Mercurial > mplayer.hg
annotate libswscale/swscale_template.c @ 23089:7b313a20939f
Declarations and code should not be mixed.
| author | diego |
|---|---|
| date | Wed, 25 Apr 2007 00:03:27 +0000 |
| parents | d581a6ba2d9c |
| children | 9528d1ebe68f |
| rev | line source |
|---|---|
| 18861 | 1 /* |
|
20094
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at> |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
3 * |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
4 * This file is part of FFmpeg. |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
5 * |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
6 * FFmpeg is free software; you can redistribute it and/or modify |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
7 * it under the terms of the GNU General Public License as published by |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
8 * the Free Software Foundation; either version 2 of the License, or |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
9 * (at your option) any later version. |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
10 * |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
11 * FFmpeg is distributed in the hope that it will be useful, |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
14 * GNU General Public License for more details. |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
15 * |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
16 * You should have received a copy of the GNU General Public License |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
17 * along with FFmpeg; if not, write to the Free Software |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
|
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
19 * |
|
21029
1f2ba24b4e47
Clarify that some of the non-SIMD code is now LGPLed.
lucabe
parents:
20946
diff
changeset
|
20 * the C code (not assembly, mmx, ...) of this file can be used |
|
1f2ba24b4e47
Clarify that some of the non-SIMD code is now LGPLed.
lucabe
parents:
20946
diff
changeset
|
21 * under the LGPL license too |
|
20094
aca9e9783f67
Change license headers to say 'FFmpeg' instead of 'this program'.
diego
parents:
20015
diff
changeset
|
22 */ |
| 18861 | 23 |
| 24 #undef REAL_MOVNTQ | |
| 25 #undef MOVNTQ | |
| 26 #undef PAVGB | |
| 27 #undef PREFETCH | |
| 28 #undef PREFETCHW | |
| 29 #undef EMMS | |
| 30 #undef SFENCE | |
| 31 | |
| 32 #ifdef HAVE_3DNOW | |
| 33 /* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */ | |
| 34 #define EMMS "femms" | |
| 35 #else | |
| 36 #define EMMS "emms" | |
| 37 #endif | |
| 38 | |
| 39 #ifdef HAVE_3DNOW | |
| 40 #define PREFETCH "prefetch" | |
| 41 #define PREFETCHW "prefetchw" | |
| 42 #elif defined ( HAVE_MMX2 ) | |
| 43 #define PREFETCH "prefetchnta" | |
| 44 #define PREFETCHW "prefetcht0" | |
| 45 #else | |
|
20724
b8fe18a742ce
Fix MacIntel build: "/nop" is illegal on Apple's older version of GAS
gpoirier
parents:
20589
diff
changeset
|
46 #define PREFETCH " # nop" |
|
b8fe18a742ce
Fix MacIntel build: "/nop" is illegal on Apple's older version of GAS
gpoirier
parents:
20589
diff
changeset
|
47 #define PREFETCHW " # nop" |
| 18861 | 48 #endif |
| 49 | |
| 50 #ifdef HAVE_MMX2 | |
| 51 #define SFENCE "sfence" | |
| 52 #else | |
|
20724
b8fe18a742ce
Fix MacIntel build: "/nop" is illegal on Apple's older version of GAS
gpoirier
parents:
20589
diff
changeset
|
53 #define SFENCE " # nop" |
| 18861 | 54 #endif |
| 55 | |
| 56 #ifdef HAVE_MMX2 | |
| 57 #define PAVGB(a,b) "pavgb " #a ", " #b " \n\t" | |
| 58 #elif defined (HAVE_3DNOW) | |
| 59 #define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t" | |
| 60 #endif | |
| 61 | |
| 62 #ifdef HAVE_MMX2 | |
| 63 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t" | |
| 64 #else | |
| 65 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t" | |
| 66 #endif | |
| 67 #define MOVNTQ(a,b) REAL_MOVNTQ(a,b) | |
| 68 | |
| 69 #ifdef HAVE_ALTIVEC | |
| 70 #include "swscale_altivec_template.c" | |
| 71 #endif | |
| 72 | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
73 #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
|
74 asm volatile(\ |
| 18861 | 75 "xor %%"REG_a", %%"REG_a" \n\t"\ |
| 76 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ | |
| 77 "movq %%mm3, %%mm4 \n\t"\ | |
| 78 "lea " offset "(%0), %%"REG_d" \n\t"\ | |
| 79 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
80 ASMALIGN(4) /* FIXME Unroll? */\ |
| 18861 | 81 "1: \n\t"\ |
| 82 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ | |
| 83 "movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\ | |
| 84 "movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm5\n\t" /* srcData */\ | |
| 85 "add $16, %%"REG_d" \n\t"\ | |
| 86 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 87 "test %%"REG_S", %%"REG_S" \n\t"\ | |
| 88 "pmulhw %%mm0, %%mm2 \n\t"\ | |
| 89 "pmulhw %%mm0, %%mm5 \n\t"\ | |
| 90 "paddw %%mm2, %%mm3 \n\t"\ | |
| 91 "paddw %%mm5, %%mm4 \n\t"\ | |
| 92 " jnz 1b \n\t"\ | |
| 93 "psraw $3, %%mm3 \n\t"\ | |
| 94 "psraw $3, %%mm4 \n\t"\ | |
| 95 "packuswb %%mm4, %%mm3 \n\t"\ | |
| 96 MOVNTQ(%%mm3, (%1, %%REGa))\ | |
| 97 "add $8, %%"REG_a" \n\t"\ | |
| 98 "cmp %2, %%"REG_a" \n\t"\ | |
| 99 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ | |
| 100 "movq %%mm3, %%mm4 \n\t"\ | |
| 101 "lea " offset "(%0), %%"REG_d" \n\t"\ | |
| 102 "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
|
103 "jb 1b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
104 :: "r" (&c->redDither),\ |
|
21325
963e85e82154
Change "p" asm constraints to "g", since "p" was a no longer necessary hack to
reimar
parents:
21029
diff
changeset
|
105 "r" (dest), "g" (width)\ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
106 : "%"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
|
107 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
108 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
109 #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
|
110 asm volatile(\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
111 "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
|
112 "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
|
113 "pxor %%mm4, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
114 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
115 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
116 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
117 "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
118 ASMALIGN(4) \ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
119 "1: \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), %%mm0\n\t" /* srcData */\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
121 "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
|
122 "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
|
123 "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
|
124 "movq %%mm0, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
125 "punpcklwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
126 "punpckhwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
127 "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
|
128 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
129 "pmaddwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
130 "paddd %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
131 "paddd %%mm3, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
132 "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
|
133 "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
|
134 "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
|
135 "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
|
136 "movq %%mm2, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
137 "punpcklwd %%mm3, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
138 "punpckhwd %%mm3, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
139 "pmaddwd %%mm1, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
140 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
141 "paddd %%mm2, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
142 "paddd %%mm0, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
143 " jnz 1b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
144 "psrad $16, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
145 "psrad $16, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
146 "psrad $16, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
147 "psrad $16, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
148 "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
|
149 "packssdw %%mm5, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
150 "packssdw %%mm7, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
151 "paddw %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
152 "paddw %%mm0, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
153 "psraw $3, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
154 "psraw $3, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
155 "packuswb %%mm6, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
156 MOVNTQ(%%mm4, (%1, %%REGa))\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
157 "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
|
158 "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
|
159 "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
|
160 "pxor %%mm4, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
161 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
162 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
163 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
164 "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
|
165 "jb 1b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
166 :: "r" (&c->redDither),\ |
|
21325
963e85e82154
Change "p" asm constraints to "g", since "p" was a no longer necessary hack to
reimar
parents:
21029
diff
changeset
|
167 "r" (dest), "g" (width)\ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
168 : "%"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
|
169 ); |
| 18861 | 170 |
| 171 #define YSCALEYUV2YV121 \ | |
| 172 "mov %2, %%"REG_a" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
173 ASMALIGN(4) /* FIXME Unroll? */\ |
| 18861 | 174 "1: \n\t"\ |
| 175 "movq (%0, %%"REG_a", 2), %%mm0 \n\t"\ | |
| 176 "movq 8(%0, %%"REG_a", 2), %%mm1\n\t"\ | |
| 177 "psraw $7, %%mm0 \n\t"\ | |
| 178 "psraw $7, %%mm1 \n\t"\ | |
| 179 "packuswb %%mm1, %%mm0 \n\t"\ | |
| 180 MOVNTQ(%%mm0, (%1, %%REGa))\ | |
| 181 "add $8, %%"REG_a" \n\t"\ | |
| 182 "jnc 1b \n\t" | |
| 183 | |
| 184 /* | |
| 185 :: "m" (-lumFilterSize), "m" (-chrFilterSize), | |
| 186 "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4), | |
| 187 "r" (dest), "m" (dstW), | |
| 188 "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize) | |
| 189 : "%eax", "%ebx", "%ecx", "%edx", "%esi" | |
| 190 */ | |
| 191 #define YSCALEYUV2PACKEDX \ | |
| 19173 | 192 asm volatile(\ |
| 18861 | 193 "xor %%"REG_a", %%"REG_a" \n\t"\ |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
194 ASMALIGN(4)\ |
| 18861 | 195 "nop \n\t"\ |
| 196 "1: \n\t"\ | |
| 197 "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ | |
| 198 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 199 "movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\ | |
| 200 "movq %%mm3, %%mm4 \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
201 ASMALIGN(4)\ |
| 18861 | 202 "2: \n\t"\ |
| 203 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ | |
| 204 "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\ | |
| 205 "movq 4096(%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\ | |
| 206 "add $16, %%"REG_d" \n\t"\ | |
| 207 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 208 "pmulhw %%mm0, %%mm2 \n\t"\ | |
| 209 "pmulhw %%mm0, %%mm5 \n\t"\ | |
| 210 "paddw %%mm2, %%mm3 \n\t"\ | |
| 211 "paddw %%mm5, %%mm4 \n\t"\ | |
| 212 "test %%"REG_S", %%"REG_S" \n\t"\ | |
| 213 " jnz 2b \n\t"\ | |
| 214 \ | |
| 215 "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\ | |
| 216 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 217 "movq "VROUNDER_OFFSET"(%0), %%mm1\n\t"\ | |
| 218 "movq %%mm1, %%mm7 \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
219 ASMALIGN(4)\ |
| 18861 | 220 "2: \n\t"\ |
| 221 "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\ | |
| 222 "movq (%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y1srcData */\ | |
| 223 "movq 8(%%"REG_S", %%"REG_a", 2), %%mm5 \n\t" /* Y2srcData */\ | |
| 224 "add $16, %%"REG_d" \n\t"\ | |
| 225 "mov (%%"REG_d"), %%"REG_S" \n\t"\ | |
| 226 "pmulhw %%mm0, %%mm2 \n\t"\ | |
| 227 "pmulhw %%mm0, %%mm5 \n\t"\ | |
| 228 "paddw %%mm2, %%mm1 \n\t"\ | |
| 229 "paddw %%mm5, %%mm7 \n\t"\ | |
| 230 "test %%"REG_S", %%"REG_S" \n\t"\ | |
| 231 " jnz 2b \n\t"\ | |
| 232 | |
| 19173 | 233 #define YSCALEYUV2PACKEDX_END\ |
| 234 :: "r" (&c->redDither), \ | |
| 235 "m" (dummy), "m" (dummy), "m" (dummy),\ | |
| 236 "r" (dest), "m" (dstW)\ | |
| 237 : "%"REG_a, "%"REG_d, "%"REG_S\ | |
| 238 ); | |
| 239 | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
240 #define YSCALEYUV2PACKEDX_ACCURATE \ |
| 19173 | 241 asm volatile(\ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
242 "xor %%"REG_a", %%"REG_a" \n\t"\ |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
243 ASMALIGN(4)\ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
244 "nop \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
245 "1: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
246 "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
|
247 "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
|
248 "pxor %%mm4, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
249 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
250 "pxor %%mm6, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
251 "pxor %%mm7, %%mm7 \n\t"\ |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
252 ASMALIGN(4)\ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
253 "2: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
254 "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
|
255 "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
|
256 "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
|
257 "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
|
258 "movq %%mm0, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
259 "punpcklwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
260 "punpckhwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
261 "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
|
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 "pmaddwd %%mm1, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
264 "paddd %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
265 "paddd %%mm3, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
266 "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
|
267 "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
|
268 "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
|
269 "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
|
270 "movq %%mm2, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
271 "punpcklwd %%mm3, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
272 "punpckhwd %%mm3, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
273 "pmaddwd %%mm1, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
274 "pmaddwd %%mm1, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
275 "paddd %%mm2, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
276 "paddd %%mm0, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
277 " jnz 2b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
278 "psrad $16, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
279 "psrad $16, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
280 "psrad $16, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
281 "psrad $16, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
282 "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
|
283 "packssdw %%mm5, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
284 "packssdw %%mm7, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
285 "paddw %%mm0, %%mm4 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
286 "paddw %%mm0, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
287 "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
|
288 "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
|
289 \ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
290 "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
|
291 "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
|
292 "pxor %%mm1, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
293 "pxor %%mm5, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
294 "pxor %%mm7, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
295 "pxor %%mm6, %%mm6 \n\t"\ |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
296 ASMALIGN(4)\ |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
297 "2: \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
298 "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
|
299 "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
|
300 "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
|
301 "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
|
302 "movq %%mm0, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
303 "punpcklwd %%mm4, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
304 "punpckhwd %%mm4, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
305 "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
|
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 "pmaddwd %%mm4, %%mm3 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
308 "paddd %%mm0, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
309 "paddd %%mm3, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
310 "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
|
311 "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
|
312 "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
|
313 "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
|
314 "movq %%mm2, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
315 "punpcklwd %%mm3, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
316 "punpckhwd %%mm3, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
317 "pmaddwd %%mm4, %%mm2 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
318 "pmaddwd %%mm4, %%mm0 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
319 "paddd %%mm2, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
320 "paddd %%mm0, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
321 " jnz 2b \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
322 "psrad $16, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
323 "psrad $16, %%mm5 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
324 "psrad $16, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
325 "psrad $16, %%mm6 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
326 "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
|
327 "packssdw %%mm5, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
328 "packssdw %%mm6, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
329 "paddw %%mm0, %%mm1 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
330 "paddw %%mm0, %%mm7 \n\t"\ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
331 "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
|
332 "movq "V_TEMP"(%0), %%mm4 \n\t"\ |
| 18861 | 333 |
| 19173 | 334 #define YSCALEYUV2RGBX \ |
| 18861 | 335 "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\ |
| 336 "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\ | |
| 337 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 338 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 339 "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\ | |
| 340 "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\ | |
| 341 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 342 "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\ | |
| 343 "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\ | |
| 344 "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 345 "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 346 "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\ | |
| 347 "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\ | |
| 348 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 349 "paddw %%mm3, %%mm4 \n\t"\ | |
| 350 "movq %%mm2, %%mm0 \n\t"\ | |
| 351 "movq %%mm5, %%mm6 \n\t"\ | |
| 352 "movq %%mm4, %%mm3 \n\t"\ | |
| 353 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 354 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 355 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 356 "paddw %%mm1, %%mm2 \n\t"\ | |
| 357 "paddw %%mm1, %%mm5 \n\t"\ | |
| 358 "paddw %%mm1, %%mm4 \n\t"\ | |
| 359 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 360 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 361 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 362 "paddw %%mm7, %%mm0 \n\t"\ | |
| 363 "paddw %%mm7, %%mm6 \n\t"\ | |
| 364 "paddw %%mm7, %%mm3 \n\t"\ | |
| 365 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 366 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 367 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 368 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 369 "pxor %%mm7, %%mm7 \n\t" | |
| 370 #if 0 | |
| 371 #define FULL_YSCALEYUV2RGB \ | |
| 372 "pxor %%mm7, %%mm7 \n\t"\ | |
| 373 "movd %6, %%mm6 \n\t" /*yalpha1*/\ | |
| 374 "punpcklwd %%mm6, %%mm6 \n\t"\ | |
| 375 "punpcklwd %%mm6, %%mm6 \n\t"\ | |
| 376 "movd %7, %%mm5 \n\t" /*uvalpha1*/\ | |
| 377 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 378 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 379 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
380 ASMALIGN(4)\ |
| 18861 | 381 "1: \n\t"\ |
| 382 "movq (%0, %%"REG_a", 2), %%mm0 \n\t" /*buf0[eax]*/\ | |
| 383 "movq (%1, %%"REG_a", 2), %%mm1 \n\t" /*buf1[eax]*/\ | |
| 384 "movq (%2, %%"REG_a",2), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 385 "movq (%3, %%"REG_a",2), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 386 "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 387 "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\ | |
| 388 "pmulhw %%mm6, %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 389 "pmulhw %%mm5, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\ | |
| 390 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 391 "movq 4096(%2, %%"REG_a",2), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 392 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 393 "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 394 "movq 4096(%3, %%"REG_a",2), %%mm0 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 395 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\ | |
| 396 "psubw %%mm0, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\ | |
| 397 "psubw "MANGLE(w80)", %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 398 "psubw "MANGLE(w400)", %%mm3 \n\t" /* 8(U-128)*/\ | |
| 399 "pmulhw "MANGLE(yCoeff)", %%mm1 \n\t"\ | |
| 400 \ | |
| 401 \ | |
| 402 "pmulhw %%mm5, %%mm4 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\ | |
| 403 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 404 "pmulhw "MANGLE(ubCoeff)", %%mm3\n\t"\ | |
| 405 "psraw $4, %%mm0 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 406 "pmulhw "MANGLE(ugCoeff)", %%mm2\n\t"\ | |
| 407 "paddw %%mm4, %%mm0 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\ | |
| 408 "psubw "MANGLE(w400)", %%mm0 \n\t" /* (V-128)8*/\ | |
| 409 \ | |
| 410 \ | |
| 411 "movq %%mm0, %%mm4 \n\t" /* (V-128)8*/\ | |
| 412 "pmulhw "MANGLE(vrCoeff)", %%mm0\n\t"\ | |
| 413 "pmulhw "MANGLE(vgCoeff)", %%mm4\n\t"\ | |
| 414 "paddw %%mm1, %%mm3 \n\t" /* B*/\ | |
| 415 "paddw %%mm1, %%mm0 \n\t" /* R*/\ | |
| 416 "packuswb %%mm3, %%mm3 \n\t"\ | |
| 417 \ | |
| 418 "packuswb %%mm0, %%mm0 \n\t"\ | |
| 419 "paddw %%mm4, %%mm2 \n\t"\ | |
| 420 "paddw %%mm2, %%mm1 \n\t" /* G*/\ | |
| 421 \ | |
| 422 "packuswb %%mm1, %%mm1 \n\t" | |
| 423 #endif | |
| 424 | |
| 425 #define REAL_YSCALEYUV2PACKED(index, c) \ | |
| 426 "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\ | |
| 427 "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1\n\t"\ | |
| 428 "psraw $3, %%mm0 \n\t"\ | |
| 429 "psraw $3, %%mm1 \n\t"\ | |
| 430 "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c")\n\t"\ | |
| 431 "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c")\n\t"\ | |
| 432 "xor "#index", "#index" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
433 ASMALIGN(4)\ |
| 18861 | 434 "1: \n\t"\ |
| 435 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 436 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 437 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 438 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 439 "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\ | |
| 440 "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\ | |
| 441 "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\ | |
| 442 "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\ | |
| 443 "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\ | |
| 444 "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 445 "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 446 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\ | |
| 447 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\ | |
| 448 "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\ | |
| 449 "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\ | |
| 450 "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\ | |
| 451 "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\ | |
| 452 "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 453 "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 454 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 455 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 456 "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 457 "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 458 "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 459 "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 460 | |
| 461 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c) | |
| 462 | |
| 463 #define REAL_YSCALEYUV2RGB(index, c) \ | |
| 464 "xor "#index", "#index" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
465 ASMALIGN(4)\ |
| 18861 | 466 "1: \n\t"\ |
| 467 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 468 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 469 "movq 4096(%2, "#index"), %%mm5\n\t" /* uvbuf0[eax+2048]*/\ | |
| 470 "movq 4096(%3, "#index"), %%mm4\n\t" /* uvbuf1[eax+2048]*/\ | |
| 471 "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\ | |
| 472 "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\ | |
| 473 "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\ | |
| 474 "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\ | |
| 475 "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\ | |
| 476 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 477 "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 478 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\ | |
| 479 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\ | |
| 480 "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\ | |
| 481 "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\ | |
| 482 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 483 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 484 "pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\ | |
| 485 "pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\ | |
| 486 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 487 "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\ | |
| 488 "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\ | |
| 489 "movq 8(%0, "#index", 2), %%mm6\n\t" /*buf0[eax]*/\ | |
| 490 "movq 8(%1, "#index", 2), %%mm7\n\t" /*buf1[eax]*/\ | |
| 491 "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 492 "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\ | |
| 493 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 494 "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\ | |
| 495 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 496 "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 497 "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 498 "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\ | |
| 499 "pmulhw "UB_COEFF"("#c"), %%mm2\n\t"\ | |
| 500 "pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\ | |
| 501 "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 502 "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 503 "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\ | |
| 504 "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\ | |
| 505 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 506 "paddw %%mm3, %%mm4 \n\t"\ | |
| 507 "movq %%mm2, %%mm0 \n\t"\ | |
| 508 "movq %%mm5, %%mm6 \n\t"\ | |
| 509 "movq %%mm4, %%mm3 \n\t"\ | |
| 510 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 511 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 512 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 513 "paddw %%mm1, %%mm2 \n\t"\ | |
| 514 "paddw %%mm1, %%mm5 \n\t"\ | |
| 515 "paddw %%mm1, %%mm4 \n\t"\ | |
| 516 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 517 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 518 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 519 "paddw %%mm7, %%mm0 \n\t"\ | |
| 520 "paddw %%mm7, %%mm6 \n\t"\ | |
| 521 "paddw %%mm7, %%mm3 \n\t"\ | |
| 522 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 523 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 524 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 525 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 526 "pxor %%mm7, %%mm7 \n\t" | |
| 527 #define YSCALEYUV2RGB(index, c) REAL_YSCALEYUV2RGB(index, c) | |
| 528 | |
| 529 #define REAL_YSCALEYUV2PACKED1(index, c) \ | |
| 530 "xor "#index", "#index" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
531 ASMALIGN(4)\ |
| 18861 | 532 "1: \n\t"\ |
| 533 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\ | |
| 534 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 535 "psraw $7, %%mm3 \n\t" \ | |
| 536 "psraw $7, %%mm4 \n\t" \ | |
| 537 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 538 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 539 "psraw $7, %%mm1 \n\t" \ | |
| 540 "psraw $7, %%mm7 \n\t" \ | |
| 541 | |
| 542 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c) | |
| 543 | |
| 544 #define REAL_YSCALEYUV2RGB1(index, c) \ | |
| 545 "xor "#index", "#index" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
546 ASMALIGN(4)\ |
| 18861 | 547 "1: \n\t"\ |
| 548 "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\ | |
| 549 "movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 550 "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\ | |
| 551 "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\ | |
| 552 "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\ | |
| 553 "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\ | |
| 554 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 555 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 556 "pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\ | |
| 557 "pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\ | |
| 558 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 559 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 560 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 561 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 562 "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 563 "pmulhw "UB_COEFF"("#c"), %%mm2\n\t"\ | |
| 564 "pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\ | |
| 565 "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 566 "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 567 "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\ | |
| 568 "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\ | |
| 569 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 570 "paddw %%mm3, %%mm4 \n\t"\ | |
| 571 "movq %%mm2, %%mm0 \n\t"\ | |
| 572 "movq %%mm5, %%mm6 \n\t"\ | |
| 573 "movq %%mm4, %%mm3 \n\t"\ | |
| 574 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 575 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 576 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 577 "paddw %%mm1, %%mm2 \n\t"\ | |
| 578 "paddw %%mm1, %%mm5 \n\t"\ | |
| 579 "paddw %%mm1, %%mm4 \n\t"\ | |
| 580 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 581 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 582 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 583 "paddw %%mm7, %%mm0 \n\t"\ | |
| 584 "paddw %%mm7, %%mm6 \n\t"\ | |
| 585 "paddw %%mm7, %%mm3 \n\t"\ | |
| 586 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 587 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 588 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 589 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 590 "pxor %%mm7, %%mm7 \n\t" | |
| 591 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c) | |
| 592 | |
| 593 #define REAL_YSCALEYUV2PACKED1b(index, c) \ | |
| 594 "xor "#index", "#index" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
595 ASMALIGN(4)\ |
| 18861 | 596 "1: \n\t"\ |
| 597 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 598 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 599 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 600 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 601 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\ | |
| 602 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\ | |
| 603 "psrlw $8, %%mm3 \n\t" \ | |
| 604 "psrlw $8, %%mm4 \n\t" \ | |
| 605 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 606 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 607 "psraw $7, %%mm1 \n\t" \ | |
| 608 "psraw $7, %%mm7 \n\t" | |
| 609 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c) | |
| 610 | |
| 611 // do vertical chrominance interpolation | |
| 612 #define REAL_YSCALEYUV2RGB1b(index, c) \ | |
| 613 "xor "#index", "#index" \n\t"\ | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
614 ASMALIGN(4)\ |
| 18861 | 615 "1: \n\t"\ |
| 616 "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\ | |
| 617 "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\ | |
| 618 "movq 4096(%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\ | |
| 619 "movq 4096(%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\ | |
| 620 "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\ | |
| 621 "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\ | |
| 622 "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\ | |
| 623 "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\ | |
| 624 "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\ | |
| 625 "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\ | |
| 626 "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\ | |
| 627 "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\ | |
| 628 "pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\ | |
| 629 "pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\ | |
| 630 /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\ | |
| 631 "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\ | |
| 632 "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\ | |
| 633 "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 634 "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\ | |
| 635 "pmulhw "UB_COEFF"("#c"), %%mm2\n\t"\ | |
| 636 "pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\ | |
| 637 "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\ | |
| 638 "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\ | |
| 639 "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\ | |
| 640 "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\ | |
| 641 /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\ | |
| 642 "paddw %%mm3, %%mm4 \n\t"\ | |
| 643 "movq %%mm2, %%mm0 \n\t"\ | |
| 644 "movq %%mm5, %%mm6 \n\t"\ | |
| 645 "movq %%mm4, %%mm3 \n\t"\ | |
| 646 "punpcklwd %%mm2, %%mm2 \n\t"\ | |
| 647 "punpcklwd %%mm5, %%mm5 \n\t"\ | |
| 648 "punpcklwd %%mm4, %%mm4 \n\t"\ | |
| 649 "paddw %%mm1, %%mm2 \n\t"\ | |
| 650 "paddw %%mm1, %%mm5 \n\t"\ | |
| 651 "paddw %%mm1, %%mm4 \n\t"\ | |
| 652 "punpckhwd %%mm0, %%mm0 \n\t"\ | |
| 653 "punpckhwd %%mm6, %%mm6 \n\t"\ | |
| 654 "punpckhwd %%mm3, %%mm3 \n\t"\ | |
| 655 "paddw %%mm7, %%mm0 \n\t"\ | |
| 656 "paddw %%mm7, %%mm6 \n\t"\ | |
| 657 "paddw %%mm7, %%mm3 \n\t"\ | |
| 658 /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\ | |
| 659 "packuswb %%mm0, %%mm2 \n\t"\ | |
| 660 "packuswb %%mm6, %%mm5 \n\t"\ | |
| 661 "packuswb %%mm3, %%mm4 \n\t"\ | |
| 662 "pxor %%mm7, %%mm7 \n\t" | |
| 663 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c) | |
| 664 | |
| 665 #define REAL_WRITEBGR32(dst, dstw, index) \ | |
| 666 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 667 "movq %%mm2, %%mm1 \n\t" /* B */\ | |
| 668 "movq %%mm5, %%mm6 \n\t" /* R */\ | |
| 669 "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\ | |
| 670 "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\ | |
| 671 "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\ | |
| 672 "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\ | |
| 673 "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\ | |
| 674 "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\ | |
| 675 "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\ | |
| 676 "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 677 "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\ | |
| 678 "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\ | |
| 679 \ | |
| 680 MOVNTQ(%%mm0, (dst, index, 4))\ | |
| 681 MOVNTQ(%%mm2, 8(dst, index, 4))\ | |
| 682 MOVNTQ(%%mm1, 16(dst, index, 4))\ | |
| 683 MOVNTQ(%%mm3, 24(dst, index, 4))\ | |
| 684 \ | |
| 685 "add $8, "#index" \n\t"\ | |
| 686 "cmp "#dstw", "#index" \n\t"\ | |
| 687 " jb 1b \n\t" | |
| 688 #define WRITEBGR32(dst, dstw, index) REAL_WRITEBGR32(dst, dstw, index) | |
| 689 | |
| 690 #define REAL_WRITEBGR16(dst, dstw, index) \ | |
| 691 "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\ | |
| 692 "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\ | |
| 693 "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\ | |
| 694 "psrlq $3, %%mm2 \n\t"\ | |
| 695 \ | |
| 696 "movq %%mm2, %%mm1 \n\t"\ | |
| 697 "movq %%mm4, %%mm3 \n\t"\ | |
| 698 \ | |
| 699 "punpcklbw %%mm7, %%mm3 \n\t"\ | |
| 700 "punpcklbw %%mm5, %%mm2 \n\t"\ | |
| 701 "punpckhbw %%mm7, %%mm4 \n\t"\ | |
| 702 "punpckhbw %%mm5, %%mm1 \n\t"\ | |
| 703 \ | |
| 704 "psllq $3, %%mm3 \n\t"\ | |
| 705 "psllq $3, %%mm4 \n\t"\ | |
| 706 \ | |
| 707 "por %%mm3, %%mm2 \n\t"\ | |
| 708 "por %%mm4, %%mm1 \n\t"\ | |
| 709 \ | |
| 710 MOVNTQ(%%mm2, (dst, index, 2))\ | |
| 711 MOVNTQ(%%mm1, 8(dst, index, 2))\ | |
| 712 \ | |
| 713 "add $8, "#index" \n\t"\ | |
| 714 "cmp "#dstw", "#index" \n\t"\ | |
| 715 " jb 1b \n\t" | |
| 716 #define WRITEBGR16(dst, dstw, index) REAL_WRITEBGR16(dst, dstw, index) | |
| 717 | |
| 718 #define REAL_WRITEBGR15(dst, dstw, index) \ | |
| 719 "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\ | |
| 720 "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\ | |
| 721 "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\ | |
| 722 "psrlq $3, %%mm2 \n\t"\ | |
| 723 "psrlq $1, %%mm5 \n\t"\ | |
| 724 \ | |
| 725 "movq %%mm2, %%mm1 \n\t"\ | |
| 726 "movq %%mm4, %%mm3 \n\t"\ | |
| 727 \ | |
| 728 "punpcklbw %%mm7, %%mm3 \n\t"\ | |
| 729 "punpcklbw %%mm5, %%mm2 \n\t"\ | |
| 730 "punpckhbw %%mm7, %%mm4 \n\t"\ | |
| 731 "punpckhbw %%mm5, %%mm1 \n\t"\ | |
| 732 \ | |
| 733 "psllq $2, %%mm3 \n\t"\ | |
| 734 "psllq $2, %%mm4 \n\t"\ | |
| 735 \ | |
| 736 "por %%mm3, %%mm2 \n\t"\ | |
| 737 "por %%mm4, %%mm1 \n\t"\ | |
| 738 \ | |
| 739 MOVNTQ(%%mm2, (dst, index, 2))\ | |
| 740 MOVNTQ(%%mm1, 8(dst, index, 2))\ | |
| 741 \ | |
| 742 "add $8, "#index" \n\t"\ | |
| 743 "cmp "#dstw", "#index" \n\t"\ | |
| 744 " jb 1b \n\t" | |
| 745 #define WRITEBGR15(dst, dstw, index) REAL_WRITEBGR15(dst, dstw, index) | |
| 746 | |
| 747 #define WRITEBGR24OLD(dst, dstw, index) \ | |
| 748 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 749 "movq %%mm2, %%mm1 \n\t" /* B */\ | |
| 750 "movq %%mm5, %%mm6 \n\t" /* R */\ | |
| 751 "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\ | |
| 752 "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\ | |
| 753 "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\ | |
| 754 "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\ | |
| 755 "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\ | |
| 756 "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\ | |
| 757 "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\ | |
| 758 "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 759 "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\ | |
| 760 "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\ | |
| 761 \ | |
| 762 "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\ | |
| 763 "psrlq $8, %%mm0 \n\t" /* 00RGB0RG 0 */\ | |
| 764 "pand "MANGLE(bm00000111)", %%mm4\n\t" /* 00000RGB 0 */\ | |
| 765 "pand "MANGLE(bm11111000)", %%mm0\n\t" /* 00RGB000 0.5 */\ | |
| 766 "por %%mm4, %%mm0 \n\t" /* 00RGBRGB 0 */\ | |
| 767 "movq %%mm2, %%mm4 \n\t" /* 0RGB0RGB 1 */\ | |
| 768 "psllq $48, %%mm2 \n\t" /* GB000000 1 */\ | |
| 769 "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\ | |
| 770 \ | |
| 771 "movq %%mm4, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 772 "psrld $16, %%mm4 \n\t" /* 000R000R 1 */\ | |
| 773 "psrlq $24, %%mm2 \n\t" /* 0000RGB0 1.5 */\ | |
| 774 "por %%mm4, %%mm2 \n\t" /* 000RRGBR 1 */\ | |
| 775 "pand "MANGLE(bm00001111)", %%mm2\n\t" /* 0000RGBR 1 */\ | |
| 776 "movq %%mm1, %%mm4 \n\t" /* 0RGB0RGB 2 */\ | |
| 777 "psrlq $8, %%mm1 \n\t" /* 00RGB0RG 2 */\ | |
| 778 "pand "MANGLE(bm00000111)", %%mm4\n\t" /* 00000RGB 2 */\ | |
| 779 "pand "MANGLE(bm11111000)", %%mm1\n\t" /* 00RGB000 2.5 */\ | |
| 780 "por %%mm4, %%mm1 \n\t" /* 00RGBRGB 2 */\ | |
| 781 "movq %%mm1, %%mm4 \n\t" /* 00RGBRGB 2 */\ | |
| 782 "psllq $32, %%mm1 \n\t" /* BRGB0000 2 */\ | |
| 783 "por %%mm1, %%mm2 \n\t" /* BRGBRGBR 1 */\ | |
| 784 \ | |
| 785 "psrlq $32, %%mm4 \n\t" /* 000000RG 2.5 */\ | |
| 786 "movq %%mm3, %%mm5 \n\t" /* 0RGB0RGB 3 */\ | |
| 787 "psrlq $8, %%mm3 \n\t" /* 00RGB0RG 3 */\ | |
| 788 "pand "MANGLE(bm00000111)", %%mm5\n\t" /* 00000RGB 3 */\ | |
| 789 "pand "MANGLE(bm11111000)", %%mm3\n\t" /* 00RGB000 3.5 */\ | |
| 790 "por %%mm5, %%mm3 \n\t" /* 00RGBRGB 3 */\ | |
| 791 "psllq $16, %%mm3 \n\t" /* RGBRGB00 3 */\ | |
| 792 "por %%mm4, %%mm3 \n\t" /* RGBRGBRG 2.5 */\ | |
| 793 \ | |
| 794 MOVNTQ(%%mm0, (dst))\ | |
| 795 MOVNTQ(%%mm2, 8(dst))\ | |
| 796 MOVNTQ(%%mm3, 16(dst))\ | |
| 797 "add $24, "#dst" \n\t"\ | |
| 798 \ | |
| 799 "add $8, "#index" \n\t"\ | |
| 800 "cmp "#dstw", "#index" \n\t"\ | |
| 801 " jb 1b \n\t" | |
| 802 | |
| 803 #define WRITEBGR24MMX(dst, dstw, index) \ | |
| 804 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 805 "movq %%mm2, %%mm1 \n\t" /* B */\ | |
| 806 "movq %%mm5, %%mm6 \n\t" /* R */\ | |
| 807 "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\ | |
| 808 "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\ | |
| 809 "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\ | |
| 810 "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\ | |
| 811 "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\ | |
| 812 "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\ | |
| 813 "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\ | |
| 814 "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\ | |
| 815 "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\ | |
| 816 "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\ | |
| 817 \ | |
| 818 "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\ | |
| 819 "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\ | |
| 820 "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\ | |
| 821 "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\ | |
| 822 \ | |
| 823 "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\ | |
| 824 "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\ | |
| 825 "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\ | |
| 826 "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\ | |
| 827 \ | |
| 828 "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\ | |
| 829 "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\ | |
| 830 "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\ | |
| 831 "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\ | |
| 832 \ | |
| 833 "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\ | |
| 834 "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\ | |
| 835 "psllq $40, %%mm2 \n\t" /* GB000000 1 */\ | |
| 836 "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\ | |
| 837 MOVNTQ(%%mm0, (dst))\ | |
| 838 \ | |
| 839 "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\ | |
| 840 "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\ | |
| 841 "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\ | |
| 842 "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\ | |
| 843 MOVNTQ(%%mm6, 8(dst))\ | |
| 844 \ | |
| 845 "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\ | |
| 846 "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\ | |
| 847 "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\ | |
| 848 MOVNTQ(%%mm5, 16(dst))\ | |
| 849 \ | |
| 850 "add $24, "#dst" \n\t"\ | |
| 851 \ | |
| 852 "add $8, "#index" \n\t"\ | |
| 853 "cmp "#dstw", "#index" \n\t"\ | |
| 854 " jb 1b \n\t" | |
| 855 | |
| 856 #define WRITEBGR24MMX2(dst, dstw, index) \ | |
| 857 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\ | |
| 858 "movq "MANGLE(M24A)", %%mm0 \n\t"\ | |
| 859 "movq "MANGLE(M24C)", %%mm7 \n\t"\ | |
| 860 "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\ | |
| 861 "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\ | |
| 862 "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\ | |
| 863 \ | |
| 864 "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\ | |
| 865 "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\ | |
| 866 "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\ | |
| 867 \ | |
| 868 "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\ | |
| 869 "por %%mm1, %%mm6 \n\t"\ | |
| 870 "por %%mm3, %%mm6 \n\t"\ | |
| 871 MOVNTQ(%%mm6, (dst))\ | |
| 872 \ | |
| 873 "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\ | |
| 874 "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\ | |
| 875 "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\ | |
| 876 "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\ | |
| 877 \ | |
| 878 "pand "MANGLE(M24B)", %%mm1 \n\t" /* B5 B4 B3 */\ | |
| 879 "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\ | |
| 880 "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\ | |
| 881 \ | |
| 882 "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\ | |
| 883 "por %%mm3, %%mm6 \n\t"\ | |
| 884 MOVNTQ(%%mm6, 8(dst))\ | |
| 885 \ | |
| 886 "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\ | |
| 887 "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\ | |
| 888 "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\ | |
| 889 \ | |
| 890 "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\ | |
| 891 "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\ | |
| 892 "pand "MANGLE(M24B)", %%mm6 \n\t" /* R7 R6 R5 */\ | |
| 893 \ | |
| 894 "por %%mm1, %%mm3 \n\t"\ | |
| 895 "por %%mm3, %%mm6 \n\t"\ | |
| 896 MOVNTQ(%%mm6, 16(dst))\ | |
| 897 \ | |
| 898 "add $24, "#dst" \n\t"\ | |
| 899 \ | |
| 900 "add $8, "#index" \n\t"\ | |
| 901 "cmp "#dstw", "#index" \n\t"\ | |
| 902 " jb 1b \n\t" | |
| 903 | |
| 904 #ifdef HAVE_MMX2 | |
| 905 #undef WRITEBGR24 | |
| 906 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX2(dst, dstw, index) | |
| 907 #else | |
| 908 #undef WRITEBGR24 | |
| 909 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index) | |
| 910 #endif | |
| 911 | |
| 912 #define REAL_WRITEYUY2(dst, dstw, index) \ | |
| 913 "packuswb %%mm3, %%mm3 \n\t"\ | |
| 914 "packuswb %%mm4, %%mm4 \n\t"\ | |
| 915 "packuswb %%mm7, %%mm1 \n\t"\ | |
| 916 "punpcklbw %%mm4, %%mm3 \n\t"\ | |
| 917 "movq %%mm1, %%mm7 \n\t"\ | |
| 918 "punpcklbw %%mm3, %%mm1 \n\t"\ | |
| 919 "punpckhbw %%mm3, %%mm7 \n\t"\ | |
| 920 \ | |
| 921 MOVNTQ(%%mm1, (dst, index, 2))\ | |
| 922 MOVNTQ(%%mm7, 8(dst, index, 2))\ | |
| 923 \ | |
| 924 "add $8, "#index" \n\t"\ | |
| 925 "cmp "#dstw", "#index" \n\t"\ | |
| 926 " jb 1b \n\t" | |
| 927 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index) | |
| 928 | |
| 929 | |
| 930 static inline void RENAME(yuv2yuvX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
| 931 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
| 932 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, long dstW, long chrDstW) | |
| 933 { | |
| 934 #ifdef HAVE_MMX | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
935 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
|
936 if(uDest){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
937 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
|
938 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
|
939 } |
| 18861 | 940 |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
941 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
|
942 }else{ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
943 if(uDest){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
944 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
|
945 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
|
946 } |
| 18861 | 947 |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
948 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
|
949 } |
| 18861 | 950 #else |
| 951 #ifdef HAVE_ALTIVEC | |
| 952 yuv2yuvX_altivec_real(lumFilter, lumSrc, lumFilterSize, | |
| 953 chrFilter, chrSrc, chrFilterSize, | |
| 954 dest, uDest, vDest, dstW, chrDstW); | |
| 955 #else //HAVE_ALTIVEC | |
| 956 yuv2yuvXinC(lumFilter, lumSrc, lumFilterSize, | |
| 957 chrFilter, chrSrc, chrFilterSize, | |
| 958 dest, uDest, vDest, dstW, chrDstW); | |
| 959 #endif //!HAVE_ALTIVEC | |
| 960 #endif | |
| 961 } | |
| 962 | |
| 963 static inline void RENAME(yuv2nv12X)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
| 964 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
| 965 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat) | |
| 966 { | |
| 967 yuv2nv12XinC(lumFilter, lumSrc, lumFilterSize, | |
| 968 chrFilter, chrSrc, chrFilterSize, | |
| 969 dest, uDest, dstW, chrDstW, dstFormat); | |
| 970 } | |
| 971 | |
| 972 static inline void RENAME(yuv2yuv1)(int16_t *lumSrc, int16_t *chrSrc, | |
| 973 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, long dstW, long chrDstW) | |
| 974 { | |
| 975 #ifdef HAVE_MMX | |
| 976 if(uDest != NULL) | |
| 977 { | |
| 978 asm volatile( | |
| 979 YSCALEYUV2YV121 | |
| 980 :: "r" (chrSrc + chrDstW), "r" (uDest + chrDstW), | |
| 981 "g" (-chrDstW) | |
| 982 : "%"REG_a | |
| 983 ); | |
| 984 | |
| 985 asm volatile( | |
| 986 YSCALEYUV2YV121 | |
| 987 :: "r" (chrSrc + 2048 + chrDstW), "r" (vDest + chrDstW), | |
| 988 "g" (-chrDstW) | |
| 989 : "%"REG_a | |
| 990 ); | |
| 991 } | |
| 992 | |
| 993 asm volatile( | |
| 994 YSCALEYUV2YV121 | |
| 995 :: "r" (lumSrc + dstW), "r" (dest + dstW), | |
| 996 "g" (-dstW) | |
| 997 : "%"REG_a | |
| 998 ); | |
| 999 #else | |
| 1000 int i; | |
| 1001 for(i=0; i<dstW; i++) | |
| 1002 { | |
| 1003 int val= lumSrc[i]>>7; | |
| 1004 | |
| 1005 if(val&256){ | |
| 1006 if(val<0) val=0; | |
| 1007 else val=255; | |
| 1008 } | |
| 1009 | |
| 1010 dest[i]= val; | |
| 1011 } | |
| 1012 | |
| 1013 if(uDest != NULL) | |
| 1014 for(i=0; i<chrDstW; i++) | |
| 1015 { | |
| 1016 int u=chrSrc[i]>>7; | |
| 1017 int v=chrSrc[i + 2048]>>7; | |
| 1018 | |
| 1019 if((u|v)&256){ | |
| 1020 if(u<0) u=0; | |
| 1021 else if (u>255) u=255; | |
| 1022 if(v<0) v=0; | |
| 1023 else if (v>255) v=255; | |
| 1024 } | |
| 1025 | |
| 1026 uDest[i]= u; | |
| 1027 vDest[i]= v; | |
| 1028 } | |
| 1029 #endif | |
| 1030 } | |
| 1031 | |
| 1032 | |
| 1033 /** | |
| 1034 * vertical scale YV12 to RGB | |
| 1035 */ | |
| 1036 static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
| 1037 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
| 1038 uint8_t *dest, long dstW, long dstY) | |
| 1039 { | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1040 #ifdef HAVE_MMX |
|
20015
d08ba4508bb0
Fix unused variable warning when compiling with MMX disabled.
diego
parents:
19872
diff
changeset
|
1041 long dummy=0; |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1042 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
|
1043 switch(c->dstFormat){ |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1044 case PIX_FMT_RGB32: |
| 19173 | 1045 YSCALEYUV2PACKEDX_ACCURATE |
| 1046 YSCALEYUV2RGBX | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1047 WRITEBGR32(%4, %5, %%REGa) |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1048 |
| 19173 | 1049 YSCALEYUV2PACKEDX_END |
| 1050 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1051 case PIX_FMT_BGR24: |
| 19173 | 1052 YSCALEYUV2PACKEDX_ACCURATE |
| 1053 YSCALEYUV2RGBX | |
| 19396 | 1054 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize |
| 1055 "add %4, %%"REG_c" \n\t" | |
| 1056 WRITEBGR24(%%REGc, %5, %%REGa) | |
| 19173 | 1057 |
| 1058 | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1059 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1060 "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
|
1061 "r" (dest), "m" (dstW) |
| 19396 | 1062 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1063 ); |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1064 return; |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1065 case PIX_FMT_BGR555: |
| 19173 | 1066 YSCALEYUV2PACKEDX_ACCURATE |
| 1067 YSCALEYUV2RGBX | |
| 1068 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1069 #ifdef DITHER1XBPP | |
| 1070 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1071 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1072 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1073 #endif | |
| 1074 | |
| 1075 WRITEBGR15(%4, %5, %%REGa) | |
| 1076 YSCALEYUV2PACKEDX_END | |
| 1077 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1078 case PIX_FMT_BGR565: |
| 19173 | 1079 YSCALEYUV2PACKEDX_ACCURATE |
| 1080 YSCALEYUV2RGBX | |
| 1081 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1082 #ifdef DITHER1XBPP | |
| 1083 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1084 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1085 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1086 #endif | |
| 1087 | |
| 1088 WRITEBGR16(%4, %5, %%REGa) | |
| 1089 YSCALEYUV2PACKEDX_END | |
| 1090 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1091 case PIX_FMT_YUYV422: |
| 19173 | 1092 YSCALEYUV2PACKEDX_ACCURATE |
| 1093 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1094 | |
| 1095 "psraw $3, %%mm3 \n\t" | |
| 1096 "psraw $3, %%mm4 \n\t" | |
| 1097 "psraw $3, %%mm1 \n\t" | |
| 1098 "psraw $3, %%mm7 \n\t" | |
| 1099 WRITEYUY2(%4, %5, %%REGa) | |
| 1100 YSCALEYUV2PACKEDX_END | |
| 1101 return; | |
| 1102 } | |
| 1103 }else{ | |
| 1104 switch(c->dstFormat) | |
| 1105 { | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1106 case PIX_FMT_RGB32: |
| 19173 | 1107 YSCALEYUV2PACKEDX |
| 1108 YSCALEYUV2RGBX | |
| 1109 WRITEBGR32(%4, %5, %%REGa) | |
| 1110 YSCALEYUV2PACKEDX_END | |
| 1111 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1112 case PIX_FMT_BGR24: |
| 19173 | 1113 YSCALEYUV2PACKEDX |
| 1114 YSCALEYUV2RGBX | |
| 19396 | 1115 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize |
| 1116 "add %4, %%"REG_c" \n\t" | |
| 1117 WRITEBGR24(%%REGc, %5, %%REGa) | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1118 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1119 :: "r" (&c->redDither), |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1120 "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
|
1121 "r" (dest), "m" (dstW) |
| 19396 | 1122 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1123 ); |
| 19173 | 1124 return; |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1125 case PIX_FMT_BGR555: |
| 19173 | 1126 YSCALEYUV2PACKEDX |
| 1127 YSCALEYUV2RGBX | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1128 /* 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
|
1129 #ifdef DITHER1XBPP |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1130 "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
|
1131 "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
|
1132 "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
|
1133 #endif |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1134 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1135 WRITEBGR15(%4, %5, %%REGa) |
| 19173 | 1136 YSCALEYUV2PACKEDX_END |
| 1137 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1138 case PIX_FMT_BGR565: |
| 19173 | 1139 YSCALEYUV2PACKEDX |
| 1140 YSCALEYUV2RGBX | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1141 /* 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
|
1142 #ifdef DITHER1XBPP |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1143 "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
|
1144 "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
|
1145 "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
|
1146 #endif |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1147 |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1148 WRITEBGR16(%4, %5, %%REGa) |
| 19173 | 1149 YSCALEYUV2PACKEDX_END |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1150 return; |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1151 case PIX_FMT_YUYV422: |
| 18861 | 1152 YSCALEYUV2PACKEDX |
| 1153 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1154 | |
| 1155 "psraw $3, %%mm3 \n\t" | |
| 1156 "psraw $3, %%mm4 \n\t" | |
| 1157 "psraw $3, %%mm1 \n\t" | |
| 1158 "psraw $3, %%mm7 \n\t" | |
| 1159 WRITEYUY2(%4, %5, %%REGa) | |
| 19173 | 1160 YSCALEYUV2PACKEDX_END |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1161 return; |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1162 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
1163 } |
| 18861 | 1164 #endif |
| 1165 #ifdef HAVE_ALTIVEC | |
| 1166 /* The following list of supported dstFormat values should | |
| 1167 match what's found in the body of altivec_yuv2packedX() */ | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1168 if(c->dstFormat==PIX_FMT_ABGR || c->dstFormat==PIX_FMT_BGRA || |
|
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1169 c->dstFormat==PIX_FMT_BGR24 || c->dstFormat==PIX_FMT_RGB24 || |
|
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1170 c->dstFormat==PIX_FMT_RGBA || c->dstFormat==PIX_FMT_ARGB) |
| 18861 | 1171 altivec_yuv2packedX (c, lumFilter, lumSrc, lumFilterSize, |
| 1172 chrFilter, chrSrc, chrFilterSize, | |
| 1173 dest, dstW, dstY); | |
| 1174 else | |
| 1175 #endif | |
| 1176 yuv2packedXinC(c, lumFilter, lumSrc, lumFilterSize, | |
| 1177 chrFilter, chrSrc, chrFilterSize, | |
| 1178 dest, dstW, dstY); | |
| 1179 } | |
| 1180 | |
| 1181 /** | |
| 1182 * vertical bilinear scale YV12 to RGB | |
| 1183 */ | |
| 1184 static inline void RENAME(yuv2packed2)(SwsContext *c, uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1, | |
| 1185 uint8_t *dest, int dstW, int yalpha, int uvalpha, int y) | |
| 1186 { | |
| 1187 int yalpha1=yalpha^4095; | |
| 1188 int uvalpha1=uvalpha^4095; | |
| 1189 int i; | |
| 1190 | |
| 1191 #if 0 //isn't used | |
| 1192 if(flags&SWS_FULL_CHR_H_INT) | |
| 1193 { | |
| 1194 switch(dstFormat) | |
| 1195 { | |
| 1196 #ifdef HAVE_MMX | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1197 case PIX_FMT_RGB32: |
| 18861 | 1198 asm volatile( |
| 1199 | |
| 1200 | |
| 1201 FULL_YSCALEYUV2RGB | |
| 1202 "punpcklbw %%mm1, %%mm3 \n\t" // BGBGBGBG | |
| 1203 "punpcklbw %%mm7, %%mm0 \n\t" // R0R0R0R0 | |
| 1204 | |
| 1205 "movq %%mm3, %%mm1 \n\t" | |
| 1206 "punpcklwd %%mm0, %%mm3 \n\t" // BGR0BGR0 | |
| 1207 "punpckhwd %%mm0, %%mm1 \n\t" // BGR0BGR0 | |
| 1208 | |
| 1209 MOVNTQ(%%mm3, (%4, %%REGa, 4)) | |
| 1210 MOVNTQ(%%mm1, 8(%4, %%REGa, 4)) | |
| 1211 | |
| 1212 "add $4, %%"REG_a" \n\t" | |
| 1213 "cmp %5, %%"REG_a" \n\t" | |
| 1214 " jb 1b \n\t" | |
| 1215 | |
| 1216 | |
| 1217 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" ((long)dstW), | |
| 1218 "m" (yalpha1), "m" (uvalpha1) | |
| 1219 : "%"REG_a | |
| 1220 ); | |
| 1221 break; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1222 case PIX_FMT_BGR24: |
| 18861 | 1223 asm volatile( |
| 1224 | |
| 1225 FULL_YSCALEYUV2RGB | |
| 1226 | |
| 1227 // lsb ... msb | |
| 1228 "punpcklbw %%mm1, %%mm3 \n\t" // BGBGBGBG | |
| 1229 "punpcklbw %%mm7, %%mm0 \n\t" // R0R0R0R0 | |
| 1230 | |
| 1231 "movq %%mm3, %%mm1 \n\t" | |
| 1232 "punpcklwd %%mm0, %%mm3 \n\t" // BGR0BGR0 | |
| 1233 "punpckhwd %%mm0, %%mm1 \n\t" // BGR0BGR0 | |
| 1234 | |
| 1235 "movq %%mm3, %%mm2 \n\t" // BGR0BGR0 | |
| 1236 "psrlq $8, %%mm3 \n\t" // GR0BGR00 | |
| 1237 "pand "MANGLE(bm00000111)", %%mm2\n\t" // BGR00000 | |
| 1238 "pand "MANGLE(bm11111000)", %%mm3\n\t" // 000BGR00 | |
| 1239 "por %%mm2, %%mm3 \n\t" // BGRBGR00 | |
| 1240 "movq %%mm1, %%mm2 \n\t" | |
| 1241 "psllq $48, %%mm1 \n\t" // 000000BG | |
| 1242 "por %%mm1, %%mm3 \n\t" // BGRBGRBG | |
| 1243 | |
| 1244 "movq %%mm2, %%mm1 \n\t" // BGR0BGR0 | |
| 1245 "psrld $16, %%mm2 \n\t" // R000R000 | |
| 1246 "psrlq $24, %%mm1 \n\t" // 0BGR0000 | |
| 1247 "por %%mm2, %%mm1 \n\t" // RBGRR000 | |
| 1248 | |
| 1249 "mov %4, %%"REG_b" \n\t" | |
| 1250 "add %%"REG_a", %%"REG_b" \n\t" | |
| 1251 | |
| 1252 #ifdef HAVE_MMX2 | |
| 1253 //FIXME Alignment | |
| 1254 "movntq %%mm3, (%%"REG_b", %%"REG_a", 2)\n\t" | |
| 1255 "movntq %%mm1, 8(%%"REG_b", %%"REG_a", 2)\n\t" | |
| 1256 #else | |
| 1257 "movd %%mm3, (%%"REG_b", %%"REG_a", 2) \n\t" | |
| 1258 "psrlq $32, %%mm3 \n\t" | |
| 1259 "movd %%mm3, 4(%%"REG_b", %%"REG_a", 2) \n\t" | |
| 1260 "movd %%mm1, 8(%%"REG_b", %%"REG_a", 2) \n\t" | |
| 1261 #endif | |
| 1262 "add $4, %%"REG_a" \n\t" | |
| 1263 "cmp %5, %%"REG_a" \n\t" | |
| 1264 " jb 1b \n\t" | |
| 1265 | |
| 1266 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "m" (dest), "m" (dstW), | |
| 1267 "m" (yalpha1), "m" (uvalpha1) | |
| 1268 : "%"REG_a, "%"REG_b | |
| 1269 ); | |
| 1270 break; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1271 case PIX_FMT_BGR555: |
| 18861 | 1272 asm volatile( |
| 1273 | |
| 1274 FULL_YSCALEYUV2RGB | |
| 1275 #ifdef DITHER1XBPP | |
| 1276 "paddusb "MANGLE(g5Dither)", %%mm1\n\t" | |
| 1277 "paddusb "MANGLE(r5Dither)", %%mm0\n\t" | |
| 1278 "paddusb "MANGLE(b5Dither)", %%mm3\n\t" | |
| 1279 #endif | |
| 1280 "punpcklbw %%mm7, %%mm1 \n\t" // 0G0G0G0G | |
| 1281 "punpcklbw %%mm7, %%mm3 \n\t" // 0B0B0B0B | |
| 1282 "punpcklbw %%mm7, %%mm0 \n\t" // 0R0R0R0R | |
| 1283 | |
| 1284 "psrlw $3, %%mm3 \n\t" | |
| 1285 "psllw $2, %%mm1 \n\t" | |
| 1286 "psllw $7, %%mm0 \n\t" | |
| 1287 "pand "MANGLE(g15Mask)", %%mm1 \n\t" | |
| 1288 "pand "MANGLE(r15Mask)", %%mm0 \n\t" | |
| 1289 | |
| 1290 "por %%mm3, %%mm1 \n\t" | |
| 1291 "por %%mm1, %%mm0 \n\t" | |
| 1292 | |
| 1293 MOVNTQ(%%mm0, (%4, %%REGa, 2)) | |
| 1294 | |
| 1295 "add $4, %%"REG_a" \n\t" | |
| 1296 "cmp %5, %%"REG_a" \n\t" | |
| 1297 " jb 1b \n\t" | |
| 1298 | |
| 1299 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstW), | |
| 1300 "m" (yalpha1), "m" (uvalpha1) | |
| 1301 : "%"REG_a | |
| 1302 ); | |
| 1303 break; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1304 case PIX_FMT_BGR565: |
| 18861 | 1305 asm volatile( |
| 1306 | |
| 1307 FULL_YSCALEYUV2RGB | |
| 1308 #ifdef DITHER1XBPP | |
| 1309 "paddusb "MANGLE(g6Dither)", %%mm1\n\t" | |
| 1310 "paddusb "MANGLE(r5Dither)", %%mm0\n\t" | |
| 1311 "paddusb "MANGLE(b5Dither)", %%mm3\n\t" | |
| 1312 #endif | |
| 1313 "punpcklbw %%mm7, %%mm1 \n\t" // 0G0G0G0G | |
| 1314 "punpcklbw %%mm7, %%mm3 \n\t" // 0B0B0B0B | |
| 1315 "punpcklbw %%mm7, %%mm0 \n\t" // 0R0R0R0R | |
| 1316 | |
| 1317 "psrlw $3, %%mm3 \n\t" | |
| 1318 "psllw $3, %%mm1 \n\t" | |
| 1319 "psllw $8, %%mm0 \n\t" | |
| 1320 "pand "MANGLE(g16Mask)", %%mm1 \n\t" | |
| 1321 "pand "MANGLE(r16Mask)", %%mm0 \n\t" | |
| 1322 | |
| 1323 "por %%mm3, %%mm1 \n\t" | |
| 1324 "por %%mm1, %%mm0 \n\t" | |
| 1325 | |
| 1326 MOVNTQ(%%mm0, (%4, %%REGa, 2)) | |
| 1327 | |
| 1328 "add $4, %%"REG_a" \n\t" | |
| 1329 "cmp %5, %%"REG_a" \n\t" | |
| 1330 " jb 1b \n\t" | |
| 1331 | |
| 1332 :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstW), | |
| 1333 "m" (yalpha1), "m" (uvalpha1) | |
| 1334 : "%"REG_a | |
| 1335 ); | |
| 1336 break; | |
| 1337 #endif | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1338 case PIX_FMT_BGR32: |
| 18861 | 1339 #ifndef HAVE_MMX |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1340 case PIX_FMT_RGB32: |
| 18861 | 1341 #endif |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1342 if(dstFormat==PIX_FMT_RGB32) |
| 18861 | 1343 { |
| 1344 int i; | |
| 1345 #ifdef WORDS_BIGENDIAN | |
| 1346 dest++; | |
| 1347 #endif | |
| 1348 for(i=0;i<dstW;i++){ | |
| 1349 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1350 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1351 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1352 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1353 dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)]; | |
| 1354 dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)]; | |
| 1355 dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)]; | |
| 1356 dest+= 4; | |
| 1357 } | |
| 1358 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1359 else if(dstFormat==PIX_FMT_BGR24) |
| 18861 | 1360 { |
| 1361 int i; | |
| 1362 for(i=0;i<dstW;i++){ | |
| 1363 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1364 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1365 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1366 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1367 dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)]; | |
| 1368 dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)]; | |
| 1369 dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)]; | |
| 1370 dest+= 3; | |
| 1371 } | |
| 1372 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1373 else if(dstFormat==PIX_FMT_BGR565) |
| 18861 | 1374 { |
| 1375 int i; | |
| 1376 for(i=0;i<dstW;i++){ | |
| 1377 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1378 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1379 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1380 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1381 | |
| 1382 ((uint16_t*)dest)[i] = | |
| 1383 clip_table16b[(Y + yuvtab_40cf[U]) >>13] | | |
| 1384 clip_table16g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] | | |
| 1385 clip_table16r[(Y + yuvtab_3343[V]) >>13]; | |
| 1386 } | |
| 1387 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1388 else if(dstFormat==PIX_FMT_BGR555) |
| 18861 | 1389 { |
| 1390 int i; | |
| 1391 for(i=0;i<dstW;i++){ | |
| 1392 // vertical linear interpolation && yuv2rgb in a single step: | |
| 1393 int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)]; | |
| 1394 int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19); | |
| 1395 int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19); | |
| 1396 | |
| 1397 ((uint16_t*)dest)[i] = | |
| 1398 clip_table15b[(Y + yuvtab_40cf[U]) >>13] | | |
| 1399 clip_table15g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] | | |
| 1400 clip_table15r[(Y + yuvtab_3343[V]) >>13]; | |
| 1401 } | |
| 1402 } | |
| 1403 }//FULL_UV_IPOL | |
| 1404 else | |
| 1405 { | |
| 1406 #endif // if 0 | |
| 1407 #ifdef HAVE_MMX | |
| 1408 switch(c->dstFormat) | |
| 1409 { | |
| 1410 //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :( | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1411 case PIX_FMT_RGB32: |
| 18861 | 1412 asm volatile( |
| 1413 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1414 "mov %4, %%"REG_b" \n\t" | |
| 1415 "push %%"REG_BP" \n\t" | |
| 1416 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1417 WRITEBGR32(%%REGb, 8280(%5), %%REGBP) | |
| 1418 "pop %%"REG_BP" \n\t" | |
| 1419 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1420 | |
| 1421 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1422 "a" (&c->redDither) | |
| 1423 ); | |
| 1424 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1425 case PIX_FMT_BGR24: |
| 18861 | 1426 asm volatile( |
| 1427 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1428 "mov %4, %%"REG_b" \n\t" | |
| 1429 "push %%"REG_BP" \n\t" | |
| 1430 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1431 WRITEBGR24(%%REGb, 8280(%5), %%REGBP) | |
| 1432 "pop %%"REG_BP" \n\t" | |
| 1433 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1434 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1435 "a" (&c->redDither) | |
| 1436 ); | |
| 1437 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1438 case PIX_FMT_BGR555: |
| 18861 | 1439 asm volatile( |
| 1440 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1441 "mov %4, %%"REG_b" \n\t" | |
| 1442 "push %%"REG_BP" \n\t" | |
| 1443 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1444 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1445 #ifdef DITHER1XBPP | |
| 1446 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1447 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1448 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1449 #endif | |
| 1450 | |
| 1451 WRITEBGR15(%%REGb, 8280(%5), %%REGBP) | |
| 1452 "pop %%"REG_BP" \n\t" | |
| 1453 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1454 | |
| 1455 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1456 "a" (&c->redDither) | |
| 1457 ); | |
| 1458 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1459 case PIX_FMT_BGR565: |
| 18861 | 1460 asm volatile( |
| 1461 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1462 "mov %4, %%"REG_b" \n\t" | |
| 1463 "push %%"REG_BP" \n\t" | |
| 1464 YSCALEYUV2RGB(%%REGBP, %5) | |
| 1465 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1466 #ifdef DITHER1XBPP | |
| 1467 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1468 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1469 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1470 #endif | |
| 1471 | |
| 1472 WRITEBGR16(%%REGb, 8280(%5), %%REGBP) | |
| 1473 "pop %%"REG_BP" \n\t" | |
| 1474 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1475 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1476 "a" (&c->redDither) | |
| 1477 ); | |
| 1478 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1479 case PIX_FMT_YUYV422: |
| 18861 | 1480 asm volatile( |
| 1481 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1482 "mov %4, %%"REG_b" \n\t" | |
| 1483 "push %%"REG_BP" \n\t" | |
| 1484 YSCALEYUV2PACKED(%%REGBP, %5) | |
| 1485 WRITEYUY2(%%REGb, 8280(%5), %%REGBP) | |
| 1486 "pop %%"REG_BP" \n\t" | |
| 1487 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1488 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1489 "a" (&c->redDither) | |
| 1490 ); | |
| 1491 return; | |
| 1492 default: break; | |
| 1493 } | |
| 1494 #endif //HAVE_MMX | |
| 1495 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C) | |
| 1496 } | |
| 1497 | |
| 1498 /** | |
| 1499 * YV12 to RGB without scaling or interpolating | |
| 1500 */ | |
| 1501 static inline void RENAME(yuv2packed1)(SwsContext *c, uint16_t *buf0, uint16_t *uvbuf0, uint16_t *uvbuf1, | |
| 1502 uint8_t *dest, int dstW, int uvalpha, int dstFormat, int flags, int y) | |
| 1503 { | |
| 1504 const int yalpha1=0; | |
| 1505 int i; | |
| 1506 | |
| 1507 uint16_t *buf1= buf0; //FIXME needed for the rgb1/bgr1 | |
| 1508 const int yalpha= 4096; //FIXME ... | |
| 1509 | |
| 1510 if(flags&SWS_FULL_CHR_H_INT) | |
| 1511 { | |
| 1512 RENAME(yuv2packed2)(c, buf0, buf0, uvbuf0, uvbuf1, dest, dstW, 0, uvalpha, y); | |
| 1513 return; | |
| 1514 } | |
| 1515 | |
| 1516 #ifdef HAVE_MMX | |
| 1517 if( uvalpha < 2048 ) // note this is not correct (shifts chrominance by 0.5 pixels) but its a bit faster | |
| 1518 { | |
| 1519 switch(dstFormat) | |
| 1520 { | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1521 case PIX_FMT_RGB32: |
| 18861 | 1522 asm volatile( |
| 1523 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1524 "mov %4, %%"REG_b" \n\t" | |
| 1525 "push %%"REG_BP" \n\t" | |
| 1526 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1527 WRITEBGR32(%%REGb, 8280(%5), %%REGBP) | |
| 1528 "pop %%"REG_BP" \n\t" | |
| 1529 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1530 | |
| 1531 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1532 "a" (&c->redDither) | |
| 1533 ); | |
| 1534 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1535 case PIX_FMT_BGR24: |
| 18861 | 1536 asm volatile( |
| 1537 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1538 "mov %4, %%"REG_b" \n\t" | |
| 1539 "push %%"REG_BP" \n\t" | |
| 1540 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1541 WRITEBGR24(%%REGb, 8280(%5), %%REGBP) | |
| 1542 "pop %%"REG_BP" \n\t" | |
| 1543 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1544 | |
| 1545 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1546 "a" (&c->redDither) | |
| 1547 ); | |
| 1548 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1549 case PIX_FMT_BGR555: |
| 18861 | 1550 asm volatile( |
| 1551 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1552 "mov %4, %%"REG_b" \n\t" | |
| 1553 "push %%"REG_BP" \n\t" | |
| 1554 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1555 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1556 #ifdef DITHER1XBPP | |
| 1557 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1558 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1559 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1560 #endif | |
| 1561 WRITEBGR15(%%REGb, 8280(%5), %%REGBP) | |
| 1562 "pop %%"REG_BP" \n\t" | |
| 1563 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1564 | |
| 1565 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1566 "a" (&c->redDither) | |
| 1567 ); | |
| 1568 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1569 case PIX_FMT_BGR565: |
| 18861 | 1570 asm volatile( |
| 1571 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1572 "mov %4, %%"REG_b" \n\t" | |
| 1573 "push %%"REG_BP" \n\t" | |
| 1574 YSCALEYUV2RGB1(%%REGBP, %5) | |
| 1575 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1576 #ifdef DITHER1XBPP | |
| 1577 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1578 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1579 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1580 #endif | |
| 1581 | |
| 1582 WRITEBGR16(%%REGb, 8280(%5), %%REGBP) | |
| 1583 "pop %%"REG_BP" \n\t" | |
| 1584 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1585 | |
| 1586 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1587 "a" (&c->redDither) | |
| 1588 ); | |
| 1589 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1590 case PIX_FMT_YUYV422: |
| 18861 | 1591 asm volatile( |
| 1592 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1593 "mov %4, %%"REG_b" \n\t" | |
| 1594 "push %%"REG_BP" \n\t" | |
| 1595 YSCALEYUV2PACKED1(%%REGBP, %5) | |
| 1596 WRITEYUY2(%%REGb, 8280(%5), %%REGBP) | |
| 1597 "pop %%"REG_BP" \n\t" | |
| 1598 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1599 | |
| 1600 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1601 "a" (&c->redDither) | |
| 1602 ); | |
| 1603 return; | |
| 1604 } | |
| 1605 } | |
| 1606 else | |
| 1607 { | |
| 1608 switch(dstFormat) | |
| 1609 { | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1610 case PIX_FMT_RGB32: |
| 18861 | 1611 asm volatile( |
| 1612 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1613 "mov %4, %%"REG_b" \n\t" | |
| 1614 "push %%"REG_BP" \n\t" | |
| 1615 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1616 WRITEBGR32(%%REGb, 8280(%5), %%REGBP) | |
| 1617 "pop %%"REG_BP" \n\t" | |
| 1618 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1619 | |
| 1620 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1621 "a" (&c->redDither) | |
| 1622 ); | |
| 1623 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1624 case PIX_FMT_BGR24: |
| 18861 | 1625 asm volatile( |
| 1626 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1627 "mov %4, %%"REG_b" \n\t" | |
| 1628 "push %%"REG_BP" \n\t" | |
| 1629 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1630 WRITEBGR24(%%REGb, 8280(%5), %%REGBP) | |
| 1631 "pop %%"REG_BP" \n\t" | |
| 1632 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1633 | |
| 1634 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1635 "a" (&c->redDither) | |
| 1636 ); | |
| 1637 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1638 case PIX_FMT_BGR555: |
| 18861 | 1639 asm volatile( |
| 1640 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1641 "mov %4, %%"REG_b" \n\t" | |
| 1642 "push %%"REG_BP" \n\t" | |
| 1643 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1644 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1645 #ifdef DITHER1XBPP | |
| 1646 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1647 "paddusb "MANGLE(g5Dither)", %%mm4\n\t" | |
| 1648 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1649 #endif | |
| 1650 WRITEBGR15(%%REGb, 8280(%5), %%REGBP) | |
| 1651 "pop %%"REG_BP" \n\t" | |
| 1652 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1653 | |
| 1654 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1655 "a" (&c->redDither) | |
| 1656 ); | |
| 1657 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1658 case PIX_FMT_BGR565: |
| 18861 | 1659 asm volatile( |
| 1660 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1661 "mov %4, %%"REG_b" \n\t" | |
| 1662 "push %%"REG_BP" \n\t" | |
| 1663 YSCALEYUV2RGB1b(%%REGBP, %5) | |
| 1664 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ | |
| 1665 #ifdef DITHER1XBPP | |
| 1666 "paddusb "MANGLE(b5Dither)", %%mm2\n\t" | |
| 1667 "paddusb "MANGLE(g6Dither)", %%mm4\n\t" | |
| 1668 "paddusb "MANGLE(r5Dither)", %%mm5\n\t" | |
| 1669 #endif | |
| 1670 | |
| 1671 WRITEBGR16(%%REGb, 8280(%5), %%REGBP) | |
| 1672 "pop %%"REG_BP" \n\t" | |
| 1673 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1674 | |
| 1675 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1676 "a" (&c->redDither) | |
| 1677 ); | |
| 1678 return; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
1679 case PIX_FMT_YUYV422: |
| 18861 | 1680 asm volatile( |
| 1681 "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t" | |
| 1682 "mov %4, %%"REG_b" \n\t" | |
| 1683 "push %%"REG_BP" \n\t" | |
| 1684 YSCALEYUV2PACKED1b(%%REGBP, %5) | |
| 1685 WRITEYUY2(%%REGb, 8280(%5), %%REGBP) | |
| 1686 "pop %%"REG_BP" \n\t" | |
| 1687 "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t" | |
| 1688 | |
| 1689 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest), | |
| 1690 "a" (&c->redDither) | |
| 1691 ); | |
| 1692 return; | |
| 1693 } | |
| 1694 } | |
| 1695 #endif | |
| 1696 if( uvalpha < 2048 ) | |
| 1697 { | |
| 1698 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1_C, YSCALE_YUV_2_PACKED1_C) | |
| 1699 }else{ | |
| 1700 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1B_C, YSCALE_YUV_2_PACKED1B_C) | |
| 1701 } | |
| 1702 } | |
| 1703 | |
| 1704 //FIXME yuy2* can read upto 7 samples to much | |
| 1705 | |
| 1706 static inline void RENAME(yuy2ToY)(uint8_t *dst, uint8_t *src, long width) | |
| 1707 { | |
| 1708 #ifdef HAVE_MMX | |
| 1709 asm volatile( | |
| 1710 "movq "MANGLE(bm01010101)", %%mm2\n\t" | |
| 1711 "mov %0, %%"REG_a" \n\t" | |
| 1712 "1: \n\t" | |
| 1713 "movq (%1, %%"REG_a",2), %%mm0 \n\t" | |
| 1714 "movq 8(%1, %%"REG_a",2), %%mm1 \n\t" | |
| 1715 "pand %%mm2, %%mm0 \n\t" | |
| 1716 "pand %%mm2, %%mm1 \n\t" | |
| 1717 "packuswb %%mm1, %%mm0 \n\t" | |
| 1718 "movq %%mm0, (%2, %%"REG_a") \n\t" | |
| 1719 "add $8, %%"REG_a" \n\t" | |
| 1720 " js 1b \n\t" | |
| 1721 : : "g" (-width), "r" (src+width*2), "r" (dst+width) | |
| 1722 : "%"REG_a | |
| 1723 ); | |
| 1724 #else | |
| 1725 int i; | |
| 1726 for(i=0; i<width; i++) | |
| 1727 dst[i]= src[2*i]; | |
| 1728 #endif | |
| 1729 } | |
| 1730 | |
| 1731 static inline void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width) | |
| 1732 { | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1733 #ifdef HAVE_MMX |
| 18861 | 1734 asm volatile( |
| 1735 "movq "MANGLE(bm01010101)", %%mm4\n\t" | |
| 1736 "mov %0, %%"REG_a" \n\t" | |
| 1737 "1: \n\t" | |
| 1738 "movq (%1, %%"REG_a",4), %%mm0 \n\t" | |
| 1739 "movq 8(%1, %%"REG_a",4), %%mm1 \n\t" | |
| 1740 "psrlw $8, %%mm0 \n\t" | |
| 1741 "psrlw $8, %%mm1 \n\t" | |
| 1742 "packuswb %%mm1, %%mm0 \n\t" | |
| 1743 "movq %%mm0, %%mm1 \n\t" | |
| 1744 "psrlw $8, %%mm0 \n\t" | |
| 1745 "pand %%mm4, %%mm1 \n\t" | |
| 1746 "packuswb %%mm0, %%mm0 \n\t" | |
| 1747 "packuswb %%mm1, %%mm1 \n\t" | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1748 "movd %%mm0, (%3, %%"REG_a") \n\t" |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1749 "movd %%mm1, (%2, %%"REG_a") \n\t" |
| 18861 | 1750 "add $4, %%"REG_a" \n\t" |
| 1751 " js 1b \n\t" | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1752 : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width) |
| 18861 | 1753 : "%"REG_a |
| 1754 ); | |
| 1755 #else | |
| 1756 int i; | |
| 1757 for(i=0; i<width; i++) | |
| 1758 { | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1759 dstU[i]= src1[4*i + 1]; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1760 dstV[i]= src1[4*i + 3]; |
| 18861 | 1761 } |
| 1762 #endif | |
| 21686 | 1763 assert(src1 == src2); |
| 18861 | 1764 } |
| 1765 | |
| 1766 //this is allmost identical to the previous, end exists only cuz yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses | |
| 1767 static inline void RENAME(uyvyToY)(uint8_t *dst, uint8_t *src, long width) | |
| 1768 { | |
| 1769 #ifdef HAVE_MMX | |
| 1770 asm volatile( | |
| 1771 "mov %0, %%"REG_a" \n\t" | |
| 1772 "1: \n\t" | |
| 1773 "movq (%1, %%"REG_a",2), %%mm0 \n\t" | |
| 1774 "movq 8(%1, %%"REG_a",2), %%mm1 \n\t" | |
| 1775 "psrlw $8, %%mm0 \n\t" | |
| 1776 "psrlw $8, %%mm1 \n\t" | |
| 1777 "packuswb %%mm1, %%mm0 \n\t" | |
| 1778 "movq %%mm0, (%2, %%"REG_a") \n\t" | |
| 1779 "add $8, %%"REG_a" \n\t" | |
| 1780 " js 1b \n\t" | |
| 1781 : : "g" (-width), "r" (src+width*2), "r" (dst+width) | |
| 1782 : "%"REG_a | |
| 1783 ); | |
| 1784 #else | |
| 1785 int i; | |
| 1786 for(i=0; i<width; i++) | |
| 1787 dst[i]= src[2*i+1]; | |
| 1788 #endif | |
| 1789 } | |
| 1790 | |
| 1791 static inline void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width) | |
| 1792 { | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1793 #ifdef HAVE_MMX |
| 18861 | 1794 asm volatile( |
| 1795 "movq "MANGLE(bm01010101)", %%mm4\n\t" | |
| 1796 "mov %0, %%"REG_a" \n\t" | |
| 1797 "1: \n\t" | |
| 1798 "movq (%1, %%"REG_a",4), %%mm0 \n\t" | |
| 1799 "movq 8(%1, %%"REG_a",4), %%mm1 \n\t" | |
| 1800 "pand %%mm4, %%mm0 \n\t" | |
| 1801 "pand %%mm4, %%mm1 \n\t" | |
| 1802 "packuswb %%mm1, %%mm0 \n\t" | |
| 1803 "movq %%mm0, %%mm1 \n\t" | |
| 1804 "psrlw $8, %%mm0 \n\t" | |
| 1805 "pand %%mm4, %%mm1 \n\t" | |
| 1806 "packuswb %%mm0, %%mm0 \n\t" | |
| 1807 "packuswb %%mm1, %%mm1 \n\t" | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1808 "movd %%mm0, (%3, %%"REG_a") \n\t" |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1809 "movd %%mm1, (%2, %%"REG_a") \n\t" |
| 18861 | 1810 "add $4, %%"REG_a" \n\t" |
| 1811 " js 1b \n\t" | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1812 : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width) |
| 18861 | 1813 : "%"REG_a |
| 1814 ); | |
| 1815 #else | |
| 1816 int i; | |
| 1817 for(i=0; i<width; i++) | |
| 1818 { | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1819 dstU[i]= src1[4*i + 0]; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1820 dstV[i]= src1[4*i + 2]; |
| 18861 | 1821 } |
| 1822 #endif | |
| 21686 | 1823 assert(src1 == src2); |
| 18861 | 1824 } |
| 1825 | |
| 1826 static inline void RENAME(bgr32ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 1827 { | |
| 1828 int i; | |
| 1829 for(i=0; i<width; i++) | |
| 1830 { | |
| 1831 int b= ((uint32_t*)src)[i]&0xFF; | |
| 1832 int g= (((uint32_t*)src)[i]>>8)&0xFF; | |
| 1833 int r= (((uint32_t*)src)[i]>>16)&0xFF; | |
| 1834 | |
| 1835 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 1836 } | |
| 1837 } | |
| 1838 | |
| 1839 static inline void RENAME(bgr32ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 1840 { | |
| 1841 int i; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1842 assert(src1 == src2); |
| 18861 | 1843 for(i=0; i<width; i++) |
| 1844 { | |
| 1845 const int a= ((uint32_t*)src1)[2*i+0]; | |
| 1846 const int e= ((uint32_t*)src1)[2*i+1]; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1847 const int l= (a&0xFF00FF) + (e&0xFF00FF); |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1848 const int h= (a&0x00FF00) + (e&0x00FF00); |
| 18861 | 1849 const int b= l&0x3FF; |
| 1850 const int g= h>>8; | |
| 1851 const int r= l>>16; | |
| 1852 | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1853 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+1)) + 128; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1854 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+1)) + 128; |
| 18861 | 1855 } |
| 1856 } | |
| 1857 | |
| 1858 static inline void RENAME(bgr24ToY)(uint8_t *dst, uint8_t *src, long width) | |
| 1859 { | |
| 1860 #ifdef HAVE_MMX | |
| 1861 asm volatile( | |
| 1862 "mov %2, %%"REG_a" \n\t" | |
| 1863 "movq "MANGLE(bgr2YCoeff)", %%mm6 \n\t" | |
| 1864 "movq "MANGLE(w1111)", %%mm5 \n\t" | |
| 1865 "pxor %%mm7, %%mm7 \n\t" | |
| 19396 | 1866 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d"\n\t" |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
1867 ASMALIGN(4) |
| 18861 | 1868 "1: \n\t" |
| 19396 | 1869 PREFETCH" 64(%0, %%"REG_d") \n\t" |
| 1870 "movd (%0, %%"REG_d"), %%mm0 \n\t" | |
| 1871 "movd 3(%0, %%"REG_d"), %%mm1 \n\t" | |
| 18861 | 1872 "punpcklbw %%mm7, %%mm0 \n\t" |
| 1873 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 19396 | 1874 "movd 6(%0, %%"REG_d"), %%mm2 \n\t" |
| 1875 "movd 9(%0, %%"REG_d"), %%mm3 \n\t" | |
| 18861 | 1876 "punpcklbw %%mm7, %%mm2 \n\t" |
| 1877 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 1878 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 1879 "pmaddwd %%mm6, %%mm1 \n\t" | |
| 1880 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 1881 "pmaddwd %%mm6, %%mm3 \n\t" | |
| 1882 #ifndef FAST_BGR2YV12 | |
| 1883 "psrad $8, %%mm0 \n\t" | |
| 1884 "psrad $8, %%mm1 \n\t" | |
| 1885 "psrad $8, %%mm2 \n\t" | |
| 1886 "psrad $8, %%mm3 \n\t" | |
| 1887 #endif | |
| 1888 "packssdw %%mm1, %%mm0 \n\t" | |
| 1889 "packssdw %%mm3, %%mm2 \n\t" | |
| 1890 "pmaddwd %%mm5, %%mm0 \n\t" | |
| 1891 "pmaddwd %%mm5, %%mm2 \n\t" | |
| 1892 "packssdw %%mm2, %%mm0 \n\t" | |
| 1893 "psraw $7, %%mm0 \n\t" | |
| 1894 | |
| 19396 | 1895 "movd 12(%0, %%"REG_d"), %%mm4 \n\t" |
| 1896 "movd 15(%0, %%"REG_d"), %%mm1 \n\t" | |
| 18861 | 1897 "punpcklbw %%mm7, %%mm4 \n\t" |
| 1898 "punpcklbw %%mm7, %%mm1 \n\t" | |
| 19396 | 1899 "movd 18(%0, %%"REG_d"), %%mm2 \n\t" |
| 1900 "movd 21(%0, %%"REG_d"), %%mm3 \n\t" | |
| 18861 | 1901 "punpcklbw %%mm7, %%mm2 \n\t" |
| 1902 "punpcklbw %%mm7, %%mm3 \n\t" | |
| 1903 "pmaddwd %%mm6, %%mm4 \n\t" | |
| 1904 "pmaddwd %%mm6, %%mm1 \n\t" | |
| 1905 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 1906 "pmaddwd %%mm6, %%mm3 \n\t" | |
| 1907 #ifndef FAST_BGR2YV12 | |
| 1908 "psrad $8, %%mm4 \n\t" | |
| 1909 "psrad $8, %%mm1 \n\t" | |
| 1910 "psrad $8, %%mm2 \n\t" | |
| 1911 "psrad $8, %%mm3 \n\t" | |
| 1912 #endif | |
| 1913 "packssdw %%mm1, %%mm4 \n\t" | |
| 1914 "packssdw %%mm3, %%mm2 \n\t" | |
| 1915 "pmaddwd %%mm5, %%mm4 \n\t" | |
| 1916 "pmaddwd %%mm5, %%mm2 \n\t" | |
| 19396 | 1917 "add $24, %%"REG_d" \n\t" |
| 18861 | 1918 "packssdw %%mm2, %%mm4 \n\t" |
| 1919 "psraw $7, %%mm4 \n\t" | |
| 1920 | |
| 1921 "packuswb %%mm4, %%mm0 \n\t" | |
| 1922 "paddusb "MANGLE(bgr2YOffset)", %%mm0 \n\t" | |
| 1923 | |
| 1924 "movq %%mm0, (%1, %%"REG_a") \n\t" | |
| 1925 "add $8, %%"REG_a" \n\t" | |
| 1926 " js 1b \n\t" | |
| 1927 : : "r" (src+width*3), "r" (dst+width), "g" (-width) | |
| 19396 | 1928 : "%"REG_a, "%"REG_d |
| 18861 | 1929 ); |
| 1930 #else | |
| 1931 int i; | |
| 1932 for(i=0; i<width; i++) | |
| 1933 { | |
| 1934 int b= src[i*3+0]; | |
| 1935 int g= src[i*3+1]; | |
| 1936 int r= src[i*3+2]; | |
| 1937 | |
| 1938 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 1939 } | |
| 1940 #endif | |
| 1941 } | |
| 1942 | |
| 1943 static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width) | |
| 1944 { | |
| 1945 #ifdef HAVE_MMX | |
| 1946 asm volatile( | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1947 "mov %3, %%"REG_a" \n\t" |
| 18861 | 1948 "movq "MANGLE(w1111)", %%mm5 \n\t" |
| 1949 "movq "MANGLE(bgr2UCoeff)", %%mm6 \n\t" | |
| 1950 "pxor %%mm7, %%mm7 \n\t" | |
| 19396 | 1951 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d" \n\t" |
| 1952 "add %%"REG_d", %%"REG_d" \n\t" | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
1953 ASMALIGN(4) |
| 18861 | 1954 "1: \n\t" |
| 19396 | 1955 PREFETCH" 64(%0, %%"REG_d") \n\t" |
| 18861 | 1956 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) |
| 19396 | 1957 "movq (%0, %%"REG_d"), %%mm0 \n\t" |
| 1958 "movq 6(%0, %%"REG_d"), %%mm2 \n\t" | |
| 18861 | 1959 "movq %%mm0, %%mm1 \n\t" |
| 1960 "movq %%mm2, %%mm3 \n\t" | |
| 1961 "psrlq $24, %%mm0 \n\t" | |
| 1962 "psrlq $24, %%mm2 \n\t" | |
| 1963 PAVGB(%%mm1, %%mm0) | |
| 1964 PAVGB(%%mm3, %%mm2) | |
| 1965 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 1966 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 1967 #else | |
| 19396 | 1968 "movd (%0, %%"REG_d"), %%mm0 \n\t" |
| 1969 "movd 3(%0, %%"REG_d"), %%mm2 \n\t" | |
| 18861 | 1970 "punpcklbw %%mm7, %%mm0 \n\t" |
| 1971 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 1972 "paddw %%mm2, %%mm0 \n\t" | |
| 19396 | 1973 "movd 6(%0, %%"REG_d"), %%mm4 \n\t" |
| 1974 "movd 9(%0, %%"REG_d"), %%mm2 \n\t" | |
| 18861 | 1975 "punpcklbw %%mm7, %%mm4 \n\t" |
| 1976 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 1977 "paddw %%mm4, %%mm2 \n\t" | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1978 "psrlw $1, %%mm0 \n\t" |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
1979 "psrlw $1, %%mm2 \n\t" |
| 18861 | 1980 #endif |
| 1981 "movq "MANGLE(bgr2VCoeff)", %%mm1 \n\t" | |
| 1982 "movq "MANGLE(bgr2VCoeff)", %%mm3 \n\t" | |
| 1983 | |
| 1984 "pmaddwd %%mm0, %%mm1 \n\t" | |
| 1985 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 1986 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 1987 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 1988 #ifndef FAST_BGR2YV12 | |
| 1989 "psrad $8, %%mm0 \n\t" | |
| 1990 "psrad $8, %%mm1 \n\t" | |
| 1991 "psrad $8, %%mm2 \n\t" | |
| 1992 "psrad $8, %%mm3 \n\t" | |
| 1993 #endif | |
| 1994 "packssdw %%mm2, %%mm0 \n\t" | |
| 1995 "packssdw %%mm3, %%mm1 \n\t" | |
| 1996 "pmaddwd %%mm5, %%mm0 \n\t" | |
| 1997 "pmaddwd %%mm5, %%mm1 \n\t" | |
| 1998 "packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0 | |
| 1999 "psraw $7, %%mm0 \n\t" | |
| 2000 | |
| 2001 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW) | |
| 19396 | 2002 "movq 12(%0, %%"REG_d"), %%mm4 \n\t" |
| 2003 "movq 18(%0, %%"REG_d"), %%mm2 \n\t" | |
| 18861 | 2004 "movq %%mm4, %%mm1 \n\t" |
| 2005 "movq %%mm2, %%mm3 \n\t" | |
| 2006 "psrlq $24, %%mm4 \n\t" | |
| 2007 "psrlq $24, %%mm2 \n\t" | |
| 2008 PAVGB(%%mm1, %%mm4) | |
| 2009 PAVGB(%%mm3, %%mm2) | |
| 2010 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 2011 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2012 #else | |
| 19396 | 2013 "movd 12(%0, %%"REG_d"), %%mm4 \n\t" |
| 2014 "movd 15(%0, %%"REG_d"), %%mm2 \n\t" | |
| 18861 | 2015 "punpcklbw %%mm7, %%mm4 \n\t" |
| 2016 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2017 "paddw %%mm2, %%mm4 \n\t" | |
| 19396 | 2018 "movd 18(%0, %%"REG_d"), %%mm5 \n\t" |
| 2019 "movd 21(%0, %%"REG_d"), %%mm2 \n\t" | |
| 18861 | 2020 "punpcklbw %%mm7, %%mm5 \n\t" |
| 2021 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2022 "paddw %%mm5, %%mm2 \n\t" | |
| 2023 "movq "MANGLE(w1111)", %%mm5 \n\t" | |
| 2024 "psrlw $2, %%mm4 \n\t" | |
| 2025 "psrlw $2, %%mm2 \n\t" | |
| 2026 #endif | |
| 2027 "movq "MANGLE(bgr2VCoeff)", %%mm1 \n\t" | |
| 2028 "movq "MANGLE(bgr2VCoeff)", %%mm3 \n\t" | |
| 2029 | |
| 2030 "pmaddwd %%mm4, %%mm1 \n\t" | |
| 2031 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2032 "pmaddwd %%mm6, %%mm4 \n\t" | |
| 2033 "pmaddwd %%mm6, %%mm2 \n\t" | |
| 2034 #ifndef FAST_BGR2YV12 | |
| 2035 "psrad $8, %%mm4 \n\t" | |
| 2036 "psrad $8, %%mm1 \n\t" | |
| 2037 "psrad $8, %%mm2 \n\t" | |
| 2038 "psrad $8, %%mm3 \n\t" | |
| 2039 #endif | |
| 2040 "packssdw %%mm2, %%mm4 \n\t" | |
| 2041 "packssdw %%mm3, %%mm1 \n\t" | |
| 2042 "pmaddwd %%mm5, %%mm4 \n\t" | |
| 2043 "pmaddwd %%mm5, %%mm1 \n\t" | |
| 19396 | 2044 "add $24, %%"REG_d" \n\t" |
| 18861 | 2045 "packssdw %%mm1, %%mm4 \n\t" // V3 V2 U3 U2 |
| 2046 "psraw $7, %%mm4 \n\t" | |
| 2047 | |
| 2048 "movq %%mm0, %%mm1 \n\t" | |
| 2049 "punpckldq %%mm4, %%mm0 \n\t" | |
| 2050 "punpckhdq %%mm4, %%mm1 \n\t" | |
| 2051 "packsswb %%mm1, %%mm0 \n\t" | |
| 2052 "paddb "MANGLE(bgr2UVOffset)", %%mm0 \n\t" | |
| 2053 | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2054 "movd %%mm0, (%1, %%"REG_a") \n\t" |
| 18861 | 2055 "punpckhdq %%mm0, %%mm0 \n\t" |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2056 "movd %%mm0, (%2, %%"REG_a") \n\t" |
| 18861 | 2057 "add $4, %%"REG_a" \n\t" |
| 2058 " js 1b \n\t" | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2059 : : "r" (src1+width*6), "r" (dstU+width), "r" (dstV+width), "g" (-width) |
| 19396 | 2060 : "%"REG_a, "%"REG_d |
| 18861 | 2061 ); |
| 2062 #else | |
| 2063 int i; | |
| 2064 for(i=0; i<width; i++) | |
| 2065 { | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2066 int b= src1[6*i + 0] + src1[6*i + 3]; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2067 int g= src1[6*i + 1] + src1[6*i + 4]; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2068 int r= src1[6*i + 2] + src1[6*i + 5]; |
| 18861 | 2069 |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2070 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+1)) + 128; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2071 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+1)) + 128; |
| 18861 | 2072 } |
| 2073 #endif | |
| 21686 | 2074 assert(src1 == src2); |
| 18861 | 2075 } |
| 2076 | |
| 2077 static inline void RENAME(bgr16ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2078 { | |
| 2079 int i; | |
| 2080 for(i=0; i<width; i++) | |
| 2081 { | |
| 2082 int d= ((uint16_t*)src)[i]; | |
| 2083 int b= d&0x1F; | |
| 2084 int g= (d>>5)&0x3F; | |
| 2085 int r= (d>>11)&0x1F; | |
| 2086 | |
| 2087 dst[i]= ((2*RY*r + GY*g + 2*BY*b)>>(RGB2YUV_SHIFT-2)) + 16; | |
| 2088 } | |
| 2089 } | |
| 2090 | |
| 2091 static inline void RENAME(bgr16ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2092 { | |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2093 int i; |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2094 assert(src1==src2); |
| 18861 | 2095 for(i=0; i<width; i++) |
| 2096 { | |
| 2097 int d0= ((uint32_t*)src1)[i]; | |
| 2098 | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2099 int dl= (d0&0x07E0F81F); |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2100 int dh= ((d0>>5)&0x07C0F83F); |
| 18861 | 2101 |
| 2102 int dh2= (dh>>11) + (dh<<21); | |
| 2103 int d= dh2 + dl; | |
| 2104 | |
| 2105 int b= d&0x7F; | |
| 2106 int r= (d>>11)&0x7F; | |
| 2107 int g= d>>21; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2108 dstU[i]= ((2*RU*r + GU*g + 2*BU*b)>>(RGB2YUV_SHIFT+1-2)) + 128; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2109 dstV[i]= ((2*RV*r + GV*g + 2*BV*b)>>(RGB2YUV_SHIFT+1-2)) + 128; |
| 18861 | 2110 } |
| 2111 } | |
| 2112 | |
| 2113 static inline void RENAME(bgr15ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2114 { | |
| 2115 int i; | |
| 2116 for(i=0; i<width; i++) | |
| 2117 { | |
| 2118 int d= ((uint16_t*)src)[i]; | |
| 2119 int b= d&0x1F; | |
| 2120 int g= (d>>5)&0x1F; | |
| 2121 int r= (d>>10)&0x1F; | |
| 2122 | |
| 2123 dst[i]= ((RY*r + GY*g + BY*b)>>(RGB2YUV_SHIFT-3)) + 16; | |
| 2124 } | |
| 2125 } | |
| 2126 | |
| 2127 static inline void RENAME(bgr15ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2128 { | |
| 2129 int i; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2130 assert(src1==src2); |
| 18861 | 2131 for(i=0; i<width; i++) |
| 2132 { | |
| 2133 int d0= ((uint32_t*)src1)[i]; | |
| 2134 | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2135 int dl= (d0&0x03E07C1F); |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2136 int dh= ((d0>>5)&0x03E0F81F); |
| 18861 | 2137 |
| 2138 int dh2= (dh>>11) + (dh<<21); | |
| 2139 int d= dh2 + dl; | |
| 2140 | |
| 2141 int b= d&0x7F; | |
| 2142 int r= (d>>10)&0x7F; | |
| 2143 int g= d>>21; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2144 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+1-3)) + 128; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2145 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+1-3)) + 128; |
| 18861 | 2146 } |
| 2147 } | |
| 2148 | |
| 2149 | |
| 2150 static inline void RENAME(rgb32ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2151 { | |
| 2152 int i; | |
| 2153 for(i=0; i<width; i++) | |
| 2154 { | |
| 2155 int r= ((uint32_t*)src)[i]&0xFF; | |
| 2156 int g= (((uint32_t*)src)[i]>>8)&0xFF; | |
| 2157 int b= (((uint32_t*)src)[i]>>16)&0xFF; | |
| 2158 | |
| 2159 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 2160 } | |
| 2161 } | |
| 2162 | |
| 2163 static inline void RENAME(rgb32ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2164 { | |
| 2165 int i; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2166 assert(src1==src2); |
| 18861 | 2167 for(i=0; i<width; i++) |
| 2168 { | |
| 2169 const int a= ((uint32_t*)src1)[2*i+0]; | |
| 2170 const int e= ((uint32_t*)src1)[2*i+1]; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2171 const int l= (a&0xFF00FF) + (e&0xFF00FF); |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2172 const int h= (a&0x00FF00) + (e&0x00FF00); |
| 18861 | 2173 const int r= l&0x3FF; |
| 2174 const int g= h>>8; | |
| 2175 const int b= l>>16; | |
| 2176 | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2177 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+1)) + 128; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2178 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+1)) + 128; |
| 18861 | 2179 } |
| 2180 } | |
| 2181 | |
| 2182 static inline void RENAME(rgb24ToY)(uint8_t *dst, uint8_t *src, int width) | |
| 2183 { | |
| 2184 int i; | |
| 2185 for(i=0; i<width; i++) | |
| 2186 { | |
| 2187 int r= src[i*3+0]; | |
| 2188 int g= src[i*3+1]; | |
| 2189 int b= src[i*3+2]; | |
| 2190 | |
| 2191 dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT); | |
| 2192 } | |
| 2193 } | |
| 2194 | |
| 2195 static inline void RENAME(rgb24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) | |
| 2196 { | |
| 2197 int i; | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2198 assert(src1==src2); |
| 18861 | 2199 for(i=0; i<width; i++) |
| 2200 { | |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2201 int r= src1[6*i + 0] + src1[6*i + 3]; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2202 int g= src1[6*i + 1] + src1[6*i + 4]; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2203 int b= src1[6*i + 2] + src1[6*i + 5]; |
| 18861 | 2204 |
|
20945
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2205 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+1)) + 128; |
|
92150c16e737
fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
michael
parents:
20724
diff
changeset
|
2206 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+1)) + 128; |
| 18861 | 2207 } |
| 2208 } | |
| 2209 | |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2210 static inline void RENAME(rgb16ToY)(uint8_t *dst, uint8_t *src, int width) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2211 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2212 int i; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2213 for(i=0; i<width; i++) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2214 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2215 int d= ((uint16_t*)src)[i]; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2216 int r= d&0x1F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2217 int g= (d>>5)&0x3F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2218 int b= (d>>11)&0x1F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2219 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2220 dst[i]= ((2*RY*r + GY*g + 2*BY*b)>>(RGB2YUV_SHIFT-2)) + 16; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2221 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2222 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2223 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2224 static inline void RENAME(rgb16ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2225 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2226 int i; |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2227 assert(src1 == src2); |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2228 for(i=0; i<width; i++) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2229 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2230 int d0= ((uint32_t*)src1)[i]; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2231 |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2232 int dl= (d0&0x07E0F81F); |
|
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2233 int dh= ((d0>>5)&0x07C0F83F); |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2234 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2235 int dh2= (dh>>11) + (dh<<21); |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2236 int d= dh2 + dl; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2237 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2238 int r= d&0x7F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2239 int b= (d>>11)&0x7F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2240 int g= d>>21; |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2241 dstU[i]= ((2*RU*r + GU*g + 2*BU*b)>>(RGB2YUV_SHIFT+1-2)) + 128; |
|
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2242 dstV[i]= ((2*RV*r + GV*g + 2*BV*b)>>(RGB2YUV_SHIFT+1-2)) + 128; |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2243 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2244 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2245 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2246 static inline void RENAME(rgb15ToY)(uint8_t *dst, uint8_t *src, int width) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2247 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2248 int i; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2249 for(i=0; i<width; i++) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2250 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2251 int d= ((uint16_t*)src)[i]; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2252 int r= d&0x1F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2253 int g= (d>>5)&0x1F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2254 int b= (d>>10)&0x1F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2255 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2256 dst[i]= ((RY*r + GY*g + BY*b)>>(RGB2YUV_SHIFT-3)) + 16; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2257 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2258 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2259 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2260 static inline void RENAME(rgb15ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2261 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2262 int i; |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2263 assert(src1 == src2); |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2264 for(i=0; i<width; i++) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2265 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2266 int d0= ((uint32_t*)src1)[i]; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2267 |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2268 int dl= (d0&0x03E07C1F); |
|
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2269 int dh= ((d0>>5)&0x03E0F81F); |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2270 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2271 int dh2= (dh>>11) + (dh<<21); |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2272 int d= dh2 + dl; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2273 |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2274 int g= d&0x7F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2275 int r= (d>>10)&0x7F; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2276 int b= d>>21; |
|
20946
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2277 dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+1-3)) + 128; |
|
bb4c952bc52c
forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
michael
parents:
20945
diff
changeset
|
2278 dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+1-3)) + 128; |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2279 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2280 } |
| 18861 | 2281 |
| 22218 | 2282 static inline void RENAME(palToY)(uint8_t *dst, uint8_t *src, int width, uint32_t *pal) |
| 2283 { | |
| 2284 int i; | |
| 2285 for(i=0; i<width; i++) | |
| 2286 { | |
| 2287 int d= src[i]; | |
| 2288 | |
|
22975
c91bb8f1cd85
convert palette to yuv instead of converting each pixel after pal->rgb
michael
parents:
22914
diff
changeset
|
2289 dst[i]= pal[d] & 0xFF; |
| 22218 | 2290 } |
| 2291 } | |
| 2292 | |
| 2293 static inline void RENAME(palToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width, uint32_t *pal) | |
| 2294 { | |
| 2295 int i; | |
| 2296 assert(src1 == src2); | |
| 2297 for(i=0; i<width; i++) | |
| 2298 { | |
| 22976 | 2299 int p= pal[src1[i]]; |
| 22218 | 2300 |
| 22976 | 2301 dstU[i]= p>>8; |
| 2302 dstV[i]= p>>16; | |
| 22218 | 2303 } |
| 2304 } | |
| 2305 | |
| 18861 | 2306 // Bilinear / Bicubic scaling |
| 2307 static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc, | |
| 2308 int16_t *filter, int16_t *filterPos, long filterSize) | |
| 2309 { | |
| 2310 #ifdef HAVE_MMX | |
| 2311 assert(filterSize % 4 == 0 && filterSize>0); | |
| 22914 | 2312 if(filterSize==4) // Always true for upscaling, sometimes for down, too. |
| 18861 | 2313 { |
| 2314 long counter= -2*dstW; | |
| 2315 filter-= counter*2; | |
| 2316 filterPos-= counter/2; | |
| 2317 dst-= counter/2; | |
| 2318 asm volatile( | |
| 19396 | 2319 #if defined(PIC) |
| 2320 "push %%"REG_b" \n\t" | |
| 2321 #endif | |
| 18861 | 2322 "pxor %%mm7, %%mm7 \n\t" |
| 2323 "movq "MANGLE(w02)", %%mm6 \n\t" | |
| 2324 "push %%"REG_BP" \n\t" // we use 7 regs here ... | |
| 2325 "mov %%"REG_a", %%"REG_BP" \n\t" | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
2326 ASMALIGN(4) |
| 18861 | 2327 "1: \n\t" |
| 2328 "movzwl (%2, %%"REG_BP"), %%eax \n\t" | |
| 2329 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t" | |
| 2330 "movq (%1, %%"REG_BP", 4), %%mm1\n\t" | |
| 2331 "movq 8(%1, %%"REG_BP", 4), %%mm3\n\t" | |
| 2332 "movd (%3, %%"REG_a"), %%mm0 \n\t" | |
| 2333 "movd (%3, %%"REG_b"), %%mm2 \n\t" | |
| 2334 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2335 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2336 "pmaddwd %%mm1, %%mm0 \n\t" | |
| 2337 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2338 "psrad $8, %%mm0 \n\t" | |
| 2339 "psrad $8, %%mm3 \n\t" | |
| 2340 "packssdw %%mm3, %%mm0 \n\t" | |
| 2341 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 2342 "packssdw %%mm0, %%mm0 \n\t" | |
| 2343 "movd %%mm0, (%4, %%"REG_BP") \n\t" | |
| 2344 "add $4, %%"REG_BP" \n\t" | |
| 2345 " jnc 1b \n\t" | |
| 2346 | |
| 2347 "pop %%"REG_BP" \n\t" | |
| 19396 | 2348 #if defined(PIC) |
| 2349 "pop %%"REG_b" \n\t" | |
| 2350 #endif | |
| 18861 | 2351 : "+a" (counter) |
| 2352 : "c" (filter), "d" (filterPos), "S" (src), "D" (dst) | |
| 19396 | 2353 #if !defined(PIC) |
| 18861 | 2354 : "%"REG_b |
| 19396 | 2355 #endif |
| 18861 | 2356 ); |
| 2357 } | |
| 2358 else if(filterSize==8) | |
| 2359 { | |
| 2360 long counter= -2*dstW; | |
| 2361 filter-= counter*4; | |
| 2362 filterPos-= counter/2; | |
| 2363 dst-= counter/2; | |
| 2364 asm volatile( | |
| 19396 | 2365 #if defined(PIC) |
| 2366 "push %%"REG_b" \n\t" | |
| 2367 #endif | |
| 18861 | 2368 "pxor %%mm7, %%mm7 \n\t" |
| 2369 "movq "MANGLE(w02)", %%mm6 \n\t" | |
| 2370 "push %%"REG_BP" \n\t" // we use 7 regs here ... | |
| 2371 "mov %%"REG_a", %%"REG_BP" \n\t" | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
2372 ASMALIGN(4) |
| 18861 | 2373 "1: \n\t" |
| 2374 "movzwl (%2, %%"REG_BP"), %%eax \n\t" | |
| 2375 "movzwl 2(%2, %%"REG_BP"), %%ebx\n\t" | |
| 2376 "movq (%1, %%"REG_BP", 8), %%mm1\n\t" | |
| 2377 "movq 16(%1, %%"REG_BP", 8), %%mm3\n\t" | |
| 2378 "movd (%3, %%"REG_a"), %%mm0 \n\t" | |
| 2379 "movd (%3, %%"REG_b"), %%mm2 \n\t" | |
| 2380 "punpcklbw %%mm7, %%mm0 \n\t" | |
| 2381 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2382 "pmaddwd %%mm1, %%mm0 \n\t" | |
| 2383 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2384 | |
| 2385 "movq 8(%1, %%"REG_BP", 8), %%mm1\n\t" | |
| 2386 "movq 24(%1, %%"REG_BP", 8), %%mm5\n\t" | |
| 2387 "movd 4(%3, %%"REG_a"), %%mm4 \n\t" | |
| 2388 "movd 4(%3, %%"REG_b"), %%mm2 \n\t" | |
| 2389 "punpcklbw %%mm7, %%mm4 \n\t" | |
| 2390 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2391 "pmaddwd %%mm1, %%mm4 \n\t" | |
| 2392 "pmaddwd %%mm2, %%mm5 \n\t" | |
| 2393 "paddd %%mm4, %%mm0 \n\t" | |
| 2394 "paddd %%mm5, %%mm3 \n\t" | |
| 2395 | |
| 2396 "psrad $8, %%mm0 \n\t" | |
| 2397 "psrad $8, %%mm3 \n\t" | |
| 2398 "packssdw %%mm3, %%mm0 \n\t" | |
| 2399 "pmaddwd %%mm6, %%mm0 \n\t" | |
| 2400 "packssdw %%mm0, %%mm0 \n\t" | |
| 2401 "movd %%mm0, (%4, %%"REG_BP") \n\t" | |
| 2402 "add $4, %%"REG_BP" \n\t" | |
| 2403 " jnc 1b \n\t" | |
| 2404 | |
| 2405 "pop %%"REG_BP" \n\t" | |
| 19396 | 2406 #if defined(PIC) |
| 2407 "pop %%"REG_b" \n\t" | |
| 2408 #endif | |
| 18861 | 2409 : "+a" (counter) |
| 2410 : "c" (filter), "d" (filterPos), "S" (src), "D" (dst) | |
| 19396 | 2411 #if !defined(PIC) |
| 18861 | 2412 : "%"REG_b |
| 19396 | 2413 #endif |
| 18861 | 2414 ); |
| 2415 } | |
| 2416 else | |
| 2417 { | |
| 2418 uint8_t *offset = src+filterSize; | |
| 2419 long counter= -2*dstW; | |
| 2420 // filter-= counter*filterSize/2; | |
| 2421 filterPos-= counter/2; | |
| 2422 dst-= counter/2; | |
| 2423 asm volatile( | |
| 2424 "pxor %%mm7, %%mm7 \n\t" | |
| 2425 "movq "MANGLE(w02)", %%mm6 \n\t" | |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
2426 ASMALIGN(4) |
| 18861 | 2427 "1: \n\t" |
| 2428 "mov %2, %%"REG_c" \n\t" | |
| 2429 "movzwl (%%"REG_c", %0), %%eax \n\t" | |
| 19396 | 2430 "movzwl 2(%%"REG_c", %0), %%edx \n\t" |
| 18861 | 2431 "mov %5, %%"REG_c" \n\t" |
| 2432 "pxor %%mm4, %%mm4 \n\t" | |
| 2433 "pxor %%mm5, %%mm5 \n\t" | |
| 2434 "2: \n\t" | |
| 2435 "movq (%1), %%mm1 \n\t" | |
| 2436 "movq (%1, %6), %%mm3 \n\t" | |
| 2437 "movd (%%"REG_c", %%"REG_a"), %%mm0\n\t" | |
| 19396 | 2438 "movd (%%"REG_c", %%"REG_d"), %%mm2\n\t" |
| 18861 | 2439 "punpcklbw %%mm7, %%mm0 \n\t" |
| 2440 "punpcklbw %%mm7, %%mm2 \n\t" | |
| 2441 "pmaddwd %%mm1, %%mm0 \n\t" | |
| 2442 "pmaddwd %%mm2, %%mm3 \n\t" | |
| 2443 "paddd %%mm3, %%mm5 \n\t" | |
| 2444 "paddd %%mm0, %%mm4 \n\t" | |
| 2445 "add $8, %1 \n\t" | |
| 2446 "add $4, %%"REG_c" \n\t" | |
| 2447 "cmp %4, %%"REG_c" \n\t" | |
| 2448 " jb 2b \n\t" | |
| 2449 "add %6, %1 \n\t" | |
| 2450 "psrad $8, %%mm4 \n\t" | |
| 2451 "psrad $8, %%mm5 \n\t" | |
| 2452 "packssdw %%mm5, %%mm4 \n\t" | |
| 2453 "pmaddwd %%mm6, %%mm4 \n\t" | |
| 2454 "packssdw %%mm4, %%mm4 \n\t" | |
| 2455 "mov %3, %%"REG_a" \n\t" | |
| 2456 "movd %%mm4, (%%"REG_a", %0) \n\t" | |
| 2457 "add $4, %0 \n\t" | |
| 2458 " jnc 1b \n\t" | |
| 2459 | |
| 2460 : "+r" (counter), "+r" (filter) | |
| 2461 : "m" (filterPos), "m" (dst), "m"(offset), | |
| 2462 "m" (src), "r" (filterSize*2) | |
| 19396 | 2463 : "%"REG_a, "%"REG_c, "%"REG_d |
| 18861 | 2464 ); |
| 2465 } | |
| 2466 #else | |
| 2467 #ifdef HAVE_ALTIVEC | |
| 2468 hScale_altivec_real(dst, dstW, src, srcW, xInc, filter, filterPos, filterSize); | |
| 2469 #else | |
| 2470 int i; | |
| 2471 for(i=0; i<dstW; i++) | |
| 2472 { | |
| 2473 int j; | |
| 2474 int srcPos= filterPos[i]; | |
| 2475 int val=0; | |
| 2476 // printf("filterPos: %d\n", filterPos[i]); | |
| 2477 for(j=0; j<filterSize; j++) | |
| 2478 { | |
| 2479 // printf("filter: %d, src: %d\n", filter[i], src[srcPos + j]); | |
| 2480 val += ((int)src[srcPos + j])*filter[filterSize*i + j]; | |
| 2481 } | |
| 2482 // filter += hFilterSize; | |
| 22321 | 2483 dst[i] = av_clip(val>>7, 0, (1<<15)-1); // the cubic equation does overflow ... |
| 18861 | 2484 // dst[i] = val>>7; |
| 2485 } | |
| 2486 #endif | |
| 2487 #endif | |
| 2488 } | |
| 2489 // *** horizontal scale Y line to temp buffer | |
| 2490 static inline void RENAME(hyscale)(uint16_t *dst, long dstWidth, uint8_t *src, int srcW, int xInc, | |
| 2491 int flags, int canMMX2BeUsed, int16_t *hLumFilter, | |
| 2492 int16_t *hLumFilterPos, int hLumFilterSize, void *funnyYCode, | |
| 2493 int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter, | |
| 22218 | 2494 int32_t *mmx2FilterPos, uint8_t *pal) |
| 18861 | 2495 { |
| 20411 | 2496 if(srcFormat==PIX_FMT_YUYV422 || srcFormat==PIX_FMT_GRAY16BE) |
| 18861 | 2497 { |
| 2498 RENAME(yuy2ToY)(formatConvBuffer, src, srcW); | |
| 2499 src= formatConvBuffer; | |
| 2500 } | |
| 20411 | 2501 else if(srcFormat==PIX_FMT_UYVY422 || srcFormat==PIX_FMT_GRAY16LE) |
| 18861 | 2502 { |
| 2503 RENAME(uyvyToY)(formatConvBuffer, src, srcW); | |
| 2504 src= formatConvBuffer; | |
| 2505 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2506 else if(srcFormat==PIX_FMT_RGB32) |
| 18861 | 2507 { |
| 2508 RENAME(bgr32ToY)(formatConvBuffer, src, srcW); | |
| 2509 src= formatConvBuffer; | |
| 2510 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2511 else if(srcFormat==PIX_FMT_BGR24) |
| 18861 | 2512 { |
| 2513 RENAME(bgr24ToY)(formatConvBuffer, src, srcW); | |
| 2514 src= formatConvBuffer; | |
| 2515 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2516 else if(srcFormat==PIX_FMT_BGR565) |
| 18861 | 2517 { |
| 2518 RENAME(bgr16ToY)(formatConvBuffer, src, srcW); | |
| 2519 src= formatConvBuffer; | |
| 2520 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2521 else if(srcFormat==PIX_FMT_BGR555) |
| 18861 | 2522 { |
| 2523 RENAME(bgr15ToY)(formatConvBuffer, src, srcW); | |
| 2524 src= formatConvBuffer; | |
| 2525 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2526 else if(srcFormat==PIX_FMT_BGR32) |
| 18861 | 2527 { |
| 2528 RENAME(rgb32ToY)(formatConvBuffer, src, srcW); | |
| 2529 src= formatConvBuffer; | |
| 2530 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2531 else if(srcFormat==PIX_FMT_RGB24) |
| 18861 | 2532 { |
| 2533 RENAME(rgb24ToY)(formatConvBuffer, src, srcW); | |
| 2534 src= formatConvBuffer; | |
| 2535 } | |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2536 else if(srcFormat==PIX_FMT_RGB565) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2537 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2538 RENAME(rgb16ToY)(formatConvBuffer, src, srcW); |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2539 src= formatConvBuffer; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2540 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2541 else if(srcFormat==PIX_FMT_RGB555) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2542 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2543 RENAME(rgb15ToY)(formatConvBuffer, src, srcW); |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2544 src= formatConvBuffer; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2545 } |
| 22226 | 2546 else if(srcFormat==PIX_FMT_RGB8 || srcFormat==PIX_FMT_BGR8 || srcFormat==PIX_FMT_PAL8 || srcFormat==PIX_FMT_BGR4_BYTE || srcFormat==PIX_FMT_RGB4_BYTE) |
| 22218 | 2547 { |
| 2548 RENAME(palToY)(formatConvBuffer, src, srcW, pal); | |
| 2549 src= formatConvBuffer; | |
| 2550 } | |
| 18861 | 2551 |
| 2552 #ifdef HAVE_MMX | |
| 2553 // use the new MMX scaler if the mmx2 can't be used (its faster than the x86asm one) | |
| 2554 if(!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed)) | |
| 2555 #else | |
| 2556 if(!(flags&SWS_FAST_BILINEAR)) | |
| 2557 #endif | |
| 2558 { | |
| 2559 RENAME(hScale)(dst, dstWidth, src, srcW, xInc, hLumFilter, hLumFilterPos, hLumFilterSize); | |
| 2560 } | |
| 2561 else // Fast Bilinear upscale / crap downscale | |
| 2562 { | |
| 20576 | 2563 #if defined(ARCH_X86) |
| 18861 | 2564 #ifdef HAVE_MMX2 |
| 2565 int i; | |
| 19396 | 2566 #if defined(PIC) |
| 2567 uint64_t ebxsave __attribute__((aligned(8))); | |
| 2568 #endif | |
| 18861 | 2569 if(canMMX2BeUsed) |
| 2570 { | |
| 2571 asm volatile( | |
| 19396 | 2572 #if defined(PIC) |
| 2573 "mov %%"REG_b", %5 \n\t" | |
| 2574 #endif | |
| 18861 | 2575 "pxor %%mm7, %%mm7 \n\t" |
| 2576 "mov %0, %%"REG_c" \n\t" | |
| 2577 "mov %1, %%"REG_D" \n\t" | |
| 2578 "mov %2, %%"REG_d" \n\t" | |
| 2579 "mov %3, %%"REG_b" \n\t" | |
| 2580 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2581 PREFETCH" (%%"REG_c") \n\t" | |
| 2582 PREFETCH" 32(%%"REG_c") \n\t" | |
| 2583 PREFETCH" 64(%%"REG_c") \n\t" | |
| 2584 | |
| 2585 #ifdef ARCH_X86_64 | |
| 2586 | |
| 2587 #define FUNNY_Y_CODE \ | |
| 2588 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2589 "call *%4 \n\t"\ | |
| 2590 "movl (%%"REG_b", %%"REG_a"), %%esi\n\t"\ | |
| 2591 "add %%"REG_S", %%"REG_c" \n\t"\ | |
| 2592 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2593 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2594 | |
| 2595 #else | |
| 2596 | |
| 2597 #define FUNNY_Y_CODE \ | |
| 2598 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2599 "call *%4 \n\t"\ | |
| 2600 "addl (%%"REG_b", %%"REG_a"), %%"REG_c"\n\t"\ | |
| 2601 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2602 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2603 | |
| 2604 #endif | |
| 2605 | |
| 2606 FUNNY_Y_CODE | |
| 2607 FUNNY_Y_CODE | |
| 2608 FUNNY_Y_CODE | |
| 2609 FUNNY_Y_CODE | |
| 2610 FUNNY_Y_CODE | |
| 2611 FUNNY_Y_CODE | |
| 2612 FUNNY_Y_CODE | |
| 2613 FUNNY_Y_CODE | |
| 2614 | |
| 19396 | 2615 #if defined(PIC) |
| 2616 "mov %5, %%"REG_b" \n\t" | |
| 2617 #endif | |
| 18861 | 2618 :: "m" (src), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos), |
| 2619 "m" (funnyYCode) | |
| 19396 | 2620 #if defined(PIC) |
| 2621 ,"m" (ebxsave) | |
| 2622 #endif | |
| 2623 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D | |
| 2624 #if !defined(PIC) | |
| 2625 ,"%"REG_b | |
| 2626 #endif | |
| 18861 | 2627 ); |
| 2628 for(i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) dst[i] = src[srcW-1]*128; | |
| 2629 } | |
| 2630 else | |
| 2631 { | |
| 2632 #endif | |
| 2633 long xInc_shr16 = xInc >> 16; | |
| 2634 uint16_t xInc_mask = xInc & 0xffff; | |
| 2635 //NO MMX just normal asm ... | |
| 2636 asm volatile( | |
| 2637 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 19396 | 2638 "xor %%"REG_d", %%"REG_d" \n\t" // xx |
| 18861 | 2639 "xorl %%ecx, %%ecx \n\t" // 2*xalpha |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
2640 ASMALIGN(4) |
| 18861 | 2641 "1: \n\t" |
| 19396 | 2642 "movzbl (%0, %%"REG_d"), %%edi \n\t" //src[xx] |
| 2643 "movzbl 1(%0, %%"REG_d"), %%esi \n\t" //src[xx+1] | |
| 18861 | 2644 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] |
| 2645 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2646 "shll $16, %%edi \n\t" | |
| 2647 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2648 "mov %1, %%"REG_D" \n\t" | |
| 2649 "shrl $9, %%esi \n\t" | |
| 2650 "movw %%si, (%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2651 "addw %4, %%cx \n\t" //2*xalpha += xInc&0xFF | |
| 19396 | 2652 "adc %3, %%"REG_d" \n\t" //xx+= xInc>>8 + carry |
| 18861 | 2653 |
| 19396 | 2654 "movzbl (%0, %%"REG_d"), %%edi \n\t" //src[xx] |
| 2655 "movzbl 1(%0, %%"REG_d"), %%esi \n\t" //src[xx+1] | |
| 18861 | 2656 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] |
| 2657 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2658 "shll $16, %%edi \n\t" | |
| 2659 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2660 "mov %1, %%"REG_D" \n\t" | |
| 2661 "shrl $9, %%esi \n\t" | |
| 2662 "movw %%si, 2(%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2663 "addw %4, %%cx \n\t" //2*xalpha += xInc&0xFF | |
| 19396 | 2664 "adc %3, %%"REG_d" \n\t" //xx+= xInc>>8 + carry |
| 18861 | 2665 |
| 2666 | |
| 2667 "add $2, %%"REG_a" \n\t" | |
| 2668 "cmp %2, %%"REG_a" \n\t" | |
| 2669 " jb 1b \n\t" | |
| 2670 | |
| 2671 | |
| 2672 :: "r" (src), "m" (dst), "m" (dstWidth), "m" (xInc_shr16), "m" (xInc_mask) | |
| 19396 | 2673 : "%"REG_a, "%"REG_d, "%ecx", "%"REG_D, "%esi" |
| 18861 | 2674 ); |
| 2675 #ifdef HAVE_MMX2 | |
| 2676 } //if MMX2 can't be used | |
| 2677 #endif | |
| 2678 #else | |
| 2679 int i; | |
| 2680 unsigned int xpos=0; | |
| 2681 for(i=0;i<dstWidth;i++) | |
| 2682 { | |
| 2683 register unsigned int xx=xpos>>16; | |
| 2684 register unsigned int xalpha=(xpos&0xFFFF)>>9; | |
| 2685 dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha; | |
| 2686 xpos+=xInc; | |
| 2687 } | |
| 2688 #endif | |
| 2689 } | |
| 2690 } | |
| 2691 | |
| 2692 inline static void RENAME(hcscale)(uint16_t *dst, long dstWidth, uint8_t *src1, uint8_t *src2, | |
| 2693 int srcW, int xInc, int flags, int canMMX2BeUsed, int16_t *hChrFilter, | |
| 2694 int16_t *hChrFilterPos, int hChrFilterSize, void *funnyUVCode, | |
| 2695 int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter, | |
| 22218 | 2696 int32_t *mmx2FilterPos, uint8_t *pal) |
| 18861 | 2697 { |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2698 if(srcFormat==PIX_FMT_YUYV422) |
| 18861 | 2699 { |
| 2700 RENAME(yuy2ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2701 src1= formatConvBuffer; | |
| 2702 src2= formatConvBuffer+2048; | |
| 2703 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2704 else if(srcFormat==PIX_FMT_UYVY422) |
| 18861 | 2705 { |
| 2706 RENAME(uyvyToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2707 src1= formatConvBuffer; | |
| 2708 src2= formatConvBuffer+2048; | |
| 2709 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2710 else if(srcFormat==PIX_FMT_RGB32) |
| 18861 | 2711 { |
| 2712 RENAME(bgr32ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2713 src1= formatConvBuffer; | |
| 2714 src2= formatConvBuffer+2048; | |
| 2715 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2716 else if(srcFormat==PIX_FMT_BGR24) |
| 18861 | 2717 { |
| 2718 RENAME(bgr24ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2719 src1= formatConvBuffer; | |
| 2720 src2= formatConvBuffer+2048; | |
| 2721 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2722 else if(srcFormat==PIX_FMT_BGR565) |
| 18861 | 2723 { |
| 2724 RENAME(bgr16ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2725 src1= formatConvBuffer; | |
| 2726 src2= formatConvBuffer+2048; | |
| 2727 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2728 else if(srcFormat==PIX_FMT_BGR555) |
| 18861 | 2729 { |
| 2730 RENAME(bgr15ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2731 src1= formatConvBuffer; | |
| 2732 src2= formatConvBuffer+2048; | |
| 2733 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2734 else if(srcFormat==PIX_FMT_BGR32) |
| 18861 | 2735 { |
| 2736 RENAME(rgb32ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2737 src1= formatConvBuffer; | |
| 2738 src2= formatConvBuffer+2048; | |
| 2739 } | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2740 else if(srcFormat==PIX_FMT_RGB24) |
| 18861 | 2741 { |
| 2742 RENAME(rgb24ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); | |
| 2743 src1= formatConvBuffer; | |
| 2744 src2= formatConvBuffer+2048; | |
| 2745 } | |
|
20589
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2746 else if(srcFormat==PIX_FMT_RGB565) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2747 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2748 RENAME(rgb16ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2749 src1= formatConvBuffer; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2750 src2= formatConvBuffer+2048; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2751 } |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2752 else if(srcFormat==PIX_FMT_RGB555) |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2753 { |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2754 RENAME(rgb15ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW); |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2755 src1= formatConvBuffer; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2756 src2= formatConvBuffer+2048; |
|
95695bfce2f0
Add support for conversions from the rgb565 and rgb555 formats
lucabe
parents:
20576
diff
changeset
|
2757 } |
| 18861 | 2758 else if(isGray(srcFormat)) |
| 2759 { | |
| 2760 return; | |
| 2761 } | |
| 22226 | 2762 else if(srcFormat==PIX_FMT_RGB8 || srcFormat==PIX_FMT_BGR8 || srcFormat==PIX_FMT_PAL8 || srcFormat==PIX_FMT_BGR4_BYTE || srcFormat==PIX_FMT_RGB4_BYTE) |
| 22218 | 2763 { |
| 2764 RENAME(palToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW, pal); | |
| 2765 src1= formatConvBuffer; | |
| 2766 src2= formatConvBuffer+2048; | |
| 2767 } | |
| 18861 | 2768 |
| 2769 #ifdef HAVE_MMX | |
| 2770 // use the new MMX scaler if the mmx2 can't be used (its faster than the x86asm one) | |
| 2771 if(!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed)) | |
| 2772 #else | |
| 2773 if(!(flags&SWS_FAST_BILINEAR)) | |
| 2774 #endif | |
| 2775 { | |
| 2776 RENAME(hScale)(dst , dstWidth, src1, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize); | |
| 2777 RENAME(hScale)(dst+2048, dstWidth, src2, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize); | |
| 2778 } | |
| 2779 else // Fast Bilinear upscale / crap downscale | |
| 2780 { | |
| 20576 | 2781 #if defined(ARCH_X86) |
| 18861 | 2782 #ifdef HAVE_MMX2 |
| 2783 int i; | |
| 19396 | 2784 #if defined(PIC) |
| 2785 uint64_t ebxsave __attribute__((aligned(8))); | |
| 2786 #endif | |
| 18861 | 2787 if(canMMX2BeUsed) |
| 2788 { | |
| 2789 asm volatile( | |
| 19396 | 2790 #if defined(PIC) |
| 2791 "mov %%"REG_b", %6 \n\t" | |
| 2792 #endif | |
| 18861 | 2793 "pxor %%mm7, %%mm7 \n\t" |
| 2794 "mov %0, %%"REG_c" \n\t" | |
| 2795 "mov %1, %%"REG_D" \n\t" | |
| 2796 "mov %2, %%"REG_d" \n\t" | |
| 2797 "mov %3, %%"REG_b" \n\t" | |
| 2798 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2799 PREFETCH" (%%"REG_c") \n\t" | |
| 2800 PREFETCH" 32(%%"REG_c") \n\t" | |
| 2801 PREFETCH" 64(%%"REG_c") \n\t" | |
| 2802 | |
| 2803 #ifdef ARCH_X86_64 | |
| 2804 | |
| 2805 #define FUNNY_UV_CODE \ | |
| 2806 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2807 "call *%4 \n\t"\ | |
| 2808 "movl (%%"REG_b", %%"REG_a"), %%esi\n\t"\ | |
| 2809 "add %%"REG_S", %%"REG_c" \n\t"\ | |
| 2810 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2811 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2812 | |
| 2813 #else | |
| 2814 | |
| 2815 #define FUNNY_UV_CODE \ | |
| 2816 "movl (%%"REG_b"), %%esi \n\t"\ | |
| 2817 "call *%4 \n\t"\ | |
| 2818 "addl (%%"REG_b", %%"REG_a"), %%"REG_c"\n\t"\ | |
| 2819 "add %%"REG_a", %%"REG_D" \n\t"\ | |
| 2820 "xor %%"REG_a", %%"REG_a" \n\t"\ | |
| 2821 | |
| 2822 #endif | |
| 2823 | |
| 2824 FUNNY_UV_CODE | |
| 2825 FUNNY_UV_CODE | |
| 2826 FUNNY_UV_CODE | |
| 2827 FUNNY_UV_CODE | |
| 2828 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 2829 "mov %5, %%"REG_c" \n\t" // src | |
| 2830 "mov %1, %%"REG_D" \n\t" // buf1 | |
| 2831 "add $4096, %%"REG_D" \n\t" | |
| 2832 PREFETCH" (%%"REG_c") \n\t" | |
| 2833 PREFETCH" 32(%%"REG_c") \n\t" | |
| 2834 PREFETCH" 64(%%"REG_c") \n\t" | |
| 2835 | |
| 2836 FUNNY_UV_CODE | |
| 2837 FUNNY_UV_CODE | |
| 2838 FUNNY_UV_CODE | |
| 2839 FUNNY_UV_CODE | |
| 2840 | |
| 19396 | 2841 #if defined(PIC) |
| 2842 "mov %6, %%"REG_b" \n\t" | |
| 2843 #endif | |
| 18861 | 2844 :: "m" (src1), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos), |
| 2845 "m" (funnyUVCode), "m" (src2) | |
| 19396 | 2846 #if defined(PIC) |
| 2847 ,"m" (ebxsave) | |
| 2848 #endif | |
|
19400
0310c3310360
Fix compilation with -no-PIC and without -fomit-frame-pointer (used by
uau
parents:
19396
diff
changeset
|
2849 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D |
| 19396 | 2850 #if !defined(PIC) |
| 2851 ,"%"REG_b | |
| 2852 #endif | |
| 18861 | 2853 ); |
| 2854 for(i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) | |
| 2855 { | |
| 2856 // printf("%d %d %d\n", dstWidth, i, srcW); | |
| 2857 dst[i] = src1[srcW-1]*128; | |
| 2858 dst[i+2048] = src2[srcW-1]*128; | |
| 2859 } | |
| 2860 } | |
| 2861 else | |
| 2862 { | |
| 2863 #endif | |
| 2864 long xInc_shr16 = (long) (xInc >> 16); | |
| 2865 uint16_t xInc_mask = xInc & 0xffff; | |
| 2866 asm volatile( | |
| 2867 "xor %%"REG_a", %%"REG_a" \n\t" // i | |
| 19396 | 2868 "xor %%"REG_d", %%"REG_d" \n\t" // xx |
| 18861 | 2869 "xorl %%ecx, %%ecx \n\t" // 2*xalpha |
|
19372
6334c14b38eb
Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
diego
parents:
19181
diff
changeset
|
2870 ASMALIGN(4) |
| 18861 | 2871 "1: \n\t" |
| 2872 "mov %0, %%"REG_S" \n\t" | |
| 19396 | 2873 "movzbl (%%"REG_S", %%"REG_d"), %%edi \n\t" //src[xx] |
| 2874 "movzbl 1(%%"REG_S", %%"REG_d"), %%esi \n\t" //src[xx+1] | |
| 18861 | 2875 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] |
| 2876 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2877 "shll $16, %%edi \n\t" | |
| 2878 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2879 "mov %1, %%"REG_D" \n\t" | |
| 2880 "shrl $9, %%esi \n\t" | |
| 2881 "movw %%si, (%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2882 | |
| 19396 | 2883 "movzbl (%5, %%"REG_d"), %%edi \n\t" //src[xx] |
| 2884 "movzbl 1(%5, %%"REG_d"), %%esi \n\t" //src[xx+1] | |
| 18861 | 2885 "subl %%edi, %%esi \n\t" //src[xx+1] - src[xx] |
| 2886 "imull %%ecx, %%esi \n\t" //(src[xx+1] - src[xx])*2*xalpha | |
| 2887 "shll $16, %%edi \n\t" | |
| 2888 "addl %%edi, %%esi \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha) | |
| 2889 "mov %1, %%"REG_D" \n\t" | |
| 2890 "shrl $9, %%esi \n\t" | |
| 2891 "movw %%si, 4096(%%"REG_D", %%"REG_a", 2)\n\t" | |
| 2892 | |
| 2893 "addw %4, %%cx \n\t" //2*xalpha += xInc&0xFF | |
| 19396 | 2894 "adc %3, %%"REG_d" \n\t" //xx+= xInc>>8 + carry |
| 18861 | 2895 "add $1, %%"REG_a" \n\t" |
| 2896 "cmp %2, %%"REG_a" \n\t" | |
| 2897 " jb 1b \n\t" | |
| 2898 | |
| 2899 /* GCC-3.3 makes MPlayer crash on IA-32 machines when using "g" operand here, | |
| 2900 which is needed to support GCC-4.0 */ | |
| 2901 #if defined(ARCH_X86_64) && ((__GNUC__ > 3) || ( __GNUC__ == 3 && __GNUC_MINOR__ >= 4)) | |
| 2902 :: "m" (src1), "m" (dst), "g" ((long)dstWidth), "m" (xInc_shr16), "m" (xInc_mask), | |
| 2903 #else | |
| 2904 :: "m" (src1), "m" (dst), "m" ((long)dstWidth), "m" (xInc_shr16), "m" (xInc_mask), | |
| 2905 #endif | |
| 2906 "r" (src2) | |
| 19396 | 2907 : "%"REG_a, "%"REG_d, "%ecx", "%"REG_D, "%esi" |
| 18861 | 2908 ); |
| 2909 #ifdef HAVE_MMX2 | |
| 2910 } //if MMX2 can't be used | |
| 2911 #endif | |
| 2912 #else | |
| 2913 int i; | |
| 2914 unsigned int xpos=0; | |
| 2915 for(i=0;i<dstWidth;i++) | |
| 2916 { | |
| 2917 register unsigned int xx=xpos>>16; | |
| 2918 register unsigned int xalpha=(xpos&0xFFFF)>>9; | |
| 2919 dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha); | |
| 2920 dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha); | |
| 2921 /* slower | |
| 2922 dst[i]= (src1[xx]<<7) + (src1[xx+1] - src1[xx])*xalpha; | |
| 2923 dst[i+2048]=(src2[xx]<<7) + (src2[xx+1] - src2[xx])*xalpha; | |
| 2924 */ | |
| 2925 xpos+=xInc; | |
| 2926 } | |
| 2927 #endif | |
| 2928 } | |
| 2929 } | |
| 2930 | |
| 2931 static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
| 2932 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
| 2933 | |
| 2934 /* load a few things into local vars to make the code more readable? and faster */ | |
| 2935 const int srcW= c->srcW; | |
| 2936 const int dstW= c->dstW; | |
| 2937 const int dstH= c->dstH; | |
| 2938 const int chrDstW= c->chrDstW; | |
| 2939 const int chrSrcW= c->chrSrcW; | |
| 2940 const int lumXInc= c->lumXInc; | |
| 2941 const int chrXInc= c->chrXInc; | |
| 2942 const int dstFormat= c->dstFormat; | |
| 2943 const int srcFormat= c->srcFormat; | |
| 2944 const int flags= c->flags; | |
| 2945 const int canMMX2BeUsed= c->canMMX2BeUsed; | |
| 2946 int16_t *vLumFilterPos= c->vLumFilterPos; | |
| 2947 int16_t *vChrFilterPos= c->vChrFilterPos; | |
| 2948 int16_t *hLumFilterPos= c->hLumFilterPos; | |
| 2949 int16_t *hChrFilterPos= c->hChrFilterPos; | |
| 2950 int16_t *vLumFilter= c->vLumFilter; | |
| 2951 int16_t *vChrFilter= c->vChrFilter; | |
| 2952 int16_t *hLumFilter= c->hLumFilter; | |
| 2953 int16_t *hChrFilter= c->hChrFilter; | |
| 2954 int32_t *lumMmxFilter= c->lumMmxFilter; | |
| 2955 int32_t *chrMmxFilter= c->chrMmxFilter; | |
| 2956 const int vLumFilterSize= c->vLumFilterSize; | |
| 2957 const int vChrFilterSize= c->vChrFilterSize; | |
| 2958 const int hLumFilterSize= c->hLumFilterSize; | |
| 2959 const int hChrFilterSize= c->hChrFilterSize; | |
| 2960 int16_t **lumPixBuf= c->lumPixBuf; | |
| 2961 int16_t **chrPixBuf= c->chrPixBuf; | |
| 2962 const int vLumBufSize= c->vLumBufSize; | |
| 2963 const int vChrBufSize= c->vChrBufSize; | |
| 2964 uint8_t *funnyYCode= c->funnyYCode; | |
| 2965 uint8_t *funnyUVCode= c->funnyUVCode; | |
| 2966 uint8_t *formatConvBuffer= c->formatConvBuffer; | |
| 2967 const int chrSrcSliceY= srcSliceY >> c->chrSrcVSubSample; | |
| 2968 const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample); | |
| 2969 int lastDstY; | |
| 22218 | 2970 uint8_t *pal=NULL; |
| 18861 | 2971 |
| 2972 /* vars whch will change and which we need to storw back in the context */ | |
| 2973 int dstY= c->dstY; | |
| 2974 int lumBufIndex= c->lumBufIndex; | |
| 2975 int chrBufIndex= c->chrBufIndex; | |
| 2976 int lastInLumBuf= c->lastInLumBuf; | |
| 2977 int lastInChrBuf= c->lastInChrBuf; | |
| 2978 | |
| 2979 if(isPacked(c->srcFormat)){ | |
| 22218 | 2980 pal= src[1]; |
| 18861 | 2981 src[0]= |
| 2982 src[1]= | |
| 2983 src[2]= src[0]; | |
| 2984 srcStride[0]= | |
| 2985 srcStride[1]= | |
| 2986 srcStride[2]= srcStride[0]; | |
| 2987 } | |
| 2988 srcStride[1]<<= c->vChrDrop; | |
| 2989 srcStride[2]<<= c->vChrDrop; | |
| 2990 | |
| 2991 // printf("swscale %X %X %X -> %X %X %X\n", (int)src[0], (int)src[1], (int)src[2], | |
| 2992 // (int)dst[0], (int)dst[1], (int)dst[2]); | |
| 2993 | |
| 2994 #if 0 //self test FIXME move to a vfilter or something | |
| 2995 { | |
| 2996 static volatile int i=0; | |
| 2997 i++; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
2998 if(srcFormat==PIX_FMT_YUV420P && i==1 && srcSliceH>= c->srcH) |
| 18861 | 2999 selfTest(src, srcStride, c->srcW, c->srcH); |
| 3000 i--; | |
| 3001 } | |
| 3002 #endif | |
| 3003 | |
| 3004 //printf("sws Strides:%d %d %d -> %d %d %d\n", srcStride[0],srcStride[1],srcStride[2], | |
| 3005 //dstStride[0],dstStride[1],dstStride[2]); | |
| 3006 | |
| 3007 if(dstStride[0]%8 !=0 || dstStride[1]%8 !=0 || dstStride[2]%8 !=0) | |
| 3008 { | |
| 3009 static int firstTime=1; //FIXME move this into the context perhaps | |
| 3010 if(flags & SWS_PRINT_INFO && firstTime) | |
| 3011 { | |
| 21981 | 3012 av_log(c, AV_LOG_WARNING, "SwScaler: Warning: dstStride is not aligned!\n" |
| 18861 | 3013 "SwScaler: ->cannot do aligned memory acesses anymore\n"); |
| 3014 firstTime=0; | |
| 3015 } | |
| 3016 } | |
| 3017 | |
| 3018 /* Note the user might start scaling the picture in the middle so this will not get executed | |
| 3019 this is not really intended but works currently, so ppl might do it */ | |
| 3020 if(srcSliceY ==0){ | |
| 3021 lumBufIndex=0; | |
| 3022 chrBufIndex=0; | |
| 3023 dstY=0; | |
| 3024 lastInLumBuf= -1; | |
| 3025 lastInChrBuf= -1; | |
| 3026 } | |
| 3027 | |
| 3028 lastDstY= dstY; | |
| 3029 | |
| 3030 for(;dstY < dstH; dstY++){ | |
| 3031 unsigned char *dest =dst[0]+dstStride[0]*dstY; | |
| 3032 const int chrDstY= dstY>>c->chrDstVSubSample; | |
| 3033 unsigned char *uDest=dst[1]+dstStride[1]*chrDstY; | |
| 3034 unsigned char *vDest=dst[2]+dstStride[2]*chrDstY; | |
| 3035 | |
| 3036 const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input | |
| 3037 const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input | |
| 3038 const int lastLumSrcY= firstLumSrcY + vLumFilterSize -1; // Last line needed as input | |
| 3039 const int lastChrSrcY= firstChrSrcY + vChrFilterSize -1; // Last line needed as input | |
| 3040 | |
| 3041 //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", | |
| 3042 // dstY, dstH, firstLumSrcY, lastInLumBuf, vLumBufSize, vChrBufSize, srcSliceY, srcSliceH, vLumFilterSize, firstChrSrcY, vChrFilterSize, c->chrSrcVSubSample); | |
| 3043 //handle holes (FAST_BILINEAR & weird filters) | |
| 3044 if(firstLumSrcY > lastInLumBuf) lastInLumBuf= firstLumSrcY-1; | |
| 3045 if(firstChrSrcY > lastInChrBuf) lastInChrBuf= firstChrSrcY-1; | |
| 3046 //printf("%d %d %d\n", firstChrSrcY, lastInChrBuf, vChrBufSize); | |
| 3047 ASSERT(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1) | |
| 3048 ASSERT(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1) | |
| 3049 | |
| 3050 // Do we have enough lines in this slice to output the dstY line | |
| 3051 if(lastLumSrcY < srcSliceY + srcSliceH && lastChrSrcY < -((-srcSliceY - srcSliceH)>>c->chrSrcVSubSample)) | |
| 3052 { | |
| 3053 //Do horizontal scaling | |
| 3054 while(lastInLumBuf < lastLumSrcY) | |
| 3055 { | |
| 3056 uint8_t *s= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0]; | |
| 3057 lumBufIndex++; | |
| 3058 // printf("%d %d %d %d\n", lumBufIndex, vLumBufSize, lastInLumBuf, lastLumSrcY); | |
| 3059 ASSERT(lumBufIndex < 2*vLumBufSize) | |
| 3060 ASSERT(lastInLumBuf + 1 - srcSliceY < srcSliceH) | |
| 3061 ASSERT(lastInLumBuf + 1 - srcSliceY >= 0) | |
| 3062 // printf("%d %d\n", lumBufIndex, vLumBufSize); | |
| 3063 RENAME(hyscale)(lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc, | |
| 3064 flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize, | |
| 3065 funnyYCode, c->srcFormat, formatConvBuffer, | |
| 22218 | 3066 c->lumMmx2Filter, c->lumMmx2FilterPos, pal); |
| 18861 | 3067 lastInLumBuf++; |
| 3068 } | |
| 3069 while(lastInChrBuf < lastChrSrcY) | |
| 3070 { | |
| 3071 uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1]; | |
| 3072 uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2]; | |
| 3073 chrBufIndex++; | |
| 3074 ASSERT(chrBufIndex < 2*vChrBufSize) | |
| 3075 ASSERT(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH)) | |
| 3076 ASSERT(lastInChrBuf + 1 - chrSrcSliceY >= 0) | |
| 3077 //FIXME replace parameters through context struct (some at least) | |
| 3078 | |
| 3079 if(!(isGray(srcFormat) || isGray(dstFormat))) | |
| 3080 RENAME(hcscale)(chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc, | |
| 3081 flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize, | |
| 3082 funnyUVCode, c->srcFormat, formatConvBuffer, | |
| 22218 | 3083 c->chrMmx2Filter, c->chrMmx2FilterPos, pal); |
| 18861 | 3084 lastInChrBuf++; |
| 3085 } | |
| 3086 //wrap buf index around to stay inside the ring buffer | |
| 3087 if(lumBufIndex >= vLumBufSize ) lumBufIndex-= vLumBufSize; | |
| 3088 if(chrBufIndex >= vChrBufSize ) chrBufIndex-= vChrBufSize; | |
| 3089 } | |
| 3090 else // not enough lines left in this slice -> load the rest in the buffer | |
| 3091 { | |
| 3092 /* printf("%d %d Last:%d %d LastInBuf:%d %d Index:%d %d Y:%d FSize: %d %d BSize: %d %d\n", | |
| 3093 firstChrSrcY,firstLumSrcY,lastChrSrcY,lastLumSrcY, | |
| 3094 lastInChrBuf,lastInLumBuf,chrBufIndex,lumBufIndex,dstY,vChrFilterSize,vLumFilterSize, | |
| 3095 vChrBufSize, vLumBufSize);*/ | |
| 3096 | |
| 3097 //Do horizontal scaling | |
| 3098 while(lastInLumBuf+1 < srcSliceY + srcSliceH) | |
| 3099 { | |
| 3100 uint8_t *s= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0]; | |
| 3101 lumBufIndex++; | |
| 3102 ASSERT(lumBufIndex < 2*vLumBufSize) | |
| 3103 ASSERT(lastInLumBuf + 1 - srcSliceY < srcSliceH) | |
| 3104 ASSERT(lastInLumBuf + 1 - srcSliceY >= 0) | |
| 3105 RENAME(hyscale)(lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc, | |
| 3106 flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize, | |
| 3107 funnyYCode, c->srcFormat, formatConvBuffer, | |
| 22218 | 3108 c->lumMmx2Filter, c->lumMmx2FilterPos, pal); |
| 18861 | 3109 lastInLumBuf++; |
| 3110 } | |
| 3111 while(lastInChrBuf+1 < (chrSrcSliceY + chrSrcSliceH)) | |
| 3112 { | |
| 3113 uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1]; | |
| 3114 uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2]; | |
| 3115 chrBufIndex++; | |
| 3116 ASSERT(chrBufIndex < 2*vChrBufSize) | |
| 3117 ASSERT(lastInChrBuf + 1 - chrSrcSliceY < chrSrcSliceH) | |
| 3118 ASSERT(lastInChrBuf + 1 - chrSrcSliceY >= 0) | |
| 3119 | |
| 3120 if(!(isGray(srcFormat) || isGray(dstFormat))) | |
| 3121 RENAME(hcscale)(chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc, | |
| 3122 flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize, | |
| 3123 funnyUVCode, c->srcFormat, formatConvBuffer, | |
| 22218 | 3124 c->chrMmx2Filter, c->chrMmx2FilterPos, pal); |
| 18861 | 3125 lastInChrBuf++; |
| 3126 } | |
| 3127 //wrap buf index around to stay inside the ring buffer | |
| 3128 if(lumBufIndex >= vLumBufSize ) lumBufIndex-= vLumBufSize; | |
| 3129 if(chrBufIndex >= vChrBufSize ) chrBufIndex-= vChrBufSize; | |
| 3130 break; //we can't output a dstY line so let's try with the next slice | |
| 3131 } | |
| 3132 | |
| 3133 #ifdef HAVE_MMX | |
| 3134 b5Dither= dither8[dstY&1]; | |
| 3135 g6Dither= dither4[dstY&1]; | |
| 3136 g5Dither= dither8[dstY&1]; | |
| 3137 r5Dither= dither8[(dstY+1)&1]; | |
| 3138 #endif | |
| 3139 if(dstY < dstH-2) | |
| 3140 { | |
| 3141 int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize; | |
| 3142 int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; | |
| 3143 #ifdef HAVE_MMX | |
| 3144 int i; | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3145 if(flags & SWS_ACCURATE_RND){ |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3146 for(i=0; i<vLumFilterSize; i+=2){ |
| 21756 | 3147 lumMmxFilter[2*i+0]= (int32_t)lumSrcPtr[i ]; |
| 3148 lumMmxFilter[2*i+1]= (int32_t)lumSrcPtr[i+(vLumFilterSize>1)]; | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3149 lumMmxFilter[2*i+2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3150 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
|
3151 + (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
|
3152 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3153 for(i=0; i<vChrFilterSize; i+=2){ |
| 21756 | 3154 chrMmxFilter[2*i+0]= (int32_t)chrSrcPtr[i ]; |
| 3155 chrMmxFilter[2*i+1]= (int32_t)chrSrcPtr[i+(vChrFilterSize>1)]; | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3156 chrMmxFilter[2*i+2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3157 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
|
3158 + (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
|
3159 } |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3160 }else{ |
| 18861 | 3161 for(i=0; i<vLumFilterSize; i++) |
| 3162 { | |
| 3163 lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i]; | |
|
22383
508e55817748
Fix a possible crash on 64 bit systems when the lumSrcPtr or chrSrcPtr
reimar
parents:
22321
diff
changeset
|
3164 lumMmxFilter[4*i+1]= (uint64_t)lumSrcPtr[i] >> 32; |
| 18861 | 3165 lumMmxFilter[4*i+2]= |
| 3166 lumMmxFilter[4*i+3]= | |
| 3167 ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001; | |
| 3168 } | |
| 3169 for(i=0; i<vChrFilterSize; i++) | |
| 3170 { | |
| 3171 chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i]; | |
|
22383
508e55817748
Fix a possible crash on 64 bit systems when the lumSrcPtr or chrSrcPtr
reimar
parents:
22321
diff
changeset
|
3172 chrMmxFilter[4*i+1]= (uint64_t)chrSrcPtr[i] >> 32; |
| 18861 | 3173 chrMmxFilter[4*i+2]= |
| 3174 chrMmxFilter[4*i+3]= | |
| 3175 ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001; | |
| 3176 } | |
|
19172
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3177 } |
| 18861 | 3178 #endif |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
3179 if(dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21){ |
| 18861 | 3180 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; |
| 3181 if(dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi | |
| 3182 RENAME(yuv2nv12X)(c, | |
| 3183 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3184 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3185 dest, uDest, dstW, chrDstW, dstFormat); | |
| 3186 } | |
| 3187 else if(isPlanarYUV(dstFormat) || isGray(dstFormat)) //YV12 like | |
| 3188 { | |
| 3189 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; | |
| 3190 if((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi | |
| 3191 if(vLumFilterSize == 1 && vChrFilterSize == 1) // Unscaled YV12 | |
| 3192 { | |
| 3193 int16_t *lumBuf = lumPixBuf[0]; | |
| 3194 int16_t *chrBuf= chrPixBuf[0]; | |
| 3195 RENAME(yuv2yuv1)(lumBuf, chrBuf, dest, uDest, vDest, dstW, chrDstW); | |
| 3196 } | |
| 3197 else //General YV12 | |
| 3198 { | |
| 3199 RENAME(yuv2yuvX)(c, | |
| 3200 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3201 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3202 dest, uDest, vDest, dstW, chrDstW); | |
| 3203 } | |
| 3204 } | |
| 3205 else | |
| 3206 { | |
| 3207 ASSERT(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2); | |
| 3208 ASSERT(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2); | |
| 3209 if(vLumFilterSize == 1 && vChrFilterSize == 2) //Unscaled RGB | |
| 3210 { | |
| 3211 int chrAlpha= vChrFilter[2*dstY+1]; | |
| 3212 RENAME(yuv2packed1)(c, *lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1), | |
| 3213 dest, dstW, chrAlpha, dstFormat, flags, dstY); | |
| 3214 } | |
| 3215 else if(vLumFilterSize == 2 && vChrFilterSize == 2) //BiLinear Upscale RGB | |
| 3216 { | |
| 3217 int lumAlpha= vLumFilter[2*dstY+1]; | |
| 3218 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
|
3219 lumMmxFilter[2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3220 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
|
3221 chrMmxFilter[2]= |
|
bae6c99a99cc
vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
michael
parents:
18861
diff
changeset
|
3222 chrMmxFilter[3]= vChrFilter[2*chrDstY]*0x10001; |
| 18861 | 3223 RENAME(yuv2packed2)(c, *lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1), |
| 3224 dest, dstW, lumAlpha, chrAlpha, dstY); | |
| 3225 } | |
| 3226 else //General RGB | |
| 3227 { | |
| 3228 RENAME(yuv2packedX)(c, | |
| 3229 vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, | |
| 3230 vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3231 dest, dstW, dstY); | |
| 3232 } | |
| 3233 } | |
| 3234 } | |
| 3235 else // hmm looks like we can't use MMX here without overwriting this array's tail | |
| 3236 { | |
| 3237 int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize; | |
| 3238 int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; | |
|
19872
8e50cba9fe03
Remove the dependency of libswscale on img_format.h
lucabe
parents:
19594
diff
changeset
|
3239 if(dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21){ |
| 18861 | 3240 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; |
| 3241 if(dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi | |
| 3242 yuv2nv12XinC( | |
| 3243 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3244 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3245 dest, uDest, dstW, chrDstW, dstFormat); | |
| 3246 } | |
| 3247 else if(isPlanarYUV(dstFormat) || isGray(dstFormat)) //YV12 | |
| 3248 { | |
| 3249 const int chrSkipMask= (1<<c->chrDstVSubSample)-1; | |
| 3250 if((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi | |
| 3251 yuv2yuvXinC( | |
| 3252 vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize, | |
| 3253 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3254 dest, uDest, vDest, dstW, chrDstW); | |
| 3255 } | |
| 3256 else | |
| 3257 { | |
| 3258 ASSERT(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2); | |
| 3259 ASSERT(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2); | |
| 3260 yuv2packedXinC(c, | |
| 3261 vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, | |
| 3262 vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, | |
| 3263 dest, dstW, dstY); | |
| 3264 } | |
| 3265 } | |
| 3266 } | |
| 3267 | |
| 3268 #ifdef HAVE_MMX | |
| 3269 __asm __volatile(SFENCE:::"memory"); | |
| 3270 __asm __volatile(EMMS:::"memory"); | |
| 3271 #endif | |
| 3272 /* store changed local vars back in the context */ | |
| 3273 c->dstY= dstY; | |
| 3274 c->lumBufIndex= lumBufIndex; | |
| 3275 c->chrBufIndex= chrBufIndex; | |
| 3276 c->lastInLumBuf= lastInLumBuf; | |
| 3277 c->lastInChrBuf= lastInChrBuf; | |
| 3278 | |
| 3279 return dstY - lastDstY; | |
| 3280 } |