Mercurial > libavcodec.hg
annotate dsputil.h @ 12417:9f06475db098 libavcodec
Move vp6_filter_diag4() from DSPContext to VP56DSPContext.
author | rbultje |
---|---|
date | Wed, 25 Aug 2010 13:42:28 +0000 |
parents | 3fc4c625b6f3 |
children | 4269ae88366b |
rev | line source |
---|---|
429 | 1 /* |
2 * DSP utils | |
8629
04423b2f6e0b
cosmetics: Remove pointless period after copyright statement non-sentences.
diego
parents:
8590
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 | 5 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3807
diff
changeset
|
6 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3807
diff
changeset
|
7 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3807
diff
changeset
|
8 * FFmpeg is free software; you can redistribute it and/or |
429 | 9 * modify it under the terms of the GNU Lesser General Public |
10 * License as published by the Free Software Foundation; either | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3807
diff
changeset
|
11 * version 2.1 of the License, or (at your option) any later version. |
429 | 12 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3807
diff
changeset
|
13 * FFmpeg is distributed in the hope that it will be useful, |
429 | 14 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 * Lesser General Public License for more details. | |
17 * | |
18 * You should have received a copy of the GNU Lesser General Public | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3807
diff
changeset
|
19 * License along with FFmpeg; if not, write to the Free Software |
3036
0b546eab515d
Update licensing information: The FSF changed postal address.
diego
parents:
3029
diff
changeset
|
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
429 | 21 */ |
1102 | 22 |
23 /** | |
11644
7dd2a45249a9
Remove explicit filename from Doxygen @file commands.
diego
parents:
11637
diff
changeset
|
24 * @file |
1106 | 25 * DSP utils. |
1213 | 26 * note, many functions in here may use MMX which trashes the FPU state, it is |
27 * absolutely necessary to call emms_c() between dsp & float/double code | |
1102 | 28 */ |
29 | |
7760 | 30 #ifndef AVCODEC_DSPUTIL_H |
31 #define AVCODEC_DSPUTIL_H | |
0 | 32 |
8573
2acf0ae7b041
Fix build: Add intreadwrite.h and bswap.h #includes where necessary.
diego
parents:
8567
diff
changeset
|
33 #include "libavutil/intreadwrite.h" |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
34 #include "avcodec.h" |
0 | 35 |
1102 | 36 |
255 | 37 //#define DEBUG |
0 | 38 /* dct code */ |
39 typedef short DCTELEM; | |
40 | |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
429
diff
changeset
|
41 void fdct_ifast (DCTELEM *data); |
1571
aa4dc16c0f18
* adding integer/floating point AAN implementations for DCT 2-4-8
romansh
parents:
1567
diff
changeset
|
42 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
|
43 void ff_jpeg_fdct_islow (DCTELEM *data); |
1567 | 44 void ff_fdct248_islow (DCTELEM *data); |
0 | 45 |
46 void j_rev_dct (DCTELEM *data); | |
2256 | 47 void j_rev_dct4 (DCTELEM *data); |
2257 | 48 void j_rev_dct2 (DCTELEM *data); |
2259 | 49 void j_rev_dct1 (DCTELEM *data); |
5887 | 50 void ff_wmv2_idct_c(DCTELEM *data); |
0 | 51 |
687
9abb13c21fbe
fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
arpi_esp
parents:
675
diff
changeset
|
52 void ff_fdct_mmx(DCTELEM *block); |
1565 | 53 void ff_fdct_mmx2(DCTELEM *block); |
1765
e31754bc5b65
SSE2 fdct by (Balatoni Denes <pnis at coder dot hu>)
michael
parents:
1739
diff
changeset
|
54 void ff_fdct_sse2(DCTELEM *block); |
0 | 55 |
2755 | 56 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
|
57 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
|
58 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
|
59 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
|
60 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
|
61 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block); |
8375
de2509cf3c44
H.264 idct functions that include the chroma, inter luma and intra16 luma loops
michael
parents:
8359
diff
changeset
|
62 void ff_h264_idct_add16_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); |
de2509cf3c44
H.264 idct functions that include the chroma, inter luma and intra16 luma loops
michael
parents:
8359
diff
changeset
|
63 void ff_h264_idct_add16intra_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); |
de2509cf3c44
H.264 idct functions that include the chroma, inter luma and intra16 luma loops
michael
parents:
8359
diff
changeset
|
64 void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); |
de2509cf3c44
H.264 idct functions that include the chroma, inter luma and intra16 luma loops
michael
parents:
8359
diff
changeset
|
65 void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); |
2272
cd43603c46f9
move h264 idct to its own file and call via function pointer in DspContext
michael
parents:
2259
diff
changeset
|
66 |
7261 | 67 void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1, |
68 const float *win, float add_bias, int len); | |
7218 | 69 void ff_float_to_int16_c(int16_t *dst, const float *src, long len); |
7446 | 70 void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels); |
3568
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
71 |
34 | 72 /* encoding scans */ |
1064 | 73 extern const uint8_t ff_alternate_horizontal_scan[64]; |
74 extern const uint8_t ff_alternate_vertical_scan[64]; | |
75 extern const uint8_t ff_zigzag_direct[64]; | |
1567 | 76 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
|
77 |
0 | 78 /* pixel operations */ |
2090 | 79 #define MAX_NEG_CROP 1024 |
0 | 80 |
81 /* temporary */ | |
4179 | 82 extern uint32_t ff_squareTbl[512]; |
4176 | 83 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP]; |
0 | 84 |
1866
1755f959ab7f
seperated out the C-based VP3 DSP functions into a different file; also
melanson
parents:
1864
diff
changeset
|
85 /* VP3 DSP functions */ |
2693 | 86 void ff_vp3_idct_c(DCTELEM *block/* align 16*/); |
87 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); | |
88 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); | |
11637 | 89 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/); |
0 | 90 |
7995 | 91 void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values); |
92 void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values); | |
93 | |
11379
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
94 /* Bink functions */ |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
95 void ff_bink_idct_c (DCTELEM *block); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
96 void ff_bink_idct_add_c(uint8_t *dest, int linesize, DCTELEM *block); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
97 void ff_bink_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
98 |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
99 /* CAVS functions */ |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
100 void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
101 void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
102 void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
103 void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
104 |
11378
f46b68960464
Move some VC1 dsp prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11376
diff
changeset
|
105 /* VC1 functions */ |
f46b68960464
Move some VC1 dsp prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11376
diff
changeset
|
106 void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd); |
f46b68960464
Move some VC1 dsp prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11376
diff
changeset
|
107 void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd); |
f46b68960464
Move some VC1 dsp prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11376
diff
changeset
|
108 |
11379
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
109 /* EA functions */ |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
110 void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block); |
4b0cc1b97def
Move some dsp func prototypes to dsputil.h; they are defined in dsputil.c
mru
parents:
11378
diff
changeset
|
111 |
3245 | 112 /* 1/2^n downscaling functions from imgconvert.c */ |
113 void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); | |
114 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); | |
115 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); | |
116 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); | |
117 | |
3248
7aa9f80e7954
mmx implementation of 3-point GMC. (5x faster than C)
lorenm
parents:
3245
diff
changeset
|
118 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
|
119 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
|
120 |
675 | 121 /* minimum alignment rules ;) |
6486 | 122 If you notice errors in the align stuff, need more alignment for some ASM code |
123 for some CPU or need to use a function with less aligned data then send a mail | |
124 to the ffmpeg-devel mailing list, ... | |
675 | 125 |
6486 | 126 !warning These alignments might not match reality, (missing attribute((align)) |
127 stuff somewhere possible). | |
6488 | 128 I (Michael) did not check them, these are just the alignments which I think |
6486 | 129 could be reached easily ... |
675 | 130 |
131 !future video codecs might need functions with less strict alignment | |
132 */ | |
0 | 133 |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
134 /* |
1064 | 135 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size); |
136 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride); | |
137 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size); | |
138 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size); | |
296 | 139 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
|
140 */ |
0 | 141 |
142 /* add and put pixel (decoding) */ | |
675 | 143 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16 |
1709 | 144 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4 |
1064 | 145 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
|
146 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 | 147 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride); |
1168 | 148 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y); |
0 | 149 |
11231 | 150 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h); |
151 | |
984 | 152 #define DEF_OLD_QPEL(name)\ |
1064 | 153 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\ |
154 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\ | |
155 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
|
156 |
984 | 157 DEF_OLD_QPEL(qpel16_mc11_old_c) |
158 DEF_OLD_QPEL(qpel16_mc31_old_c) | |
159 DEF_OLD_QPEL(qpel16_mc12_old_c) | |
160 DEF_OLD_QPEL(qpel16_mc32_old_c) | |
161 DEF_OLD_QPEL(qpel16_mc13_old_c) | |
162 DEF_OLD_QPEL(qpel16_mc33_old_c) | |
163 DEF_OLD_QPEL(qpel8_mc11_old_c) | |
164 DEF_OLD_QPEL(qpel8_mc31_old_c) | |
165 DEF_OLD_QPEL(qpel8_mc12_old_c) | |
166 DEF_OLD_QPEL(qpel8_mc32_old_c) | |
167 DEF_OLD_QPEL(qpel8_mc13_old_c) | |
168 DEF_OLD_QPEL(qpel8_mc33_old_c) | |
651 | 169 |
170 #define CALL_2X_PIXELS(a, b, n)\ | |
171 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ | |
172 b(block , pixels , line_size, h);\ | |
173 b(block+n, pixels+n, line_size, h);\ | |
174 } | |
255 | 175 |
0 | 176 /* motion estimation */ |
1709 | 177 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2 |
5127 | 178 // although currently h<4 is not used as functions with width <8 are neither used nor implemented |
1708 | 179 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 | 180 |
1102 | 181 /** |
6438 | 182 * Scantable. |
183 */ | |
184 typedef struct ScanTable{ | |
185 const uint8_t *scantable; | |
186 uint8_t permutated[64]; | |
187 uint8_t raster_end[64]; | |
8590 | 188 #if ARCH_PPC |
6438 | 189 /** Used by dct_quantize_altivec to find last-non-zero */ |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10944
diff
changeset
|
190 DECLARE_ALIGNED(16, uint8_t, inverse)[64]; |
6438 | 191 #endif |
192 } ScanTable; | |
193 | |
194 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable); | |
195 | |
11784 | 196 void ff_emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize, |
6445 | 197 int block_w, int block_h, |
198 int src_x, int src_y, int w, int h); | |
199 | |
6438 | 200 /** |
1102 | 201 * DSPContext. |
202 */ | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
203 typedef struct DSPContext { |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
204 /* pixel ops : interface with DCT */ |
1064 | 205 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size); |
206 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride); | |
207 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
|
208 void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); |
11231 | 209 void (*put_pixels_nonclamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); |
1064 | 210 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); |
2763 | 211 void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size); |
212 void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size); | |
4988
689490842cf5
factor sum_abs_dctelem out of dct_sad, and simd it.
lorenm
parents:
4962
diff
changeset
|
213 int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/); |
1136 | 214 /** |
215 * translational global motion compensation. | |
216 */ | |
1064 | 217 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder); |
1136 | 218 /** |
219 * global motion compensation. | |
220 */ | |
1064 | 221 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy, |
2979 | 222 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height); |
8288 | 223 void (*clear_block)(DCTELEM *block/*align 16*/); |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
224 void (*clear_blocks)(DCTELEM *blocks/*align 16*/); |
1064 | 225 int (*pix_sum)(uint8_t * pix, int line_size); |
226 int (*pix_norm1)(uint8_t * pix, int line_size); | |
1708 | 227 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4 |
2967 | 228 |
8976
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
229 me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */ |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
230 me_cmp_func sse[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
231 me_cmp_func hadamard8_diff[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
232 me_cmp_func dct_sad[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
233 me_cmp_func quant_psnr[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
234 me_cmp_func bit[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
235 me_cmp_func rd[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
236 me_cmp_func vsad[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
237 me_cmp_func vsse[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
238 me_cmp_func nsse[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
239 me_cmp_func w53[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
240 me_cmp_func w97[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
241 me_cmp_func dct_max[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
242 me_cmp_func dct264_sad[6]; |
936 | 243 |
8976
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
244 me_cmp_func me_pre_cmp[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
245 me_cmp_func me_cmp[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
246 me_cmp_func me_sub_cmp[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
247 me_cmp_func mb_cmp[6]; |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
248 me_cmp_func ildct_cmp[6]; //only width 16 used |
e7d87561b42b
Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents:
8785
diff
changeset
|
249 me_cmp_func frame_skip_cmp[6]; //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
|
250 |
5255 | 251 int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2, |
252 int size); | |
4749 | 253 |
1136 | 254 /** |
255 * Halfpel motion compensation with rounding (a+b+1)>>1. | |
4751 | 256 * this is an array[4][4] of motion compensation functions for 4 |
1713 | 257 * horizontal blocksizes (8,16) and the 4 halfpel positions<br> |
1213 | 258 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ] |
1136 | 259 * @param block destination where the result is stored |
260 * @param pixels source | |
261 * @param line_size number of bytes in a horizontal line of block | |
262 * @param h height | |
263 */ | |
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
|
264 op_pixels_func put_pixels_tab[4][4]; |
1136 | 265 |
266 /** | |
267 * Halfpel motion compensation with rounding (a+b+1)>>1. | |
2967 | 268 * This is an array[4][4] of motion compensation functions for 4 |
1713 | 269 * horizontal blocksizes (8,16) and the 4 halfpel positions<br> |
1213 | 270 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ] |
1136 | 271 * @param block destination into which the result is averaged (a+b+1)>>1 |
272 * @param pixels source | |
273 * @param line_size number of bytes in a horizontal line of block | |
274 * @param h height | |
275 */ | |
1319 | 276 op_pixels_func avg_pixels_tab[4][4]; |
1136 | 277 |
278 /** | |
279 * Halfpel motion compensation with no rounding (a+b)>>1. | |
4751 | 280 * this is an array[2][4] of motion compensation functions for 2 |
1225 | 281 * horizontal blocksizes (8,16) and the 4 halfpel positions<br> |
1213 | 282 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ] |
1136 | 283 * @param block destination where the result is stored |
284 * @param pixels source | |
285 * @param line_size number of bytes in a horizontal line of block | |
286 * @param h height | |
287 */ | |
2075 | 288 op_pixels_func put_no_rnd_pixels_tab[4][4]; |
1136 | 289 |
290 /** | |
291 * Halfpel motion compensation with no rounding (a+b)>>1. | |
4751 | 292 * this is an array[2][4] of motion compensation functions for 2 |
1225 | 293 * horizontal blocksizes (8,16) and the 4 halfpel positions<br> |
1213 | 294 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ] |
1136 | 295 * @param block destination into which the result is averaged (a+b)>>1 |
296 * @param pixels source | |
297 * @param line_size number of bytes in a horizontal line of block | |
298 * @param h height | |
299 */ | |
2075 | 300 op_pixels_func avg_no_rnd_pixels_tab[4][4]; |
2967 | 301 |
1864 | 302 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 | 303 |
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
|
304 /** |
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
|
305 * Thirdpel motion compensation with rounding (a+b+1)>>1. |
4751 | 306 * this is an array[12] of motion compensation functions for the 9 thirdpe |
307 * positions<br> | |
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
|
308 * *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
|
309 * @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
|
310 * @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
|
311 * @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
|
312 * @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
|
313 */ |
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
|
314 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width? |
1319 | 315 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width? |
316 | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
317 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
|
318 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
|
319 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
|
320 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]; |
936 | 321 qpel_mc_func put_mspel_pixels_tab[8]; |
2967 | 322 |
1168 | 323 /** |
4751 | 324 * h264 Chroma MC |
1168 | 325 */ |
326 h264_chroma_mc_func put_h264_chroma_pixels_tab[3]; | |
9439
ef3a7b711cc0
Rename put_no_rnd_h264_chroma* to reflect its usage in VC1 only
conrad
parents:
9437
diff
changeset
|
327 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3]; |
3663 | 328 /* This is really one func used in VC-1 decoding */ |
9439
ef3a7b711cc0
Rename put_no_rnd_h264_chroma* to reflect its usage in VC1 only
conrad
parents:
9437
diff
changeset
|
329 h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3]; |
9440 | 330 h264_chroma_mc_func avg_no_rnd_vc1_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
|
331 |
3020
c75fb0747e74
use h264 MC functions for 2xX Xx2 blocks in snow too
michael
parents:
3010
diff
changeset
|
332 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
|
333 qpel_mc_func avg_h264_qpel_pixels_tab[4][16]; |
2967 | 334 |
3807
6a40092eb9e6
approximate qpel functions: sacrifice some quality for some decoding speed. enabled on B-frames with -lavdopts fast.
lorenm
parents:
3746
diff
changeset
|
335 qpel_mc_func put_2tap_qpel_pixels_tab[4][16]; |
6a40092eb9e6
approximate qpel functions: sacrifice some quality for some decoding speed. enabled on B-frames with -lavdopts fast.
lorenm
parents:
3746
diff
changeset
|
336 qpel_mc_func avg_2tap_qpel_pixels_tab[4][16]; |
6a40092eb9e6
approximate qpel functions: sacrifice some quality for some decoding speed. enabled on B-frames with -lavdopts fast.
lorenm
parents:
3746
diff
changeset
|
337 |
1708 | 338 me_cmp_func pix_abs[2][4]; |
2967 | 339 |
866 | 340 /* huffyuv specific */ |
341 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w); | |
6384 | 342 void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w); |
936 | 343 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w); |
1527 | 344 /** |
345 * subtract huffyuv's variant of median prediction | |
346 * note, this might read from src1[-1], src2[-1] | |
347 */ | |
10431 | 348 void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top); |
349 void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top); | |
10430 | 350 int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left); |
10878
a8620b001ed3
Implement alpha channel decoding for BGR HuffYUV.
astrange
parents:
10827
diff
changeset
|
351 void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha); |
6384 | 352 /* this might write to dst[w] */ |
353 void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp); | |
6241 | 354 void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w); |
2633 | 355 |
1644 | 356 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale); |
357 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale); | |
358 | |
2045 | 359 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
|
360 |
5887 | 361 void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale); |
362 void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale); | |
363 | |
11637 | 364 void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/); |
7995 | 365 void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values); |
366 void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values); | |
367 | |
3568
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
368 /* assume len is a multiple of 4, and arrays are 16-byte aligned */ |
3536
545a15c19c91
sse & sse2 implementations of vorbis channel coupling.
lorenm
parents:
3526
diff
changeset
|
369 void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize); |
7563 | 370 void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len); |
5737 | 371 /* no alignment needed */ |
10424
94595d0e617c
Move autocorrelation function from flacenc.c to lpc.c. Also rename the
jbr
parents:
10420
diff
changeset
|
372 void (*lpc_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc); |
3574 | 373 /* assume len is a multiple of 8, and arrays are 16-byte aligned */ |
3568
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
374 void (*vector_fmul)(float *dst, const float *src, int len); |
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
375 void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len); |
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
376 /* assume len is a multiple of 8, and src arrays are 16-byte aligned */ |
10300
4d1b9ca628fc
Drop unused args from vector_fmul_add_add, simpify code, and rename
mru
parents:
10219
diff
changeset
|
377 void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len); |
7261 | 378 /* assume len is a multiple of 4, and arrays are 16-byte aligned */ |
379 void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len); | |
7564 | 380 /* assume len is a multiple of 8, and arrays are 16-byte aligned */ |
381 void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len); | |
10105 | 382 void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */); |
10219 | 383 /** |
384 * Multiply a vector of floats by a scalar float. Source and | |
385 * destination vectors must overlap exactly or not at all. | |
386 * @param dst result vector, 16-byte aligned | |
387 * @param src input vector, 16-byte aligned | |
388 * @param mul scalar value | |
389 * @param len length of vector, multiple of 4 | |
390 */ | |
391 void (*vector_fmul_scalar)(float *dst, const float *src, float mul, | |
392 int len); | |
393 /** | |
394 * Multiply a vector of floats by concatenated short vectors of | |
395 * floats and by a scalar float. Source and destination vectors | |
396 * must overlap exactly or not at all. | |
397 * [0]: short vectors of length 2, 8-byte aligned | |
398 * [1]: short vectors of length 4, 16-byte aligned | |
399 * @param dst output vector, 16-byte aligned | |
400 * @param src input vector, 16-byte aligned | |
401 * @param sv array of pointers to short vectors | |
402 * @param mul scalar value | |
403 * @param len number of elements in src and dst, multiple of 4 | |
404 */ | |
405 void (*vector_fmul_sv_scalar[2])(float *dst, const float *src, | |
406 const float **sv, float mul, int len); | |
407 /** | |
408 * Multiply short vectors of floats by a scalar float, store | |
409 * concatenated result. | |
410 * [0]: short vectors of length 2, 8-byte aligned | |
411 * [1]: short vectors of length 4, 16-byte aligned | |
412 * @param dst output vector, 16-byte aligned | |
413 * @param sv array of pointers to short vectors | |
414 * @param mul scalar value | |
415 * @param len number of output elements, multiple of 4 | |
416 */ | |
417 void (*sv_fmul_scalar[2])(float *dst, const float **sv, | |
418 float mul, int len); | |
419 /** | |
420 * Calculate the scalar product of two vectors of floats. | |
421 * @param v1 first vector, 16-byte aligned | |
422 * @param v2 second vector, 16-byte aligned | |
423 * @param len length of vectors, multiple of 4 | |
424 */ | |
425 float (*scalarproduct_float)(const float *v1, const float *v2, int len); | |
426 /** | |
427 * Calculate the sum and difference of two vectors of floats. | |
428 * @param v1 first input vector, sum output, 16-byte aligned | |
429 * @param v2 second input vector, difference output, 16-byte aligned | |
430 * @param len length of vectors, multiple of 4 | |
431 */ | |
432 void (*butterflies_float)(float *restrict v1, float *restrict v2, int len); | |
3568
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
433 |
945caa35ee9a
sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents:
3555
diff
changeset
|
434 /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767] |
3660
7e1ee254a3ee
Align the input buffer in ffplay, introduce a public macro for aligned declarations
lu_zero
parents:
3656
diff
changeset
|
435 * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */ |
7218 | 436 void (*float_to_int16)(int16_t *dst, const float *src, long len); |
7286
e267f2519248
float_to_int16_interleave: change src to an array of pointers instead of assuming it's contiguous.
lorenm
parents:
7263
diff
changeset
|
437 void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels); |
3536
545a15c19c91
sse & sse2 implementations of vorbis channel coupling.
lorenm
parents:
3526
diff
changeset
|
438 |
1092 | 439 /* (I)DCT */ |
440 void (*fdct)(DCTELEM *block/* align 16*/); | |
1567 | 441 void (*fdct248)(DCTELEM *block/* align 16*/); |
2967 | 442 |
1324
7d328fd9d8a5
the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
michaelni
parents:
1320
diff
changeset
|
443 /* IDCT really*/ |
7d328fd9d8a5
the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
michaelni
parents:
1320
diff
changeset
|
444 void (*idct)(DCTELEM *block/* align 16*/); |
2967 | 445 |
1102 | 446 /** |
1104 | 447 * block -> idct -> clip to unsigned 8 bit -> dest. |
1102 | 448 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...) |
5728 | 449 * @param line_size size in bytes of a horizontal line of dest |
1102 | 450 */ |
1092 | 451 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); |
2967 | 452 |
1102 | 453 /** |
454 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest. | |
5728 | 455 * @param line_size size in bytes of a horizontal line of dest |
1102 | 456 */ |
1092 | 457 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); |
2967 | 458 |
1102 | 459 /** |
1104 | 460 * idct input permutation. |
1241 | 461 * several optimized IDCTs need a permutated input (relative to the normal order of the reference |
462 * IDCT) | |
463 * this permutation must be performed before the idct_put/add, note, normally this can be merged | |
464 * with the zigzag/alternate scan<br> | |
1102 | 465 * an example to avoid confusion: |
466 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...) | |
467 * - (x -> referece dct -> reference idct -> x) | |
468 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x) | |
469 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...) | |
470 */ | |
1092 | 471 uint8_t idct_permutation[64]; |
472 int idct_permutation_type; | |
473 #define FF_NO_IDCT_PERM 1 | |
474 #define FF_LIBMPEG2_IDCT_PERM 2 | |
475 #define FF_SIMPLE_IDCT_PERM 3 | |
476 #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
|
477 #define FF_PARTTRANS_IDCT_PERM 5 |
6600
c3213c91124c
Add a new IDCT permutation, used in xvid_sse2 and possibly future similar IDCTs.
astrange
parents:
6488
diff
changeset
|
478 #define FF_SSE2_IDCT_PERM 6 |
1092 | 479 |
1784 | 480 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale); |
481 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale); | |
482 #define BASIS_SHIFT 16 | |
483 #define RECON_SHIFT 6 | |
2967 | 484 |
6437 | 485 void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w); |
6449
208074826b34
move EDGE_WIDTH definition allong with draw_edges where it belongs
aurel
parents:
6445
diff
changeset
|
486 #define EDGE_WIDTH 16 |
6437 | 487 |
3215
06f98047ff26
prefetch pixels for future motion compensation. 2-5% faster h264.
lorenm
parents:
3198
diff
changeset
|
488 void (*prefetch)(void *mem, int stride, int h); |
3245 | 489 |
490 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); | |
3526 | 491 |
9585 | 492 /* mlp/truehd functions */ |
9647
d0fe5dc427f0
mlp: Simplify adressing of state and coeffs arrays for both filters by making
ramiro
parents:
9585
diff
changeset
|
493 void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff, |
d0fe5dc427f0
mlp: Simplify adressing of state and coeffs arrays for both filters by making
ramiro
parents:
9585
diff
changeset
|
494 int firorder, int iirorder, |
9585 | 495 unsigned int filter_shift, int32_t mask, int blocksize, |
496 int32_t *sample_buffer); | |
497 | |
3526 | 498 /* vc1 functions */ |
499 void (*vc1_inv_trans_8x8)(DCTELEM *b); | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5907
diff
changeset
|
500 void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block); |
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5907
diff
changeset
|
501 void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block); |
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5907
diff
changeset
|
502 void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block); |
9859
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9658
diff
changeset
|
503 void (*vc1_inv_trans_8x8_dc)(uint8_t *dest, int line_size, DCTELEM *block); |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9658
diff
changeset
|
504 void (*vc1_inv_trans_8x4_dc)(uint8_t *dest, int line_size, DCTELEM *block); |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9658
diff
changeset
|
505 void (*vc1_inv_trans_4x8_dc)(uint8_t *dest, int line_size, DCTELEM *block); |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9658
diff
changeset
|
506 void (*vc1_inv_trans_4x4_dc)(uint8_t *dest, int line_size, DCTELEM *block); |
4239 | 507 void (*vc1_v_overlap)(uint8_t* src, int stride); |
508 void (*vc1_h_overlap)(uint8_t* src, int stride); | |
9443
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
509 void (*vc1_v_loop_filter4)(uint8_t *src, int stride, int pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
510 void (*vc1_h_loop_filter4)(uint8_t *src, int stride, int pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
511 void (*vc1_v_loop_filter8)(uint8_t *src, int stride, int pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
512 void (*vc1_h_loop_filter8)(uint8_t *src, int stride, int pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
513 void (*vc1_v_loop_filter16)(uint8_t *src, int stride, int pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
514 void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq); |
3526 | 515 /* put 8x8 block with bicubic interpolation and quarterpel precision |
516 * last argument is actually round value instead of height | |
517 */ | |
518 op_pixels_func put_vc1_mspel_pixels_tab[16]; | |
9437 | 519 op_pixels_func avg_vc1_mspel_pixels_tab[16]; |
5887 | 520 |
521 /* intrax8 functions */ | |
5907
fbd10e6dfbe1
cosmetics: Fix spacial --> spatial typo in function names.
diego
parents:
5887
diff
changeset
|
522 void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize); |
fbd10e6dfbe1
cosmetics: Fix spacial --> spatial typo in function names.
diego
parents:
5887
diff
changeset
|
523 void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize, |
5887 | 524 int * range, int * sum, int edges); |
525 | |
7203
87b1dfb5a98d
Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents:
7135
diff
changeset
|
526 /** |
87b1dfb5a98d
Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents:
7135
diff
changeset
|
527 * Calculate scalar product of two vectors. |
7232
cc55dd004819
Monkey's Audio decoder vector functions work on input with length
kostya
parents:
7218
diff
changeset
|
528 * @param len length of vectors, should be multiple of 16 |
7203
87b1dfb5a98d
Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents:
7135
diff
changeset
|
529 * @param shift number of bits to discard from product |
87b1dfb5a98d
Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents:
7135
diff
changeset
|
530 */ |
11981 | 531 int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len, int shift); |
10644 | 532 /* ape functions */ |
533 /** | |
534 * Calculate scalar product of v1 and v2, | |
535 * and v1[i] += v3[i] * mul | |
536 * @param len length of vectors, should be multiple of 16 | |
537 */ | |
11981 | 538 int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul); |
8232 | 539 |
8233 | 540 /* rv30 functions */ |
541 qpel_mc_func put_rv30_tpel_pixels_tab[4][16]; | |
542 qpel_mc_func avg_rv30_tpel_pixels_tab[4][16]; | |
543 | |
8232 | 544 /* rv40 functions */ |
545 qpel_mc_func put_rv40_qpel_pixels_tab[4][16]; | |
546 qpel_mc_func avg_rv40_qpel_pixels_tab[4][16]; | |
547 h264_chroma_mc_func put_rv40_chroma_pixels_tab[3]; | |
548 h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3]; | |
11231 | 549 |
550 /* bink functions */ | |
551 op_fill_func fill_block_tab[2]; | |
552 void (*scale_block)(const uint8_t src[64]/*align 8*/, uint8_t *dst/*align 8*/, int linesize); | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
553 } DSPContext; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
554 |
4197 | 555 void dsputil_static_init(void); |
1092 | 556 void dsputil_init(DSPContext* p, AVCodecContext *avctx); |
0 | 557 |
4281
de525a2b41db
ff_check_alignment to warn the user about a missaligned stack
michael
parents:
4279
diff
changeset
|
558 int ff_check_alignment(void); |
de525a2b41db
ff_check_alignment to warn the user about a missaligned stack
michael
parents:
4279
diff
changeset
|
559 |
764 | 560 /** |
561 * permute block according to permuatation. | |
562 * @param last last non zero element in scantable order | |
563 */ | |
1064 | 564 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last); |
34 | 565 |
1729 | 566 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type); |
567 | |
2979 | 568 #define BYTE_VEC32(c) ((c)*0x01010101UL) |
1264 | 569 |
570 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b) | |
571 { | |
572 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1); | |
573 } | |
574 | |
575 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b) | |
576 { | |
577 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1); | |
578 } | |
579 | |
2184 | 580 static inline int get_penalty_factor(int lambda, int lambda2, int type){ |
581 switch(type&0xFF){ | |
582 default: | |
583 case FF_CMP_SAD: | |
584 return lambda>>FF_LAMBDA_SHIFT; | |
585 case FF_CMP_DCT: | |
586 return (3*lambda)>>(FF_LAMBDA_SHIFT+1); | |
587 case FF_CMP_W53: | |
588 return (4*lambda)>>(FF_LAMBDA_SHIFT); | |
589 case FF_CMP_W97: | |
590 return (2*lambda)>>(FF_LAMBDA_SHIFT); | |
591 case FF_CMP_SATD: | |
3010
533c6386eca9
8x8 integer dct from x264 as cmp function (under CONFIG_GPL)
michael
parents:
2979
diff
changeset
|
592 case FF_CMP_DCT264: |
2184 | 593 return (2*lambda)>>FF_LAMBDA_SHIFT; |
594 case FF_CMP_RD: | |
595 case FF_CMP_PSNR: | |
596 case FF_CMP_SSE: | |
597 case FF_CMP_NSSE: | |
598 return lambda2>>FF_LAMBDA_SHIFT; | |
599 case FF_CMP_BIT: | |
600 return 1; | |
601 } | |
602 } | |
603 | |
1102 | 604 /** |
1104 | 605 * Empty mmx state. |
1102 | 606 * this must be called between any dsp function and float/double code. |
607 * for example sin(); dsp->idct_put(); emms_c(); cos() | |
608 */ | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
609 #define emms_c() |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
610 |
995
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
611 /* should be defined by architectures supporting |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
612 one or more MultiMedia extension */ |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
613 int mm_support(void); |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
614 |
5149 | 615 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx); |
8359 | 616 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx); |
5149 | 617 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx); |
618 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx); | |
619 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx); | |
620 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx); | |
621 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx); | |
622 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx); | |
623 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx); | |
624 | |
11485 | 625 void ff_dsputil_init_dwt(DSPContext *c); |
11376
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
626 void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx); |
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
627 void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx); |
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
628 void ff_vc1dsp_init(DSPContext* c, AVCodecContext *avctx); |
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
629 void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx); |
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
630 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx); |
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
631 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx); |
0b220468ba0d
Move prototypes for various dsputil init functions to dsputil.h
mru
parents:
11370
diff
changeset
|
632 |
8590 | 633 #if HAVE_MMX |
0 | 634 |
862 | 635 #undef emms_c |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
636 |
0 | 637 static inline void emms(void) |
638 { | |
8031 | 639 __asm__ volatile ("emms;":::"memory"); |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
640 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
641 |
12414 | 642 #define emms_c() emms() |
0 | 643 |
8590 | 644 #elif ARCH_ARM |
62 | 645 |
8590 | 646 #if HAVE_NEON |
7687
e5b5a9af1b68
ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents:
7678
diff
changeset
|
647 # define STRIDE_ALIGN 16 |
e5b5a9af1b68
ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents:
7678
diff
changeset
|
648 #endif |
e5b5a9af1b68
ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents:
7678
diff
changeset
|
649 |
8590 | 650 #elif ARCH_PPC |
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
|
651 |
2324 | 652 #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
|
653 |
8590 | 654 #elif HAVE_MMI |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
655 |
2324 | 656 #define STRIDE_ALIGN 16 |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
657 |
7135
6bd6a2da306e
Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents:
7094
diff
changeset
|
658 #else |
6bd6a2da306e
Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents:
7094
diff
changeset
|
659 |
6bd6a2da306e
Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents:
7094
diff
changeset
|
660 #define mm_support() 0 |
6bd6a2da306e
Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents:
7094
diff
changeset
|
661 |
6363
7ebd1cdb2142
clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents:
6241
diff
changeset
|
662 #endif |
3728 | 663 |
6363
7ebd1cdb2142
clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents:
6241
diff
changeset
|
664 #ifndef STRIDE_ALIGN |
7ebd1cdb2142
clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents:
6241
diff
changeset
|
665 # define STRIDE_ALIGN 8 |
0 | 666 #endif |
667 | |
11194
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
668 #define LOCAL_ALIGNED(a, t, v, s, ...) \ |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
669 uint8_t la_##v[sizeof(t s __VA_ARGS__) + (a)]; \ |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
670 t (*v) __VA_ARGS__ = (void *)FFALIGN((uintptr_t)la_##v, a) |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
671 |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
672 #if HAVE_LOCAL_ALIGNED_8 |
11369 | 673 # define LOCAL_ALIGNED_8(t, v, s, ...) DECLARE_ALIGNED(8, t, v) s __VA_ARGS__ |
11194
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
674 #else |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
675 # define LOCAL_ALIGNED_8(t, v, s, ...) LOCAL_ALIGNED(8, t, v, s, __VA_ARGS__) |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
676 #endif |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
677 |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
678 #if HAVE_LOCAL_ALIGNED_16 |
11369 | 679 # define LOCAL_ALIGNED_16(t, v, s, ...) DECLARE_ALIGNED(16, t, v) s __VA_ARGS__ |
11194
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
680 #else |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
681 # define LOCAL_ALIGNED_16(t, v, s, ...) LOCAL_ALIGNED(16, t, v, s, __VA_ARGS__) |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
682 #endif |
7b3f6955462b
Add LOCAL_ALIGNED() macro for declaring aligned local arrays
mru
parents:
10961
diff
changeset
|
683 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
684 /* PSNR */ |
1064 | 685 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
|
686 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
|
687 AVCodecContext *avctx); |
781 | 688 |
6056
558c1fd0ee72
Fix typo in macro name: WARPER8_16_SQ --> WRAPPER8_16_SQ.
diego
parents:
5997
diff
changeset
|
689 #define WRAPPER8_16(name8, name16)\ |
1708 | 690 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\ |
691 return name8(s, dst , src , stride, h)\ | |
692 +name8(s, dst+8 , src+8 , stride, h);\ | |
693 } | |
694 | |
6056
558c1fd0ee72
Fix typo in macro name: WARPER8_16_SQ --> WRAPPER8_16_SQ.
diego
parents:
5997
diff
changeset
|
695 #define WRAPPER8_16_SQ(name8, name16)\ |
1708 | 696 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\ |
697 int score=0;\ | |
698 score +=name8(s, dst , src , stride, 8);\ | |
699 score +=name8(s, dst+8 , src+8 , stride, 8);\ | |
700 if(h==16){\ | |
701 dst += 8*stride;\ | |
702 src += 8*stride;\ | |
703 score +=name8(s, dst , src , stride, 8);\ | |
704 score +=name8(s, dst+8 , src+8 , stride, 8);\ | |
705 }\ | |
706 return score;\ | |
936 | 707 } |
708 | |
4240 | 709 |
9436 | 710 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h) |
4240 | 711 { |
712 int i; | |
713 for(i=0; i<h; i++) | |
714 { | |
5520
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
715 AV_WN16(dst , AV_RN16(src )); |
4240 | 716 dst+=dstStride; |
717 src+=srcStride; | |
718 } | |
719 } | |
720 | |
9436 | 721 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h) |
4240 | 722 { |
723 int i; | |
724 for(i=0; i<h; i++) | |
725 { | |
5520
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
726 AV_WN32(dst , AV_RN32(src )); |
4240 | 727 dst+=dstStride; |
728 src+=srcStride; | |
729 } | |
730 } | |
731 | |
9436 | 732 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h) |
4240 | 733 { |
734 int i; | |
735 for(i=0; i<h; i++) | |
736 { | |
5520
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
737 AV_WN32(dst , AV_RN32(src )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
738 AV_WN32(dst+4 , AV_RN32(src+4 )); |
4240 | 739 dst+=dstStride; |
740 src+=srcStride; | |
741 } | |
742 } | |
743 | |
9436 | 744 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h) |
4240 | 745 { |
746 int i; | |
747 for(i=0; i<h; i++) | |
748 { | |
5520
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
749 AV_WN32(dst , AV_RN32(src )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
750 AV_WN32(dst+4 , AV_RN32(src+4 )); |
4240 | 751 dst[8]= src[8]; |
752 dst+=dstStride; | |
753 src+=srcStride; | |
754 } | |
755 } | |
756 | |
9436 | 757 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h) |
4240 | 758 { |
759 int i; | |
760 for(i=0; i<h; i++) | |
761 { | |
5520
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
762 AV_WN32(dst , AV_RN32(src )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
763 AV_WN32(dst+4 , AV_RN32(src+4 )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
764 AV_WN32(dst+8 , AV_RN32(src+8 )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
765 AV_WN32(dst+12, AV_RN32(src+12)); |
4240 | 766 dst+=dstStride; |
767 src+=srcStride; | |
768 } | |
769 } | |
770 | |
9436 | 771 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h) |
4240 | 772 { |
773 int i; | |
774 for(i=0; i<h; i++) | |
775 { | |
5520
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
776 AV_WN32(dst , AV_RN32(src )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
777 AV_WN32(dst+4 , AV_RN32(src+4 )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
778 AV_WN32(dst+8 , AV_RN32(src+8 )); |
c16a59ef6a86
* renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents:
5291
diff
changeset
|
779 AV_WN32(dst+12, AV_RN32(src+12)); |
4240 | 780 dst[16]= src[16]; |
781 dst+=dstStride; | |
782 src+=srcStride; | |
783 } | |
784 } | |
785 | |
7760 | 786 #endif /* AVCODEC_DSPUTIL_H */ |