Mercurial > libavcodec.hg
annotate dsputil.h @ 1092:f59c3f66363b libavcodec
MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
bitexact cleanup
author | michaelni |
---|---|
date | Mon, 03 Mar 2003 14:54:00 +0000 |
parents | 92fb44eae6b6 |
children | eeaa1e837efb |
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; | |
1008 | 28 //typedef int DCTELEM; |
0 | 29 |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
429
diff
changeset
|
30 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
|
31 void ff_jpeg_fdct_islow (DCTELEM *data); |
0 | 32 |
33 void j_rev_dct (DCTELEM *data); | |
34 | |
687
9abb13c21fbe
fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
arpi_esp
parents:
675
diff
changeset
|
35 void ff_fdct_mmx(DCTELEM *block); |
0 | 36 |
34 | 37 /* encoding scans */ |
1064 | 38 extern const uint8_t ff_alternate_horizontal_scan[64]; |
39 extern const uint8_t ff_alternate_vertical_scan[64]; | |
40 extern const uint8_t 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
|
41 |
0 | 42 /* pixel operations */ |
43 #define MAX_NEG_CROP 384 | |
44 | |
45 /* temporary */ | |
1064 | 46 extern uint32_t squareTbl[512]; |
47 extern uint8_t cropTbl[256 + 2 * MAX_NEG_CROP]; | |
0 | 48 |
49 | |
675 | 50 /* minimum alignment rules ;) |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
51 if u notice errors in the align stuff, need more alignment for some asm code for some cpu |
675 | 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 |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
60 /* |
1064 | 61 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size); |
62 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride); | |
63 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size); | |
64 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size); | |
296 | 65 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
|
66 */ |
0 | 67 |
68 /* add and put pixel (decoding) */ | |
675 | 69 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16 |
1064 | 70 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h); |
71 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride); | |
0 | 72 |
984 | 73 #define DEF_OLD_QPEL(name)\ |
1064 | 74 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\ |
75 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\ | |
76 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
|
77 |
984 | 78 DEF_OLD_QPEL(qpel16_mc11_old_c) |
79 DEF_OLD_QPEL(qpel16_mc31_old_c) | |
80 DEF_OLD_QPEL(qpel16_mc12_old_c) | |
81 DEF_OLD_QPEL(qpel16_mc32_old_c) | |
82 DEF_OLD_QPEL(qpel16_mc13_old_c) | |
83 DEF_OLD_QPEL(qpel16_mc33_old_c) | |
84 DEF_OLD_QPEL(qpel8_mc11_old_c) | |
85 DEF_OLD_QPEL(qpel8_mc31_old_c) | |
86 DEF_OLD_QPEL(qpel8_mc12_old_c) | |
87 DEF_OLD_QPEL(qpel8_mc32_old_c) | |
88 DEF_OLD_QPEL(qpel8_mc13_old_c) | |
89 DEF_OLD_QPEL(qpel8_mc33_old_c) | |
651 | 90 |
91 #define CALL_2X_PIXELS(a, b, n)\ | |
92 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ | |
93 b(block , pixels , line_size, h);\ | |
94 b(block+n, pixels+n, line_size, h);\ | |
95 } | |
255 | 96 |
0 | 97 /* motion estimation */ |
98 | |
1064 | 99 typedef int (*op_pixels_abs_func)(uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size)/* __attribute__ ((const))*/; |
936 | 100 |
1064 | 101 typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size)/* __attribute__ ((const))*/; |
936 | 102 |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
103 typedef struct DSPContext { |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
104 /* pixel ops : interface with DCT */ |
1064 | 105 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size); |
106 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride); | |
107 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); | |
108 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); | |
109 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder); | |
110 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy, | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
111 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height); |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
112 void (*clear_blocks)(DCTELEM *blocks/*align 16*/); |
1064 | 113 int (*pix_sum)(uint8_t * pix, int line_size); |
114 int (*pix_norm1)(uint8_t * pix, int line_size); | |
936 | 115 me_cmp_func sad[2]; /* identical to pix_absAxA except additional void * */ |
116 me_cmp_func sse[2]; | |
117 me_cmp_func hadamard8_diff[2]; | |
118 me_cmp_func dct_sad[2]; | |
119 me_cmp_func quant_psnr[2]; | |
1007 | 120 me_cmp_func bit[2]; |
121 me_cmp_func rd[2]; | |
936 | 122 int (*hadamard8_abs )(uint8_t *src, int stride, int mean); |
123 | |
954 | 124 me_cmp_func me_pre_cmp[11]; |
936 | 125 me_cmp_func me_cmp[11]; |
126 me_cmp_func me_sub_cmp[11]; | |
127 me_cmp_func mb_cmp[11]; | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
128 |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
129 /* maybe create an array for 16/8 functions */ |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
130 op_pixels_func put_pixels_tab[2][4]; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
131 op_pixels_func avg_pixels_tab[2][4]; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
132 op_pixels_func put_no_rnd_pixels_tab[2][4]; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
133 op_pixels_func avg_no_rnd_pixels_tab[2][4]; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
134 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
|
135 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
|
136 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
|
137 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]; |
936 | 138 qpel_mc_func put_mspel_pixels_tab[8]; |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
139 |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
140 op_pixels_abs_func pix_abs16x16; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
141 op_pixels_abs_func pix_abs16x16_x2; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
142 op_pixels_abs_func pix_abs16x16_y2; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
143 op_pixels_abs_func pix_abs16x16_xy2; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
144 op_pixels_abs_func pix_abs8x8; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
145 op_pixels_abs_func pix_abs8x8_x2; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
146 op_pixels_abs_func pix_abs8x8_y2; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
147 op_pixels_abs_func pix_abs8x8_xy2; |
866 | 148 |
149 /* huffyuv specific */ | |
150 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w); | |
936 | 151 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w); |
1092 | 152 |
153 /* (I)DCT */ | |
154 void (*fdct)(DCTELEM *block/* align 16*/); | |
155 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); | |
156 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); | |
157 uint8_t idct_permutation[64]; | |
158 int idct_permutation_type; | |
159 #define FF_NO_IDCT_PERM 1 | |
160 #define FF_LIBMPEG2_IDCT_PERM 2 | |
161 #define FF_SIMPLE_IDCT_PERM 3 | |
162 #define FF_TRANSPOSE_IDCT_PERM 4 | |
163 | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
164 } DSPContext; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
165 |
1092 | 166 void dsputil_init(DSPContext* p, AVCodecContext *avctx); |
0 | 167 |
764 | 168 /** |
169 * permute block according to permuatation. | |
170 * @param last last non zero element in scantable order | |
171 */ | |
1064 | 172 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last); |
34 | 173 |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
174 #define emms_c() |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
175 |
995
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
176 /* should be defined by architectures supporting |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
177 one or more MultiMedia extension */ |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
178 int mm_support(void); |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
179 |
62 | 180 #if defined(HAVE_MMX) |
0 | 181 |
862 | 182 #undef emms_c |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
183 |
0 | 184 #define MM_MMX 0x0001 /* standard MMX */ |
185 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
186 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
187 #define MM_SSE 0x0008 /* SSE functions */ | |
188 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
189 | |
190 extern int mm_flags; | |
191 | |
1064 | 192 void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size); |
193 void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size); | |
0 | 194 |
195 static inline void emms(void) | |
196 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
197 __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
|
198 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
199 |
936 | 200 |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
201 #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
|
202 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
203 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
|
204 emms();\ |
0 | 205 } |
206 | |
207 #define __align8 __attribute__ ((aligned (8))) | |
208 | |
1092 | 209 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx); |
210 void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx); | |
1065 | 211 |
62 | 212 #elif defined(ARCH_ARMV4L) |
213 | |
214 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
215 line ptimizations */ | |
216 #define __align8 __attribute__ ((aligned (4))) | |
217 | |
1092 | 218 void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx); |
62 | 219 |
88 | 220 #elif defined(HAVE_MLIB) |
221 | |
222 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
223 #define __align8 __attribute__ ((aligned (8))) | |
224 | |
1092 | 225 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx); |
88 | 226 |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
227 #elif defined(ARCH_ALPHA) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
228 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
229 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
230 |
1092 | 231 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx); |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
232 |
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
|
233 #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
|
234 |
894
a408778eff87
altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
884
diff
changeset
|
235 #define MM_ALTIVEC 0x0001 /* standard AltiVec */ |
a408778eff87
altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
884
diff
changeset
|
236 |
a408778eff87
altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
884
diff
changeset
|
237 extern int mm_flags; |
a408778eff87
altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
884
diff
changeset
|
238 |
1033
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
239 #if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN) |
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
240 #include <altivec.h> |
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
241 #endif |
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
242 |
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
|
243 #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
|
244 |
1092 | 245 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx); |
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
|
246 |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
247 #elif defined(HAVE_MMI) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
248 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
249 #define __align8 __attribute__ ((aligned (16))) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
250 |
1092 | 251 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx); |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
252 |
0 | 253 #else |
254 | |
255 #define __align8 | |
256 | |
257 #endif | |
258 | |
493
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
259 #ifdef __GNUC__ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
260 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
261 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
|
262 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
|
263 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
264 #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
|
265 #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
|
266 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
267 #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
|
268 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
269 #else /* __GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
270 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
271 #define LD32(a) (*((uint32_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
272 #define LD64(a) (*((uint64_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
273 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
274 #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
|
275 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
276 #endif /* !__GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
277 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
278 /* PSNR */ |
1064 | 279 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
|
280 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
|
281 AVCodecContext *avctx); |
781 | 282 |
283 /* FFT computation */ | |
284 | |
285 /* NOTE: soon integer code will be added, so you must use the | |
286 FFTSample type */ | |
287 typedef float FFTSample; | |
288 | |
289 typedef struct FFTComplex { | |
290 FFTSample re, im; | |
291 } FFTComplex; | |
292 | |
293 typedef struct FFTContext { | |
294 int nbits; | |
295 int inverse; | |
296 uint16_t *revtab; | |
297 FFTComplex *exptab; | |
298 FFTComplex *exptab1; /* only used by SSE code */ | |
299 void (*fft_calc)(struct FFTContext *s, FFTComplex *z); | |
300 } FFTContext; | |
301 | |
302 int fft_init(FFTContext *s, int nbits, int inverse); | |
303 void fft_permute(FFTContext *s, FFTComplex *z); | |
304 void fft_calc_c(FFTContext *s, FFTComplex *z); | |
305 void fft_calc_sse(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
|
306 void fft_calc_altivec(FFTContext *s, FFTComplex *z); |
e05d525505c5
fft altivec by Romain Dolbeau - simplified patch, test it on PPC with fft-test and wma decoding
bellard
parents:
954
diff
changeset
|
307 |
781 | 308 static inline void fft_calc(FFTContext *s, FFTComplex *z) |
309 { | |
310 s->fft_calc(s, z); | |
311 } | |
312 void fft_end(FFTContext *s); | |
313 | |
314 /* MDCT computation */ | |
315 | |
316 typedef struct MDCTContext { | |
317 int n; /* size of MDCT (i.e. number of input data * 2) */ | |
318 int nbits; /* n = 2^nbits */ | |
319 /* pre/post rotation tables */ | |
320 FFTSample *tcos; | |
321 FFTSample *tsin; | |
322 FFTContext fft; | |
323 } MDCTContext; | |
324 | |
794 | 325 int ff_mdct_init(MDCTContext *s, int nbits, int inverse); |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
326 void ff_imdct_calc(MDCTContext *s, FFTSample *output, |
781 | 327 const FFTSample *input, FFTSample *tmp); |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
328 void ff_mdct_calc(MDCTContext *s, FFTSample *out, |
781 | 329 const FFTSample *input, FFTSample *tmp); |
794 | 330 void ff_mdct_end(MDCTContext *s); |
781 | 331 |
936 | 332 #define WARPER88_1616(name8, name16)\ |
333 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride){\ | |
334 return name8(s, dst , src , stride)\ | |
335 +name8(s, dst+8 , src+8 , stride)\ | |
336 +name8(s, dst +8*stride, src +8*stride, stride)\ | |
337 +name8(s, dst+8+8*stride, src+8+8*stride, stride);\ | |
338 } | |
339 | |
838
b78812db886f
lrintf detection (based upon a patch by Franois Revol <revol at free dot fr>)
michaelni
parents:
802
diff
changeset
|
340 #ifndef HAVE_LRINTF |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
341 /* XXX: add ISOC specific test to avoid specific BSD testing. */ |
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
342 /* better than nothing implementation. */ |
802 | 343 /* btw, rintf() is existing on fbsd too -- alex */ |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
344 static inline long int lrintf(float x) |
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
345 { |
1040
998d5035b15b
win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
bellard
parents:
1033
diff
changeset
|
346 #ifdef CONFIG_WIN32 |
998d5035b15b
win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
bellard
parents:
1033
diff
changeset
|
347 /* XXX: incorrect, but make it compile */ |
998d5035b15b
win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
bellard
parents:
1033
diff
changeset
|
348 return (int)(x); |
998d5035b15b
win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
bellard
parents:
1033
diff
changeset
|
349 #else |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
350 return (int)(rint(x)); |
1040
998d5035b15b
win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
bellard
parents:
1033
diff
changeset
|
351 #endif |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
352 } |
0 | 353 #endif |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
354 |
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
355 #endif |