Mercurial > libavcodec.hg
annotate dsputil.h @ 623:92e99e506920 libavcodec
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
author | michaelni |
---|---|
date | Wed, 28 Aug 2002 13:14:36 +0000 |
parents | c0005de2be59 |
children | bb6a69f9d409 |
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); |
612 | 65 extern int (*pix_sum)(UINT8 * pix, int line_size); |
66 extern int (*pix_norm1)(UINT8 * pix, int line_size); | |
67 | |
255 | 68 |
0 | 69 |
70 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size); | |
324 | 71 void diff_pixels_c(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride); |
0 | 72 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); |
73 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
296 | 74 void clear_blocks_c(DCTELEM *blocks); |
0 | 75 |
76 /* add and put pixel (decoding) */ | |
77 typedef void (*op_pixels_func)(UINT8 *block, const UINT8 *pixels, int line_size, int h); | |
255 | 78 typedef void (*qpel_mc_func)(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my); |
0 | 79 |
80 extern op_pixels_func put_pixels_tab[4]; | |
81 extern op_pixels_func avg_pixels_tab[4]; | |
82 extern op_pixels_func put_no_rnd_pixels_tab[4]; | |
83 extern op_pixels_func avg_no_rnd_pixels_tab[4]; | |
255 | 84 extern qpel_mc_func qpel_mc_rnd_tab[16]; |
85 extern qpel_mc_func qpel_mc_no_rnd_tab[16]; | |
86 | |
0 | 87 /* motion estimation */ |
88 | |
294 | 89 typedef int (*op_pixels_abs_func)(UINT8 *blk1, UINT8 *blk2, int line_size); |
0 | 90 |
91 extern op_pixels_abs_func pix_abs16x16; | |
92 extern op_pixels_abs_func pix_abs16x16_x2; | |
93 extern op_pixels_abs_func pix_abs16x16_y2; | |
94 extern op_pixels_abs_func pix_abs16x16_xy2; | |
294 | 95 extern op_pixels_abs_func pix_abs8x8; |
96 extern op_pixels_abs_func pix_abs8x8_x2; | |
97 extern op_pixels_abs_func pix_abs8x8_y2; | |
98 extern op_pixels_abs_func pix_abs8x8_xy2; | |
0 | 99 |
294 | 100 int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx); |
101 int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
102 int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
103 int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
0 | 104 |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
105 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
|
106 { |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
107 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
|
108 } |
34 | 109 |
110 void block_permute(INT16 *block); | |
111 | |
62 | 112 #if defined(HAVE_MMX) |
0 | 113 |
114 #define MM_MMX 0x0001 /* standard MMX */ | |
115 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
116 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
117 #define MM_SSE 0x0008 /* SSE functions */ | |
118 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
119 | |
120 extern int mm_flags; | |
121 | |
122 int mm_support(void); | |
123 | |
124 static inline void emms(void) | |
125 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
126 __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
|
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 |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
129 #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
|
130 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
131 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
|
132 emms();\ |
0 | 133 } |
134 | |
135 #define __align8 __attribute__ ((aligned (8))) | |
136 | |
137 void dsputil_init_mmx(void); | |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
324
diff
changeset
|
138 void dsputil_set_bit_exact_mmx(void); |
0 | 139 |
62 | 140 #elif defined(ARCH_ARMV4L) |
141 | |
142 #define emms_c() | |
143 | |
144 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
145 line ptimizations */ | |
146 #define __align8 __attribute__ ((aligned (4))) | |
147 | |
148 void dsputil_init_armv4l(void); | |
149 | |
88 | 150 #elif defined(HAVE_MLIB) |
151 | |
152 #define emms_c() | |
153 | |
154 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
155 #define __align8 __attribute__ ((aligned (8))) | |
156 | |
157 void dsputil_init_mlib(void); | |
158 | |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
159 #elif defined(ARCH_ALPHA) |
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 #define emms_c() |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
162 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
163 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
164 void dsputil_init_alpha(void); |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
165 |
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
|
166 #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
|
167 |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
168 #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
|
169 #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
|
170 |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
171 void dsputil_init_altivec(void); |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
172 |
0 | 173 #else |
174 | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
175 #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
|
176 |
0 | 177 #define __align8 |
178 | |
179 #endif | |
180 | |
493
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
181 #ifdef __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 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
|
184 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
|
185 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
186 #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
|
187 #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
|
188 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
189 #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
|
190 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
191 #else /* __GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
192 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
193 #define LD32(a) (*((uint32_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
194 #define LD64(a) (*((uint64_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
195 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
196 #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
|
197 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
198 #endif /* !__GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
199 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
200 /* PSNR */ |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
201 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
|
202 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
|
203 AVCodecContext *avctx); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
204 |
0 | 205 #endif |