annotate dsputil.h @ 3350:eff63ac2b545 libavcodec

Make gcc-3 happy again, thanks to Olivier Castan <castan.o@free.fr> for pointing the issue
author lu_zero
date Tue, 20 Jun 2006 10:35:30 +0000
parents 647a677c00a4
children adccbf4a1040
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
429
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
1 /*
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
2 * DSP utils
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
1739
07a484280a82 copyright year update of the files i touched and remembered, things look annoyingly unmaintained otherwise
michael
parents: 1729
diff changeset
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
429
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
5 *
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
6 * This library is free software; you can redistribute it and/or
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
7 * modify it under the terms of the GNU Lesser General Public
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
8 * License as published by the Free Software Foundation; either
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
9 * version 2 of the License, or (at your option) any later version.
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
10 *
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
11 * This library is distributed in the hope that it will be useful,
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
14 * Lesser General Public License for more details.
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
15 *
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
16 * You should have received a copy of the GNU Lesser General Public
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
17 * License along with this library; if not, write to the Free Software
3036
0b546eab515d Update licensing information: The FSF changed postal address.
diego
parents: 3029
diff changeset
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
429
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
19 */
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
20
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
21 /**
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
22 * @file dsputil.h
1106
1e39f273ecd6 per file doxy
michaelni
parents: 1104
diff changeset
23 * DSP utils.
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
24 * note, many functions in here may use MMX which trashes the FPU state, it is
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
25 * absolutely necessary to call emms_c() between dsp & float/double code
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
26 */
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
27
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
28 #ifndef DSPUTIL_H
986e461dc072 Initial revision
glantau
parents:
diff changeset
29 #define DSPUTIL_H
986e461dc072 Initial revision
glantau
parents:
diff changeset
30
986e461dc072 Initial revision
glantau
parents:
diff changeset
31 #include "common.h"
252
ddb1a0e94cf4 - Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents: 214
diff changeset
32 #include "avcodec.h"
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
33
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
34
255
db20b987c32d divx5-gmc support
michaelni
parents: 252
diff changeset
35 //#define DEBUG
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
36 /* dct code */
986e461dc072 Initial revision
glantau
parents:
diff changeset
37 typedef short DCTELEM;
3198
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
38 typedef int DWTELEM;
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
39
474
11dbd00682fc avoid name clash with libjpeg - added missing externs
bellard
parents: 429
diff changeset
40 void fdct_ifast (DCTELEM *data);
1571
aa4dc16c0f18 * adding integer/floating point AAN implementations for DCT 2-4-8
romansh
parents: 1567
diff changeset
41 void fdct_ifast248 (DCTELEM *data);
625
bb6a69f9d409 slow but accurate integer dct from IJG (should be ok with the LGPL as the old DCT is the fast integer DCT from IJG)
michaelni
parents: 623
diff changeset
42 void ff_jpeg_fdct_islow (DCTELEM *data);
1567
e08df4d22d27 * introducing dct248 into the DSP context.
romansh
parents: 1565
diff changeset
43 void ff_fdct248_islow (DCTELEM *data);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
44
986e461dc072 Initial revision
glantau
parents:
diff changeset
45 void j_rev_dct (DCTELEM *data);
2256
7e0b2e86afa9 1/2 resolution decoding
michael
parents: 2185
diff changeset
46 void j_rev_dct4 (DCTELEM *data);
2257
5f64a30339e5 1/4 resolution decoding
michael
parents: 2256
diff changeset
47 void j_rev_dct2 (DCTELEM *data);
2259
12e75af1d44c 1/8 resolution decoding
michael
parents: 2257
diff changeset
48 void j_rev_dct1 (DCTELEM *data);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
49
687
9abb13c21fbe fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
arpi_esp
parents: 675
diff changeset
50 void ff_fdct_mmx(DCTELEM *block);
1565
1a9a63f59849 minor mmx2 optimization if the dct
michael
parents: 1527
diff changeset
51 void ff_fdct_mmx2(DCTELEM *block);
1765
e31754bc5b65 SSE2 fdct by (Balatoni Denes <pnis at coder dot hu>)
michael
parents: 1739
diff changeset
52 void ff_fdct_sse2(DCTELEM *block);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
53
2755
975074f04b95 decode H.264 with 8x8 transform.
lorenm
parents: 2725
diff changeset
54 void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride);
2272
cd43603c46f9 move h264 idct to its own file and call via function pointer in DspContext
michael
parents: 2259
diff changeset
55 void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride);
3105
2d35fb3cb940 h264: special case dc-only idct. ~1% faster overall
lorenm
parents: 3089
diff changeset
56 void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
2d35fb3cb940 h264: special case dc-only idct. ~1% faster overall
lorenm
parents: 3089
diff changeset
57 void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
2272
cd43603c46f9 move h264 idct to its own file and call via function pointer in DspContext
michael
parents: 2259
diff changeset
58 void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block);
cd43603c46f9 move h264 idct to its own file and call via function pointer in DspContext
michael
parents: 2259
diff changeset
59 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block);
cd43603c46f9 move h264 idct to its own file and call via function pointer in DspContext
michael
parents: 2259
diff changeset
60
34
2733a4c1c693 added block permutation functions
glantau
parents: 19
diff changeset
61 /* encoding scans */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
62 extern const uint8_t ff_alternate_horizontal_scan[64];
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
63 extern const uint8_t ff_alternate_vertical_scan[64];
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
64 extern const uint8_t ff_zigzag_direct[64];
1567
e08df4d22d27 * introducing dct248 into the DSP context.
romansh
parents: 1565
diff changeset
65 extern const uint8_t ff_zigzag248_direct[64];
190
9e0e56869d05 fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents: 174
diff changeset
66
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
67 /* pixel operations */
2090
869805505b30 increase size of crop table
michael
parents: 2075
diff changeset
68 #define MAX_NEG_CROP 1024
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
69
986e461dc072 Initial revision
glantau
parents:
diff changeset
70 /* temporary */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
71 extern uint32_t squareTbl[512];
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
72 extern uint8_t cropTbl[256 + 2 * MAX_NEG_CROP];
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
73
1866
1755f959ab7f seperated out the C-based VP3 DSP functions into a different file; also
melanson
parents: 1864
diff changeset
74 /* VP3 DSP functions */
2693
02925a3903b6 porting vp3 idct over to lavc idct api
michael
parents: 2633
diff changeset
75 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
02925a3903b6 porting vp3 idct over to lavc idct api
michael
parents: 2633
diff changeset
76 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
02925a3903b6 porting vp3 idct over to lavc idct api
michael
parents: 2633
diff changeset
77 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
78
3245
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
79 /* 1/2^n downscaling functions from imgconvert.c */
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
80 void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
81 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
82 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
83 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
84
3248
7aa9f80e7954 mmx implementation of 3-point GMC. (5x faster than C)
lorenm
parents: 3245
diff changeset
85 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
7aa9f80e7954 mmx implementation of 3-point GMC. (5x faster than C)
lorenm
parents: 3245
diff changeset
86 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
7aa9f80e7954 mmx implementation of 3-point GMC. (5x faster than C)
lorenm
parents: 3245
diff changeset
87
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
88 /* minimum alignment rules ;)
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
89 if u notice errors in the align stuff, need more alignment for some asm code for some cpu
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
90 or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ...
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
91
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
92 !warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible)
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
93 i (michael) didnt check them, these are just the alignents which i think could be reached easily ...
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
94
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
95 !future video codecs might need functions with less strict alignment
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
96 */
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
97
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
98 /*
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
99 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
100 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
101 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
102 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
296
c1a8a1b4a24b sizeof(s->block) isnt 64*6*2 anymore bugfix
michaelni
parents: 294
diff changeset
103 void clear_blocks_c(DCTELEM *blocks);
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
104 */
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
105
986e461dc072 Initial revision
glantau
parents:
diff changeset
106 /* add and put pixel (decoding) */
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
107 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
1709
39a209b0c82c info about h and tpel
michael
parents: 1708
diff changeset
108 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
109 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
1267
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
110 typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
111 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
1168
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
112 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
2415
db2cf6005d19 H.264 weighted prediction.
lorenm
parents: 2388
diff changeset
113 typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
3029
7a92269a3ccd tweak h264_biweight
lorenm
parents: 3020
diff changeset
114 typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
115
984
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
116 #define DEF_OLD_QPEL(name)\
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
117 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
118 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
119 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
120
984
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
121 DEF_OLD_QPEL(qpel16_mc11_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
122 DEF_OLD_QPEL(qpel16_mc31_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
123 DEF_OLD_QPEL(qpel16_mc12_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
124 DEF_OLD_QPEL(qpel16_mc32_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
125 DEF_OLD_QPEL(qpel16_mc13_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
126 DEF_OLD_QPEL(qpel16_mc33_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
127 DEF_OLD_QPEL(qpel8_mc11_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
128 DEF_OLD_QPEL(qpel8_mc31_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
129 DEF_OLD_QPEL(qpel8_mc12_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
130 DEF_OLD_QPEL(qpel8_mc32_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
131 DEF_OLD_QPEL(qpel8_mc13_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
132 DEF_OLD_QPEL(qpel8_mc33_old_c)
651
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
133
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
134 #define CALL_2X_PIXELS(a, b, n)\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
135 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
136 b(block , pixels , line_size, h);\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
137 b(block+n, pixels+n, line_size, h);\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
138 }
255
db20b987c32d divx5-gmc support
michaelni
parents: 252
diff changeset
139
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
140 /* motion estimation */
1709
39a209b0c82c info about h and tpel
michael
parents: 1708
diff changeset
141 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
39a209b0c82c info about h and tpel
michael
parents: 1708
diff changeset
142 // allthough currently h<4 is not used as functions with width <8 are not used and neither implemented
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
143 typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
144
1168
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
145
3198
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
146 // for snow slices
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
147 typedef struct slice_buffer_s slice_buffer;
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
148
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
149 /**
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
150 * DSPContext.
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
151 */
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
152 typedef struct DSPContext {
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
153 /* pixel ops : interface with DCT */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
154 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
155 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
156 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
1984
ef919e9ef73e separate out put_signed_pixels_clamped() into its own function and
melanson
parents: 1977
diff changeset
157 void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
158 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
2763
01ed2c3b8f08 H.264 lossless mode
lorenm
parents: 2755
diff changeset
159 void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
01ed2c3b8f08 H.264 lossless mode
lorenm
parents: 2755
diff changeset
160 void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
161 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
162 * translational global motion compensation.
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
163 */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
164 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
165 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
166 * global motion compensation.
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
167 */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
168 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
169 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
170 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
171 int (*pix_sum)(uint8_t * pix, int line_size);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
172 int (*pix_norm1)(uint8_t * pix, int line_size);
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
173 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
174
1729
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
175 me_cmp_func sad[5]; /* identical to pix_absAxA except additional void * */
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
176 me_cmp_func sse[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
177 me_cmp_func hadamard8_diff[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
178 me_cmp_func dct_sad[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
179 me_cmp_func quant_psnr[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
180 me_cmp_func bit[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
181 me_cmp_func rd[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
182 me_cmp_func vsad[5];
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
183 me_cmp_func vsse[5];
2065
9e4bebc39ade noise preserving sum of squares comparission function
michael
parents: 2045
diff changeset
184 me_cmp_func nsse[5];
2184
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
185 me_cmp_func w53[5];
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
186 me_cmp_func w97[5];
2382
23e7af611c34 more flexible frame skip decission
michael
parents: 2324
diff changeset
187 me_cmp_func dct_max[5];
3010
533c6386eca9 8x8 integer dct from x264 as cmp function (under CONFIG_GPL)
michael
parents: 2979
diff changeset
188 me_cmp_func dct264_sad[5];
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
189
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
190 me_cmp_func me_pre_cmp[5];
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
191 me_cmp_func me_cmp[5];
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
192 me_cmp_func me_sub_cmp[5];
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
193 me_cmp_func mb_cmp[5];
1729
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
194 me_cmp_func ildct_cmp[5]; //only width 16 used
2382
23e7af611c34 more flexible frame skip decission
michael
parents: 2324
diff changeset
195 me_cmp_func frame_skip_cmp[5]; //only width 8 used
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
196
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
197 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
198 * Halfpel motion compensation with rounding (a+b+1)>>1.
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
199 * this is an array[4][4] of motion compensation funcions for 4
1713
49e46374baa4 put_pixels width comment fix
michael
parents: 1709
diff changeset
200 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
201 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
202 * @param block destination where the result is stored
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
203 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
204 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
205 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
206 */
1267
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
207 op_pixels_func put_pixels_tab[4][4];
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
208
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
209 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
210 * Halfpel motion compensation with rounding (a+b+1)>>1.
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
211 * This is an array[4][4] of motion compensation functions for 4
1713
49e46374baa4 put_pixels width comment fix
michael
parents: 1709
diff changeset
212 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
213 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
214 * @param block destination into which the result is averaged (a+b+1)>>1
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
215 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
216 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
217 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
218 */
1319
449f6e32b425 added support for B-frames and multiple slices
tmmm
parents: 1273
diff changeset
219 op_pixels_func avg_pixels_tab[4][4];
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
220
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
221 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
222 * Halfpel motion compensation with no rounding (a+b)>>1.
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
223 * this is an array[2][4] of motion compensation funcions for 2
1225
michaelni
parents: 1213
diff changeset
224 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
225 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
226 * @param block destination where the result is stored
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
227 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
228 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
229 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
230 */
2075
aae3b4e66560 chroma me fix
michael
parents: 2065
diff changeset
231 op_pixels_func put_no_rnd_pixels_tab[4][4];
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
232
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
233 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
234 * Halfpel motion compensation with no rounding (a+b)>>1.
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
235 * this is an array[2][4] of motion compensation funcions for 2
1225
michaelni
parents: 1213
diff changeset
236 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
237 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
238 * @param block destination into which the result is averaged (a+b)>>1
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
239 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
240 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
241 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
242 */
2075
aae3b4e66560 chroma me fix
michael
parents: 2065
diff changeset
243 op_pixels_func avg_no_rnd_pixels_tab[4][4];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
244
1864
9a73e6dc5d68 fix motion compensation with (x+1/2,y+1/2) MVs
michael
parents: 1784
diff changeset
245 void (*put_no_rnd_pixels_l2[2])(uint8_t *block/*align width (8 or 16)*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
246
1267
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
247 /**
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
248 * Thirdpel motion compensation with rounding (a+b+1)>>1.
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
249 * this is an array[12] of motion compensation funcions for the 9 thirdpel positions<br>
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
250 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
251 * @param block destination where the result is stored
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
252 * @param pixels source
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
253 * @param line_size number of bytes in a horizontal line of block
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
254 * @param h height
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
255 */
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
256 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
1319
449f6e32b425 added support for B-frames and multiple slices
tmmm
parents: 1273
diff changeset
257 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
449f6e32b425 added support for B-frames and multiple slices
tmmm
parents: 1273
diff changeset
258
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
259 qpel_mc_func put_qpel_pixels_tab[2][16];
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
260 qpel_mc_func avg_qpel_pixels_tab[2][16];
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
261 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
262 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
263 qpel_mc_func put_mspel_pixels_tab[8];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
264
1168
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
265 /**
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
266 * h264 Chram MC
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
267 */
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
268 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
269 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
270
3020
c75fb0747e74 use h264 MC functions for 2xX Xx2 blocks in snow too
michael
parents: 3010
diff changeset
271 qpel_mc_func put_h264_qpel_pixels_tab[4][16];
c75fb0747e74 use h264 MC functions for 2xX Xx2 blocks in snow too
michael
parents: 3010
diff changeset
272 qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
273
2415
db2cf6005d19 H.264 weighted prediction.
lorenm
parents: 2388
diff changeset
274 h264_weight_func weight_h264_pixels_tab[10];
db2cf6005d19 H.264 weighted prediction.
lorenm
parents: 2388
diff changeset
275 h264_biweight_func biweight_h264_pixels_tab[10];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
276
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
277 me_cmp_func pix_abs[2][4];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
278
866
725ef4ea3ecc huffyuv
michaelni
parents: 862
diff changeset
279 /* huffyuv specific */
725ef4ea3ecc huffyuv
michaelni
parents: 862
diff changeset
280 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
281 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
1527
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
282 /**
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
283 * subtract huffyuv's variant of median prediction
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
284 * note, this might read from src1[-1], src2[-1]
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
285 */
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
286 void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
1273
a979fab41ed8 ASV1 codec
michaelni
parents: 1267
diff changeset
287 void (*bswap_buf)(uint32_t *dst, uint32_t *src, int w);
2633
72e6ffa1f3a5 MMX for H.264 deblocking filter
lorenm
parents: 2415
diff changeset
288
2707
360024d31dab H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
lorenm
parents: 2696
diff changeset
289 void (*h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
360024d31dab H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
lorenm
parents: 2696
diff changeset
290 void (*h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
360024d31dab H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
lorenm
parents: 2696
diff changeset
291 void (*h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
360024d31dab H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
lorenm
parents: 2696
diff changeset
292 void (*h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
360024d31dab H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
lorenm
parents: 2696
diff changeset
293 void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta);
360024d31dab H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
lorenm
parents: 2696
diff changeset
294 void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
295
1644
835cf346975e h263 loop filter
michael
parents: 1603
diff changeset
296 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
835cf346975e h263 loop filter
michael
parents: 1603
diff changeset
297 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
835cf346975e h263 loop filter
michael
parents: 1603
diff changeset
298
2045
9447bbd8a7e9 rewrite h261 loop filter
michael
parents: 2044
diff changeset
299 void (*h261_loop_filter)(uint8_t *src, int stride);
2044
b6f2add2511e h261 decoder by (Maarten Daniels <maarten.daniels at student dot luc dot ac dot be>)
michael
parents: 2024
diff changeset
300
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
301 /* (I)DCT */
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
302 void (*fdct)(DCTELEM *block/* align 16*/);
1567
e08df4d22d27 * introducing dct248 into the DSP context.
romansh
parents: 1565
diff changeset
303 void (*fdct248)(DCTELEM *block/* align 16*/);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
304
1324
7d328fd9d8a5 the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
michaelni
parents: 1320
diff changeset
305 /* IDCT really*/
7d328fd9d8a5 the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
michaelni
parents: 1320
diff changeset
306 void (*idct)(DCTELEM *block/* align 16*/);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
307
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
308 /**
1104
6a01ffc5351c cosmetics
michaelni
parents: 1102
diff changeset
309 * block -> idct -> clip to unsigned 8 bit -> dest.
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
310 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
1111
michaelni
parents: 1106
diff changeset
311 * @param line_size size in bytes of a horizotal line of dest
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
312 */
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
313 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
314
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
315 /**
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
316 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
1111
michaelni
parents: 1106
diff changeset
317 * @param line_size size in bytes of a horizotal line of dest
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
318 */
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
319 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
320
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
321 /**
1104
6a01ffc5351c cosmetics
michaelni
parents: 1102
diff changeset
322 * idct input permutation.
1241
6a873cad413e more idct_permute docs
michaelni
parents: 1225
diff changeset
323 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
6a873cad413e more idct_permute docs
michaelni
parents: 1225
diff changeset
324 * IDCT)
6a873cad413e more idct_permute docs
michaelni
parents: 1225
diff changeset
325 * this permutation must be performed before the idct_put/add, note, normally this can be merged
6a873cad413e more idct_permute docs
michaelni
parents: 1225
diff changeset
326 * with the zigzag/alternate scan<br>
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
327 * an example to avoid confusion:
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
328 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
329 * - (x -> referece dct -> reference idct -> x)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
330 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
331 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
332 */
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
333 uint8_t idct_permutation[64];
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
334 int idct_permutation_type;
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
335 #define FF_NO_IDCT_PERM 1
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
336 #define FF_LIBMPEG2_IDCT_PERM 2
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
337 #define FF_SIMPLE_IDCT_PERM 3
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
338 #define FF_TRANSPOSE_IDCT_PERM 4
2696
9699d325049d porting the mmx&sse2 (sse2 untested) vp3 idcts to the lavc idct API
michael
parents: 2693
diff changeset
339 #define FF_PARTTRANS_IDCT_PERM 5
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
340
1784
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
341 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
342 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
343 #define BASIS_SHIFT 16
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
344 #define RECON_SHIFT 6
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
345
2272
cd43603c46f9 move h264 idct to its own file and call via function pointer in DspContext
michael
parents: 2259
diff changeset
346 void (*h264_idct_add)(uint8_t *dst, DCTELEM *block, int stride);
2755
975074f04b95 decode H.264 with 8x8 transform.
lorenm
parents: 2725
diff changeset
347 void (*h264_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
3105
2d35fb3cb940 h264: special case dc-only idct. ~1% faster overall
lorenm
parents: 3089
diff changeset
348 void (*h264_idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
2d35fb3cb940 h264: special case dc-only idct. ~1% faster overall
lorenm
parents: 3089
diff changeset
349 void (*h264_idct8_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
3198
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
350
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
351 /* snow wavelet */
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
352 void (*vertical_compose97i)(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width);
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
353 void (*horizontal_compose97i)(DWTELEM *b, int width);
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
354 void (*inner_add_yblock)(uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
3215
06f98047ff26 prefetch pixels for future motion compensation. 2-5% faster h264.
lorenm
parents: 3198
diff changeset
355
06f98047ff26 prefetch pixels for future motion compensation. 2-5% faster h264.
lorenm
parents: 3198
diff changeset
356 void (*prefetch)(void *mem, int stride, int h);
3245
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
357
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
358 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
359 } DSPContext;
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
360
1201
e0fc95a6eb4e fixed static init
bellard
parents: 1168
diff changeset
361 void dsputil_static_init(void);
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
362 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
363
764
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
364 /**
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
365 * permute block according to permuatation.
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
366 * @param last last non zero element in scantable order
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
367 */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
368 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
34
2733a4c1c693 added block permutation functions
glantau
parents: 19
diff changeset
369
1729
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
370 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
371
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
372 #define BYTE_VEC32(c) ((c)*0x01010101UL)
1264
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
373
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
374 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
375 {
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
376 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
377 }
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
378
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
379 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
380 {
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
381 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
382 }
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
383
2184
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
384 static inline int get_penalty_factor(int lambda, int lambda2, int type){
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
385 switch(type&0xFF){
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
386 default:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
387 case FF_CMP_SAD:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
388 return lambda>>FF_LAMBDA_SHIFT;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
389 case FF_CMP_DCT:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
390 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
391 case FF_CMP_W53:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
392 return (4*lambda)>>(FF_LAMBDA_SHIFT);
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
393 case FF_CMP_W97:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
394 return (2*lambda)>>(FF_LAMBDA_SHIFT);
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
395 case FF_CMP_SATD:
3010
533c6386eca9 8x8 integer dct from x264 as cmp function (under CONFIG_GPL)
michael
parents: 2979
diff changeset
396 case FF_CMP_DCT264:
2184
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
397 return (2*lambda)>>FF_LAMBDA_SHIFT;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
398 case FF_CMP_RD:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
399 case FF_CMP_PSNR:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
400 case FF_CMP_SSE:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
401 case FF_CMP_NSSE:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
402 return lambda2>>FF_LAMBDA_SHIFT;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
403 case FF_CMP_BIT:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
404 return 1;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
405 }
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
406 }
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
407
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
408 /**
1104
6a01ffc5351c cosmetics
michaelni
parents: 1102
diff changeset
409 * Empty mmx state.
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
410 * this must be called between any dsp function and float/double code.
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
411 * for example sin(); dsp->idct_put(); emms_c(); cos()
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
412 */
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
413 #define emms_c()
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
414
995
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
415 /* should be defined by architectures supporting
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
416 one or more MultiMedia extension */
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
417 int mm_support(void);
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
418
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
419 #ifdef __GNUC__
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
420 #define DECLARE_ALIGNED_16(t,v) t v __attribute__ ((aligned (16)))
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
421 #else
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
422 #define DECLARE_ALIGNED_16(t,v) __declspec(align(16)) t v
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
423 #endif
1974
8c5489b2cf3e move __align16 some place where non-MMX machines can see it
melanson
parents: 1972
diff changeset
424
62
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
425 #if defined(HAVE_MMX)
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
426
862
058194d7ade6 * fixing some minor const warnings
kabi
parents: 853
diff changeset
427 #undef emms_c
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
428
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
429 #define MM_MMX 0x0001 /* standard MMX */
986e461dc072 Initial revision
glantau
parents:
diff changeset
430 #define MM_3DNOW 0x0004 /* AMD 3DNOW */
986e461dc072 Initial revision
glantau
parents:
diff changeset
431 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
986e461dc072 Initial revision
glantau
parents:
diff changeset
432 #define MM_SSE 0x0008 /* SSE functions */
986e461dc072 Initial revision
glantau
parents:
diff changeset
433 #define MM_SSE2 0x0010 /* PIV SSE2 functions */
2388
55a72627a2c5 x86 cpu capabilities detection rewrite / cleanup
michael
parents: 2382
diff changeset
434 #define MM_3DNOWEXT 0x0020 /* AMD 3DNowExt */
3279
647a677c00a4 Remove unused and unsupported Cyrix's "Extended MMX",
gpoirier
parents: 3248
diff changeset
435 #define MM_SSE3 0x0040 /* Prescott SSE3 functions */
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
436
986e461dc072 Initial revision
glantau
parents:
diff changeset
437 extern int mm_flags;
986e461dc072 Initial revision
glantau
parents:
diff changeset
438
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
439 void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
440 void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
1984
ef919e9ef73e separate out put_signed_pixels_clamped() into its own function and
melanson
parents: 1977
diff changeset
441 void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
442
986e461dc072 Initial revision
glantau
parents:
diff changeset
443 static inline void emms(void)
986e461dc072 Initial revision
glantau
parents:
diff changeset
444 {
6
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
445 __asm __volatile ("emms;":::"memory");
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
446 }
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
447
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
448
6
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
449 #define emms_c() \
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
450 {\
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
451 if (mm_flags & MM_MMX)\
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
452 emms();\
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
453 }
986e461dc072 Initial revision
glantau
parents:
diff changeset
454
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
455 #ifdef __GNUC__
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
456 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
457 #else
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
458 #define DECLARE_ALIGNED_8(t,v) __declspec(align(8)) t v
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
459 #endif
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
460
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
461 #define STRIDE_ALIGN 8
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
462
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
463 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
464 void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
1065
92fb44eae6b6 * prototypes at one place
kabi
parents: 1064
diff changeset
465
62
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
466 #elif defined(ARCH_ARMV4L)
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
467
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
468 /* This is to use 4 bytes read to the IDCT pointers for some 'zero'
1974
8c5489b2cf3e move __align16 some place where non-MMX machines can see it
melanson
parents: 1972
diff changeset
469 line optimizations */
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
470 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (4)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
471 #define STRIDE_ALIGN 4
62
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
472
2776
930e56f92c57 IWMMXT configure support + runtime selection patch by (Gildas Bazin, gbazin : altern org)
michael
parents: 2763
diff changeset
473 #define MM_IWMMXT 0x0100 /* XScale IWMMXT */
930e56f92c57 IWMMXT configure support + runtime selection patch by (Gildas Bazin, gbazin : altern org)
michael
parents: 2763
diff changeset
474
930e56f92c57 IWMMXT configure support + runtime selection patch by (Gildas Bazin, gbazin : altern org)
michael
parents: 2763
diff changeset
475 extern int mm_flags;
930e56f92c57 IWMMXT configure support + runtime selection patch by (Gildas Bazin, gbazin : altern org)
michael
parents: 2763
diff changeset
476
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
477 void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx);
62
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
478
88
06f63b58d2a8 mlib merge
glantau
parents: 62
diff changeset
479 #elif defined(HAVE_MLIB)
06f63b58d2a8 mlib merge
glantau
parents: 62
diff changeset
480
06f63b58d2a8 mlib merge
glantau
parents: 62
diff changeset
481 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
482 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
483 #define STRIDE_ALIGN 8
88
06f63b58d2a8 mlib merge
glantau
parents: 62
diff changeset
484
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
485 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
88
06f63b58d2a8 mlib merge
glantau
parents: 62
diff changeset
486
1959
55b7435c59b8 VIS optimized motion compensation code. by (David S. Miller <davem at redhat dot com>)
michael
parents: 1879
diff changeset
487 #elif defined(ARCH_SPARC)
55b7435c59b8 VIS optimized motion compensation code. by (David S. Miller <davem at redhat dot com>)
michael
parents: 1879
diff changeset
488
55b7435c59b8 VIS optimized motion compensation code. by (David S. Miller <davem at redhat dot com>)
michael
parents: 1879
diff changeset
489 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
490 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
491 #define STRIDE_ALIGN 8
1959
55b7435c59b8 VIS optimized motion compensation code. by (David S. Miller <davem at redhat dot com>)
michael
parents: 1879
diff changeset
492 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
55b7435c59b8 VIS optimized motion compensation code. by (David S. Miller <davem at redhat dot com>)
michael
parents: 1879
diff changeset
493
214
73df666cacc7 Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents: 190
diff changeset
494 #elif defined(ARCH_ALPHA)
73df666cacc7 Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents: 190
diff changeset
495
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
496 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
497 #define STRIDE_ALIGN 8
214
73df666cacc7 Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents: 190
diff changeset
498
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
499 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
214
73df666cacc7 Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents: 190
diff changeset
500
623
92e99e506920 first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 612
diff changeset
501 #elif defined(ARCH_POWERPC)
92e99e506920 first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 612
diff changeset
502
894
a408778eff87 altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 884
diff changeset
503 #define MM_ALTIVEC 0x0001 /* standard AltiVec */
a408778eff87 altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 884
diff changeset
504
a408778eff87 altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 884
diff changeset
505 extern int mm_flags;
a408778eff87 altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 884
diff changeset
506
1033
b4172ff70d27 Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents: 1008
diff changeset
507 #if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN)
1653
11433ade9e06 altivec.h bug workaround (untested)
michael
parents: 1644
diff changeset
508 #define pixel altivec_pixel
1033
b4172ff70d27 Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents: 1008
diff changeset
509 #include <altivec.h>
1653
11433ade9e06 altivec.h bug workaround (untested)
michael
parents: 1644
diff changeset
510 #undef pixel
1033
b4172ff70d27 Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents: 1008
diff changeset
511 #endif
b4172ff70d27 Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents: 1008
diff changeset
512
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
513 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (16)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
514 #define STRIDE_ALIGN 16
623
92e99e506920 first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 612
diff changeset
515
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
516 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
623
92e99e506920 first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents: 612
diff changeset
517
689
efcbfbd18864 ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents: 687
diff changeset
518 #elif defined(HAVE_MMI)
efcbfbd18864 ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents: 687
diff changeset
519
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
520 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (16)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
521 #define STRIDE_ALIGN 16
689
efcbfbd18864 ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents: 687
diff changeset
522
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
523 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
689
efcbfbd18864 ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents: 687
diff changeset
524
1259
e8c3884f2c7e sh4 optimized idct & bswap patch by (BERO <bero at geocities dot co dot jp>)
michaelni
parents: 1241
diff changeset
525 #elif defined(ARCH_SH4)
e8c3884f2c7e sh4 optimized idct & bswap patch by (BERO <bero at geocities dot co dot jp>)
michaelni
parents: 1241
diff changeset
526
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
527 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
528 #define STRIDE_ALIGN 8
1259
e8c3884f2c7e sh4 optimized idct & bswap patch by (BERO <bero at geocities dot co dot jp>)
michaelni
parents: 1241
diff changeset
529
e8c3884f2c7e sh4 optimized idct & bswap patch by (BERO <bero at geocities dot co dot jp>)
michaelni
parents: 1241
diff changeset
530 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
e8c3884f2c7e sh4 optimized idct & bswap patch by (BERO <bero at geocities dot co dot jp>)
michaelni
parents: 1241
diff changeset
531
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
532 #else
986e461dc072 Initial revision
glantau
parents:
diff changeset
533
3089
072dbc669253 MSVC-compatible __align8/__align16 declaration
diego
parents: 3036
diff changeset
534 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
535 #define STRIDE_ALIGN 8
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
536
986e461dc072 Initial revision
glantau
parents:
diff changeset
537 #endif
986e461dc072 Initial revision
glantau
parents:
diff changeset
538
493
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
539 #ifdef __GNUC__
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
540
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
541 struct unaligned_64 { uint64_t l; } __attribute__((packed));
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
542 struct unaligned_32 { uint32_t l; } __attribute__((packed));
1267
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
543 struct unaligned_16 { uint16_t l; } __attribute__((packed));
493
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
544
1267
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
545 #define LD16(a) (((const struct unaligned_16 *) (a))->l)
493
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
546 #define LD32(a) (((const struct unaligned_32 *) (a))->l)
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
547 #define LD64(a) (((const struct unaligned_64 *) (a))->l)
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
548
3020
c75fb0747e74 use h264 MC functions for 2xX Xx2 blocks in snow too
michael
parents: 3010
diff changeset
549 #define ST16(a, b) (((struct unaligned_16 *) (a))->l) = (b)
493
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
550 #define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
551
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
552 #else /* __GNUC__ */
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
553
1267
85b71f9f7450 moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
michaelni
parents: 1264
diff changeset
554 #define LD16(a) (*((uint16_t*)(a)))
493
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
555 #define LD32(a) (*((uint32_t*)(a)))
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
556 #define LD64(a) (*((uint64_t*)(a)))
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
557
3217
d9eceb8313c2 ST16 is undefined for non-GNU compilers.
mru
parents: 3215
diff changeset
558 #define ST16(a, b) *((uint16_t*)(a)) = (b)
493
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
559 #define ST32(a, b) *((uint32_t*)(a)) = (b)
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
560
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
561 #endif /* !__GNUC__ */
873b9075d853 move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents: 480
diff changeset
562
252
ddb1a0e94cf4 - Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents: 214
diff changeset
563 /* PSNR */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
564 void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
252
ddb1a0e94cf4 - Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents: 214
diff changeset
565 int orig_linesize[3], int coded_linesize,
ddb1a0e94cf4 - Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents: 214
diff changeset
566 AVCodecContext *avctx);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
567
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
568 /* FFT computation */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
569
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
570 /* NOTE: soon integer code will be added, so you must use the
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
571 FFTSample type */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
572 typedef float FFTSample;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
573
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
574 typedef struct FFTComplex {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
575 FFTSample re, im;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
576 } FFTComplex;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
577
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
578 typedef struct FFTContext {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
579 int nbits;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
580 int inverse;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
581 uint16_t *revtab;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
582 FFTComplex *exptab;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
583 FFTComplex *exptab1; /* only used by SSE code */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
584 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
585 } FFTContext;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
586
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
587 int ff_fft_init(FFTContext *s, int nbits, int inverse);
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
588 void ff_fft_permute(FFTContext *s, FFTComplex *z);
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
589 void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
590 void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
3175
c20c181e0eca 3DNow! & Extended 3DNow! versions of FFT
corey
parents: 3105
diff changeset
591 void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z);
c20c181e0eca 3DNow! & Extended 3DNow! versions of FFT
corey
parents: 3105
diff changeset
592 void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z);
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
593 void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z);
975
e05d525505c5 fft altivec by Romain Dolbeau - simplified patch, test it on PPC with fft-test and wma decoding
bellard
parents: 954
diff changeset
594
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
595 static inline void ff_fft_calc(FFTContext *s, FFTComplex *z)
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
596 {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
597 s->fft_calc(s, z);
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
598 }
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
599 void ff_fft_end(FFTContext *s);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
600
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
601 /* MDCT computation */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
602
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
603 typedef struct MDCTContext {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
604 int n; /* size of MDCT (i.e. number of input data * 2) */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
605 int nbits; /* n = 2^nbits */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
606 /* pre/post rotation tables */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
607 FFTSample *tcos;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
608 FFTSample *tsin;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
609 FFTContext fft;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
610 } MDCTContext;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
611
794
670009af4fc2 avoid name clash
bellard
parents: 781
diff changeset
612 int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
613 void ff_imdct_calc(MDCTContext *s, FFTSample *output,
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
614 const FFTSample *input, FFTSample *tmp);
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
615 void ff_mdct_calc(MDCTContext *s, FFTSample *out,
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
616 const FFTSample *input, FFTSample *tmp);
794
670009af4fc2 avoid name clash
bellard
parents: 781
diff changeset
617 void ff_mdct_end(MDCTContext *s);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
618
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
619 #define WARPER8_16(name8, name16)\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
620 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
621 return name8(s, dst , src , stride, h)\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
622 +name8(s, dst+8 , src+8 , stride, h);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
623 }
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
624
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
625 #define WARPER8_16_SQ(name8, name16)\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
626 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
627 int score=0;\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
628 score +=name8(s, dst , src , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
629 score +=name8(s, dst+8 , src+8 , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
630 if(h==16){\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
631 dst += 8*stride;\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
632 src += 8*stride;\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
633 score +=name8(s, dst , src , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
634 score +=name8(s, dst+8 , src+8 , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
635 }\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
636 return score;\
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
637 }
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
638
1040
998d5035b15b win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
bellard
parents: 1033
diff changeset
639 #endif