Mercurial > libavcodec.hg
annotate dsputil.h @ 790:b9156f8e6747 libavcodec
-fomit-frame-pointer is needed for the PIC code
author | michael |
---|---|
date | Tue, 29 Oct 2002 11:21:43 +0000 |
parents | 6f5e87957bcb |
children | 670009af4fc2 |
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 */ |
706
e65798d228ea
idct permutation cleanup, idct can be selected per context now
michaelni
parents:
689
diff
changeset
|
37 extern const UINT8 ff_alternate_horizontal_scan[64]; |
e65798d228ea
idct permutation cleanup, idct can be selected per context now
michaelni
parents:
689
diff
changeset
|
38 extern const UINT8 ff_alternate_vertical_scan[64]; |
e65798d228ea
idct permutation cleanup, idct can be selected per context now
michaelni
parents:
689
diff
changeset
|
39 extern const UINT8 ff_zigzag_direct[64]; |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
40 |
0 | 41 /* pixel operations */ |
42 #define MAX_NEG_CROP 384 | |
43 | |
44 /* temporary */ | |
45 extern UINT32 squareTbl[512]; | |
50 | 46 extern UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; |
0 | 47 |
48 void dsputil_init(void); | |
49 | |
675 | 50 /* minimum alignment rules ;) |
51 if u notice errors in the align stuff, need more alignment for some asm code for some cpu | |
52 or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ... | |
53 | |
54 !warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible) | |
55 i (michael) didnt check them, these are just the alignents which i think could be reached easily ... | |
56 | |
57 !future video codecs might need functions with less strict alignment | |
58 */ | |
0 | 59 |
675 | 60 /* pixel ops : interface with DCT */ |
61 extern void (*get_pixels)(DCTELEM *block/*align 16*/, const UINT8 *pixels/*align 8*/, int line_size); | |
62 extern void (*diff_pixels)(DCTELEM *block/*align 16*/, const UINT8 *s1/*align 8*/, const UINT8 *s2/*align 8*/, int stride); | |
63 extern void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size); | |
64 extern void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size); | |
753 | 65 extern void (*ff_gmc1)(UINT8 *dst/*align 8*/, UINT8 *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder); |
66 extern void (*ff_gmc )(UINT8 *dst/*align 8*/, UINT8 *src/*align 1*/, int stride, int h, int ox, int oy, | |
67 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height); | |
675 | 68 extern void (*clear_blocks)(DCTELEM *blocks/*align 16*/); |
612 | 69 extern int (*pix_sum)(UINT8 * pix, int line_size); |
70 extern int (*pix_norm1)(UINT8 * pix, int line_size); | |
71 | |
255 | 72 |
0 | 73 |
74 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size); | |
324 | 75 void diff_pixels_c(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride); |
0 | 76 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); |
77 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
296 | 78 void clear_blocks_c(DCTELEM *blocks); |
0 | 79 |
80 /* add and put pixel (decoding) */ | |
675 | 81 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16 |
82 typedef void (*op_pixels_func)(UINT8 *block/*align width (8 or 16)*/, const UINT8 *pixels/*align 1*/, int line_size, int h); | |
83 typedef void (*qpel_mc_func)(UINT8 *dst/*align width (8 or 16)*/, UINT8 *src/*align 1*/, int stride); | |
0 | 84 |
651 | 85 extern op_pixels_func put_pixels_tab[2][4]; |
86 extern op_pixels_func avg_pixels_tab[2][4]; | |
87 extern op_pixels_func put_no_rnd_pixels_tab[2][4]; | |
88 extern op_pixels_func avg_no_rnd_pixels_tab[2][4]; | |
89 extern qpel_mc_func put_qpel_pixels_tab[2][16]; | |
90 extern qpel_mc_func avg_qpel_pixels_tab[2][16]; | |
91 extern qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16]; | |
92 extern qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]; | |
93 | |
94 #define CALL_2X_PIXELS(a, b, n)\ | |
95 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ | |
96 b(block , pixels , line_size, h);\ | |
97 b(block+n, pixels+n, line_size, h);\ | |
98 } | |
255 | 99 |
0 | 100 /* motion estimation */ |
101 | |
675 | 102 typedef int (*op_pixels_abs_func)(UINT8 *blk1/*align width (8 or 16)*/, UINT8 *blk2/*align 1*/, int line_size); |
0 | 103 |
104 extern op_pixels_abs_func pix_abs16x16; | |
105 extern op_pixels_abs_func pix_abs16x16_x2; | |
106 extern op_pixels_abs_func pix_abs16x16_y2; | |
107 extern op_pixels_abs_func pix_abs16x16_xy2; | |
294 | 108 extern op_pixels_abs_func pix_abs8x8; |
109 extern op_pixels_abs_func pix_abs8x8_x2; | |
110 extern op_pixels_abs_func pix_abs8x8_y2; | |
111 extern op_pixels_abs_func pix_abs8x8_xy2; | |
0 | 112 |
294 | 113 int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx); |
114 int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
115 int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
116 int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
0 | 117 |
764 | 118 /** |
119 * permute block according to permuatation. | |
120 * @param last last non zero element in scantable order | |
121 */ | |
122 void ff_block_permute(INT16 *block, UINT8 *permutation, const UINT8 *scantable, int last); | |
34 | 123 |
62 | 124 #if defined(HAVE_MMX) |
0 | 125 |
126 #define MM_MMX 0x0001 /* standard MMX */ | |
127 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
128 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
129 #define MM_SSE 0x0008 /* SSE functions */ | |
130 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
131 | |
132 extern int mm_flags; | |
133 | |
134 int mm_support(void); | |
135 | |
136 static inline void emms(void) | |
137 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
138 __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
|
139 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
140 |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
141 #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
|
142 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
143 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
|
144 emms();\ |
0 | 145 } |
146 | |
147 #define __align8 __attribute__ ((aligned (8))) | |
148 | |
149 void dsputil_init_mmx(void); | |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
324
diff
changeset
|
150 void dsputil_set_bit_exact_mmx(void); |
0 | 151 |
62 | 152 #elif defined(ARCH_ARMV4L) |
153 | |
154 #define emms_c() | |
155 | |
156 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
157 line ptimizations */ | |
158 #define __align8 __attribute__ ((aligned (4))) | |
159 | |
160 void dsputil_init_armv4l(void); | |
161 | |
88 | 162 #elif defined(HAVE_MLIB) |
163 | |
164 #define emms_c() | |
165 | |
166 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
167 #define __align8 __attribute__ ((aligned (8))) | |
168 | |
169 void dsputil_init_mlib(void); | |
170 | |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
171 #elif defined(ARCH_ALPHA) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
172 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
173 #define emms_c() |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
174 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
175 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
176 void dsputil_init_alpha(void); |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
177 |
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
|
178 #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
|
179 |
92e99e506920
first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
612
diff
changeset
|
180 #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
|
181 #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
|
182 |
638
0012f75c92bb
altivec build tidyup patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
626
diff
changeset
|
183 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
|
184 |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
185 #elif defined(HAVE_MMI) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
186 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
187 #define emms_c() |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
188 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
189 #define __align8 __attribute__ ((aligned (16))) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
190 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
191 void dsputil_init_mmi(void); |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
192 |
0 | 193 #else |
194 | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
195 #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
|
196 |
0 | 197 #define __align8 |
198 | |
199 #endif | |
200 | |
493
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
201 #ifdef __GNUC__ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
202 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
203 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
|
204 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
|
205 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
206 #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
|
207 #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
|
208 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
209 #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
|
210 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
211 #else /* __GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
212 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
213 #define LD32(a) (*((uint32_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
214 #define LD64(a) (*((uint64_t*)(a))) |
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 #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
|
217 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
218 #endif /* !__GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
219 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
220 /* PSNR */ |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
221 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
|
222 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
|
223 AVCodecContext *avctx); |
781 | 224 |
225 /* FFT computation */ | |
226 | |
227 /* NOTE: soon integer code will be added, so you must use the | |
228 FFTSample type */ | |
229 typedef float FFTSample; | |
230 | |
231 typedef struct FFTComplex { | |
232 FFTSample re, im; | |
233 } FFTComplex; | |
234 | |
235 typedef struct FFTContext { | |
236 int nbits; | |
237 int inverse; | |
238 uint16_t *revtab; | |
239 FFTComplex *exptab; | |
240 FFTComplex *exptab1; /* only used by SSE code */ | |
241 void (*fft_calc)(struct FFTContext *s, FFTComplex *z); | |
242 } FFTContext; | |
243 | |
244 int fft_init(FFTContext *s, int nbits, int inverse); | |
245 void fft_permute(FFTContext *s, FFTComplex *z); | |
246 void fft_calc_c(FFTContext *s, FFTComplex *z); | |
247 void fft_calc_sse(FFTContext *s, FFTComplex *z); | |
248 static inline void fft_calc(FFTContext *s, FFTComplex *z) | |
249 { | |
250 s->fft_calc(s, z); | |
251 } | |
252 void fft_end(FFTContext *s); | |
253 | |
254 /* MDCT computation */ | |
255 | |
256 typedef struct MDCTContext { | |
257 int n; /* size of MDCT (i.e. number of input data * 2) */ | |
258 int nbits; /* n = 2^nbits */ | |
259 /* pre/post rotation tables */ | |
260 FFTSample *tcos; | |
261 FFTSample *tsin; | |
262 FFTContext fft; | |
263 } MDCTContext; | |
264 | |
265 int mdct_init(MDCTContext *s, int nbits, int inverse); | |
266 void imdct_calc(MDCTContext *s, FFTSample *output, | |
267 const FFTSample *input, FFTSample *tmp); | |
268 void mdct_calc(MDCTContext *s, FFTSample *out, | |
269 const FFTSample *input, FFTSample *tmp); | |
270 void mdct_end(MDCTContext *s); | |
271 | |
0 | 272 #endif |