Mercurial > libavcodec.hg
annotate dsputil.h @ 1104:6a01ffc5351c libavcodec
cosmetics
author | michaelni |
---|---|
date | Thu, 06 Mar 2003 10:26:19 +0000 |
parents | eeaa1e837efb |
children | 1e39f273ecd6 |
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 */ | |
1102 | 19 |
20 /** | |
21 * @file dsputil.h | |
22 * @brief DSP utils | |
23 * | |
24 */ | |
25 | |
0 | 26 #ifndef DSPUTIL_H |
27 #define DSPUTIL_H | |
28 | |
29 #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
|
30 #include "avcodec.h" |
0 | 31 |
1102 | 32 |
255 | 33 //#define DEBUG |
0 | 34 /* dct code */ |
35 typedef short DCTELEM; | |
1008 | 36 //typedef int DCTELEM; |
0 | 37 |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
429
diff
changeset
|
38 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
|
39 void ff_jpeg_fdct_islow (DCTELEM *data); |
0 | 40 |
41 void j_rev_dct (DCTELEM *data); | |
42 | |
687
9abb13c21fbe
fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
arpi_esp
parents:
675
diff
changeset
|
43 void ff_fdct_mmx(DCTELEM *block); |
0 | 44 |
34 | 45 /* encoding scans */ |
1064 | 46 extern const uint8_t ff_alternate_horizontal_scan[64]; |
47 extern const uint8_t ff_alternate_vertical_scan[64]; | |
48 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
|
49 |
0 | 50 /* pixel operations */ |
51 #define MAX_NEG_CROP 384 | |
52 | |
53 /* temporary */ | |
1064 | 54 extern uint32_t squareTbl[512]; |
55 extern uint8_t cropTbl[256 + 2 * MAX_NEG_CROP]; | |
0 | 56 |
57 | |
675 | 58 /* minimum alignment rules ;) |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
59 if u notice errors in the align stuff, need more alignment for some asm code for some cpu |
675 | 60 or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ... |
61 | |
62 !warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible) | |
63 i (michael) didnt check them, these are just the alignents which i think could be reached easily ... | |
64 | |
65 !future video codecs might need functions with less strict alignment | |
66 */ | |
0 | 67 |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
68 /* |
1064 | 69 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size); |
70 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride); | |
71 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size); | |
72 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size); | |
296 | 73 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
|
74 */ |
0 | 75 |
76 /* add and put pixel (decoding) */ | |
675 | 77 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16 |
1064 | 78 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h); |
79 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride); | |
0 | 80 |
984 | 81 #define DEF_OLD_QPEL(name)\ |
1064 | 82 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\ |
83 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\ | |
84 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
|
85 |
984 | 86 DEF_OLD_QPEL(qpel16_mc11_old_c) |
87 DEF_OLD_QPEL(qpel16_mc31_old_c) | |
88 DEF_OLD_QPEL(qpel16_mc12_old_c) | |
89 DEF_OLD_QPEL(qpel16_mc32_old_c) | |
90 DEF_OLD_QPEL(qpel16_mc13_old_c) | |
91 DEF_OLD_QPEL(qpel16_mc33_old_c) | |
92 DEF_OLD_QPEL(qpel8_mc11_old_c) | |
93 DEF_OLD_QPEL(qpel8_mc31_old_c) | |
94 DEF_OLD_QPEL(qpel8_mc12_old_c) | |
95 DEF_OLD_QPEL(qpel8_mc32_old_c) | |
96 DEF_OLD_QPEL(qpel8_mc13_old_c) | |
97 DEF_OLD_QPEL(qpel8_mc33_old_c) | |
651 | 98 |
99 #define CALL_2X_PIXELS(a, b, n)\ | |
100 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ | |
101 b(block , pixels , line_size, h);\ | |
102 b(block+n, pixels+n, line_size, h);\ | |
103 } | |
255 | 104 |
0 | 105 /* motion estimation */ |
106 | |
1064 | 107 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 | 108 |
1064 | 109 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 | 110 |
1102 | 111 /** |
112 * DSPContext. | |
113 */ | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
114 typedef struct DSPContext { |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
115 /* pixel ops : interface with DCT */ |
1064 | 116 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size); |
117 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride); | |
118 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); | |
119 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size); | |
120 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder); | |
121 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
|
122 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
|
123 void (*clear_blocks)(DCTELEM *blocks/*align 16*/); |
1064 | 124 int (*pix_sum)(uint8_t * pix, int line_size); |
125 int (*pix_norm1)(uint8_t * pix, int line_size); | |
936 | 126 me_cmp_func sad[2]; /* identical to pix_absAxA except additional void * */ |
127 me_cmp_func sse[2]; | |
128 me_cmp_func hadamard8_diff[2]; | |
129 me_cmp_func dct_sad[2]; | |
130 me_cmp_func quant_psnr[2]; | |
1007 | 131 me_cmp_func bit[2]; |
132 me_cmp_func rd[2]; | |
936 | 133 int (*hadamard8_abs )(uint8_t *src, int stride, int mean); |
134 | |
954 | 135 me_cmp_func me_pre_cmp[11]; |
936 | 136 me_cmp_func me_cmp[11]; |
137 me_cmp_func me_sub_cmp[11]; | |
138 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
|
139 |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
140 /* 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
|
141 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
|
142 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
|
143 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
|
144 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
|
145 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
|
146 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
|
147 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
|
148 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]; |
936 | 149 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
|
150 |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
151 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
|
152 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
|
153 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
|
154 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
|
155 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
|
156 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
|
157 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
|
158 op_pixels_abs_func pix_abs8x8_xy2; |
866 | 159 |
160 /* huffyuv specific */ | |
161 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w); | |
936 | 162 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w); |
1092 | 163 |
164 /* (I)DCT */ | |
165 void (*fdct)(DCTELEM *block/* align 16*/); | |
1102 | 166 |
167 /** | |
1104 | 168 * block -> idct -> clip to unsigned 8 bit -> dest. |
1102 | 169 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...) |
170 * @param line_size size in pixels of a horizotal line of dest | |
171 */ | |
1092 | 172 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); |
1102 | 173 |
174 /** | |
175 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest. | |
176 * @param line_size size in pixels of a horizotal line of dest | |
177 */ | |
1092 | 178 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/); |
1102 | 179 |
180 /** | |
1104 | 181 * idct input permutation. |
1102 | 182 * an example to avoid confusion: |
183 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...) | |
184 * - (x -> referece dct -> reference idct -> x) | |
185 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x) | |
186 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...) | |
187 */ | |
1092 | 188 uint8_t idct_permutation[64]; |
189 int idct_permutation_type; | |
190 #define FF_NO_IDCT_PERM 1 | |
191 #define FF_LIBMPEG2_IDCT_PERM 2 | |
192 #define FF_SIMPLE_IDCT_PERM 3 | |
193 #define FF_TRANSPOSE_IDCT_PERM 4 | |
194 | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
195 } DSPContext; |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
196 |
1092 | 197 void dsputil_init(DSPContext* p, AVCodecContext *avctx); |
0 | 198 |
764 | 199 /** |
200 * permute block according to permuatation. | |
201 * @param last last non zero element in scantable order | |
202 */ | |
1064 | 203 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last); |
34 | 204 |
1102 | 205 /** |
1104 | 206 * Empty mmx state. |
1102 | 207 * this must be called between any dsp function and float/double code. |
208 * for example sin(); dsp->idct_put(); emms_c(); cos() | |
209 */ | |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
210 #define emms_c() |
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
211 |
995
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
212 /* should be defined by architectures supporting |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
213 one or more MultiMedia extension */ |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
214 int mm_support(void); |
edc10966b081
altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
michaelni
parents:
984
diff
changeset
|
215 |
62 | 216 #if defined(HAVE_MMX) |
0 | 217 |
862 | 218 #undef emms_c |
853
eacc2dd8fd9d
* using DSPContext - so each codec could use its local (sub)set of CPU extension
kabi
parents:
838
diff
changeset
|
219 |
0 | 220 #define MM_MMX 0x0001 /* standard MMX */ |
221 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
222 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
223 #define MM_SSE 0x0008 /* SSE functions */ | |
224 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
225 | |
226 extern int mm_flags; | |
227 | |
1064 | 228 void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size); |
229 void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size); | |
0 | 230 |
231 static inline void emms(void) | |
232 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
233 __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
|
234 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
235 |
936 | 236 |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
237 #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
|
238 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
239 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
|
240 emms();\ |
0 | 241 } |
242 | |
243 #define __align8 __attribute__ ((aligned (8))) | |
244 | |
1092 | 245 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx); |
246 void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx); | |
1065 | 247 |
62 | 248 #elif defined(ARCH_ARMV4L) |
249 | |
250 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
251 line ptimizations */ | |
252 #define __align8 __attribute__ ((aligned (4))) | |
253 | |
1092 | 254 void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx); |
62 | 255 |
88 | 256 #elif defined(HAVE_MLIB) |
257 | |
258 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
259 #define __align8 __attribute__ ((aligned (8))) | |
260 | |
1092 | 261 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx); |
88 | 262 |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
263 #elif defined(ARCH_ALPHA) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
264 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
265 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
266 |
1092 | 267 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
|
268 |
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
|
269 #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
|
270 |
894
a408778eff87
altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
884
diff
changeset
|
271 #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
|
272 |
a408778eff87
altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
michaelni
parents:
884
diff
changeset
|
273 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
|
274 |
1033
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
275 #if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN) |
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
276 #include <altivec.h> |
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
277 #endif |
b4172ff70d27
Altivec on non darwin systems patch by Romain Dolbeau
bellard
parents:
1008
diff
changeset
|
278 |
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
|
279 #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
|
280 |
1092 | 281 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
|
282 |
689
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
283 #elif defined(HAVE_MMI) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
284 |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
285 #define __align8 __attribute__ ((aligned (16))) |
efcbfbd18864
ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
687
diff
changeset
|
286 |
1092 | 287 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
|
288 |
0 | 289 #else |
290 | |
291 #define __align8 | |
292 | |
293 #endif | |
294 | |
493
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
295 #ifdef __GNUC__ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
296 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
297 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
|
298 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
|
299 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
300 #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
|
301 #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
|
302 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
303 #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
|
304 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
305 #else /* __GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
306 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
307 #define LD32(a) (*((uint32_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
308 #define LD64(a) (*((uint64_t*)(a))) |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
309 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
310 #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
|
311 |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
312 #endif /* !__GNUC__ */ |
873b9075d853
move unaligned access macros to dsputil.h - added unaligned 32 bit store
bellard
parents:
480
diff
changeset
|
313 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
314 /* PSNR */ |
1064 | 315 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
|
316 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
|
317 AVCodecContext *avctx); |
781 | 318 |
319 /* FFT computation */ | |
320 | |
321 /* NOTE: soon integer code will be added, so you must use the | |
322 FFTSample type */ | |
323 typedef float FFTSample; | |
324 | |
325 typedef struct FFTComplex { | |
326 FFTSample re, im; | |
327 } FFTComplex; | |
328 | |
329 typedef struct FFTContext { | |
330 int nbits; | |
331 int inverse; | |
332 uint16_t *revtab; | |
333 FFTComplex *exptab; | |
334 FFTComplex *exptab1; /* only used by SSE code */ | |
335 void (*fft_calc)(struct FFTContext *s, FFTComplex *z); | |
336 } FFTContext; | |
337 | |
338 int fft_init(FFTContext *s, int nbits, int inverse); | |
339 void fft_permute(FFTContext *s, FFTComplex *z); | |
340 void fft_calc_c(FFTContext *s, FFTComplex *z); | |
341 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
|
342 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
|
343 |
781 | 344 static inline void fft_calc(FFTContext *s, FFTComplex *z) |
345 { | |
346 s->fft_calc(s, z); | |
347 } | |
348 void fft_end(FFTContext *s); | |
349 | |
350 /* MDCT computation */ | |
351 | |
352 typedef struct MDCTContext { | |
353 int n; /* size of MDCT (i.e. number of input data * 2) */ | |
354 int nbits; /* n = 2^nbits */ | |
355 /* pre/post rotation tables */ | |
356 FFTSample *tcos; | |
357 FFTSample *tsin; | |
358 FFTContext fft; | |
359 } MDCTContext; | |
360 | |
794 | 361 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
|
362 void ff_imdct_calc(MDCTContext *s, FFTSample *output, |
781 | 363 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
|
364 void ff_mdct_calc(MDCTContext *s, FFTSample *out, |
781 | 365 const FFTSample *input, FFTSample *tmp); |
794 | 366 void ff_mdct_end(MDCTContext *s); |
781 | 367 |
936 | 368 #define WARPER88_1616(name8, name16)\ |
369 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride){\ | |
370 return name8(s, dst , src , stride)\ | |
371 +name8(s, dst+8 , src+8 , stride)\ | |
372 +name8(s, dst +8*stride, src +8*stride, stride)\ | |
373 +name8(s, dst+8+8*stride, src+8+8*stride, stride);\ | |
374 } | |
375 | |
838
b78812db886f
lrintf detection (based upon a patch by Franois Revol <revol at free dot fr>)
michaelni
parents:
802
diff
changeset
|
376 #ifndef HAVE_LRINTF |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
377 /* 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
|
378 /* better than nothing implementation. */ |
802 | 379 /* 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
|
380 static inline long int lrintf(float x) |
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
381 { |
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
|
382 #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
|
383 /* 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
|
384 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
|
385 #else |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
386 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
|
387 #endif |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
388 } |
0 | 389 #endif |
796
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
390 |
8a5b70c68fbd
added lrintf for non ISOC libcs (fixme: find a better test)
bellard
parents:
794
diff
changeset
|
391 #endif |