Mercurial > libavcodec.hg
annotate dsputil.h @ 498:0b4450c15067 libavcodec
dc scale simplification/optimization
author | michaelni |
---|---|
date | Tue, 18 Jun 2002 00:46:02 +0000 |
parents | 873b9075d853 |
children | c0005de2be59 |
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); |
0 | 30 |
31 void j_rev_dct (DCTELEM *data); | |
32 | |
33 void fdct_mmx(DCTELEM *block); | |
34 | |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
429
diff
changeset
|
35 extern void (*av_fdct)(DCTELEM *block); |
0 | 36 |
34 | 37 /* encoding scans */ |
38 extern UINT8 ff_alternate_horizontal_scan[64]; | |
39 extern UINT8 ff_alternate_vertical_scan[64]; | |
40 extern UINT8 zigzag_direct[64]; | |
41 | |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
42 /* permutation table */ |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
43 extern UINT8 permutation[64]; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
44 |
0 | 45 /* pixel operations */ |
46 #define MAX_NEG_CROP 384 | |
47 | |
48 /* temporary */ | |
49 extern UINT32 squareTbl[512]; | |
50 | 50 extern UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; |
0 | 51 |
52 void dsputil_init(void); | |
53 | |
54 /* pixel ops : interface with DCT */ | |
55 | |
19
82d4c9be9873
MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
arpi_esp
parents:
6
diff
changeset
|
56 extern void (*ff_idct)(DCTELEM *block); |
480 | 57 extern void (*ff_idct_put)(UINT8 *dest, int line_size, DCTELEM *block); |
58 extern void (*ff_idct_add)(UINT8 *dest, int line_size, DCTELEM *block); | |
0 | 59 extern void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size); |
324 | 60 extern void (*diff_pixels)(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride); |
0 | 61 extern void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); |
62 extern void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); | |
255 | 63 extern void (*gmc1)(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder); |
296 | 64 extern void (*clear_blocks)(DCTELEM *blocks); |
255 | 65 |
0 | 66 |
67 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size); | |
324 | 68 void diff_pixels_c(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride); |
0 | 69 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); |
70 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
296 | 71 void clear_blocks_c(DCTELEM *blocks); |
0 | 72 |
73 /* add and put pixel (decoding) */ | |
74 typedef void (*op_pixels_func)(UINT8 *block, const UINT8 *pixels, int line_size, int h); | |
255 | 75 typedef void (*qpel_mc_func)(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my); |
0 | 76 |
77 extern op_pixels_func put_pixels_tab[4]; | |
78 extern op_pixels_func avg_pixels_tab[4]; | |
79 extern op_pixels_func put_no_rnd_pixels_tab[4]; | |
80 extern op_pixels_func avg_no_rnd_pixels_tab[4]; | |
255 | 81 extern qpel_mc_func qpel_mc_rnd_tab[16]; |
82 extern qpel_mc_func qpel_mc_no_rnd_tab[16]; | |
83 | |
0 | 84 /* motion estimation */ |
85 | |
294 | 86 typedef int (*op_pixels_abs_func)(UINT8 *blk1, UINT8 *blk2, int line_size); |
0 | 87 |
88 extern op_pixels_abs_func pix_abs16x16; | |
89 extern op_pixels_abs_func pix_abs16x16_x2; | |
90 extern op_pixels_abs_func pix_abs16x16_y2; | |
91 extern op_pixels_abs_func pix_abs16x16_xy2; | |
294 | 92 extern op_pixels_abs_func pix_abs8x8; |
93 extern op_pixels_abs_func pix_abs8x8_x2; | |
94 extern op_pixels_abs_func pix_abs8x8_y2; | |
95 extern op_pixels_abs_func pix_abs8x8_xy2; | |
0 | 96 |
294 | 97 int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx); |
98 int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
99 int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
100 int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
0 | 101 |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
102 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
|
103 { |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
104 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
|
105 } |
34 | 106 |
107 void block_permute(INT16 *block); | |
108 | |
62 | 109 #if defined(HAVE_MMX) |
0 | 110 |
111 #define MM_MMX 0x0001 /* standard MMX */ | |
112 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
113 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
114 #define MM_SSE 0x0008 /* SSE functions */ | |
115 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
116 | |
117 extern int mm_flags; | |
118 | |
119 int mm_support(void); | |
120 | |
121 static inline void emms(void) | |
122 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
123 __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
|
124 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
125 |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
126 #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
|
127 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
128 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
|
129 emms();\ |
0 | 130 } |
131 | |
132 #define __align8 __attribute__ ((aligned (8))) | |
133 | |
134 void dsputil_init_mmx(void); | |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
324
diff
changeset
|
135 void dsputil_set_bit_exact_mmx(void); |
0 | 136 |
62 | 137 #elif defined(ARCH_ARMV4L) |
138 | |
139 #define emms_c() | |
140 | |
141 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
142 line ptimizations */ | |
143 #define __align8 __attribute__ ((aligned (4))) | |
144 | |
145 void dsputil_init_armv4l(void); | |
146 | |
88 | 147 #elif defined(HAVE_MLIB) |
148 | |
149 #define emms_c() | |
150 | |
151 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
152 #define __align8 __attribute__ ((aligned (8))) | |
153 | |
154 void dsputil_init_mlib(void); | |
155 | |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
156 #elif defined(ARCH_ALPHA) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
157 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
158 #define emms_c() |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
159 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
160 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
161 void dsputil_init_alpha(void); |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
162 |
0 | 163 #else |
164 | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
165 #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
|
166 |
0 | 167 #define __align8 |
168 | |
169 #endif | |
170 | |
493
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
171 #ifdef __GNUC__ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
172 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
173 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
|
174 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
|
175 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
176 #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
|
177 #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
|
178 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
179 #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
|
180 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
181 #else /* __GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
182 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
183 #define LD32(a) (*((uint32_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
184 #define LD64(a) (*((uint64_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
185 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
186 #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
|
187 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
188 #endif /* !__GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
189 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
190 /* PSNR */ |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
191 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
|
192 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
|
193 AVCodecContext *avctx); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
194 |
0 | 195 #endif |