annotate dsputil.h @ 9477:94edfc2a9b8b libavcodec

Remove a pointless right-shift in xan decoder.
author reimar
date Fri, 17 Apr 2009 17:54:55 +0000
parents 3970fe47fea3
children 5e1d9508b62f
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
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
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
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
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
9 * modify it under the terms of the GNU Lesser General Public
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
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
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
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
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
16 * Lesser General Public License for more details.
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
17 *
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
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
718a22dc121f license/copyright change
glantau
parents: 403
diff changeset
21 */
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
22
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
23 /**
8718
e9d9d946f213 Use full internal pathname in doxygen @file directives.
diego
parents: 8694
diff changeset
24 * @file libavcodec/dsputil.h
1106
1e39f273ecd6 per file doxy
michaelni
parents: 1104
diff changeset
25 * DSP utils.
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
26 * 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
27 * absolutely necessary to call emms_c() between dsp & float/double code
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
28 */
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
29
7760
c4a4495715dd Globally rename the header inclusion guard names.
stefano
parents: 7687
diff changeset
30 #ifndef AVCODEC_DSPUTIL_H
c4a4495715dd Globally rename the header inclusion guard names.
stefano
parents: 7687
diff changeset
31 #define AVCODEC_DSPUTIL_H
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
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
986e461dc072 Initial revision
glantau
parents:
diff changeset
35
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
36
255
db20b987c32d divx5-gmc support
michaelni
parents: 252
diff changeset
37 //#define DEBUG
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
38 /* dct code */
986e461dc072 Initial revision
glantau
parents:
diff changeset
39 typedef short DCTELEM;
3198
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
40 typedef int DWTELEM;
5587
3ae03eacbe9f use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents: 5573
diff changeset
41 typedef short IDWTELEM;
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
42
474
11dbd00682fc avoid name clash with libjpeg - added missing externs
bellard
parents: 429
diff changeset
43 void fdct_ifast (DCTELEM *data);
1571
aa4dc16c0f18 * adding integer/floating point AAN implementations for DCT 2-4-8
romansh
parents: 1567
diff changeset
44 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
45 void ff_jpeg_fdct_islow (DCTELEM *data);
1567
e08df4d22d27 * introducing dct248 into the DSP context.
romansh
parents: 1565
diff changeset
46 void ff_fdct248_islow (DCTELEM *data);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
47
986e461dc072 Initial revision
glantau
parents:
diff changeset
48 void j_rev_dct (DCTELEM *data);
2256
7e0b2e86afa9 1/2 resolution decoding
michael
parents: 2185
diff changeset
49 void j_rev_dct4 (DCTELEM *data);
2257
5f64a30339e5 1/4 resolution decoding
michael
parents: 2256
diff changeset
50 void j_rev_dct2 (DCTELEM *data);
2259
12e75af1d44c 1/8 resolution decoding
michael
parents: 2257
diff changeset
51 void j_rev_dct1 (DCTELEM *data);
5887
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
52 void ff_wmv2_idct_c(DCTELEM *data);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
53
687
9abb13c21fbe fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
arpi_esp
parents: 675
diff changeset
54 void ff_fdct_mmx(DCTELEM *block);
1565
1a9a63f59849 minor mmx2 optimization if the dct
michael
parents: 1527
diff changeset
55 void ff_fdct_mmx2(DCTELEM *block);
1765
e31754bc5b65 SSE2 fdct by (Balatoni Denes <pnis at coder dot hu>)
michael
parents: 1739
diff changeset
56 void ff_fdct_sse2(DCTELEM *block);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
57
2755
975074f04b95 decode H.264 with 8x8 transform.
lorenm
parents: 2725
diff changeset
58 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
59 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
60 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
61 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
62 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
63 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
64 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
65 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
66 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
67 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
68
3568
945caa35ee9a sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents: 3555
diff changeset
69 void ff_vector_fmul_add_add_c(float *dst, const float *src0, const float *src1,
945caa35ee9a sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents: 3555
diff changeset
70 const float *src2, int src3, int blocksize, int step);
7261
032a49f033e8 simplify vorbis windowing
lorenm
parents: 7232
diff changeset
71 void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
032a49f033e8 simplify vorbis windowing
lorenm
parents: 7232
diff changeset
72 const float *win, float add_bias, int len);
7218
7f3d6509628b Fix x86-64
michael
parents: 7203
diff changeset
73 void ff_float_to_int16_c(int16_t *dst, const float *src, long len);
7446
a4ebced5d4a2 r14205 broke vorbis float_to_int on ppc
lorenm
parents: 7327
diff changeset
74 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
75
34
2733a4c1c693 added block permutation functions
glantau
parents: 19
diff changeset
76 /* encoding scans */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
77 extern const uint8_t ff_alternate_horizontal_scan[64];
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
78 extern const uint8_t ff_alternate_vertical_scan[64];
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
79 extern const uint8_t ff_zigzag_direct[64];
1567
e08df4d22d27 * introducing dct248 into the DSP context.
romansh
parents: 1565
diff changeset
80 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
81
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
82 /* pixel operations */
2090
869805505b30 increase size of crop table
michael
parents: 2075
diff changeset
83 #define MAX_NEG_CROP 1024
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
84
986e461dc072 Initial revision
glantau
parents:
diff changeset
85 /* temporary */
4179
46d38e3d7038 rename squareTbl -> ff_squareTbl
mru
parents: 4176
diff changeset
86 extern uint32_t ff_squareTbl[512];
4176
23da44e8fd05 rename cropTbl -> ff_cropTbl
mru
parents: 3947
diff changeset
87 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
88
1866
1755f959ab7f seperated out the C-based VP3 DSP functions into a different file; also
melanson
parents: 1864
diff changeset
89 /* VP3 DSP functions */
2693
02925a3903b6 porting vp3 idct over to lavc idct api
michael
parents: 2633
diff changeset
90 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
02925a3903b6 porting vp3 idct over to lavc idct api
michael
parents: 2633
diff changeset
91 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
92 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
93
7995
1fbfce20cb79 Move VP3 loop filter to DSPContext
conrad
parents: 7760
diff changeset
94 void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
1fbfce20cb79 Move VP3 loop filter to DSPContext
conrad
parents: 7760
diff changeset
95 void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
1fbfce20cb79 Move VP3 loop filter to DSPContext
conrad
parents: 7760
diff changeset
96
8785
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
97 /* VP6 DSP functions */
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
98 void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, int stride,
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
99 const int16_t *h_weights, const int16_t *v_weights);
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
100
3245
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
101 /* 1/2^n downscaling functions from imgconvert.c */
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
102 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
103 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
104 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
105 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
106
3248
7aa9f80e7954 mmx implementation of 3-point GMC. (5x faster than C)
lorenm
parents: 3245
diff changeset
107 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
108 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
109
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
110 /* minimum alignment rules ;)
6486
3f96dc62c4b9 misc spelling fixes
diego
parents: 6449
diff changeset
111 If you notice errors in the align stuff, need more alignment for some ASM code
3f96dc62c4b9 misc spelling fixes
diego
parents: 6449
diff changeset
112 for some CPU or need to use a function with less aligned data then send a mail
3f96dc62c4b9 misc spelling fixes
diego
parents: 6449
diff changeset
113 to the ffmpeg-devel mailing list, ...
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
114
6486
3f96dc62c4b9 misc spelling fixes
diego
parents: 6449
diff changeset
115 !warning These alignments might not match reality, (missing attribute((align))
3f96dc62c4b9 misc spelling fixes
diego
parents: 6449
diff changeset
116 stuff somewhere possible).
6488
ecbe9565dcae typo fixes
diego
parents: 6486
diff changeset
117 I (Michael) did not check them, these are just the alignments which I think
6486
3f96dc62c4b9 misc spelling fixes
diego
parents: 6449
diff changeset
118 could be reached easily ...
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
119
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
120 !future video codecs might need functions with less strict alignment
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
121 */
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
122
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
123 /*
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
124 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
125 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
126 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
127 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
128 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
129 */
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
130
986e461dc072 Initial revision
glantau
parents:
diff changeset
131 /* add and put pixel (decoding) */
675
9008cbe229ce alignment info
michaelni
parents: 651
diff changeset
132 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
1709
39a209b0c82c info about h and tpel
michael
parents: 1708
diff changeset
133 //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
134 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
135 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
136 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
137 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
138 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
139 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
140
984
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
141 #define DEF_OLD_QPEL(name)\
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
142 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
143 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
144 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
145
984
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
146 DEF_OLD_QPEL(qpel16_mc11_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
147 DEF_OLD_QPEL(qpel16_mc31_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
148 DEF_OLD_QPEL(qpel16_mc12_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
149 DEF_OLD_QPEL(qpel16_mc32_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
150 DEF_OLD_QPEL(qpel16_mc13_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
151 DEF_OLD_QPEL(qpel16_mc33_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
152 DEF_OLD_QPEL(qpel8_mc11_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
153 DEF_OLD_QPEL(qpel8_mc31_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
154 DEF_OLD_QPEL(qpel8_mc12_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
155 DEF_OLD_QPEL(qpel8_mc32_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
156 DEF_OLD_QPEL(qpel8_mc13_old_c)
e162c09efbe7 qpel fix
michaelni
parents: 975
diff changeset
157 DEF_OLD_QPEL(qpel8_mc33_old_c)
651
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
158
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
159 #define CALL_2X_PIXELS(a, b, n)\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
160 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
161 b(block , pixels , line_size, h);\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
162 b(block+n, pixels+n, line_size, h);\
45e8f39fda50 put/avg_pixels16
michaelni
parents: 638
diff changeset
163 }
255
db20b987c32d divx5-gmc support
michaelni
parents: 252
diff changeset
164
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
165 /* motion estimation */
1709
39a209b0c82c info about h and tpel
michael
parents: 1708
diff changeset
166 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
5127
4dbe6578f811 misc spelling fixes
diego
parents: 4999
diff changeset
167 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
168 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
169
1168
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
170
3198
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
171 // for snow slices
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
172 typedef struct slice_buffer_s slice_buffer;
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
173
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
174 /**
6438
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
175 * Scantable.
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
176 */
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
177 typedef struct ScanTable{
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
178 const uint8_t *scantable;
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
179 uint8_t permutated[64];
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
180 uint8_t raster_end[64];
8590
7a463923ecd1 Change semantic of CONFIG_*, HAVE_* and ARCH_*.
aurel
parents: 8573
diff changeset
181 #if ARCH_PPC
6438
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
182 /** Used by dct_quantize_altivec to find last-non-zero */
6440
b6d05f8827cd Fix compilation on powerpc.
aurel
parents: 6438
diff changeset
183 DECLARE_ALIGNED(16, uint8_t, inverse[64]);
6438
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
184 #endif
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
185 } ScanTable;
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
186
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
187 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
188
6445
2b553c57ec51 move ff_emulated_edge_mc() to dsputil
aurel
parents: 6440
diff changeset
189 void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize,
2b553c57ec51 move ff_emulated_edge_mc() to dsputil
aurel
parents: 6440
diff changeset
190 int block_w, int block_h,
2b553c57ec51 move ff_emulated_edge_mc() to dsputil
aurel
parents: 6440
diff changeset
191 int src_x, int src_y, int w, int h);
2b553c57ec51 move ff_emulated_edge_mc() to dsputil
aurel
parents: 6440
diff changeset
192
6438
e1dd408a7864 move ff_init_scantable() into dsputil
aurel
parents: 6437
diff changeset
193 /**
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
194 * DSPContext.
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
195 */
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
196 typedef struct DSPContext {
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
197 /* pixel ops : interface with DCT */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
198 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
199 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
200 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
201 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
202 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
203 void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
01ed2c3b8f08 H.264 lossless mode
lorenm
parents: 2755
diff changeset
204 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
205 int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
206 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
207 * translational global motion compensation.
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
208 */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
209 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
210 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
211 * global motion compensation.
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
212 */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
213 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
214 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
8288
800444234375 clear_block mmx
lorenm
parents: 8233
diff changeset
215 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
216 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
217 int (*pix_sum)(uint8_t * pix, int line_size);
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
218 int (*pix_norm1)(uint8_t * pix, int line_size);
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
219 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
220
8976
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
221 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
222 me_cmp_func sse[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
223 me_cmp_func hadamard8_diff[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
224 me_cmp_func dct_sad[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
225 me_cmp_func quant_psnr[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
226 me_cmp_func bit[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
227 me_cmp_func rd[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
228 me_cmp_func vsad[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
229 me_cmp_func vsse[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
230 me_cmp_func nsse[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
231 me_cmp_func w53[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
232 me_cmp_func w97[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
233 me_cmp_func dct_max[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
234 me_cmp_func dct264_sad[6];
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
235
8976
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
236 me_cmp_func me_pre_cmp[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
237 me_cmp_func me_cmp[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
238 me_cmp_func me_sub_cmp[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
239 me_cmp_func mb_cmp[6];
e7d87561b42b Making the arrays accomodate an extra intra 8x8 cmp function
romansh
parents: 8785
diff changeset
240 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
241 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
242
5255
669a97223dc7 make arguments to ssd_int8_vs_int16() const
mru
parents: 5149
diff changeset
243 int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
669a97223dc7 make arguments to ssd_int8_vs_int16() const
mru
parents: 5149
diff changeset
244 int size);
4749
7011f597e473 mmx 16-bit ssd. 2.3x faster svq1 encoding.
lorenm
parents: 4436
diff changeset
245
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
246 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
247 * Halfpel motion compensation with rounding (a+b+1)>>1.
4751
0d48b61e655d Typos in comments
lu_zero
parents: 4749
diff changeset
248 * this is an array[4][4] of motion compensation functions for 4
1713
49e46374baa4 put_pixels width comment fix
michael
parents: 1709
diff changeset
249 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
250 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
251 * @param block destination where the result is stored
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
252 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
253 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
254 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
255 */
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
256 op_pixels_func put_pixels_tab[4][4];
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
257
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
258 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
259 * Halfpel motion compensation with rounding (a+b+1)>>1.
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
260 * This is an array[4][4] of motion compensation functions for 4
1713
49e46374baa4 put_pixels width comment fix
michael
parents: 1709
diff changeset
261 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
262 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
263 * @param block destination into which the result is averaged (a+b+1)>>1
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
264 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
265 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
266 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
267 */
1319
449f6e32b425 added support for B-frames and multiple slices
tmmm
parents: 1273
diff changeset
268 op_pixels_func avg_pixels_tab[4][4];
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
269
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
270 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
271 * Halfpel motion compensation with no rounding (a+b)>>1.
4751
0d48b61e655d Typos in comments
lu_zero
parents: 4749
diff changeset
272 * this is an array[2][4] of motion compensation functions for 2
1225
michaelni
parents: 1213
diff changeset
273 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
274 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
275 * @param block destination where the result is stored
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
276 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
277 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
278 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
279 */
2075
aae3b4e66560 chroma me fix
michael
parents: 2065
diff changeset
280 op_pixels_func put_no_rnd_pixels_tab[4][4];
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
281
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
282 /**
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
283 * Halfpel motion compensation with no rounding (a+b)>>1.
4751
0d48b61e655d Typos in comments
lu_zero
parents: 4749
diff changeset
284 * this is an array[2][4] of motion compensation functions for 2
1225
michaelni
parents: 1213
diff changeset
285 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
1213
a71fe70b428b doc fixes (forgot to commit)
michaelni
parents: 1201
diff changeset
286 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
1136
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
287 * @param block destination into which the result is averaged (a+b)>>1
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
288 * @param pixels source
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
289 * @param line_size number of bytes in a horizontal line of block
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
290 * @param h height
9eb49e9e8dd9 halfpel MC doxy
michaelni
parents: 1111
diff changeset
291 */
2075
aae3b4e66560 chroma me fix
michael
parents: 2065
diff changeset
292 op_pixels_func avg_no_rnd_pixels_tab[4][4];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
293
1864
9a73e6dc5d68 fix motion compensation with (x+1/2,y+1/2) MVs
michael
parents: 1784
diff changeset
294 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
295
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
296 /**
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
297 * Thirdpel motion compensation with rounding (a+b+1)>>1.
4751
0d48b61e655d Typos in comments
lu_zero
parents: 4749
diff changeset
298 * this is an array[12] of motion compensation functions for the 9 thirdpe
0d48b61e655d Typos in comments
lu_zero
parents: 4749
diff changeset
299 * 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
300 * *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
301 * @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
302 * @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
303 * @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
304 * @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
305 */
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
306 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
307 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
308
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
309 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
310 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
311 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
312 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
313 qpel_mc_func put_mspel_pixels_tab[8];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
314
1168
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
315 /**
4751
0d48b61e655d Typos in comments
lu_zero
parents: 4749
diff changeset
316 * h264 Chroma MC
1168
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
317 */
5af9aeadbdc3 H264 decoder & demuxer
michaelni
parents: 1136
diff changeset
318 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
319 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
3663
8e180a3276fe New function for chroma MC (will be used in VC-1)
kostya
parents: 3660
diff changeset
320 /* 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
321 h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3];
9440
daee921fb6bb VC1: add and use avg_no_rnd chroma MC functions
conrad
parents: 9439
diff changeset
322 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
323
3020
c75fb0747e74 use h264 MC functions for 2xX Xx2 blocks in snow too
michael
parents: 3010
diff changeset
324 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
325 qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
326
3807
6a40092eb9e6 approximate qpel functions: sacrifice some quality for some decoding speed. enabled on B-frames with -lavdopts fast.
lorenm
parents: 3746
diff changeset
327 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
328 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
329
2415
db2cf6005d19 H.264 weighted prediction.
lorenm
parents: 2388
diff changeset
330 h264_weight_func weight_h264_pixels_tab[10];
db2cf6005d19 H.264 weighted prediction.
lorenm
parents: 2388
diff changeset
331 h264_biweight_func biweight_h264_pixels_tab[10];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
332
3395
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
333 /* AVS specific */
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
334 qpel_mc_func put_cavs_qpel_pixels_tab[2][16];
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
335 qpel_mc_func avg_cavs_qpel_pixels_tab[2][16];
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
336 void (*cavs_filter_lv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
337 void (*cavs_filter_lh)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
338 void (*cavs_filter_cv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
339 void (*cavs_filter_ch)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
340 void (*cavs_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
adccbf4a1040 CAVS decoder by (Stefan Gehrer stefan.gehrer gmx.de)
michael
parents: 3279
diff changeset
341
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
342 me_cmp_func pix_abs[2][4];
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
343
866
725ef4ea3ecc huffyuv
michaelni
parents: 862
diff changeset
344 /* huffyuv specific */
725ef4ea3ecc huffyuv
michaelni
parents: 862
diff changeset
345 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
6384
0a403ade8c81 simd and unroll png_filter_row
lorenm
parents: 6363
diff changeset
346 void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
347 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
348 /**
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
349 * subtract huffyuv's variant of median prediction
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
350 * note, this might read from src1[-1], src2[-1]
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
351 */
8ffd0c00e6df mmx2 optimization of huffyuv median encoding
michael
parents: 1446
diff changeset
352 void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
8760
31138c296ac6 ff_add_hfyu_median_prediction_mmx2
lorenm
parents: 8718
diff changeset
353 void (*add_hfyu_median_prediction)(uint8_t *dst, uint8_t *top, uint8_t *diff, int w, int *left, int *left_top);
6384
0a403ade8c81 simd and unroll png_filter_row
lorenm
parents: 6363
diff changeset
354 /* this might write to dst[w] */
0a403ade8c81 simd and unroll png_filter_row
lorenm
parents: 6363
diff changeset
355 void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
6241
2799f65a24de const src for bswap
michael
parents: 6142
diff changeset
356 void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
2633
72e6ffa1f3a5 MMX for H.264 deblocking filter
lorenm
parents: 2415
diff changeset
357
7579
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
358 void (*h264_v_loop_filter_luma)(uint8_t *pix/*align 16*/, int stride, int alpha, int beta, int8_t *tc0);
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
359 void (*h264_h_loop_filter_luma)(uint8_t *pix/*align 4 */, int stride, int alpha, int beta, int8_t *tc0);
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
360 /* v/h_loop_filter_luma_intra: align 16 */
8395
195cba8f6257 Move filter_luma_intra into dsputil for later addition of asm.
darkshikari
parents: 8375
diff changeset
361 void (*h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
195cba8f6257 Move filter_luma_intra into dsputil for later addition of asm.
darkshikari
parents: 8375
diff changeset
362 void (*h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
7579
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
363 void (*h264_v_loop_filter_chroma)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta, int8_t *tc0);
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
364 void (*h264_h_loop_filter_chroma)(uint8_t *pix/*align 4*/, int stride, int alpha, int beta, int8_t *tc0);
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
365 void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
30bc7ba91daa document some dsp alignments
lorenm
parents: 7573
diff changeset
366 void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
3645
47821be55b6c mmx implementation of deblocking strength decision.
lorenm
parents: 3574
diff changeset
367 // h264_loop_filter_strength: simd only. the C version is inlined in h264.c
47821be55b6c mmx implementation of deblocking strength decision.
lorenm
parents: 3574
diff changeset
368 void (*h264_loop_filter_strength)(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],
7327
483421b11d98 Fix h264_loop_filter_strength_mmx2() so it works with PAFF.
michael
parents: 7286
diff changeset
369 int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
370
1644
835cf346975e h263 loop filter
michael
parents: 1603
diff changeset
371 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
835cf346975e h263 loop filter
michael
parents: 1603
diff changeset
372 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
835cf346975e h263 loop filter
michael
parents: 1603
diff changeset
373
2045
9447bbd8a7e9 rewrite h261 loop filter
michael
parents: 2044
diff changeset
374 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
375
5887
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
376 void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
377 void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
378
7995
1fbfce20cb79 Move VP3 loop filter to DSPContext
conrad
parents: 7760
diff changeset
379 void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
1fbfce20cb79 Move VP3 loop filter to DSPContext
conrad
parents: 7760
diff changeset
380 void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
1fbfce20cb79 Move VP3 loop filter to DSPContext
conrad
parents: 7760
diff changeset
381
8785
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
382 void (*vp6_filter_diag4)(uint8_t *dst, uint8_t *src, int stride,
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
383 const int16_t *h_weights,const int16_t *v_weights);
bee83b3f9a6b move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
aurel
parents: 8760
diff changeset
384
3568
945caa35ee9a sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents: 3555
diff changeset
385 /* 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
386 void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
7563
8390efaa0c03 simd downmix
lorenm
parents: 7547
diff changeset
387 void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
5737
efa3c1f9259a sse2 version of compute_autocorr().
lorenm
parents: 5728
diff changeset
388 /* no alignment needed */
efa3c1f9259a sse2 version of compute_autocorr().
lorenm
parents: 5728
diff changeset
389 void (*flac_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc);
3574
f549d1e685f7 vorbis simd tweaks
lorenm
parents: 3568
diff changeset
390 /* 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
391 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
392 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
393 /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
945caa35ee9a sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents: 3555
diff changeset
394 void (*vector_fmul_add_add)(float *dst, const float *src0, const float *src1, const float *src2, int src3, int len, int step);
7261
032a49f033e8 simplify vorbis windowing
lorenm
parents: 7232
diff changeset
395 /* assume len is a multiple of 4, and arrays are 16-byte aligned */
032a49f033e8 simplify vorbis windowing
lorenm
parents: 7232
diff changeset
396 void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len);
7564
7cf793954871 simd int->float
lorenm
parents: 7563
diff changeset
397 /* assume len is a multiple of 8, and arrays are 16-byte aligned */
7cf793954871 simd int->float
lorenm
parents: 7563
diff changeset
398 void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);
3568
945caa35ee9a sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents: 3555
diff changeset
399
945caa35ee9a sse and 3dnow implementations of float->int conversion and mdct windowing.
lorenm
parents: 3555
diff changeset
400 /* 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
401 * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
7218
7f3d6509628b Fix x86-64
michael
parents: 7203
diff changeset
402 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
403 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
404
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
405 /* (I)DCT */
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
406 void (*fdct)(DCTELEM *block/* align 16*/);
1567
e08df4d22d27 * introducing dct248 into the DSP context.
romansh
parents: 1565
diff changeset
407 void (*fdct248)(DCTELEM *block/* align 16*/);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
408
1324
7d328fd9d8a5 the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
michaelni
parents: 1320
diff changeset
409 /* IDCT really*/
7d328fd9d8a5 the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
michaelni
parents: 1320
diff changeset
410 void (*idct)(DCTELEM *block/* align 16*/);
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
411
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
412 /**
1104
6a01ffc5351c cosmetics
michaelni
parents: 1102
diff changeset
413 * block -> idct -> clip to unsigned 8 bit -> dest.
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
414 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
5728
d7970c9e3049 cosmetics: typos
takis
parents: 5703
diff changeset
415 * @param line_size size in bytes of a horizontal line of dest
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
416 */
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
417 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
418
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
419 /**
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
420 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
5728
d7970c9e3049 cosmetics: typos
takis
parents: 5703
diff changeset
421 * @param line_size size in bytes of a horizontal line of dest
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
422 */
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
423 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
424
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
425 /**
1104
6a01ffc5351c cosmetics
michaelni
parents: 1102
diff changeset
426 * idct input permutation.
1241
6a873cad413e more idct_permute docs
michaelni
parents: 1225
diff changeset
427 * 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
428 * IDCT)
6a873cad413e more idct_permute docs
michaelni
parents: 1225
diff changeset
429 * 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
430 * with the zigzag/alternate scan<br>
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
431 * an example to avoid confusion:
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
432 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
433 * - (x -> referece dct -> reference idct -> x)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
434 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
435 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
436 */
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
437 uint8_t idct_permutation[64];
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
438 int idct_permutation_type;
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
439 #define FF_NO_IDCT_PERM 1
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
440 #define FF_LIBMPEG2_IDCT_PERM 2
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
441 #define FF_SIMPLE_IDCT_PERM 3
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
442 #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
443 #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
444 #define FF_SSE2_IDCT_PERM 6
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
445
1784
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
446 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
447 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
448 #define BASIS_SHIFT 16
65f7bd09f37b quantizer noise shaping optimization
michael
parents: 1765
diff changeset
449 #define RECON_SHIFT 6
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 2887
diff changeset
450
6437
5154ab444372 move draw_edges() into dsputil
aurel
parents: 6391
diff changeset
451 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
452 #define EDGE_WIDTH 16
6437
5154ab444372 move draw_edges() into dsputil
aurel
parents: 6391
diff changeset
453
4268
3a4d4be05618 Add comment denoting group of H.264 functions
takis
parents: 4263
diff changeset
454 /* h264 functions */
8403
c9918d072bdb Document new idct requirements.
michael
parents: 8395
diff changeset
455 /* NOTE!!! if you implement any of h264_idct8_add, h264_idct8_add4 then you must implement all of them
c9918d072bdb Document new idct requirements.
michael
parents: 8395
diff changeset
456 NOTE!!! if you implement any of h264_idct_add, h264_idct_add16, h264_idct_add16intra, h264_idct_add8 then you must implement all of them
c9918d072bdb Document new idct requirements.
michael
parents: 8395
diff changeset
457 The reason for above, is that no 2 out of one list may use a different permutation.
c9918d072bdb Document new idct requirements.
michael
parents: 8395
diff changeset
458 */
7678
081c54b62e56 document some dsp alignments
lorenm
parents: 7579
diff changeset
459 void (*h264_idct_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
081c54b62e56 document some dsp alignments
lorenm
parents: 7579
diff changeset
460 void (*h264_idct8_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
081c54b62e56 document some dsp alignments
lorenm
parents: 7579
diff changeset
461 void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
081c54b62e56 document some dsp alignments
lorenm
parents: 7579
diff changeset
462 void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
4279
1f84e4f18aa1 Added entry for H.264 DCT to the DSP context.
takis
parents: 4268
diff changeset
463 void (*h264_dct)(DCTELEM block[4][4]);
8375
de2509cf3c44 H.264 idct functions that include the chroma, inter luma and intra16 luma loops
michael
parents: 8359
diff changeset
464 void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, 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
465 void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, 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
466 void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, 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
467 void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
3198
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
468
6b9f0c4fbdbe First part of a series of speed-enchancing patches.
gpoirier
parents: 3175
diff changeset
469 /* snow wavelet */
5587
3ae03eacbe9f use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents: 5573
diff changeset
470 void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
3ae03eacbe9f use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents: 5573
diff changeset
471 void (*horizontal_compose97i)(IDWTELEM *b, int width);
4436
d3e389536b0a Add the const specifier as needed to reduce the number of warnings.
takis
parents: 4311
diff changeset
472 void (*inner_add_yblock)(const 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
473
06f98047ff26 prefetch pixels for future motion compensation. 2-5% faster h264.
lorenm
parents: 3198
diff changeset
474 void (*prefetch)(void *mem, int stride, int h);
3245
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
475
c2c29be6282e remove mpegvideo.c img resample dependancy
michael
parents: 3217
diff changeset
476 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
3526
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
477
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
478 /* vc1 functions */
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
479 void (*vc1_inv_trans_8x8)(DCTELEM *b);
5997
90de28dfd8d6 Switch VC-1 decoder to output decoded residual immediately.
kostya
parents: 5907
diff changeset
480 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
481 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
482 void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
4239
30b14d0f2324 Correct rounding values in overlap filtering
kostya
parents: 4197
diff changeset
483 void (*vc1_v_overlap)(uint8_t* src, int stride);
30b14d0f2324 Correct rounding values in overlap filtering
kostya
parents: 4197
diff changeset
484 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
485 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
486 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
487 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
488 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
489 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
490 void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq);
3526
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
491 /* put 8x8 block with bicubic interpolation and quarterpel precision
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
492 * last argument is actually round value instead of height
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
493 */
7dc8e4a12105 New functions in DSPContext for VC-1 decoding
kostya
parents: 3395
diff changeset
494 op_pixels_func put_vc1_mspel_pixels_tab[16];
9437
8aa2e86549cd VC1: Do qpel when needed for both MVs in a B frame
conrad
parents: 9436
diff changeset
495 op_pixels_func avg_vc1_mspel_pixels_tab[16];
5887
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
496
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
497 /* intrax8 functions */
5907
fbd10e6dfbe1 cosmetics: Fix spacial --> spatial typo in function names.
diego
parents: 5887
diff changeset
498 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
499 void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
5887
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
500 int * range, int * sum, int edges);
83ac4620c6ed intrax8 decoder patch by "someone"
michael
parents: 5830
diff changeset
501
7203
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
502 /* ape functions */
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
503 /**
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
504 * Add contents of the second vector to the first one.
7232
cc55dd004819 Monkey's Audio decoder vector functions work on input with length
kostya
parents: 7218
diff changeset
505 * @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
506 */
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
507 void (*add_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
508 /**
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
509 * Add contents of the second vector to the first one.
7232
cc55dd004819 Monkey's Audio decoder vector functions work on input with length
kostya
parents: 7218
diff changeset
510 * @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
511 */
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
512 void (*sub_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
513 /**
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
514 * Calculate scalar product of two vectors.
7232
cc55dd004819 Monkey's Audio decoder vector functions work on input with length
kostya
parents: 7218
diff changeset
515 * @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
516 * @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
517 */
87b1dfb5a98d Add several vector functions used by Monkey's Audio decoder to dsputil
kostya
parents: 7135
diff changeset
518 int32_t (*scalarproduct_int16)(int16_t *v1, int16_t *v2/*align 16*/, int len, int shift);
8232
520a6e56b8c0 RV40 MC functions
kostya
parents: 8104
diff changeset
519
8233
ababfa151ced enable RV40 decoder
kostya
parents: 8232
diff changeset
520 /* rv30 functions */
ababfa151ced enable RV40 decoder
kostya
parents: 8232
diff changeset
521 qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
ababfa151ced enable RV40 decoder
kostya
parents: 8232
diff changeset
522 qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];
ababfa151ced enable RV40 decoder
kostya
parents: 8232
diff changeset
523
8232
520a6e56b8c0 RV40 MC functions
kostya
parents: 8104
diff changeset
524 /* rv40 functions */
520a6e56b8c0 RV40 MC functions
kostya
parents: 8104
diff changeset
525 qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
520a6e56b8c0 RV40 MC functions
kostya
parents: 8104
diff changeset
526 qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
520a6e56b8c0 RV40 MC functions
kostya
parents: 8104
diff changeset
527 h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
520a6e56b8c0 RV40 MC functions
kostya
parents: 8104
diff changeset
528 h264_chroma_mc_func avg_rv40_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
529 } DSPContext;
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
530
4197
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
531 void dsputil_static_init(void);
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
532 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
533
4281
de525a2b41db ff_check_alignment to warn the user about a missaligned stack
michael
parents: 4279
diff changeset
534 int ff_check_alignment(void);
de525a2b41db ff_check_alignment to warn the user about a missaligned stack
michael
parents: 4279
diff changeset
535
764
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
536 /**
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
537 * permute block according to permuatation.
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
538 * @param last last non zero element in scantable order
d4726182dfd2 optimize block_permute()
michaelni
parents: 753
diff changeset
539 */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
540 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
541
1729
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
542 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
a4a5e7521339 interlaced dct decision cleanup
michael
parents: 1713
diff changeset
543
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
544 #define BYTE_VEC32(c) ((c)*0x01010101UL)
1264
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
545
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
546 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
547 {
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
548 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
549 }
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
550
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
551 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
552 {
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
553 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
554 }
2fa34e615c76 cleanup
michaelni
parents: 1259
diff changeset
555
2184
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
556 static inline int get_penalty_factor(int lambda, int lambda2, int type){
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
557 switch(type&0xFF){
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
558 default:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
559 case FF_CMP_SAD:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
560 return lambda>>FF_LAMBDA_SHIFT;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
561 case FF_CMP_DCT:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
562 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
563 case FF_CMP_W53:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
564 return (4*lambda)>>(FF_LAMBDA_SHIFT);
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
565 case FF_CMP_W97:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
566 return (2*lambda)>>(FF_LAMBDA_SHIFT);
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
567 case FF_CMP_SATD:
3010
533c6386eca9 8x8 integer dct from x264 as cmp function (under CONFIG_GPL)
michael
parents: 2979
diff changeset
568 case FF_CMP_DCT264:
2184
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
569 return (2*lambda)>>FF_LAMBDA_SHIFT;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
570 case FF_CMP_RD:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
571 case FF_CMP_PSNR:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
572 case FF_CMP_SSE:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
573 case FF_CMP_NSSE:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
574 return lambda2>>FF_LAMBDA_SHIFT;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
575 case FF_CMP_BIT:
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
576 return 1;
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
577 }
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
578 }
3378d0677903 4x4 SSE compare function
michael
parents: 2090
diff changeset
579
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
580 /**
1104
6a01ffc5351c cosmetics
michaelni
parents: 1102
diff changeset
581 * Empty mmx state.
1102
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
582 * this must be called between any dsp function and float/double code.
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
583 * for example sin(); dsp->idct_put(); emms_c(); cos()
eeaa1e837efb more dox
michaelni
parents: 1092
diff changeset
584 */
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
585 #define emms_c()
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
586
995
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
587 /* should be defined by architectures supporting
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
588 one or more MultiMedia extension */
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
589 int mm_support(void);
edc10966b081 altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents: 984
diff changeset
590
5149
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
591 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
8359
9281a8a9387a ARM: replace "armv4l" with "arm"
mru
parents: 8288
diff changeset
592 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
5149
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
593 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
594 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
595 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
596 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
597 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
598 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
599 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
b7bae4ce8b07 move dsputil_init_foo() out from #ifdef
mru
parents: 5127
diff changeset
600
4999
0c4bf6b7d1c6 use DECLARE_ALIGNED from mem.h (which is included through common.h)
alex
parents: 4988
diff changeset
601 #define DECLARE_ALIGNED_16(t, v) DECLARE_ALIGNED(16, t, v)
1974
8c5489b2cf3e move __align16 some place where non-MMX machines can see it
melanson
parents: 1972
diff changeset
602
8590
7a463923ecd1 Change semantic of CONFIG_*, HAVE_* and ARCH_*.
aurel
parents: 8573
diff changeset
603 #if HAVE_MMX
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
604
862
058194d7ade6 * fixing some minor const warnings
kabi
parents: 853
diff changeset
605 #undef emms_c
853
eacc2dd8fd9d * using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents: 838
diff changeset
606
4197
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
607 extern int mm_flags;
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
608
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
609 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
610 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
611 void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
612
986e461dc072 Initial revision
glantau
parents:
diff changeset
613 static inline void emms(void)
986e461dc072 Initial revision
glantau
parents:
diff changeset
614 {
8031
eebc7209c47f Convert asm keyword into __asm__.
flameeyes
parents: 7995
diff changeset
615 __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
616 }
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
617
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
618
6
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
619 #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
620 {\
8104
0d108ec85620 Remove duplicated MM_* macros for CPU capabilities from dsputil.h.
rathann
parents: 8031
diff changeset
621 if (mm_flags & FF_MM_MMX)\
6
ec4642daa6fe added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents: 2
diff changeset
622 emms();\
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
623 }
986e461dc072 Initial revision
glantau
parents:
diff changeset
624
1092
f59c3f66363b MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
michaelni
parents: 1065
diff changeset
625 void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
1065
92fb44eae6b6 * prototypes at one place
kabi
parents: 1064
diff changeset
626
8590
7a463923ecd1 Change semantic of CONFIG_*, HAVE_* and ARCH_*.
aurel
parents: 8573
diff changeset
627 #elif ARCH_ARM
62
4bfc845cdfea arm optimizations
glantau
parents: 50
diff changeset
628
4197
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
629 extern int mm_flags;
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
630
8590
7a463923ecd1 Change semantic of CONFIG_*, HAVE_* and ARCH_*.
aurel
parents: 8573
diff changeset
631 #if HAVE_NEON
7687
e5b5a9af1b68 ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents: 7678
diff changeset
632 # define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
e5b5a9af1b68 ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents: 7678
diff changeset
633 # define STRIDE_ALIGN 16
e5b5a9af1b68 ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents: 7678
diff changeset
634 #endif
e5b5a9af1b68 ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
mru
parents: 7678
diff changeset
635
8590
7a463923ecd1 Change semantic of CONFIG_*, HAVE_* and ARCH_*.
aurel
parents: 8573
diff changeset
636 #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
637
4197
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
638 extern int mm_flags;
bbe0bc387a19 revert bad checkin
mru
parents: 4196
diff changeset
639
4999
0c4bf6b7d1c6 use DECLARE_ALIGNED from mem.h (which is included through common.h)
alex
parents: 4988
diff changeset
640 #define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
641 #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
642
8590
7a463923ecd1 Change semantic of CONFIG_*, HAVE_* and ARCH_*.
aurel
parents: 8573
diff changeset
643 #elif HAVE_MMI
689
efcbfbd18864 ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents: 687
diff changeset
644
4999
0c4bf6b7d1c6 use DECLARE_ALIGNED from mem.h (which is included through common.h)
alex
parents: 4988
diff changeset
645 #define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
2324
ba91c976c7aa stride align cleanup
michael
parents: 2272
diff changeset
646 #define STRIDE_ALIGN 16
689
efcbfbd18864 ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents: 687
diff changeset
647
7135
6bd6a2da306e Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents: 7094
diff changeset
648 #else
6bd6a2da306e Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents: 7094
diff changeset
649
6bd6a2da306e Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents: 7094
diff changeset
650 #define mm_flags 0
6bd6a2da306e Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents: 7094
diff changeset
651 #define mm_support() 0
6bd6a2da306e Define mm_flags/support to be 0 on architectures where they don't exist.
astrange
parents: 7094
diff changeset
652
6363
7ebd1cdb2142 clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents: 6241
diff changeset
653 #endif
3728
4d346abd723c ADI Blackfin optimizations
diego
parents: 3665
diff changeset
654
6363
7ebd1cdb2142 clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents: 6241
diff changeset
655 #ifndef DECLARE_ALIGNED_8
7ebd1cdb2142 clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents: 6241
diff changeset
656 # define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(8, t, v)
7ebd1cdb2142 clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents: 6241
diff changeset
657 #endif
3728
4d346abd723c ADI Blackfin optimizations
diego
parents: 3665
diff changeset
658
6363
7ebd1cdb2142 clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents: 6241
diff changeset
659 #ifndef STRIDE_ALIGN
7ebd1cdb2142 clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
mru
parents: 6241
diff changeset
660 # define STRIDE_ALIGN 8
0
986e461dc072 Initial revision
glantau
parents:
diff changeset
661 #endif
986e461dc072 Initial revision
glantau
parents:
diff changeset
662
252
ddb1a0e94cf4 - Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents: 214
diff changeset
663 /* PSNR */
1064
b32afefe7d33 * UINTX -> uintx_t INTX -> intx_t
kabi
parents: 1040
diff changeset
664 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
665 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
666 AVCodecContext *avctx);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
667
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
668 /* FFT computation */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
669
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
670 /* 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
671 FFTSample type */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
672 typedef float FFTSample;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
673
3555
5ea82888103e 3dnow2 implementation of imdct.
lorenm
parents: 3536
diff changeset
674 struct MDCTContext;
5ea82888103e 3dnow2 implementation of imdct.
lorenm
parents: 3536
diff changeset
675
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
676 typedef struct FFTComplex {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
677 FFTSample re, im;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
678 } FFTComplex;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
679
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
680 typedef struct FFTContext {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
681 int nbits;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
682 int inverse;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
683 uint16_t *revtab;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
684 FFTComplex *exptab;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
685 FFTComplex *exptab1; /* only used by SSE code */
7542
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
686 FFTComplex *tmp_buf;
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
687 void (*fft_permute)(struct FFTContext *s, FFTComplex *z);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
688 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
7546
97383e012cb9 remove mdct tmp buffer
lorenm
parents: 7544
diff changeset
689 void (*imdct_calc)(struct MDCTContext *s, FFTSample *output, const FFTSample *input);
97383e012cb9 remove mdct tmp buffer
lorenm
parents: 7544
diff changeset
690 void (*imdct_half)(struct MDCTContext *s, FFTSample *output, const FFTSample *input);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
691 } FFTContext;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
692
8694
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
693 extern FFTSample* ff_cos_tabs[13];
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
694
8636
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
695 /**
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
696 * Sets up a complex FFT.
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
697 * @param nbits log2 of the length of the input array
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
698 * @param inverse if 0 perform the forward transform, if 1 perform the inverse
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
699 */
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
700 int ff_fft_init(FFTContext *s, int nbits, int inverse);
7542
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
701 void ff_fft_permute_c(FFTContext *s, FFTComplex *z);
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
702 void ff_fft_permute_sse(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
703 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
704 void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
3175
c20c181e0eca 3DNow! & Extended 3DNow! versions of FFT
corey
parents: 3105
diff changeset
705 void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z);
c20c181e0eca 3DNow! & Extended 3DNow! versions of FFT
corey
parents: 3105
diff changeset
706 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
707 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
708
8636
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
709 /**
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
710 * Do the permutation needed BEFORE calling ff_fft_calc().
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
711 */
7542
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
712 static inline void ff_fft_permute(FFTContext *s, FFTComplex *z)
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
713 {
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
714 s->fft_permute(s, z);
a8a8205a9081 split-radix FFT
lorenm
parents: 7446
diff changeset
715 }
8636
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
716 /**
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
717 * Do a complex FFT with the parameters defined in ff_fft_init(). The
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
718 * input data must be permuted before. No 1.0/sqrt(n) normalization is done.
9766c268bc9f Clean up FFT related comments.
alexc
parents: 8629
diff changeset
719 */
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
720 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
721 {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
722 s->fft_calc(s, z);
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
723 }
1879
dd63cb7e5080 fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
michael
parents: 1866
diff changeset
724 void ff_fft_end(FFTContext *s);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
725
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
726 /* MDCT computation */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
727
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
728 typedef struct MDCTContext {
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
729 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
730 int nbits; /* n = 2^nbits */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
731 /* pre/post rotation tables */
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
732 FFTSample *tcos;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
733 FFTSample *tsin;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
734 FFTContext fft;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
735 } MDCTContext;
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
736
7547
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
737 static inline void ff_imdct_calc(MDCTContext *s, FFTSample *output, const FFTSample *input)
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
738 {
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
739 s->fft.imdct_calc(s, output, input);
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
740 }
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
741 static inline void ff_imdct_half(MDCTContext *s, FFTSample *output, const FFTSample *input)
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
742 {
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
743 s->fft.imdct_half(s, output, input);
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
744 }
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
745
6139
5077d1562573 Make the Kaiser-Bessel window generator a common function
andoma
parents: 6056
diff changeset
746 /**
5077d1562573 Make the Kaiser-Bessel window generator a common function
andoma
parents: 6056
diff changeset
747 * Generate a Kaiser-Bessel Derived Window.
5077d1562573 Make the Kaiser-Bessel window generator a common function
andoma
parents: 6056
diff changeset
748 * @param window pointer to half window
6142
a35b838ab955 Add variable alpha and size of half window for Kaiser-Bessel Derived window
superdump
parents: 6139
diff changeset
749 * @param alpha determines window shape
a35b838ab955 Add variable alpha and size of half window for Kaiser-Bessel Derived window
superdump
parents: 6139
diff changeset
750 * @param n size of half window
6139
5077d1562573 Make the Kaiser-Bessel window generator a common function
andoma
parents: 6056
diff changeset
751 */
6142
a35b838ab955 Add variable alpha and size of half window for Kaiser-Bessel Derived window
superdump
parents: 6139
diff changeset
752 void ff_kbd_window_init(float *window, float alpha, int n);
6139
5077d1562573 Make the Kaiser-Bessel window generator a common function
andoma
parents: 6056
diff changeset
753
7094
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
754 /**
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
755 * Generate a sine window.
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
756 * @param window pointer to half window
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
757 * @param n size of half window
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
758 */
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
759 void ff_sine_window_init(float *window, int n);
7573
7802295cae6f Add declarations for the sine tables used in wma.c (half window sizes: 128,
superdump
parents: 7564
diff changeset
760 extern float ff_sine_128 [ 128];
7802295cae6f Add declarations for the sine tables used in wma.c (half window sizes: 128,
superdump
parents: 7564
diff changeset
761 extern float ff_sine_256 [ 256];
7802295cae6f Add declarations for the sine tables used in wma.c (half window sizes: 128,
superdump
parents: 7564
diff changeset
762 extern float ff_sine_512 [ 512];
7802295cae6f Add declarations for the sine tables used in wma.c (half window sizes: 128,
superdump
parents: 7564
diff changeset
763 extern float ff_sine_1024[1024];
7802295cae6f Add declarations for the sine tables used in wma.c (half window sizes: 128,
superdump
parents: 7564
diff changeset
764 extern float ff_sine_2048[2048];
8567
0d5b2b0e7a87 Add size that is needed for the wmapro codec
banan
parents: 8496
diff changeset
765 extern float ff_sine_4096[4096];
0d5b2b0e7a87 Add size that is needed for the wmapro codec
banan
parents: 8496
diff changeset
766 extern float *ff_sine_windows[6];
7094
b0820b8bd4dd Add generic ff_sine_window_init function and implement in codecs appropriately
superdump
parents: 6600
diff changeset
767
794
670009af4fc2 avoid name clash
bellard
parents: 781
diff changeset
768 int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
7547
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
769 void ff_imdct_calc_c(MDCTContext *s, FFTSample *output, const FFTSample *input);
8226017a65ae mdct wrapper function to match fft
lorenm
parents: 7546
diff changeset
770 void ff_imdct_half_c(MDCTContext *s, FFTSample *output, const FFTSample *input);
7546
97383e012cb9 remove mdct tmp buffer
lorenm
parents: 7544
diff changeset
771 void ff_imdct_calc_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input);
7544
ee1cb5ab9f99 optimize imdct_half:
lorenm
parents: 7542
diff changeset
772 void ff_imdct_half_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input);
7546
97383e012cb9 remove mdct tmp buffer
lorenm
parents: 7544
diff changeset
773 void ff_imdct_calc_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input);
7544
ee1cb5ab9f99 optimize imdct_half:
lorenm
parents: 7542
diff changeset
774 void ff_imdct_half_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input);
7546
97383e012cb9 remove mdct tmp buffer
lorenm
parents: 7544
diff changeset
775 void ff_imdct_calc_sse(MDCTContext *s, FFTSample *output, const FFTSample *input);
7544
ee1cb5ab9f99 optimize imdct_half:
lorenm
parents: 7542
diff changeset
776 void ff_imdct_half_sse(MDCTContext *s, FFTSample *output, const FFTSample *input);
7546
97383e012cb9 remove mdct tmp buffer
lorenm
parents: 7544
diff changeset
777 void ff_mdct_calc(MDCTContext *s, FFTSample *out, const FFTSample *input);
794
670009af4fc2 avoid name clash
bellard
parents: 781
diff changeset
778 void ff_mdct_end(MDCTContext *s);
781
6f5e87957bcb new generic FFT/MDCT code for audio codecs
bellard
parents: 764
diff changeset
779
8694
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
780 /* Real Discrete Fourier Transform */
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
781
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
782 enum RDFTransformType {
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
783 RDFT,
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
784 IRDFT,
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
785 RIDFT,
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
786 IRIDFT,
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
787 };
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
788
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
789 typedef struct {
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
790 int nbits;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
791 int inverse;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
792 int sign_convention;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
793
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
794 /* pre/post rotation tables */
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
795 FFTSample *tcos;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
796 FFTSample *tsin;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
797 FFTContext fft;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
798 } RDFTContext;
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
799
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
800 /**
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
801 * Sets up a real FFT.
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
802 * @param nbits log2 of the length of the input array
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
803 * @param trans the type of transform
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
804 */
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
805 int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans);
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
806 void ff_rdft_calc(RDFTContext *s, FFTSample *data);
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
807 void ff_rdft_end(RDFTContext *s);
68fd157bab48 Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
alexc
parents: 8636
diff changeset
808
6056
558c1fd0ee72 Fix typo in macro name: WARPER8_16_SQ --> WRAPPER8_16_SQ.
diego
parents: 5997
diff changeset
809 #define WRAPPER8_16(name8, name16)\
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
810 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
811 return name8(s, dst , src , stride, h)\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
812 +name8(s, dst+8 , src+8 , stride, h);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
813 }
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
814
6056
558c1fd0ee72 Fix typo in macro name: WARPER8_16_SQ --> WRAPPER8_16_SQ.
diego
parents: 5997
diff changeset
815 #define WRAPPER8_16_SQ(name8, name16)\
1708
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
816 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
817 int score=0;\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
818 score +=name8(s, dst , src , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
819 score +=name8(s, dst+8 , src+8 , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
820 if(h==16){\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
821 dst += 8*stride;\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
822 src += 8*stride;\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
823 score +=name8(s, dst , src , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
824 score +=name8(s, dst+8 , src+8 , stride, 8);\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
825 }\
dea5b2946999 interlaced motion estimation
michael
parents: 1653
diff changeset
826 return score;\
936
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
827 }
caa77cd960c0 qpel encoding
michaelni
parents: 894
diff changeset
828
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
829
9436
dfed6a243bab Mark src constant in copy_block*().
alexc
parents: 8976
diff changeset
830 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
831 {
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
832 int i;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
833 for(i=0; i<h; i++)
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
834 {
5520
c16a59ef6a86 * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents: 5291
diff changeset
835 AV_WN16(dst , AV_RN16(src ));
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
836 dst+=dstStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
837 src+=srcStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
838 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
839 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
840
9436
dfed6a243bab Mark src constant in copy_block*().
alexc
parents: 8976
diff changeset
841 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
842 {
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
843 int i;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
844 for(i=0; i<h; i++)
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
845 {
5520
c16a59ef6a86 * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents: 5291
diff changeset
846 AV_WN32(dst , AV_RN32(src ));
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
847 dst+=dstStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
848 src+=srcStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
849 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
850 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
851
9436
dfed6a243bab Mark src constant in copy_block*().
alexc
parents: 8976
diff changeset
852 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
853 {
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
854 int i;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
855 for(i=0; i<h; i++)
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
856 {
5520
c16a59ef6a86 * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents: 5291
diff changeset
857 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
858 AV_WN32(dst+4 , AV_RN32(src+4 ));
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
859 dst+=dstStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
860 src+=srcStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
861 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
862 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
863
9436
dfed6a243bab Mark src constant in copy_block*().
alexc
parents: 8976
diff changeset
864 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
865 {
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
866 int i;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
867 for(i=0; i<h; i++)
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
868 {
5520
c16a59ef6a86 * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents: 5291
diff changeset
869 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
870 AV_WN32(dst+4 , AV_RN32(src+4 ));
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
871 dst[8]= src[8];
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
872 dst+=dstStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
873 src+=srcStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
874 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
875 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
876
9436
dfed6a243bab Mark src constant in copy_block*().
alexc
parents: 8976
diff changeset
877 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
878 {
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
879 int i;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
880 for(i=0; i<h; i++)
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
881 {
5520
c16a59ef6a86 * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents: 5291
diff changeset
882 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
883 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
884 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
885 AV_WN32(dst+12, AV_RN32(src+12));
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
886 dst+=dstStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
887 src+=srcStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
888 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
889 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
890
9436
dfed6a243bab Mark src constant in copy_block*().
alexc
parents: 8976
diff changeset
891 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
892 {
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
893 int i;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
894 for(i=0; i<h; i++)
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
895 {
5520
c16a59ef6a86 * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
romansh
parents: 5291
diff changeset
896 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
897 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
898 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
899 AV_WN32(dst+12, AV_RN32(src+12));
4240
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
900 dst[16]= src[16];
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
901 dst+=dstStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
902 src+=srcStride;
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
903 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
904 }
e880555ad2c6 Move copy_block* functions to dsp dsputil.h
gpoirier
parents: 4239
diff changeset
905
7760
c4a4495715dd Globally rename the header inclusion guard names.
stefano
parents: 7687
diff changeset
906 #endif /* AVCODEC_DSPUTIL_H */