Mercurial > libavcodec.hg
annotate dsputil.h @ 692:852b5a416161 libavcodec
fixing ac prediction encoding with adaptive quantization
author | michaelni |
---|---|
date | Tue, 24 Sep 2002 09:15:46 +0000 |
parents | efcbfbd18864 |
children | e65798d228ea |
rev | line source |
---|---|
429 | 1 /* |
2 * DSP utils | |
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard. | |
4 * | |
5 * This library is free software; you can redistribute it and/or | |
6 * modify it under the terms of the GNU Lesser General Public | |
7 * License as published by the Free Software Foundation; either | |
8 * version 2 of the License, or (at your option) any later version. | |
9 * | |
10 * This library is distributed in the hope that it will be useful, | |
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
13 * Lesser General Public License for more details. | |
14 * | |
15 * You should have received a copy of the GNU Lesser General Public | |
16 * License along with this library; if not, write to the Free Software | |
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
18 */ | |
0 | 19 #ifndef DSPUTIL_H |
20 #define DSPUTIL_H | |
21 | |
22 #include "common.h" | |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
23 #include "avcodec.h" |
0 | 24 |
255 | 25 //#define DEBUG |
0 | 26 /* dct code */ |
27 typedef short DCTELEM; | |
28 | |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
429
diff
changeset
|
29 void fdct_ifast (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
|
30 void ff_jpeg_fdct_islow (DCTELEM *data); |
0 | 31 |
32 void j_rev_dct (DCTELEM *data); | |
33 | |
687
9abb13c21fbe
fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
arpi_esp
parents:
675
diff
changeset
|
34 void ff_fdct_mmx(DCTELEM *block); |
0 | 35 |
34 | 36 /* encoding scans */ |
37 extern UINT8 ff_alternate_horizontal_scan[64]; | |
38 extern UINT8 ff_alternate_vertical_scan[64]; | |
39 extern UINT8 zigzag_direct[64]; | |
40 | |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
41 /* permutation table */ |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
42 extern UINT8 permutation[64]; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
43 |
0 | 44 /* pixel operations */ |
45 #define MAX_NEG_CROP 384 | |
46 | |
47 /* temporary */ | |
48 extern UINT32 squareTbl[512]; | |
50 | 49 extern UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; |
0 | 50 |
51 void dsputil_init(void); | |
52 | |
675 | 53 /* minimum alignment rules ;) |
54 if u notice errors in the align stuff, need more alignment for some asm code for some cpu | |
55 or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ... | |
56 | |
57 !warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible) | |
58 i (michael) didnt check them, these are just the alignents which i think could be reached easily ... | |
59 | |
60 !future video codecs might need functions with less strict alignment | |
61 */ | |
0 | 62 |
675 | 63 /* pixel ops : interface with DCT */ |
64 extern void (*ff_idct)(DCTELEM *block/*align 16*/); | |
65 extern void (*ff_idct_put)(UINT8 *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); | |
66 extern void (*ff_idct_add)(UINT8 *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); | |
67 extern void (*get_pixels)(DCTELEM *block/*align 16*/, const UINT8 *pixels/*align 8*/, int line_size); | |
68 extern void (*diff_pixels)(DCTELEM *block/*align 16*/, const UINT8 *s1/*align 8*/, const UINT8 *s2/*align 8*/, int stride); | |
69 extern void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size); | |
70 extern void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size); | |
71 extern void (*gmc1)(UINT8 *dst/*align 8*/, UINT8 *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder); | |
72 extern void (*clear_blocks)(DCTELEM *blocks/*align 16*/); | |
612 | 73 extern int (*pix_sum)(UINT8 * pix, int line_size); |
74 extern int (*pix_norm1)(UINT8 * pix, int line_size); | |
75 | |
255 | 76 |
0 | 77 |
78 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size); | |
324 | 79 void diff_pixels_c(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride); |
0 | 80 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); |
81 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
296 | 82 void clear_blocks_c(DCTELEM *blocks); |
0 | 83 |
84 /* add and put pixel (decoding) */ | |
675 | 85 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16 |
86 typedef void (*op_pixels_func)(UINT8 *block/*align width (8 or 16)*/, const UINT8 *pixels/*align 1*/, int line_size, int h); | |
87 typedef void (*qpel_mc_func)(UINT8 *dst/*align width (8 or 16)*/, UINT8 *src/*align 1*/, int stride); | |
0 | 88 |
651 | 89 extern op_pixels_func put_pixels_tab[2][4]; |
90 extern op_pixels_func avg_pixels_tab[2][4]; | |
91 extern op_pixels_func put_no_rnd_pixels_tab[2][4]; | |
92 extern op_pixels_func avg_no_rnd_pixels_tab[2][4]; | |
93 extern qpel_mc_func put_qpel_pixels_tab[2][16]; | |
94 extern qpel_mc_func avg_qpel_pixels_tab[2][16]; | |
95 extern qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16]; | |
96 extern qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]; | |
97 | |
98 #define CALL_2X_PIXELS(a, b, n)\ | |
99 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ | |
100 b(block , pixels , line_size, h);\ | |
101 b(block+n, pixels+n, line_size, h);\ | |
102 } | |
255 | 103 |
0 | 104 /* motion estimation */ |
105 | |
675 | 106 typedef int (*op_pixels_abs_func)(UINT8 *blk1/*align width (8 or 16)*/, UINT8 *blk2/*align 1*/, int line_size); |
0 | 107 |
108 extern op_pixels_abs_func pix_abs16x16; | |
109 extern op_pixels_abs_func pix_abs16x16_x2; | |
110 extern op_pixels_abs_func pix_abs16x16_y2; | |
111 extern op_pixels_abs_func pix_abs16x16_xy2; | |
294 | 112 extern op_pixels_abs_func pix_abs8x8; |
113 extern op_pixels_abs_func pix_abs8x8_x2; | |
114 extern op_pixels_abs_func pix_abs8x8_y2; | |
115 extern op_pixels_abs_func pix_abs8x8_xy2; | |
0 | 116 |
294 | 117 int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx); |
118 int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
119 int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
120 int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
0 | 121 |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
122 static inline int block_permute_op(int j) |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
123 { |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
124 return permutation[j]; |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
125 } |
34 | 126 |
127 void block_permute(INT16 *block); | |
128 | |
62 | 129 #if defined(HAVE_MMX) |
0 | 130 |
131 #define MM_MMX 0x0001 /* standard MMX */ | |
132 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
133 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
134 #define MM_SSE 0x0008 /* SSE functions */ | |
135 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
136 | |
137 extern int mm_flags; | |
138 | |
139 int mm_support(void); | |
140 | |
141 static inline void emms(void) | |
142 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
143 __asm __volatile ("emms;":::"memory"); |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
144 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
145 |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
146 #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
|
147 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
148 if (mm_flags & MM_MMX)\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
149 emms();\ |
0 | 150 } |
151 | |
152 #define __align8 __attribute__ ((aligned (8))) | |
153 | |
154 void dsputil_init_mmx(void); | |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
324
diff
changeset
|
155 void dsputil_set_bit_exact_mmx(void); |
0 | 156 |
62 | 157 #elif defined(ARCH_ARMV4L) |
158 | |
159 #define emms_c() | |
160 | |
161 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
162 line ptimizations */ | |
163 #define __align8 __attribute__ ((aligned (4))) | |
164 | |
165 void dsputil_init_armv4l(void); | |
166 | |
88 | 167 #elif defined(HAVE_MLIB) |
168 | |
169 #define emms_c() | |
170 | |
171 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
172 #define __align8 __attribute__ ((aligned (8))) | |
173 | |
174 void dsputil_init_mlib(void); | |
175 | |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
176 #elif defined(ARCH_ALPHA) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
177 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
178 #define emms_c() |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
179 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
180 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
181 void dsputil_init_alpha(void); |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
182 |
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
|
183 #elif defined(ARCH_POWERPC) |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
184 |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
185 #define emms_c() |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
186 #define __align8 __attribute__ ((aligned (16))) |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
187 |
638
0012f75c92bb
altivec build tidyup patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
626
diff
changeset
|
188 void dsputil_init_ppc(void); |
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
|
189 |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
190 #elif defined(HAVE_MMI) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
191 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
192 #define emms_c() |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
193 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
194 #define __align8 __attribute__ ((aligned (16))) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
195 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
196 void dsputil_init_mmi(void); |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
197 |
0 | 198 #else |
199 | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
200 #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
|
201 |
0 | 202 #define __align8 |
203 | |
204 #endif | |
205 | |
493
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
206 #ifdef __GNUC__ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
207 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
208 struct unaligned_64 { uint64_t l; } __attribute__((packed)); |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
209 struct unaligned_32 { uint32_t l; } __attribute__((packed)); |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
210 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
211 #define LD32(a) (((const struct unaligned_32 *) (a))->l) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
212 #define LD64(a) (((const struct unaligned_64 *) (a))->l) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
213 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
214 #define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
215 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
216 #else /* __GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
217 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
218 #define LD32(a) (*((uint32_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
219 #define LD64(a) (*((uint64_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
220 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
221 #define ST32(a, b) *((uint32_t*)(a)) = (b) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
222 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
223 #endif /* !__GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
224 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
225 /* PSNR */ |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
226 void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3], |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
227 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
|
228 AVCodecContext *avctx); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
229 |
0 | 230 #endif |