]; | |
| 2448 } | |
| 2449 // filter += hFilterSize; | |
| 2450 dst[i] = MIN(MAX(0, val>>7), (1<<15)-1); // the cubic equation does overflow ... | |
| 2451 // dst[i] = val>>7; | |
| 2452 } | |
| 2453 #endif | |
| 2454 #endif | |
| 2455 } | |
| 2456 // *** horizontal scale Y line to temp buffer | |
| 2457 static inline void RENAME(hyscale)(uint16_t *dst, long dstWidth, uint8_t *src, int srcW, int xInc, | |
| 2458 int flags, int canMMX2BeUsed, int16_t *hLumFilter, | |
| 2459 int16_t *hLumFilterPos, int hLumFilterSize, void *funnyYCode, | |
| 2460 int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter, | |
| 2461 int32_t *mmx2FilterPos) | |
| 2462 { | |
| 2463 if(srcFormat==IMGFMT_YUY2) | |
| 2464 { | |
| 2465 RENAME(yuy2ToY)(formatConvBuffer, src, srcW); | |
| 2466 src= formatConvBuffer; | |
| 2467 } | |
| 2468 else if(srcFormat==IMGFMT_UYVY) | |
| 2469 { | |
| 2470 RENAME(uyvyToY)(formatConvBuffer, src, srcW); | |
| 2471 src= formatConvBuffer; | |
| 2472 } | |
| 2473 else if(srcFormat==IMGFMT_BGR32) | |
| 2474 { | |
| 2475 RENAME(bgr32ToY)(formatConvBuffer, src, srcW); | |
| 2476 src= formatConvBuffer; | |
| 2477 } | |
| 2478 else if(srcFormat==IMGFMT_BGR24) | |
| 2479 { | |
| 2480 RENAME(bgr24ToY)(formatConvBuffer, src, srcW); | |
| 2481 src= formatConvBuffer; | |
| 2482 } | |
| 2483 else if(srcFormat==IMGFMT_BGR16) | |
| 2484 { | |
| 2485 RENAME(bgr16ToY)(formatConvBuffer, src, srcW); | |
| 2486 src= formatConvBuffer; | |
| 2487 } | |
| 2488 else if(srcFormat==IMGFMT_BGR15) | |
| 2489 { | |
| 2490 RENAME(bgr15ToY)(formatConvBuffer, src, srcW); | |
| 2491 src= formatConvBuffer; | |
| 2492 } | |
| 2493 else if(srcFormat==IMGFMT_RGB32) | |
| 2494 { | |
| 2495 RENAME(rgb32ToY)(formatConvBuffer, src, srcW); | |
| 2496 src= formatConvBuffer; | |
| 2497 } | |
| 2498 else if(srcFormat==IMGFMT_RGB24) | |
| 2499 { | |
| 2500 RENAME(rgb24ToY)(formatConvBuffer, src, srcW); | |
| 2501 src= formatConvBuffer; | |
| 2502 } | |
| 2503 | |
| 2504 #ifdef HAVE_MMX | |
| 2505 // use the new MMX scaler if the mmx2 can't be used (its faster than the x86asm one) | |
| 2506 if(!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed)) | |
| 2507 #else | |
| 2508 if(!(flags&SWS_FAST_BILINEAR)) | |
| 2509 #endif | |
| 2510 { | |
| 2511 RENAME(hScale)(dst, dstWidth, src, srcW, xInc, hLumFilter, hLumFilterPos, hLumFilterSize); | |
| 2512 } | |
| 2513 else // Fast Bilinear upscale / crap downscale | |
| 2514 { | |
| 2515 #if defined(ARCH_X86) || defined(ARCH_X86_64) | |
| 2516 #ifdef HAVE_MMX2 | |
| 2517 int i; | |
| 2518 if(canMMX2BeUsed) | |
| 2519 { | |
| 2520 asm volatile( | |
| 2521 "pxor %%mm7, %%mm7 \n\t" | |
| 2522 "mov %0, %%"REG_c" \n\t" | |
| 2523 "mov %1, %%"REG_D" \n\t" | |
| 2524 "mov %2, %%"REG_d" \n\t" | |
| 2525 "mov %3, %%"REG_b" \n\t" | |
| 2526 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2527 PREFETCH" (%%"REG_c") \n\t" | |
| 2528 PREFETCH" 32(%%"REG_c") \n\t" | |
| 2529 PREFETCH" 64(%%"REG_c") \n\t" | |
| 2530 | |
| 2531 #ifdef ARCH_X86_64 | |
| 2532 | |
| 2533 #define FUNNY_Y_CODE \ | |
| 2534 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2535 "call *%4 \n\t"\ | |
| 2536 "movl (%%"REG_b", %%"REG_a"), %%esi\n\t"\ | |
| 2537 "add %%"REG_S", %%"REG_c" \n\t"\ | |
| 2538 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2539 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2540 | |
| 2541 #else | |
| 2542 | |
| 2543 #define FUNNY_Y_CODE \ | |
| 2544 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2545 "call *%4 \n\t"\ | |
| 2546 "addl (%%"REG_b", %%"REG_a"), %%"REG_c"\n\t"\ | |
| 2547 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2548 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2549 | |
| 2550 #endif | |
| 2551 | |
| 2552 FUNNY_Y_CODE | |
| 2553 FUNNY_Y_CODE | |
| 2554 FUNNY_Y_CODE | |
| 2555 FUNNY_Y_CODE | |
| 2556 FUNNY_Y_CODE | |
| 2557 FUNNY_Y_CODE | |
| 2558 FUNNY_Y_CODE | |
| 2559 FUNNY_Y_CODE | |
| 2560 | |
| 2561 :: "m" (src), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos), | |
| 2562 "m" (funnyYCode) | |
| 2563 : "%"REG_a, "%"REG_b, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D | |
| 2564 ); | |
| 2565 for(i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) dst[i] = src[srcW-1]*128; | |
| 2566 } | |
| 2567 else | |
| 2568 { | |
| 2569 #endif | |
| 2570 long xInc_shr16 = xInc >> 16; | |
| 2571 uint16_t xInc_mask = xInc & 0xffff; | |
| 2572 //NO MMX just normal asm ... | |
| 2573 asm volatile( | |
| 2574 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2575 "xor %%"REG_b", %%"REG_b" \n\t" // xx | |
| 2576 "xorl %%ecx, %%ecx \n\t" // 2*xalpha | |
| 2577 ASMALIGN16 | |
| 2578 "1: \n\t" | |
| 2579 "movzbl (%0, %%"REG_b"), %%edi \n\t" //src[xx] | |
| 2580 "movzbl 1(%0, %%"REG_b"), %%esi \n\t" //src[xx+1] | |
| 2581 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] | |
| 2582 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2583 "shll $16, %%edi \n\t" | |
| 2584 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2585 "mov %1, %%"REG_D" \n\t" | |
| 2586 "shrl $9, %%esi \n\t" | |
| 2587 "movw %%si, (%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2588 "addw %4, %%cx \n\t" //2*xalpha += xInc&0xFF | |
| 2589 "adc %3, %%"REG_b" \n\t" //xx+= xInc>>8 + carry | |
| 2590 | |
| 2591 "movzbl (%0, %%"REG_b"), %%edi \n\t" //src[xx] | |
| 2592 "movzbl 1(%0, %%"REG_b"), %%esi \n\t" //src[xx+1] | |
| 2593 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] | |
| 2594 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2595 "shll $16, %%edi \n\t" | |
| 2596 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2597 "mov %1, %%"REG_D" \n\t" | |
| 2598 "shrl $9, %%esi \n\t" | |
| 2599 "movw %%si, 2(%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2600 "addw %4, %%cx \n\t" //2*xalpha += xInc&0xFF | |
| 2601 "adc %3, %%"REG_b" \n\t" //xx+= xInc>>8 + carry | |
| 2602 | |
| 2603 | |
| 2604 "add $2, %%"REG_a" \n\t" | |
| 2605 "cmp %2, %%"REG_a" \n\t" | |
| 2606 " jb 1b \n\t" | |
| 2607 | |
| 2608 | |
| 2609 :: "r" (src), "m" (dst), "m" (dstWidth), "m" (xInc_shr16), "m" (xInc_mask) | |
| 2610 : "%"REG_a, "%"REG_b, "%ecx", "%"REG_D, "%esi" | |
| 2611 ); | |
| 2612 #ifdef HAVE_MMX2 | |
| 2613 } //if MMX2 can't be used | |
| 2614 #endif | |
| 2615 #else | |
| 2616 int i; | |
| 2617 unsigned int xpos=0; | |
| 2618 for(i=0;i<dstWidth;i++) | |
| 2619 { | |
| 2620 register unsigned int xx=xpos>>16; | |
| 2621 register unsigned int xalpha=(xpos&0xFFFF)>>9; | |
| 2622 dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha; | |
| 2623 xpos+=xInc; | |
| 2624 } | |
| 2625 #endif | |
| 2626 } | |
| 2627 } | |
| 2628 | |
| 2629 inline static void RENAME(hcscale)(uint16_t *dst, long dstWidth, uint8_t *src1, uint8_t *src2, | |
| 2630 int srcW, int xInc, int flags, int canMMX2BeUsed, int16_t *hChrFilter, | |
| 2631 int16_t *hChrFilterPos, int hChrFilterSize, void *funnyUVCode, | |
| 2632 int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter, | |
| 2633 int32_t *mmx2FilterPos) | |
| 2634 { | |
| 2635 if(srcFormat==IMGFMT_YUY2) | |
| 2636 { | |
| 2637 RENAME(yuy2ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2638 src1= formatConvBuffer; | |
| 2639 src2= formatConvBuffer+2048; | |
| 2640 } | |
| 2641 else if(srcFormat==IMGFMT_UYVY) | |
| 2642 { | |
| 2643 RENAME(uyvyToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2644 src1= formatConvBuffer; | |
| 2645 src2= formatConvBuffer+2048; | |
| 2646 } | |
| 2647 else if(srcFormat==IMGFMT_BGR32) | |
| 2648 { | |
| 2649 RENAME(bgr32ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2650 src1= formatConvBuffer; | |
| 2651 src2= formatConvBuffer+2048; | |
| 2652 } | |
| 2653 else if(srcFormat==IMGFMT_BGR24) | |
| 2654 { | |
| 2655 RENAME(bgr24ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2656 src1= formatConvBuffer; | |
| 2657 src2= formatConvBuffer+2048; | |
| 2658 } | |
| 2659 else if(srcFormat==IMGFMT_BGR16) | |
| 2660 { | |
| 2661 RENAME(bgr16ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2662 src1= formatConvBuffer; | |
| 2663 src2= formatConvBuffer+2048; | |
| 2664 } | |
| 2665 else if(srcFormat==IMGFMT_BGR15) | |
| 2666 { | |
| 2667 RENAME(bgr15ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2668 src1= formatConvBuffer; | |
| 2669 src2= formatConvBuffer+2048; | |
| 2670 } | |
| 2671 else if(srcFormat==IMGFMT_RGB32) | |
| 2672 { | |
| 2673 RENAME(rgb32ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2674 src1= formatConvBuffer; | |
| 2675 src2= formatConvBuffer+2048; | |
| 2676 } | |
| 2677 else if(srcFormat==IMGFMT_RGB24) | |
| 2678 { | |
| 2679 RENAME(rgb24ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2680 src1= formatConvBuffer; | |
| 2681 src2= formatConvBuffer+2048; | |
| 2682 } | |
| 2683 else if(isGray(srcFormat)) | |
| 2684 { | |
| 2685 return; | |
| 2686 } | |
| 2687 | |
| 2688 #ifdef HAVE_MMX | |
| 2689 // use the new MMX scaler if the mmx2 can't be used (its faster than the x86asm one) | |
| 2690 if(!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed)) | |
| 2691 #else | |
| 2692 if(!(flags&SWS_FAST_BILINEAR)) | |
| 2693 #endif | |
| 2694 { | |
| 2695 RENAME(hScale)(dst , dstWidth, src1, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize); | |
| 2696 RENAME(hScale)(dst+2048, dstWidth, src2, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize); | |
| 2697 } | |
| 2698 else // Fast Bilinear upscale / crap downscale | |
| 2699 { | |
| 2700 #if defined(ARCH_X86) || defined(ARCH_X86_64) | |
| 2701 #ifdef HAVE_MMX2 | |
| 2702 int i; | |
| 2703 if(canMMX2BeUsed) | |
| 2704 { | |
| 2705 asm volatile( | |
| 2706 "pxor %%mm7, %%mm7 \n\t" | |
| 2707 "mov %0, %%"REG_c" \n\t" | |
| 2708 "mov %1, %%"REG_D" \n\t" | |
| 2709 "mov %2, %%"REG_d" \n\t" | |
| 2710 "mov %3, %%"REG_b" \n\t" | |
| 2711 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2712 PREFETCH" (%%"REG_c") \n\t" | |
| 2713 PREFETCH" 32(%%"REG_c") \n\t" | |
| 2714 PREFETCH" 64(%%"REG_c") \n\t" | |
| 2715 | |
| 2716 #ifdef ARCH_X86_64 | |
| 2717 | |
| 2718 #define FUNNY_UV_CODE \ | |
| 2719 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2720 "call *%4 \n\t"\ | |
| 2721 "movl (%%"REG_b", %%"REG_a"), %%esi\n\t"\ | |
| 2722 "add %%"REG_S", %%"REG_c" \n\t"\ | |
| 2723 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2724 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2725 | |
| 2726 #else | |
| 2727 | |
| 2728 #define FUNNY_UV_CODE \ | |
| 2729 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2730 "call *%4 \n\t"\ | |
| 2731 "addl (%%"REG_b", %%"REG_a"), %%"REG_c"\n\t"\ | |
| 2732 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2733 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2734 | |
| 2735 #endif | |
| 2736 | |
| 2737 FUNNY_UV_CODE | |
| 2738 FUNNY_UV_CODE | |
| 2739 FUNNY_UV_CODE | |
| 2740 FUNNY_UV_CODE | |
| 2741 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2742 "mov %5, %%"REG_c" \n\t" // src | |
| 2743 "mov %1, %%"REG_D" \n\t" // buf1 | |
| 2744 "add $4096, %%"REG_D" \n\t" | |
| 2745 PREFETCH" (%%"REG_c") \n\t" | |
| 2746 PREFETCH" 32(%%"REG_c") \n\t" | |
| 2747 PREFETCH" 64(%%"REG_c") \n\t" | |
| 2748 | |
| 2749 FUNNY_UV_CODE | |
| 2750 FUNNY_UV_CODE | |
| 2751 FUNNY_UV_CODE | |
| 2752 FUNNY_UV_CODE | |
| 2753 | |
| 2754 :: "m" (src1), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos), | |
| 2755 "m" (funnyUVCode), "m" (src2) | |
| 2756 : "%"REG_a, "%"REG_b, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D | |
| 2757 ); | |
| 2758 for(i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) | |
| 2759 { | |
| 2760 // printf("%d %d %d\n", dstWidth, i, srcW); | |
| 2761 dst[i] = src1[srcW-1]*128; | |
| 2762 dst[i+2048] = src2[srcW-1]*128; | |
| 2763 } | |
| 2764 } | |
| 2765 else | |
| 2766 { | |
| 2767 #endif | |
| 2768 long xInc_shr16 = (long) (xInc >> 16); | |
| 2769 uint16_t xInc_mask = xInc & 0xffff; | |
| 2770 asm volatile( | |
| 2771 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2772 "xor %%"REG_b", %%"REG_b" \n\t" // xx | |
| 2773 "xorl %%ecx, %%ecx \n\t" // 2*xalpha | |
| 2774 ASMALIGN16 | |
| 2775 "1: \n\t" | |
| 2776 "mov %0, %%"REG_S" \n\t" | |
| 2777 "movzbl (%%"REG_S", %%"REG_b"), %%edi \n\t" //src[xx] | |
| 2778 "movzbl 1(%%"REG_S", %%"REG_b"), %%esi \n\t" //src[xx+1] | |
| 2779 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] | |
| 2780 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2781 "shll $16, %%edi \n\t" | |
| 2782 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2783 "mov %1, %%"REG_D" \n\t" | |
| 2784 "shrl $9, %%esi \n\t" | |
| 2785 "movw %%si, (%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2786 | |
| 2787 "movzbl (%5, %%"REG_b"), %%edi \n\t" //src[xx] | |
| 2788 "movzbl 1(%5, %%"REG_b"), %%esi \n\t" //src[xx+1] | |
| 2789 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] | |
| 2790 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2791 "shll $16, %%edi \n\t" | |
| 2792 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2793 "mov %1, %%"REG_D" \n\t" | |
| 2794 "shrl $9, %%esi \n\t" | |
| 2795 "movw %%si, 4096(%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2796 | |
| 2797 "addw %4, %%cx \n\t" //2*xalpha += xInc&0xFF | |
| 2798 "adc %3, %%"REG_b" \n\t" //xx+= xInc>>8 + carry | |
| 2799 "add $1, %%"REG_a" \n\t" | |
| 2800 "cmp %2, %%"REG_a" \n\t" | |
| 2801 " jb 1b \n\t" | |
| 2802 | |
| 2803 /* GCC-3.3 makes MPlayer crash on IA-32 machines when using "g" operand here, | |
| 2804 which is needed to support GCC-4.0 */ | |
| 2805 #if defined(ARCH_X86_64) && ((__GNUC__ > 3) || ( __GNUC__ == 3 && __GNUC_MINOR__ >= 4)) | |
| 2806 :: "m" (src1), "m" (dst), "g" ((long)dstWidth), "m" (xInc_shr16), "m" (xInc_mask), | |
| 2807 #else | |
| 2808 :: "m" (src1), "m" (dst), "m" ((long)dstWidth), "m" (xInc_shr16), "m" (xInc_mask), | |
| 2809 #endif | |
| 2810 "r" (src2) | |
| 2811 : "%"REG_a, "%"REG_b, "%ecx", "%"REG_D, "%esi" | |
| 2812 ); | |
| 2813 #ifdef HAVE_MMX2 | |
| 2814 } //if MMX2 can't be used | |
| 2815 #endif | |
| 2816 #else | |
| 2817 int i; | |
| 2818 unsigned int xpos=0; | |
| 2819 for(i=0;i<dstWidth;i++) | |
| 2820 { | |
| 2821 register unsigned int xx=xpos>>16; | |
| 2822 register unsigned int xalpha=(xpos&0xFFFF)>>9; | |
| 2823 dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha); | |
| 2824 dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha); | |
| 2825 /* slower | |
| 2826 dst[i]= (src1[xx]<<7) + (src1[xx+1] - src1[xx])*xalpha; | |
| 2827 dst[i+2048]=(src2[xx]<<7) + (src2[xx+1] - src2[xx])*xalpha; | |
| 2828 */ | |
| 2829 xpos+=xInc; | |
| 2830 } | |
| 2831 #endif | |
| 2832 } | |
| 2833 } | |
| 2834 | |
| 2835 static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
| 2836 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
| 2837 | |
| 2838 /* load a few things into local vars to make the code more readable? and faster */ | |
| 2839 const int srcW= c->srcW; | |
| 2840 const int dstW= c->dstW; | |
| 2841 const int dstH= c->dstH; | |
| 2842 const int chrDstW= c->chrDstW; | |
| 2843 const int chrSrcW= c->chrSrcW; | |
| 2844 const int lumXInc= c->lumXInc; | |
| 2845 const int chrXInc= c->chrXInc; | |
| 2846 const int dstFormat= c->dstFormat; | |
| 2847 const int srcFormat= c->srcFormat; | |
| 2848 const int flags= c->flags; | |
| 2849 const int canMMX2BeUsed= c->canMMX2BeUsed; | |
| 2850 int16_t *vLumFilterPos= c->vLumFilterPos; | |
| 2851 int16_t *vChrFilterPos= c->vChrFilterPos; | |
| 2852 int16_t *hLumFilterPos= c->hLumFilterPos; | |
| 2853 int16_t *hChrFilterPos= c->hChrFilterPos; | |
| 2854 int16_t *vLumFilter= c->vLumFilter; | |
| 2855 int16_t *vChrFilter= c->vChrFilter; | |
| 2856 int16_t *hLumFilter= c->hLumFilter; | |
| 2857 int16_t *hChrFilter= c->hChrFilter; | |
| 2858 int32_t *lumMmxFilter= c->lumMmxFilter; | |
| 2859 int32_t *chrMmxFilter= c->chrMmxFilter; | |
| 2860 const int vLumFilterSize= c->vLumFilterSize; | |
| 2861 const int vChrFilterSize= c->vChrFilterSize; | |
| 2862 const int hLumFilterSize= c->hLumFilterSize; | |
| 2863 const int hChrFilterSize= c->hChrFilterSize; | |
| 2864 int16_t **lumPixBuf= c->lumPixBuf; | |
| 2865 int16_t **chrPixBuf= c->chrPixBuf; | |
| 2866 const int vLumBufSize= c->vLumBufSize; | |
| 2867 const int vChrBufSize= c->vChrBufSize; | |
| 2868 uint8_t *funnyYCode= c->funnyYCode; | |
| 2869 uint8_t *funnyUVCode= c->funnyUVCode; | |
| 2870 uint8_t *formatConvBuffer= c->formatConvBuffer; | |
| 2871 const int chrSrcSliceY= srcSliceY >> c->chrSrcVSubSample; | |
| 2872 const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample); | |
| 2873 int lastDstY; | |
| 2874 | |
| 2875 /* vars whch will change and which we need to storw back in the context */ | |
| 2876 int dstY= c->dstY; | |
| 2877 int lumBufIndex= c->lumBufIndex; | |
| 2878 int chrBufIndex= c->chrBufIndex; | |
| 2879 int lastInLumBuf= c->lastInLumBuf; | |
| 2880 int lastInChrBuf= c->lastInChrBuf; | |
| 2881 | |
| 2882 if(isPacked(c->srcFormat)){ | |
| 2883 src[0]= | |
| 2884 src[1]= | |
| 2885 src[2]= src[0]; | |
| 2886 srcStride[0]= | |
| 2887 srcStride[1]= | |
| 2888 srcStride[2]= srcStride[0]; | |
| 2889 } | |
| 2890 srcStride[1]<<= c->vChrDrop; | |
| 2891 srcStride[2]<<= c->vChrDrop; | |
| 2892 | |
| 2893 // printf("swscale %X %X %X -> %X %X %X\n", (int)src[0], (int)src[1], (int)src[2], | |
| 2894 // (int)dst[0], (int)dst[1], (int)dst[2]); | |
| 2895 | |
| 2896 #if 0 //self test FIXME move to a vfilter or something | |
| 2897 { | |
| 2898 static volatile int i=0; | |
| 2899 i++; | |
| 2900 if(srcFormat==IMGFMT_YV12 && i==1 && srcSliceH>= c->srcH) | |
| 2901 selfTest(src, srcStride, c->srcW, c->srcH); | |
| 2902 i--; | |
| 2903 } | |
| 2904 #endif | |
| 2905 | |
| 2906 //printf("sws Strides:%d %d %d -> %d %d %d\n", srcStride[0],srcStride[1],srcStride[2], | |
| 2907 //dstStride[0],dstStride[1],dstStride[2]); | |
| 2908 | |
| 2909 if(dstStride[0]%8 !=0 || dstStride[1]%8 !=0 || dstStride[2]%8 !=0) | |
| 2910 { | |
| 2911 static int firstTime=1; //FIXME move this into the context perhaps | |
| 2912 if(flags & SWS_PRINT_INFO && firstTime) | |
| 2913 { | |
| 2914 MSG_WARN("SwScaler: Warning: dstStride is not aligned!\n" | |
| 2915 "SwScaler: ->cannot do aligned memory acesses anymore\n"); | |
| 2916 firstTime=0; | |
| 2917 } | |
| 2918 } | |
| 2919 | |
| 2920 /* Note the user might start scaling the picture in the middle so this will not get executed | |
| 2921 this is not really intended but works currently, so ppl might do it */ | |
| 2922 if(srcSliceY ==0){ | |
| 2923 lumBufIndex=0; | |
| 2924 chrBufIndex=0; | |
| 2925 dstY=0; | |
| 2926 lastInLumBuf= -1; | |
| 2927 lastInChrBuf= -1; | |
| 2928 } | |
| 2929 | |
| 2930 lastDstY= dstY; | |
| 2931 | |
| 2932 for(;dstY < dstH; dstY++){ | |
| 2933 unsigned char *dest =dst[0]+dstStride[0]*dstY; | |
| 2934 const int chrDstY= dstY>>c->chrDstVSubSample; | |
| 2935 unsigned char *uDest=dst[1]+dstStride[1]*chrDstY; | |
| 2936 unsigned char *vDest=dst[2]+dstStride[2]*chrDstY; | |
| 2937 | |
| 2938 const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input | |
| 2939 const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input | |
| 2940 const int lastLumSrcY= firstLumSrcY + vLumFilterSize -1; // Last line needed as input | |
| 2941 const int lastChrSrcY= firstChrSrcY + vChrFilterSize -1; // Last line needed as input | |
| 2942 | |
| 2943 //printf("dstY:%d dstH:%d firstLumSrcY:%d lastInLumBuf:%d vLumBufSize: %d vChrBufSize: %d slice: %d %d vLumFilterSize: %d firstChrSrcY: %d vChrFilterSize: %d c->chrSrcVSubSample: %d\n", | |
| 2944 // dstY, dstH, firstLumSrcY, lastInLumBuf, vLumBufSize, vChrBufSize, srcSliceY, srcSliceH, vLumFilterSize, firstChrSrcY, vChrFilterSize, c->chrSrcVSubSample); | |
| 2945 //handle holes (FAST_BILINEAR & weird filters) | |
| 2946 if(firstLumSrcY > lastInLumBuf) lastInLumBuf= firstLumSrcY-1; | |
| 2947 if(firstChrSrcY > lastInChrBuf) lastInChrBuf= firstChrSrcY-1; | |
| 2948 //printf("%d %d %d\n", firstChrSrcY, lastInChrBuf, vChrBufSize); | |
| 2949 ASSERT(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1) | |
| 2950 ASSERT(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1) | |
| 2951 | |
| 2952 // Do we have enough lines in this slice to output the dstY line | |
| 2953 if(lastLumSrcY < srcSliceY + srcSliceH && lastChrSrcY < -((-srcSliceY - srcSliceH)>>c->chrSrcVSubSample)) | |
| 2954 { | |
| 2955 //Do horizontal scaling | |
| 2956 while(lastInLumBuf < lastLumSrcY) | |
| 2957 { | |
| 2958 uint8_t *s= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0]; | |
| 2959 lumBufIndex++; | |
| 2960 // printf("%d %d %d %d\n", lumBufIndex, vLumBufSize, lastInLumBuf, lastLumSrcY); | |
| 2961 ASSERT(lumBufIndex < 2*vLumBufSize) | |
| 2962 ASSERT(lastInLumBuf + 1 - srcSliceY < srcSliceH) | |
| 2963 ASSERT(lastInLumBuf + 1 - srcSliceY >= 0) | |
| 2964 // printf("%d %d\n", lumBufIndex, vLumBufSize); | |
| 2965 RENAME(hyscale)(lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc, | |
| 2966 flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize, | |
| 2967 funnyYCode, c->srcFormat, formatConvBuffer, | |
| 2968 c->lumMmx2Filter, c->lumMmx2FilterPos); | |
| 2969 lastInLumBuf++; | |
| 2970 } | |
| 2971 while(lastInChrBuf < lastChrSrcY) | |
| 2972 { | |
| 2973 uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1]; | |
| 2974 uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2]; | |
| 2975 chrBufIndex++; | |
| 2976 ASSERT(chrBufIndex < 2*vChrBufSize) | |
| 2977 ASSERT(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH)) | |
| 2978 ASSERT(lastInChrBuf + 1 - chrSrcSliceY >= 0) | |
| 2979 //FIXME replace parameters through context struct (some at least) | |
| 2980 | |
| 2981 if(!(isGray(srcFormat) || isGray(dstFormat))) | |
| 2982 RENAME(hcscale)(chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc, | |
| 2983 flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize, | |
| 2984 funnyUVCode, c->srcFormat, formatConvBuffer, | |
| 2985 c->chrMmx2Filter, c->chrMmx2FilterPos); | |
| 2986 lastInChrBuf++; | |
| 2987 } | |
| 2988 //wrap buf index around to stay inside the ring buffer | |
| 2989 if(lumBufIndex >= vLumBufSize ) lumBufIndex-= vLumBufSize; | |
| 2990 if(chrBufIndex >= vChrBufSize ) chrBufIndex-= vChrBufSize; | |
| 2991 } | |
| 2992 else // not enough lines left in this slice -> load the rest in the buffer | |
| 2993 { | |
| 2994 /* printf("%d %d Last:%d %d LastInBuf:%d %d Index:%d %d Y:%d FSize: %d %d BSize: %d %d\n", | |
| 2995 firstChrSrcY,firstLumSrcY,lastChrSrcY,lastLumSrcY, | |
| 2996 lastInChrBuf,lastInLumBuf,chrBufIndex,lumBufIndex,dstY,vChrFilterSize,vLumFilterSize, | |
| 2997 vChrBufSize, vLumBufSize);*/ | |
| 2998 | |
| 2999 //Do horizontal scaling | |
| 3000 while(lastInLumBuf+1 < srcSliceY + srcSliceH) | |
| 3001 { | |
| 3002 uint8_t *s= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0]; | |
| 3003 lumBufIndex++; | |
| 3004 ASSERT(lumBufIndex < 2*vLumBufSize) | |
| 3005 ASSERT(lastInLumBuf + 1 - srcSliceY < srcSliceH) | |
| 3006 ASSERT(lastInLumBuf + 1 - srcSliceY >= 0) | |
| 3007 RENAME(hyscale)(lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc, | |
| 3008 flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize, | |
| 3009 funnyYCode, c->srcFormat, formatConvBuffer, | |
| 3010 c->lumMmx2Filter, c->lumMmx2FilterPos); | |
| 3011 lastInLumBuf++; | |
| 3012 } | |
| 3013 while(lastInChrBuf+1 < (chrSrcSliceY + chrSrcSliceH)) | |
| 3014 { | |
| 3015 uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1]; | |
| 3016 uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2]; | |
| 3017 chrBufIndex++; | |
| 3018 ASSERT(chrBufIndex < 2*vChrBufSize) | |
| 3019 ASSERT(lastInChrBuf + 1 - chrSrcSliceY < chrSrcSliceH) | |
| 3020 ASSERT(lastInChrBuf + 1 - chrSrcSliceY >= 0) | |
| 3021 | |
| 3022 if(!(isGray(srcFormat) || isGray(dstFormat))) | |
| 3023 RENAME(hcscale)(chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc, | |
| 3024 flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize, | |
| 3025 funnyUVCode, c->srcFormat, formatConvBuffer, | |
| 3026 c->chrMmx2Filter, c->chrMmx2FilterPos); | |
| 3027 lastInChrBuf++; | |
| 3028 } | |
| 3029 //wrap buf index around to stay inside the ring buffer | |
| 3030 if(lumBufIndex >= vLumBufSize ) lumBufIndex-= vLumBufSize; | |
| 3031 if(chrBufIndex >= vChrBufSize ) chrBufIndex-= vChrBufSize; | |
| 3032 break; //we can't output a dstY line so let's try with the next slice | |
| 3033 } | |
| 3034 | |
| 3035 #ifdef HAVE_MMX | |
| 3036 b5Dither= dither8[dstY&1]; | |
| 3037 g6Dither= dither4[dstY&1]; | |
| 3038 g5Dither= dither8[dstY&1]; | |
| 3039 r5Dither= dither8[(dstY+1)&1]; | |
| 3040 #endif | |
| 3041 if(dstY < dstH-2) | |
| 3042 { | |
| 3043 int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize; | |
| 3044 int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; | |
| 3045 #ifdef HAVE_MMX | |
| 3046 int i; | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3047 if(flags & SWS_ACCURATE_RND){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3048 for(i=0; i<vLumFilterSize; i+=2){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3049 lumMmxFilter[2*i+0]= lumSrcPtr[i ]; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3050 lumMmxFilter[2*i+1]= lumSrcPtr[i+(vLumFilterSize>1)]; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3051 lumMmxFilter[2*i+2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3052 lumMmxFilter[2*i+3]= vLumFilter[dstY*vLumFilterSize + i ] |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3053 + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3054 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3055 for(i=0; i<vChrFilterSize; i+=2){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3056 chrMmxFilter[2*i+0]= chrSrcPtr[i ]; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3057 chrMmxFilter[2*i+1]= chrSrcPtr[i+(vChrFilterSize>1)]; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3058 chrMmxFilter[2*i+2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3059 chrMmxFilter[2*i+3]= vChrFilter[chrDstY*vChrFilterSize + i ] |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3060 + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3061 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3062 }else{ |
| 18861 | 3063 for(i=0; i<vLumFilterSize; i++) |
| 3064 { | |
| 3065 lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i]; | |
| 3066 lumMmxFilter[4*i+2]= | |
| 3067 lumMmxFilter[4*i+3]= | |
| 3068 ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001; | |
| 3069 } | |
| 3070 for(i=0; i<vChrFilterSize; i++) | |
| 3071 { | |
| 3072 chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i]; | |
| 3073 chrMmxFilter[4*i+2]= | |
| 3074 chrMmxFilter[4*i+3]= | |
| 3075 ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001; | |
| 3076 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3077 } |
| 18861 | 3078 #endif |
| 3079 if(dstFormat == IMGFMT_NV12 || dstFormat == IMGFMT_NV21){ | |
| 3080 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; | |
| 3081 if(dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi | |
| 3082 RENAME(yuv2nv12X)(c, | |
| 3083 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3084 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3085 dest, uDest, dstW, chrDstW, dstFormat); | |
| 3086 } | |
| 3087 else if(isPlanarYUV(dstFormat) || isGray(dstFormat)) //YV12 like | |
| 3088 { | |
| 3089 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; | |
| 3090 if((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi | |
| 3091 if(vLumFilterSize == 1 && vChrFilterSize == 1) // Unscaled YV12 | |
| 3092 { | |
| 3093 int16_t *lumBuf = lumPixBuf[0]; | |
| 3094 int16_t *chrBuf= chrPixBuf[0]; | |
| 3095 RENAME(yuv2yuv1)(lumBuf, chrBuf, dest, uDest, vDest, dstW, chrDstW); | |
| 3096 } | |
| 3097 else //General YV12 | |
| 3098 { | |
| 3099 RENAME(yuv2yuvX)(c, | |
| 3100 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3101 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3102 dest, uDest, vDest, dstW, chrDstW); | |
| 3103 } | |
| 3104 } | |
| 3105 else | |
| 3106 { | |
| 3107 ASSERT(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2); | |
| 3108 ASSERT(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2); | |
| 3109 if(vLumFilterSize == 1 && vChrFilterSize == 2) //Unscaled RGB | |
| 3110 { | |
| 3111 int chrAlpha= vChrFilter[2*dstY+1]; | |
| 3112 RENAME(yuv2packed1)(c, *lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1), | |
| 3113 dest, dstW, chrAlpha, dstFormat, flags, dstY); | |
| 3114 } | |
| 3115 else if(vLumFilterSize == 2 && vChrFilterSize == 2) //BiLinear Upscale RGB | |
| 3116 { | |
| 3117 int lumAlpha= vLumFilter[2*dstY+1]; | |
| 3118 int chrAlpha= vChrFilter[2*dstY+1]; | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3119 lumMmxFilter[2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3120 lumMmxFilter[3]= vLumFilter[2*dstY ]*0x10001; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3121 chrMmxFilter[2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3122 chrMmxFilter[3]= vChrFilter[2*chrDstY]*0x10001; |
| 18861 | 3123 RENAME(yuv2packed2)(c, *lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1), |
| 3124 dest, dstW, lumAlpha, chrAlpha, dstY); | |
| 3125 } | |
| 3126 else //General RGB | |
| 3127 { | |
| 3128 RENAME(yuv2packedX)(c, | |
| 3129 vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, | |
| 3130 vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3131 dest, dstW, dstY); | |
| 3132 } | |
| 3133 } | |
| 3134 } | |
| 3135 else // hmm looks like we can't use MMX here without overwriting this array's tail | |
| 3136 { | |
| 3137 int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize; | |
| 3138 int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; | |
| 3139 if(dstFormat == IMGFMT_NV12 || dstFormat == IMGFMT_NV21){ | |
| 3140 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; | |
| 3141 if(dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi | |
| 3142 yuv2nv12XinC( | |
| 3143 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3144 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3145 dest, uDest, dstW, chrDstW, dstFormat); | |
| 3146 } | |
| 3147 else if(isPlanarYUV(dstFormat) || isGray(dstFormat)) //YV12 | |
| 3148 { | |
| 3149 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; | |
| 3150 if((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi | |
| 3151 yuv2yuvXinC( | |
| 3152 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3153 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3154 dest, uDest, vDest, dstW, chrDstW); | |
| 3155 } | |
| 3156 else | |
| 3157 { | |
| 3158 ASSERT(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2); | |
| 3159 ASSERT(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2); | |
| 3160 yuv2packedXinC(c, | |
| 3161 vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, | |
| 3162 vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3163 dest, dstW, dstY); | |
| 3164 } | |
| 3165 } | |
| 3166 } | |
| 3167 | |
| 3168 #ifdef HAVE_MMX | |
| 3169 __asm __volatile(SFENCE:::"memory"); | |
| 3170 __asm __volatile(EMMS:::"memory"); | |
| 3171 #endif | |
| 3172 /* store changed local vars back in the context */ | |
| 3173 c->dstY= dstY; | |
| 3174 c->lumBufIndex= lumBufIndex; | |
| 3175 c->chrBufIndex= chrBufIndex; | |
| 3176 c->lastInLumBuf= lastInLumBuf; | |
| 3177 c->lastInChrBuf= lastInChrBuf; | |
| 3178 | |
| 3179 return dstY - lastDstY; | |
| 3180 } |