Mercurial > libavcodec.hg
annotate dsputil.c @ 516:0cb7ab35ccb2 libavcodec
* Detect "retrict" keyword
* Use "restrict" in some dsputil routines (yields large speedup)
author | mellum |
---|---|
date | Thu, 04 Jul 2002 01:47:32 +0000 |
parents | 873b9075d853 |
children | 389e30fe7269 |
rev | line source |
---|---|
0 | 1 /* |
2 * DSP utils | |
429 | 3 * Copyright (c) 2000, 2001 Fabrice Bellard. |
0 | 4 * |
429 | 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. | |
0 | 9 * |
429 | 10 * This library is distributed in the hope that it will be useful, |
0 | 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
429 | 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
13 * Lesser General Public License for more details. | |
0 | 14 * |
429 | 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 | |
256 | 18 * |
385 | 19 * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at> |
0 | 20 */ |
21 #include "avcodec.h" | |
22 #include "dsputil.h" | |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
23 #include "simple_idct.h" |
516 | 24 #include "config.h" |
25 | |
26 /* Suppress restrict if it was not defined in config.h */ | |
27 #ifndef restrict | |
28 #define restrict | |
29 #endif | |
0 | 30 |
19
82d4c9be9873
MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
arpi_esp
parents:
6
diff
changeset
|
31 void (*ff_idct)(DCTELEM *block); |
480 | 32 void (*ff_idct_put)(UINT8 *dest, int line_size, DCTELEM *block); |
33 void (*ff_idct_add)(UINT8 *dest, int line_size, DCTELEM *block); | |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
468
diff
changeset
|
34 void (*av_fdct)(DCTELEM *block); |
0 | 35 void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size); |
324 | 36 void (*diff_pixels)(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride); |
0 | 37 void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); |
38 void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); | |
255 | 39 void (*gmc1)(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder); |
296 | 40 void (*clear_blocks)(DCTELEM *blocks); |
0 | 41 |
42 op_pixels_abs_func pix_abs16x16; | |
43 op_pixels_abs_func pix_abs16x16_x2; | |
44 op_pixels_abs_func pix_abs16x16_y2; | |
45 op_pixels_abs_func pix_abs16x16_xy2; | |
46 | |
294 | 47 op_pixels_abs_func pix_abs8x8; |
48 op_pixels_abs_func pix_abs8x8_x2; | |
49 op_pixels_abs_func pix_abs8x8_y2; | |
50 op_pixels_abs_func pix_abs8x8_xy2; | |
51 | |
50 | 52 UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; |
0 | 53 UINT32 squareTbl[512]; |
54 | |
435
9247ad420889
* compatibilized declaration with its original definition
kabi
parents:
429
diff
changeset
|
55 extern INT16 default_intra_matrix[64]; |
9247ad420889
* compatibilized declaration with its original definition
kabi
parents:
429
diff
changeset
|
56 extern INT16 default_non_intra_matrix[64]; |
9247ad420889
* compatibilized declaration with its original definition
kabi
parents:
429
diff
changeset
|
57 extern INT16 ff_mpeg4_default_intra_matrix[64]; |
9247ad420889
* compatibilized declaration with its original definition
kabi
parents:
429
diff
changeset
|
58 extern INT16 ff_mpeg4_default_non_intra_matrix[64]; |
34 | 59 |
60 UINT8 zigzag_direct[64] = { | |
61 0, 1, 8, 16, 9, 2, 3, 10, | |
62 17, 24, 32, 25, 18, 11, 4, 5, | |
63 12, 19, 26, 33, 40, 48, 41, 34, | |
64 27, 20, 13, 6, 7, 14, 21, 28, | |
65 35, 42, 49, 56, 57, 50, 43, 36, | |
66 29, 22, 15, 23, 30, 37, 44, 51, | |
67 58, 59, 52, 45, 38, 31, 39, 46, | |
68 53, 60, 61, 54, 47, 55, 62, 63 | |
69 }; | |
70 | |
220 | 71 /* not permutated inverse zigzag_direct + 1 for MMX quantizer */ |
72 UINT16 __align8 inv_zigzag_direct16[64]; | |
73 | |
74 /* not permutated zigzag_direct for MMX quantizer */ | |
75 UINT8 zigzag_direct_noperm[64]; | |
76 | |
34 | 77 UINT8 ff_alternate_horizontal_scan[64] = { |
78 0, 1, 2, 3, 8, 9, 16, 17, | |
79 10, 11, 4, 5, 6, 7, 15, 14, | |
80 13, 12, 19, 18, 24, 25, 32, 33, | |
81 26, 27, 20, 21, 22, 23, 28, 29, | |
82 30, 31, 34, 35, 40, 41, 48, 49, | |
83 42, 43, 36, 37, 38, 39, 44, 45, | |
84 46, 47, 50, 51, 56, 57, 58, 59, | |
85 52, 53, 54, 55, 60, 61, 62, 63, | |
86 }; | |
87 | |
88 UINT8 ff_alternate_vertical_scan[64] = { | |
89 0, 8, 16, 24, 1, 9, 2, 10, | |
90 17, 25, 32, 40, 48, 56, 57, 49, | |
91 41, 33, 26, 18, 3, 11, 4, 12, | |
92 19, 27, 34, 42, 50, 58, 35, 43, | |
93 51, 59, 20, 28, 5, 13, 6, 14, | |
94 21, 29, 36, 44, 52, 60, 37, 45, | |
95 53, 61, 22, 30, 7, 15, 23, 31, | |
96 38, 46, 54, 62, 39, 47, 55, 63, | |
97 }; | |
98 | |
320
cda7d0857baf
- ME setting moved to AVCodecContext/MpegEncContext, no longer a global.
pulento
parents:
312
diff
changeset
|
99 #ifdef SIMPLE_IDCT |
cda7d0857baf
- ME setting moved to AVCodecContext/MpegEncContext, no longer a global.
pulento
parents:
312
diff
changeset
|
100 |
209 | 101 /* Input permutation for the simple_idct_mmx */ |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
102 static UINT8 simple_mmx_permutation[64]={ |
209 | 103 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D, |
104 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D, | |
105 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D, | |
106 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F, | |
107 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F, | |
108 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D, | |
109 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F, | |
110 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F, | |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
111 }; |
320
cda7d0857baf
- ME setting moved to AVCodecContext/MpegEncContext, no longer a global.
pulento
parents:
312
diff
changeset
|
112 #endif |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
113 |
220 | 114 /* a*inverse[b]>>32 == a/b for all 0<=a<=65536 && 2<=b<=255 */ |
115 UINT32 inverse[256]={ | |
116 0, 4294967295U,2147483648U,1431655766, 1073741824, 858993460, 715827883, 613566757, | |
117 536870912, 477218589, 429496730, 390451573, 357913942, 330382100, 306783379, 286331154, | |
118 268435456, 252645136, 238609295, 226050911, 214748365, 204522253, 195225787, 186737709, | |
119 178956971, 171798692, 165191050, 159072863, 153391690, 148102321, 143165577, 138547333, | |
120 134217728, 130150525, 126322568, 122713352, 119304648, 116080198, 113025456, 110127367, | |
121 107374183, 104755300, 102261127, 99882961, 97612894, 95443718, 93368855, 91382283, | |
122 89478486, 87652394, 85899346, 84215046, 82595525, 81037119, 79536432, 78090315, | |
123 76695845, 75350304, 74051161, 72796056, 71582789, 70409300, 69273667, 68174085, | |
124 67108864, 66076420, 65075263, 64103990, 63161284, 62245903, 61356676, 60492498, | |
125 59652324, 58835169, 58040099, 57266231, 56512728, 55778797, 55063684, 54366675, | |
126 53687092, 53024288, 52377650, 51746594, 51130564, 50529028, 49941481, 49367441, | |
127 48806447, 48258060, 47721859, 47197443, 46684428, 46182445, 45691142, 45210183, | |
128 44739243, 44278014, 43826197, 43383509, 42949673, 42524429, 42107523, 41698712, | |
129 41297763, 40904451, 40518560, 40139882, 39768216, 39403370, 39045158, 38693400, | |
130 38347923, 38008561, 37675152, 37347542, 37025581, 36709123, 36398028, 36092163, | |
131 35791395, 35495598, 35204650, 34918434, 34636834, 34359739, 34087043, 33818641, | |
132 33554432, 33294321, 33038210, 32786010, 32537632, 32292988, 32051995, 31814573, | |
133 31580642, 31350127, 31122952, 30899046, 30678338, 30460761, 30246249, 30034737, | |
134 29826162, 29620465, 29417585, 29217465, 29020050, 28825284, 28633116, 28443493, | |
135 28256364, 28071682, 27889399, 27709467, 27531842, 27356480, 27183338, 27012373, | |
136 26843546, 26676816, 26512144, 26349493, 26188825, 26030105, 25873297, 25718368, | |
137 25565282, 25414008, 25264514, 25116768, 24970741, 24826401, 24683721, 24542671, | |
138 24403224, 24265352, 24129030, 23994231, 23860930, 23729102, 23598722, 23469767, | |
139 23342214, 23216040, 23091223, 22967740, 22845571, 22724695, 22605092, 22486740, | |
140 22369622, 22253717, 22139007, 22025474, 21913099, 21801865, 21691755, 21582751, | |
141 21474837, 21367997, 21262215, 21157475, 21053762, 20951060, 20849356, 20748635, | |
142 20648882, 20550083, 20452226, 20355296, 20259280, 20164166, 20069941, 19976593, | |
143 19884108, 19792477, 19701685, 19611723, 19522579, 19434242, 19346700, 19259944, | |
144 19173962, 19088744, 19004281, 18920561, 18837576, 18755316, 18673771, 18592933, | |
145 18512791, 18433337, 18354562, 18276457, 18199014, 18122225, 18046082, 17970575, | |
146 17895698, 17821442, 17747799, 17674763, 17602325, 17530479, 17459217, 17388532, | |
147 17318417, 17248865, 17179870, 17111424, 17043522, 16976156, 16909321, 16843010, | |
148 }; | |
149 | |
200 | 150 /* used to skip zeros at the end */ |
151 UINT8 zigzag_end[64]; | |
152 | |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
153 UINT8 permutation[64]; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
154 //UINT8 invPermutation[64]; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
155 |
468 | 156 static void build_zigzag_end(void) |
200 | 157 { |
158 int lastIndex; | |
159 int lastIndexAfterPerm=0; | |
160 for(lastIndex=0; lastIndex<64; lastIndex++) | |
161 { | |
162 if(zigzag_direct[lastIndex] > lastIndexAfterPerm) | |
163 lastIndexAfterPerm= zigzag_direct[lastIndex]; | |
164 zigzag_end[lastIndex]= lastIndexAfterPerm + 1; | |
165 } | |
166 } | |
167 | |
516 | 168 void get_pixels_c(DCTELEM *restrict block, const UINT8 *pixels, int line_size) |
0 | 169 { |
170 int i; | |
171 | |
172 /* read the pixels */ | |
173 for(i=0;i<8;i++) { | |
516 | 174 block[0] = pixels[0]; |
175 block[1] = pixels[1]; | |
176 block[2] = pixels[2]; | |
177 block[3] = pixels[3]; | |
178 block[4] = pixels[4]; | |
179 block[5] = pixels[5]; | |
180 block[6] = pixels[6]; | |
181 block[7] = pixels[7]; | |
182 pixels += line_size; | |
183 block += 8; | |
0 | 184 } |
185 } | |
186 | |
516 | 187 void diff_pixels_c(DCTELEM *restrict block, const UINT8 *s1, const UINT8 *s2, |
188 int stride){ | |
324 | 189 int i; |
190 | |
191 /* read the pixels */ | |
192 for(i=0;i<8;i++) { | |
516 | 193 block[0] = s1[0] - s2[0]; |
194 block[1] = s1[1] - s2[1]; | |
195 block[2] = s1[2] - s2[2]; | |
196 block[3] = s1[3] - s2[3]; | |
197 block[4] = s1[4] - s2[4]; | |
198 block[5] = s1[5] - s2[5]; | |
199 block[6] = s1[6] - s2[6]; | |
200 block[7] = s1[7] - s2[7]; | |
324 | 201 s1 += stride; |
202 s2 += stride; | |
516 | 203 block += 8; |
324 | 204 } |
205 } | |
206 | |
207 | |
516 | 208 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels, |
209 int line_size) | |
0 | 210 { |
211 int i; | |
212 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
213 | |
214 /* read the pixels */ | |
215 for(i=0;i<8;i++) { | |
516 | 216 pixels[0] = cm[block[0]]; |
217 pixels[1] = cm[block[1]]; | |
218 pixels[2] = cm[block[2]]; | |
219 pixels[3] = cm[block[3]]; | |
220 pixels[4] = cm[block[4]]; | |
221 pixels[5] = cm[block[5]]; | |
222 pixels[6] = cm[block[6]]; | |
223 pixels[7] = cm[block[7]]; | |
224 | |
225 pixels += line_size; | |
226 block += 8; | |
0 | 227 } |
228 } | |
229 | |
516 | 230 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels, |
231 int line_size) | |
0 | 232 { |
233 int i; | |
234 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
235 | |
236 /* read the pixels */ | |
237 for(i=0;i<8;i++) { | |
516 | 238 pixels[0] = cm[pixels[0] + block[0]]; |
239 pixels[1] = cm[pixels[1] + block[1]]; | |
240 pixels[2] = cm[pixels[2] + block[2]]; | |
241 pixels[3] = cm[pixels[3] + block[3]]; | |
242 pixels[4] = cm[pixels[4] + block[4]]; | |
243 pixels[5] = cm[pixels[5] + block[5]]; | |
244 pixels[6] = cm[pixels[6] + block[6]]; | |
245 pixels[7] = cm[pixels[7] + block[7]]; | |
246 pixels += line_size; | |
247 block += 8; | |
0 | 248 } |
249 } | |
250 | |
385 | 251 #if 0 |
252 | |
253 #define PIXOP2(OPNAME, OP) \ | |
254 void OPNAME ## _pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
255 {\ | |
256 int i;\ | |
257 for(i=0; i<h; i++){\ | |
258 OP(*((uint64_t*)block), LD64(pixels));\ | |
259 pixels+=line_size;\ | |
260 block +=line_size;\ | |
261 }\ | |
262 }\ | |
263 \ | |
264 void OPNAME ## _no_rnd_pixels_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
265 {\ | |
266 int i;\ | |
267 for(i=0; i<h; i++){\ | |
268 const uint64_t a= LD64(pixels );\ | |
269 const uint64_t b= LD64(pixels+1);\ | |
270 OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\ | |
271 pixels+=line_size;\ | |
272 block +=line_size;\ | |
273 }\ | |
274 }\ | |
275 \ | |
276 void OPNAME ## _pixels_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
277 {\ | |
278 int i;\ | |
279 for(i=0; i<h; i++){\ | |
280 const uint64_t a= LD64(pixels );\ | |
281 const uint64_t b= LD64(pixels+1);\ | |
282 OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\ | |
283 pixels+=line_size;\ | |
284 block +=line_size;\ | |
285 }\ | |
286 }\ | |
287 \ | |
288 void OPNAME ## _no_rnd_pixels_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
289 {\ | |
290 int i;\ | |
291 for(i=0; i<h; i++){\ | |
292 const uint64_t a= LD64(pixels );\ | |
293 const uint64_t b= LD64(pixels+line_size);\ | |
294 OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\ | |
295 pixels+=line_size;\ | |
296 block +=line_size;\ | |
297 }\ | |
298 }\ | |
299 \ | |
300 void OPNAME ## _pixels_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
301 {\ | |
302 int i;\ | |
303 for(i=0; i<h; i++){\ | |
304 const uint64_t a= LD64(pixels );\ | |
305 const uint64_t b= LD64(pixels+line_size);\ | |
306 OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\ | |
307 pixels+=line_size;\ | |
308 block +=line_size;\ | |
309 }\ | |
310 }\ | |
311 \ | |
312 void OPNAME ## _pixels_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
313 {\ | |
314 int i;\ | |
315 const uint64_t a= LD64(pixels );\ | |
316 const uint64_t b= LD64(pixels+1);\ | |
317 uint64_t l0= (a&0x0303030303030303ULL)\ | |
318 + (b&0x0303030303030303ULL)\ | |
319 + 0x0202020202020202ULL;\ | |
320 uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\ | |
321 + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\ | |
322 uint64_t l1,h1;\ | |
323 \ | |
324 pixels+=line_size;\ | |
325 for(i=0; i<h; i+=2){\ | |
326 uint64_t a= LD64(pixels );\ | |
327 uint64_t b= LD64(pixels+1);\ | |
328 l1= (a&0x0303030303030303ULL)\ | |
329 + (b&0x0303030303030303ULL);\ | |
330 h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\ | |
331 + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\ | |
332 OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\ | |
333 pixels+=line_size;\ | |
334 block +=line_size;\ | |
335 a= LD64(pixels );\ | |
336 b= LD64(pixels+1);\ | |
337 l0= (a&0x0303030303030303ULL)\ | |
338 + (b&0x0303030303030303ULL)\ | |
339 + 0x0202020202020202ULL;\ | |
340 h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\ | |
341 + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\ | |
342 OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\ | |
343 pixels+=line_size;\ | |
344 block +=line_size;\ | |
345 }\ | |
346 }\ | |
347 \ | |
348 void OPNAME ## _no_rnd_pixels_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
349 {\ | |
350 int i;\ | |
351 const uint64_t a= LD64(pixels );\ | |
352 const uint64_t b= LD64(pixels+1);\ | |
353 uint64_t l0= (a&0x0303030303030303ULL)\ | |
354 + (b&0x0303030303030303ULL)\ | |
355 + 0x0101010101010101ULL;\ | |
356 uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\ | |
357 + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\ | |
358 uint64_t l1,h1;\ | |
359 \ | |
360 pixels+=line_size;\ | |
361 for(i=0; i<h; i+=2){\ | |
362 uint64_t a= LD64(pixels );\ | |
363 uint64_t b= LD64(pixels+1);\ | |
364 l1= (a&0x0303030303030303ULL)\ | |
365 + (b&0x0303030303030303ULL);\ | |
366 h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\ | |
367 + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\ | |
368 OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\ | |
369 pixels+=line_size;\ | |
370 block +=line_size;\ | |
371 a= LD64(pixels );\ | |
372 b= LD64(pixels+1);\ | |
373 l0= (a&0x0303030303030303ULL)\ | |
374 + (b&0x0303030303030303ULL)\ | |
375 + 0x0101010101010101ULL;\ | |
376 h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\ | |
377 + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\ | |
378 OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\ | |
379 pixels+=line_size;\ | |
380 block +=line_size;\ | |
381 }\ | |
382 }\ | |
383 \ | |
384 void (*OPNAME ## _pixels_tab[4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\ | |
385 OPNAME ## _pixels,\ | |
386 OPNAME ## _pixels_x2,\ | |
387 OPNAME ## _pixels_y2,\ | |
388 OPNAME ## _pixels_xy2,\ | |
389 };\ | |
390 \ | |
391 void (*OPNAME ## _no_rnd_pixels_tab[4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\ | |
392 OPNAME ## _pixels,\ | |
393 OPNAME ## _no_rnd_pixels_x2,\ | |
394 OPNAME ## _no_rnd_pixels_y2,\ | |
395 OPNAME ## _no_rnd_pixels_xy2,\ | |
396 }; | |
397 | |
398 #define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEFEFEFEFEULL)>>1) ) | |
399 #else // 64 bit variant | |
400 | |
401 #define PIXOP2(OPNAME, OP) \ | |
402 void OPNAME ## _pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
403 {\ | |
404 int i;\ | |
405 for(i=0; i<h; i++){\ | |
406 OP(*((uint32_t*)(block )), LD32(pixels ));\ | |
407 OP(*((uint32_t*)(block+4)), LD32(pixels+4));\ | |
408 pixels+=line_size;\ | |
409 block +=line_size;\ | |
410 }\ | |
411 }\ | |
412 \ | |
413 void OPNAME ## _no_rnd_pixels_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
414 {\ | |
415 int i;\ | |
416 for(i=0; i<h; i++){\ | |
417 int j;\ | |
418 for(j=0; j<2; j++){\ | |
419 const uint32_t a= LD32(pixels );\ | |
420 const uint32_t b= LD32(pixels+1);\ | |
421 OP(*((uint32_t*)block), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\ | |
422 pixels+=4;\ | |
423 block +=4;\ | |
424 }\ | |
425 pixels+=line_size-8;\ | |
426 block +=line_size-8;\ | |
427 }\ | |
428 }\ | |
429 \ | |
430 void OPNAME ## _pixels_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
431 {\ | |
432 int i;\ | |
433 for(i=0; i<h; i++){\ | |
434 int j;\ | |
435 for(j=0; j<2; j++){\ | |
436 const uint32_t a= LD32(pixels );\ | |
437 const uint32_t b= LD32(pixels+1);\ | |
438 OP(*((uint32_t*)block), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\ | |
439 pixels+=4;\ | |
440 block +=4;\ | |
441 }\ | |
442 pixels+=line_size-8;\ | |
443 block +=line_size-8;\ | |
444 }\ | |
445 }\ | |
446 \ | |
447 void OPNAME ## _no_rnd_pixels_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
448 {\ | |
449 int i;\ | |
450 for(i=0; i<h; i++){\ | |
451 int j;\ | |
452 for(j=0; j<2; j++){\ | |
453 const uint32_t a= LD32(pixels );\ | |
454 const uint32_t b= LD32(pixels+line_size);\ | |
455 OP(*((uint32_t*)block), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\ | |
456 pixels+=4;\ | |
457 block +=4;\ | |
458 }\ | |
459 pixels+=line_size-8;\ | |
460 block +=line_size-8;\ | |
461 }\ | |
462 }\ | |
463 \ | |
464 void OPNAME ## _pixels_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
465 {\ | |
466 int i;\ | |
467 for(i=0; i<h; i++){\ | |
468 int j;\ | |
469 for(j=0; j<2; j++){\ | |
470 const uint32_t a= LD32(pixels );\ | |
471 const uint32_t b= LD32(pixels+line_size);\ | |
472 OP(*((uint32_t*)block), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\ | |
473 pixels+=4;\ | |
474 block +=4;\ | |
475 }\ | |
476 pixels+=line_size-8;\ | |
477 block +=line_size-8;\ | |
478 }\ | |
479 }\ | |
480 \ | |
481 void OPNAME ## _pixels_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
482 {\ | |
483 int j;\ | |
484 for(j=0; j<2; j++){\ | |
485 int i;\ | |
486 const uint32_t a= LD32(pixels );\ | |
487 const uint32_t b= LD32(pixels+1);\ | |
488 uint32_t l0= (a&0x03030303UL)\ | |
489 + (b&0x03030303UL)\ | |
490 + 0x02020202UL;\ | |
491 uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ | |
492 + ((b&0xFCFCFCFCUL)>>2);\ | |
493 uint32_t l1,h1;\ | |
494 \ | |
495 pixels+=line_size;\ | |
496 for(i=0; i<h; i+=2){\ | |
497 uint32_t a= LD32(pixels );\ | |
498 uint32_t b= LD32(pixels+1);\ | |
499 l1= (a&0x03030303UL)\ | |
500 + (b&0x03030303UL);\ | |
501 h1= ((a&0xFCFCFCFCUL)>>2)\ | |
502 + ((b&0xFCFCFCFCUL)>>2);\ | |
503 OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ | |
504 pixels+=line_size;\ | |
505 block +=line_size;\ | |
506 a= LD32(pixels );\ | |
507 b= LD32(pixels+1);\ | |
508 l0= (a&0x03030303UL)\ | |
509 + (b&0x03030303UL)\ | |
510 + 0x02020202UL;\ | |
511 h0= ((a&0xFCFCFCFCUL)>>2)\ | |
512 + ((b&0xFCFCFCFCUL)>>2);\ | |
513 OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ | |
514 pixels+=line_size;\ | |
515 block +=line_size;\ | |
516 }\ | |
517 pixels+=4-line_size*(h+1);\ | |
518 block +=4-line_size*h;\ | |
519 }\ | |
520 }\ | |
521 \ | |
522 void OPNAME ## _no_rnd_pixels_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ | |
523 {\ | |
524 int j;\ | |
525 for(j=0; j<2; j++){\ | |
526 int i;\ | |
527 const uint32_t a= LD32(pixels );\ | |
528 const uint32_t b= LD32(pixels+1);\ | |
529 uint32_t l0= (a&0x03030303UL)\ | |
530 + (b&0x03030303UL)\ | |
531 + 0x01010101UL;\ | |
532 uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ | |
533 + ((b&0xFCFCFCFCUL)>>2);\ | |
534 uint32_t l1,h1;\ | |
535 \ | |
536 pixels+=line_size;\ | |
537 for(i=0; i<h; i+=2){\ | |
538 uint32_t a= LD32(pixels );\ | |
539 uint32_t b= LD32(pixels+1);\ | |
540 l1= (a&0x03030303UL)\ | |
541 + (b&0x03030303UL);\ | |
542 h1= ((a&0xFCFCFCFCUL)>>2)\ | |
543 + ((b&0xFCFCFCFCUL)>>2);\ | |
544 OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ | |
545 pixels+=line_size;\ | |
546 block +=line_size;\ | |
547 a= LD32(pixels );\ | |
548 b= LD32(pixels+1);\ | |
549 l0= (a&0x03030303UL)\ | |
550 + (b&0x03030303UL)\ | |
551 + 0x01010101UL;\ | |
552 h0= ((a&0xFCFCFCFCUL)>>2)\ | |
553 + ((b&0xFCFCFCFCUL)>>2);\ | |
554 OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ | |
555 pixels+=line_size;\ | |
556 block +=line_size;\ | |
557 }\ | |
558 pixels+=4-line_size*(h+1);\ | |
559 block +=4-line_size*h;\ | |
560 }\ | |
561 }\ | |
562 \ | |
563 void (*OPNAME ## _pixels_tab[4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\ | |
564 OPNAME ## _pixels,\ | |
565 OPNAME ## _pixels_x2,\ | |
566 OPNAME ## _pixels_y2,\ | |
567 OPNAME ## _pixels_xy2,\ | |
568 };\ | |
569 \ | |
570 void (*OPNAME ## _no_rnd_pixels_tab[4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\ | |
571 OPNAME ## _pixels,\ | |
572 OPNAME ## _no_rnd_pixels_x2,\ | |
573 OPNAME ## _no_rnd_pixels_y2,\ | |
574 OPNAME ## _no_rnd_pixels_xy2,\ | |
575 }; | |
576 #define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEUL)>>1) ) | |
577 #endif | |
578 | |
579 #define op_put(a, b) a = b | |
580 | |
581 PIXOP2(avg, op_avg) | |
582 PIXOP2(put, op_put) | |
583 #undef op_avg | |
584 #undef op_put | |
585 | |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
586 #if 0 |
385 | 587 /* FIXME this stuff could be removed as its ot really used anymore */ |
0 | 588 #define PIXOP(BTYPE, OPNAME, OP, INCR) \ |
589 \ | |
590 static void OPNAME ## _pixels(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
591 { \ | |
592 BTYPE *p; \ | |
593 const UINT8 *pix; \ | |
594 \ | |
595 p = block; \ | |
596 pix = pixels; \ | |
597 do { \ | |
598 OP(p[0], pix[0]); \ | |
599 OP(p[1], pix[1]); \ | |
600 OP(p[2], pix[2]); \ | |
601 OP(p[3], pix[3]); \ | |
602 OP(p[4], pix[4]); \ | |
603 OP(p[5], pix[5]); \ | |
604 OP(p[6], pix[6]); \ | |
605 OP(p[7], pix[7]); \ | |
606 pix += line_size; \ | |
607 p += INCR; \ | |
608 } while (--h);; \ | |
609 } \ | |
610 \ | |
611 static void OPNAME ## _pixels_x2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
612 { \ | |
613 BTYPE *p; \ | |
614 const UINT8 *pix; \ | |
615 \ | |
616 p = block; \ | |
617 pix = pixels; \ | |
618 do { \ | |
619 OP(p[0], avg2(pix[0], pix[1])); \ | |
620 OP(p[1], avg2(pix[1], pix[2])); \ | |
621 OP(p[2], avg2(pix[2], pix[3])); \ | |
622 OP(p[3], avg2(pix[3], pix[4])); \ | |
623 OP(p[4], avg2(pix[4], pix[5])); \ | |
624 OP(p[5], avg2(pix[5], pix[6])); \ | |
625 OP(p[6], avg2(pix[6], pix[7])); \ | |
626 OP(p[7], avg2(pix[7], pix[8])); \ | |
627 pix += line_size; \ | |
628 p += INCR; \ | |
629 } while (--h); \ | |
630 } \ | |
631 \ | |
632 static void OPNAME ## _pixels_y2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
633 { \ | |
634 BTYPE *p; \ | |
635 const UINT8 *pix; \ | |
636 const UINT8 *pix1; \ | |
637 \ | |
638 p = block; \ | |
639 pix = pixels; \ | |
640 pix1 = pixels + line_size; \ | |
641 do { \ | |
642 OP(p[0], avg2(pix[0], pix1[0])); \ | |
643 OP(p[1], avg2(pix[1], pix1[1])); \ | |
644 OP(p[2], avg2(pix[2], pix1[2])); \ | |
645 OP(p[3], avg2(pix[3], pix1[3])); \ | |
646 OP(p[4], avg2(pix[4], pix1[4])); \ | |
647 OP(p[5], avg2(pix[5], pix1[5])); \ | |
648 OP(p[6], avg2(pix[6], pix1[6])); \ | |
649 OP(p[7], avg2(pix[7], pix1[7])); \ | |
650 pix += line_size; \ | |
651 pix1 += line_size; \ | |
652 p += INCR; \ | |
653 } while(--h); \ | |
654 } \ | |
655 \ | |
656 static void OPNAME ## _pixels_xy2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
657 { \ | |
658 BTYPE *p; \ | |
659 const UINT8 *pix; \ | |
660 const UINT8 *pix1; \ | |
661 \ | |
662 p = block; \ | |
663 pix = pixels; \ | |
664 pix1 = pixels + line_size; \ | |
665 do { \ | |
666 OP(p[0], avg4(pix[0], pix[1], pix1[0], pix1[1])); \ | |
667 OP(p[1], avg4(pix[1], pix[2], pix1[1], pix1[2])); \ | |
668 OP(p[2], avg4(pix[2], pix[3], pix1[2], pix1[3])); \ | |
669 OP(p[3], avg4(pix[3], pix[4], pix1[3], pix1[4])); \ | |
670 OP(p[4], avg4(pix[4], pix[5], pix1[4], pix1[5])); \ | |
671 OP(p[5], avg4(pix[5], pix[6], pix1[5], pix1[6])); \ | |
672 OP(p[6], avg4(pix[6], pix[7], pix1[6], pix1[7])); \ | |
673 OP(p[7], avg4(pix[7], pix[8], pix1[7], pix1[8])); \ | |
674 pix += line_size; \ | |
675 pix1 += line_size; \ | |
676 p += INCR; \ | |
677 } while(--h); \ | |
678 } \ | |
679 \ | |
680 void (*OPNAME ## _pixels_tab[4])(BTYPE *block, const UINT8 *pixels, int line_size, int h) = { \ | |
681 OPNAME ## _pixels, \ | |
682 OPNAME ## _pixels_x2, \ | |
683 OPNAME ## _pixels_y2, \ | |
684 OPNAME ## _pixels_xy2, \ | |
685 }; | |
686 | |
687 /* rounding primitives */ | |
688 #define avg2(a,b) ((a+b+1)>>1) | |
689 #define avg4(a,b,c,d) ((a+b+c+d+2)>>2) | |
690 | |
691 #define op_avg(a, b) a = avg2(a, b) | |
692 #define op_sub(a, b) a -= b | |
693 | |
694 PIXOP(DCTELEM, sub, op_sub, 8) | |
695 | |
696 /* not rounding primitives */ | |
697 #undef avg2 | |
698 #undef avg4 | |
699 #define avg2(a,b) ((a+b)>>1) | |
700 #define avg4(a,b,c,d) ((a+b+c+d+1)>>2) | |
701 | |
702 /* motion estimation */ | |
703 | |
704 #undef avg2 | |
705 #undef avg4 | |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
706 #endif |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
707 |
0 | 708 #define avg2(a,b) ((a+b+1)>>1) |
709 #define avg4(a,b,c,d) ((a+b+c+d+2)>>2) | |
710 | |
255 | 711 static void gmc1_c(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder) |
712 { | |
713 const int A=(16-x16)*(16-y16); | |
714 const int B=( x16)*(16-y16); | |
715 const int C=(16-x16)*( y16); | |
716 const int D=( x16)*( y16); | |
717 int i; | |
718 rounder= 128 - rounder; | |
719 | |
720 for(i=0; i<h; i++) | |
721 { | |
722 dst[0]= (A*src[0] + B*src[1] + C*src[srcStride+0] + D*src[srcStride+1] + rounder)>>8; | |
723 dst[1]= (A*src[1] + B*src[2] + C*src[srcStride+1] + D*src[srcStride+2] + rounder)>>8; | |
724 dst[2]= (A*src[2] + B*src[3] + C*src[srcStride+2] + D*src[srcStride+3] + rounder)>>8; | |
725 dst[3]= (A*src[3] + B*src[4] + C*src[srcStride+3] + D*src[srcStride+4] + rounder)>>8; | |
726 dst[4]= (A*src[4] + B*src[5] + C*src[srcStride+4] + D*src[srcStride+5] + rounder)>>8; | |
727 dst[5]= (A*src[5] + B*src[6] + C*src[srcStride+5] + D*src[srcStride+6] + rounder)>>8; | |
728 dst[6]= (A*src[6] + B*src[7] + C*src[srcStride+6] + D*src[srcStride+7] + rounder)>>8; | |
729 dst[7]= (A*src[7] + B*src[8] + C*src[srcStride+7] + D*src[srcStride+8] + rounder)>>8; | |
730 dst+= srcStride; | |
731 src+= srcStride; | |
732 } | |
733 } | |
734 | |
735 static void qpel_h_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h, int r) | |
736 { | |
737 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
738 int i; | |
739 for(i=0; i<h; i++) | |
740 { | |
294 | 741 dst[0]= cm[(((src[0]+src[1])*20 - (src[0]+src[2])*6 + (src[1]+src[3])*3 - (src[2]+src[4]) + r)>>5)]; |
742 dst[1]= cm[(((src[1]+src[2])*20 - (src[0]+src[3])*6 + (src[0]+src[4])*3 - (src[1]+src[5]) + r)>>5)]; | |
743 dst[2]= cm[(((src[2]+src[3])*20 - (src[1]+src[4])*6 + (src[0]+src[5])*3 - (src[0]+src[6]) + r)>>5)]; | |
744 dst[3]= cm[(((src[3]+src[4])*20 - (src[2]+src[5])*6 + (src[1]+src[6])*3 - (src[0]+src[7]) + r)>>5)]; | |
745 dst[4]= cm[(((src[4]+src[5])*20 - (src[3]+src[6])*6 + (src[2]+src[7])*3 - (src[1]+src[8]) + r)>>5)]; | |
746 dst[5]= cm[(((src[5]+src[6])*20 - (src[4]+src[7])*6 + (src[3]+src[8])*3 - (src[2]+src[8]) + r)>>5)]; | |
747 dst[6]= cm[(((src[6]+src[7])*20 - (src[5]+src[8])*6 + (src[4]+src[8])*3 - (src[3]+src[7]) + r)>>5)]; | |
748 dst[7]= cm[(((src[7]+src[8])*20 - (src[6]+src[8])*6 + (src[5]+src[7])*3 - (src[4]+src[6]) + r)>>5)]; | |
255 | 749 dst+=dstStride; |
750 src+=srcStride; | |
751 } | |
752 } | |
753 | |
754 static void qpel_v_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int w, int r) | |
755 { | |
756 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
757 int i; | |
758 for(i=0; i<w; i++) | |
759 { | |
760 const int src0= src[0*srcStride]; | |
761 const int src1= src[1*srcStride]; | |
762 const int src2= src[2*srcStride]; | |
763 const int src3= src[3*srcStride]; | |
764 const int src4= src[4*srcStride]; | |
765 const int src5= src[5*srcStride]; | |
766 const int src6= src[6*srcStride]; | |
767 const int src7= src[7*srcStride]; | |
768 const int src8= src[8*srcStride]; | |
294 | 769 dst[0*dstStride]= cm[(((src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4) + r)>>5)]; |
770 dst[1*dstStride]= cm[(((src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5) + r)>>5)]; | |
771 dst[2*dstStride]= cm[(((src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6) + r)>>5)]; | |
772 dst[3*dstStride]= cm[(((src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7) + r)>>5)]; | |
773 dst[4*dstStride]= cm[(((src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8) + r)>>5)]; | |
774 dst[5*dstStride]= cm[(((src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8) + r)>>5)]; | |
775 dst[6*dstStride]= cm[(((src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7) + r)>>5)]; | |
776 dst[7*dstStride]= cm[(((src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6) + r)>>5)]; | |
255 | 777 dst++; |
778 src++; | |
779 } | |
780 } | |
781 | |
782 static inline void put_block(UINT8 *dst, UINT8 *src, int dstStride, int srcStride) | |
783 { | |
784 int i; | |
785 for(i=0; i<8; i++) | |
786 { | |
787 dst[0]= src[0]; | |
788 dst[1]= src[1]; | |
789 dst[2]= src[2]; | |
790 dst[3]= src[3]; | |
791 dst[4]= src[4]; | |
792 dst[5]= src[5]; | |
793 dst[6]= src[6]; | |
794 dst[7]= src[7]; | |
795 dst+=dstStride; | |
796 src+=srcStride; | |
797 } | |
798 } | |
799 | |
800 static inline void avg2_block(UINT8 *dst, UINT8 *src1, UINT8 *src2, int dstStride, int srcStride, int r) | |
801 { | |
802 int i; | |
803 for(i=0; i<8; i++) | |
804 { | |
805 dst[0]= (src1[0] + src2[0] + r)>>1; | |
806 dst[1]= (src1[1] + src2[1] + r)>>1; | |
807 dst[2]= (src1[2] + src2[2] + r)>>1; | |
808 dst[3]= (src1[3] + src2[3] + r)>>1; | |
809 dst[4]= (src1[4] + src2[4] + r)>>1; | |
810 dst[5]= (src1[5] + src2[5] + r)>>1; | |
811 dst[6]= (src1[6] + src2[6] + r)>>1; | |
812 dst[7]= (src1[7] + src2[7] + r)>>1; | |
813 dst+=dstStride; | |
814 src1+=srcStride; | |
815 src2+=8; | |
816 } | |
817 } | |
818 | |
819 static inline void avg4_block(UINT8 *dst, UINT8 *src1, UINT8 *src2, UINT8 *src3, UINT8 *src4, int dstStride, int srcStride, int r) | |
820 { | |
821 int i; | |
822 for(i=0; i<8; i++) | |
823 { | |
824 dst[0]= (src1[0] + src2[0] + src3[0] + src4[0] + r)>>2; | |
825 dst[1]= (src1[1] + src2[1] + src3[1] + src4[1] + r)>>2; | |
826 dst[2]= (src1[2] + src2[2] + src3[2] + src4[2] + r)>>2; | |
827 dst[3]= (src1[3] + src2[3] + src3[3] + src4[3] + r)>>2; | |
828 dst[4]= (src1[4] + src2[4] + src3[4] + src4[4] + r)>>2; | |
829 dst[5]= (src1[5] + src2[5] + src3[5] + src4[5] + r)>>2; | |
830 dst[6]= (src1[6] + src2[6] + src3[6] + src4[6] + r)>>2; | |
831 dst[7]= (src1[7] + src2[7] + src3[7] + src4[7] + r)>>2; | |
832 dst+=dstStride; | |
833 src1+=srcStride; | |
834 src2+=8; | |
256 | 835 src3+=8; |
255 | 836 src4+=8; |
837 } | |
838 } | |
839 | |
840 #define QPEL_MC(r, name) \ | |
841 static void qpel_mc00_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
842 {\ | |
843 put_block(dst, src, dstStride, srcStride);\ | |
844 }\ | |
845 \ | |
846 static void qpel_mc10_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
847 {\ | |
848 UINT8 half[64];\ | |
294 | 849 qpel_h_lowpass(half, src, 8, srcStride, 8, 16-r);\ |
255 | 850 avg2_block(dst, src, half, dstStride, srcStride, 1-r);\ |
851 }\ | |
852 \ | |
853 static void qpel_mc20_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
854 {\ | |
294 | 855 qpel_h_lowpass(dst, src, dstStride, srcStride, 8, 16-r);\ |
255 | 856 }\ |
857 \ | |
858 static void qpel_mc30_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
859 {\ | |
860 UINT8 half[64];\ | |
294 | 861 qpel_h_lowpass(half, src, 8, srcStride, 8, 16-r);\ |
255 | 862 avg2_block(dst, src+1, half, dstStride, srcStride, 1-r);\ |
863 }\ | |
864 \ | |
865 static void qpel_mc01_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
866 {\ | |
867 UINT8 half[64];\ | |
294 | 868 qpel_v_lowpass(half, src, 8, srcStride, 8, 16-r);\ |
255 | 869 avg2_block(dst, src, half, dstStride, srcStride, 1-r);\ |
870 }\ | |
871 \ | |
872 static void qpel_mc02_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
873 {\ | |
294 | 874 qpel_v_lowpass(dst, src, dstStride, srcStride, 8, 16-r);\ |
255 | 875 }\ |
876 \ | |
877 static void qpel_mc03_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
878 {\ | |
879 UINT8 half[64];\ | |
294 | 880 qpel_v_lowpass(half, src, 8, srcStride, 8, 16-r);\ |
255 | 881 avg2_block(dst, src+srcStride, half, dstStride, srcStride, 1-r);\ |
882 }\ | |
883 static void qpel_mc11_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
884 {\ | |
885 UINT8 halfH[72];\ | |
256 | 886 UINT8 halfV[64];\ |
255 | 887 UINT8 halfHV[64];\ |
294 | 888 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
889 qpel_v_lowpass(halfV, src, 8, srcStride, 8, 16-r);\ | |
890 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
255 | 891 avg4_block(dst, src, halfH, halfV, halfHV, dstStride, srcStride, 2-r);\ |
892 }\ | |
893 static void qpel_mc31_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
894 {\ | |
895 UINT8 halfH[72];\ | |
256 | 896 UINT8 halfV[64];\ |
255 | 897 UINT8 halfHV[64];\ |
294 | 898 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
899 qpel_v_lowpass(halfV, src+1, 8, srcStride, 8, 16-r);\ | |
900 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
255 | 901 avg4_block(dst, src+1, halfH, halfV, halfHV, dstStride, srcStride, 2-r);\ |
902 }\ | |
903 static void qpel_mc13_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
904 {\ | |
905 UINT8 halfH[72];\ | |
256 | 906 UINT8 halfV[64];\ |
255 | 907 UINT8 halfHV[64];\ |
294 | 908 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
909 qpel_v_lowpass(halfV, src, 8, srcStride, 8, 16-r);\ | |
910 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
256 | 911 avg4_block(dst, src+srcStride, halfH+8, halfV, halfHV, dstStride, srcStride, 2-r);\ |
255 | 912 }\ |
913 static void qpel_mc33_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
914 {\ | |
915 UINT8 halfH[72];\ | |
256 | 916 UINT8 halfV[64];\ |
255 | 917 UINT8 halfHV[64];\ |
294 | 918 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
919 qpel_v_lowpass(halfV, src+1, 8, srcStride, 8, 16-r);\ | |
920 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
256 | 921 avg4_block(dst, src+srcStride+1, halfH+8, halfV, halfHV, dstStride, srcStride, 2-r);\ |
255 | 922 }\ |
923 static void qpel_mc21_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
924 {\ | |
925 UINT8 halfH[72];\ | |
926 UINT8 halfHV[64];\ | |
294 | 927 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
928 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
255 | 929 avg2_block(dst, halfH, halfHV, dstStride, 8, 1-r);\ |
930 }\ | |
931 static void qpel_mc23_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
932 {\ | |
933 UINT8 halfH[72];\ | |
934 UINT8 halfHV[64];\ | |
294 | 935 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
936 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
255 | 937 avg2_block(dst, halfH+8, halfHV, dstStride, 8, 1-r);\ |
938 }\ | |
939 static void qpel_mc12_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
940 {\ | |
941 UINT8 halfH[72];\ | |
256 | 942 UINT8 halfV[64];\ |
255 | 943 UINT8 halfHV[64];\ |
294 | 944 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
945 qpel_v_lowpass(halfV, src, 8, srcStride, 8, 16-r);\ | |
946 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
256 | 947 avg2_block(dst, halfV, halfHV, dstStride, 8, 1-r);\ |
255 | 948 }\ |
949 static void qpel_mc32_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
950 {\ | |
951 UINT8 halfH[72];\ | |
256 | 952 UINT8 halfV[64];\ |
255 | 953 UINT8 halfHV[64];\ |
294 | 954 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
955 qpel_v_lowpass(halfV, src+1, 8, srcStride, 8, 16-r);\ | |
956 qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\ | |
256 | 957 avg2_block(dst, halfV, halfHV, dstStride, 8, 1-r);\ |
255 | 958 }\ |
959 static void qpel_mc22_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\ | |
960 {\ | |
961 UINT8 halfH[72];\ | |
294 | 962 qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\ |
963 qpel_v_lowpass(dst, halfH, dstStride, 8, 8, 16-r);\ | |
255 | 964 }\ |
965 qpel_mc_func qpel_mc ## name ## _tab[16]={ \ | |
966 qpel_mc00_c ## name, \ | |
967 qpel_mc10_c ## name, \ | |
968 qpel_mc20_c ## name, \ | |
969 qpel_mc30_c ## name, \ | |
970 qpel_mc01_c ## name, \ | |
971 qpel_mc11_c ## name, \ | |
972 qpel_mc21_c ## name, \ | |
973 qpel_mc31_c ## name, \ | |
974 qpel_mc02_c ## name, \ | |
975 qpel_mc12_c ## name, \ | |
976 qpel_mc22_c ## name, \ | |
977 qpel_mc32_c ## name, \ | |
978 qpel_mc03_c ## name, \ | |
979 qpel_mc13_c ## name, \ | |
980 qpel_mc23_c ## name, \ | |
981 qpel_mc33_c ## name, \ | |
982 }; | |
983 | |
984 QPEL_MC(0, _rnd) | |
985 QPEL_MC(1, _no_rnd) | |
986 | |
294 | 987 int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size) |
0 | 988 { |
989 int s, i; | |
990 | |
991 s = 0; | |
294 | 992 for(i=0;i<16;i++) { |
0 | 993 s += abs(pix1[0] - pix2[0]); |
994 s += abs(pix1[1] - pix2[1]); | |
995 s += abs(pix1[2] - pix2[2]); | |
996 s += abs(pix1[3] - pix2[3]); | |
997 s += abs(pix1[4] - pix2[4]); | |
998 s += abs(pix1[5] - pix2[5]); | |
999 s += abs(pix1[6] - pix2[6]); | |
1000 s += abs(pix1[7] - pix2[7]); | |
1001 s += abs(pix1[8] - pix2[8]); | |
1002 s += abs(pix1[9] - pix2[9]); | |
1003 s += abs(pix1[10] - pix2[10]); | |
1004 s += abs(pix1[11] - pix2[11]); | |
1005 s += abs(pix1[12] - pix2[12]); | |
1006 s += abs(pix1[13] - pix2[13]); | |
1007 s += abs(pix1[14] - pix2[14]); | |
1008 s += abs(pix1[15] - pix2[15]); | |
1009 pix1 += line_size; | |
1010 pix2 += line_size; | |
1011 } | |
1012 return s; | |
1013 } | |
1014 | |
294 | 1015 int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size) |
0 | 1016 { |
1017 int s, i; | |
1018 | |
1019 s = 0; | |
294 | 1020 for(i=0;i<16;i++) { |
0 | 1021 s += abs(pix1[0] - avg2(pix2[0], pix2[1])); |
1022 s += abs(pix1[1] - avg2(pix2[1], pix2[2])); | |
1023 s += abs(pix1[2] - avg2(pix2[2], pix2[3])); | |
1024 s += abs(pix1[3] - avg2(pix2[3], pix2[4])); | |
1025 s += abs(pix1[4] - avg2(pix2[4], pix2[5])); | |
1026 s += abs(pix1[5] - avg2(pix2[5], pix2[6])); | |
1027 s += abs(pix1[6] - avg2(pix2[6], pix2[7])); | |
1028 s += abs(pix1[7] - avg2(pix2[7], pix2[8])); | |
1029 s += abs(pix1[8] - avg2(pix2[8], pix2[9])); | |
1030 s += abs(pix1[9] - avg2(pix2[9], pix2[10])); | |
1031 s += abs(pix1[10] - avg2(pix2[10], pix2[11])); | |
1032 s += abs(pix1[11] - avg2(pix2[11], pix2[12])); | |
1033 s += abs(pix1[12] - avg2(pix2[12], pix2[13])); | |
1034 s += abs(pix1[13] - avg2(pix2[13], pix2[14])); | |
1035 s += abs(pix1[14] - avg2(pix2[14], pix2[15])); | |
1036 s += abs(pix1[15] - avg2(pix2[15], pix2[16])); | |
1037 pix1 += line_size; | |
1038 pix2 += line_size; | |
1039 } | |
1040 return s; | |
1041 } | |
1042 | |
294 | 1043 int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size) |
0 | 1044 { |
1045 int s, i; | |
1046 UINT8 *pix3 = pix2 + line_size; | |
1047 | |
1048 s = 0; | |
294 | 1049 for(i=0;i<16;i++) { |
0 | 1050 s += abs(pix1[0] - avg2(pix2[0], pix3[0])); |
1051 s += abs(pix1[1] - avg2(pix2[1], pix3[1])); | |
1052 s += abs(pix1[2] - avg2(pix2[2], pix3[2])); | |
1053 s += abs(pix1[3] - avg2(pix2[3], pix3[3])); | |
1054 s += abs(pix1[4] - avg2(pix2[4], pix3[4])); | |
1055 s += abs(pix1[5] - avg2(pix2[5], pix3[5])); | |
1056 s += abs(pix1[6] - avg2(pix2[6], pix3[6])); | |
1057 s += abs(pix1[7] - avg2(pix2[7], pix3[7])); | |
1058 s += abs(pix1[8] - avg2(pix2[8], pix3[8])); | |
1059 s += abs(pix1[9] - avg2(pix2[9], pix3[9])); | |
1060 s += abs(pix1[10] - avg2(pix2[10], pix3[10])); | |
1061 s += abs(pix1[11] - avg2(pix2[11], pix3[11])); | |
1062 s += abs(pix1[12] - avg2(pix2[12], pix3[12])); | |
1063 s += abs(pix1[13] - avg2(pix2[13], pix3[13])); | |
1064 s += abs(pix1[14] - avg2(pix2[14], pix3[14])); | |
1065 s += abs(pix1[15] - avg2(pix2[15], pix3[15])); | |
1066 pix1 += line_size; | |
1067 pix2 += line_size; | |
1068 pix3 += line_size; | |
1069 } | |
1070 return s; | |
1071 } | |
1072 | |
294 | 1073 int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size) |
0 | 1074 { |
1075 int s, i; | |
1076 UINT8 *pix3 = pix2 + line_size; | |
1077 | |
1078 s = 0; | |
294 | 1079 for(i=0;i<16;i++) { |
0 | 1080 s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1])); |
1081 s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2])); | |
1082 s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3])); | |
1083 s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4])); | |
1084 s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5])); | |
1085 s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6])); | |
1086 s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7])); | |
1087 s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8])); | |
1088 s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9])); | |
1089 s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10])); | |
1090 s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11])); | |
1091 s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12])); | |
1092 s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13])); | |
1093 s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14])); | |
1094 s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15])); | |
1095 s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16])); | |
1096 pix1 += line_size; | |
1097 pix2 += line_size; | |
1098 pix3 += line_size; | |
1099 } | |
1100 return s; | |
1101 } | |
1102 | |
294 | 1103 int pix_abs8x8_c(UINT8 *pix1, UINT8 *pix2, int line_size) |
1104 { | |
1105 int s, i; | |
1106 | |
1107 s = 0; | |
1108 for(i=0;i<8;i++) { | |
1109 s += abs(pix1[0] - pix2[0]); | |
1110 s += abs(pix1[1] - pix2[1]); | |
1111 s += abs(pix1[2] - pix2[2]); | |
1112 s += abs(pix1[3] - pix2[3]); | |
1113 s += abs(pix1[4] - pix2[4]); | |
1114 s += abs(pix1[5] - pix2[5]); | |
1115 s += abs(pix1[6] - pix2[6]); | |
1116 s += abs(pix1[7] - pix2[7]); | |
1117 pix1 += line_size; | |
1118 pix2 += line_size; | |
1119 } | |
1120 return s; | |
1121 } | |
1122 | |
1123 int pix_abs8x8_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size) | |
1124 { | |
1125 int s, i; | |
1126 | |
1127 s = 0; | |
1128 for(i=0;i<8;i++) { | |
1129 s += abs(pix1[0] - avg2(pix2[0], pix2[1])); | |
1130 s += abs(pix1[1] - avg2(pix2[1], pix2[2])); | |
1131 s += abs(pix1[2] - avg2(pix2[2], pix2[3])); | |
1132 s += abs(pix1[3] - avg2(pix2[3], pix2[4])); | |
1133 s += abs(pix1[4] - avg2(pix2[4], pix2[5])); | |
1134 s += abs(pix1[5] - avg2(pix2[5], pix2[6])); | |
1135 s += abs(pix1[6] - avg2(pix2[6], pix2[7])); | |
1136 s += abs(pix1[7] - avg2(pix2[7], pix2[8])); | |
1137 pix1 += line_size; | |
1138 pix2 += line_size; | |
1139 } | |
1140 return s; | |
1141 } | |
1142 | |
1143 int pix_abs8x8_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size) | |
1144 { | |
1145 int s, i; | |
1146 UINT8 *pix3 = pix2 + line_size; | |
1147 | |
1148 s = 0; | |
1149 for(i=0;i<8;i++) { | |
1150 s += abs(pix1[0] - avg2(pix2[0], pix3[0])); | |
1151 s += abs(pix1[1] - avg2(pix2[1], pix3[1])); | |
1152 s += abs(pix1[2] - avg2(pix2[2], pix3[2])); | |
1153 s += abs(pix1[3] - avg2(pix2[3], pix3[3])); | |
1154 s += abs(pix1[4] - avg2(pix2[4], pix3[4])); | |
1155 s += abs(pix1[5] - avg2(pix2[5], pix3[5])); | |
1156 s += abs(pix1[6] - avg2(pix2[6], pix3[6])); | |
1157 s += abs(pix1[7] - avg2(pix2[7], pix3[7])); | |
1158 pix1 += line_size; | |
1159 pix2 += line_size; | |
1160 pix3 += line_size; | |
1161 } | |
1162 return s; | |
1163 } | |
1164 | |
1165 int pix_abs8x8_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size) | |
1166 { | |
1167 int s, i; | |
1168 UINT8 *pix3 = pix2 + line_size; | |
1169 | |
1170 s = 0; | |
1171 for(i=0;i<8;i++) { | |
1172 s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1])); | |
1173 s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2])); | |
1174 s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3])); | |
1175 s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4])); | |
1176 s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5])); | |
1177 s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6])); | |
1178 s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7])); | |
1179 s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8])); | |
1180 pix1 += line_size; | |
1181 pix2 += line_size; | |
1182 pix3 += line_size; | |
1183 } | |
1184 return s; | |
1185 } | |
1186 | |
34 | 1187 /* permute block according so that it corresponds to the MMX idct |
1188 order */ | |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1189 #ifdef SIMPLE_IDCT |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1190 /* general permutation, but perhaps slightly slower */ |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1191 void block_permute(INT16 *block) |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1192 { |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1193 int i; |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1194 INT16 temp[64]; |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1195 |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1196 for(i=0; i<64; i++) temp[ block_permute_op(i) ] = block[i]; |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1197 |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1198 for(i=0; i<64; i++) block[i] = temp[i]; |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1199 } |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1200 #else |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1201 |
34 | 1202 void block_permute(INT16 *block) |
1203 { | |
1204 int tmp1, tmp2, tmp3, tmp4, tmp5, tmp6; | |
1205 int i; | |
1206 | |
1207 for(i=0;i<8;i++) { | |
1208 tmp1 = block[1]; | |
1209 tmp2 = block[2]; | |
1210 tmp3 = block[3]; | |
1211 tmp4 = block[4]; | |
1212 tmp5 = block[5]; | |
1213 tmp6 = block[6]; | |
1214 block[1] = tmp2; | |
1215 block[2] = tmp4; | |
1216 block[3] = tmp6; | |
1217 block[4] = tmp1; | |
1218 block[5] = tmp3; | |
1219 block[6] = tmp5; | |
1220 block += 8; | |
1221 } | |
1222 } | |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1223 #endif |
34 | 1224 |
296 | 1225 void clear_blocks_c(DCTELEM *blocks) |
1226 { | |
1227 memset(blocks, 0, sizeof(DCTELEM)*6*64); | |
1228 } | |
1229 | |
480 | 1230 /* XXX: those functions should be suppressed ASAP when all IDCTs are |
1231 converted */ | |
1232 void gen_idct_put(UINT8 *dest, int line_size, DCTELEM *block) | |
1233 { | |
1234 ff_idct (block); | |
1235 put_pixels_clamped(block, dest, line_size); | |
1236 } | |
1237 | |
1238 void gen_idct_add(UINT8 *dest, int line_size, DCTELEM *block) | |
1239 { | |
1240 ff_idct (block); | |
1241 add_pixels_clamped(block, dest, line_size); | |
1242 } | |
1243 | |
0 | 1244 void dsputil_init(void) |
1245 { | |
34 | 1246 int i, j; |
88 | 1247 int use_permuted_idct; |
0 | 1248 |
1249 for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i; | |
1250 for(i=0;i<MAX_NEG_CROP;i++) { | |
1251 cropTbl[i] = 0; | |
1252 cropTbl[i + MAX_NEG_CROP + 256] = 255; | |
1253 } | |
1254 | |
1255 for(i=0;i<512;i++) { | |
1256 squareTbl[i] = (i - 256) * (i - 256); | |
1257 } | |
1258 | |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1259 #ifdef SIMPLE_IDCT |
480 | 1260 ff_idct = NULL; |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1261 #else |
19
82d4c9be9873
MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
arpi_esp
parents:
6
diff
changeset
|
1262 ff_idct = j_rev_dct; |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1263 #endif |
0 | 1264 get_pixels = get_pixels_c; |
324 | 1265 diff_pixels = diff_pixels_c; |
0 | 1266 put_pixels_clamped = put_pixels_clamped_c; |
1267 add_pixels_clamped = add_pixels_clamped_c; | |
255 | 1268 gmc1= gmc1_c; |
296 | 1269 clear_blocks= clear_blocks_c; |
0 | 1270 |
294 | 1271 pix_abs16x16 = pix_abs16x16_c; |
1272 pix_abs16x16_x2 = pix_abs16x16_x2_c; | |
1273 pix_abs16x16_y2 = pix_abs16x16_y2_c; | |
0 | 1274 pix_abs16x16_xy2 = pix_abs16x16_xy2_c; |
294 | 1275 pix_abs8x8 = pix_abs8x8_c; |
1276 pix_abs8x8_x2 = pix_abs8x8_x2_c; | |
1277 pix_abs8x8_y2 = pix_abs8x8_y2_c; | |
1278 pix_abs8x8_xy2 = pix_abs8x8_xy2_c; | |
474
11dbd00682fc
avoid name clash with libjpeg - added missing externs
bellard
parents:
468
diff
changeset
|
1279 av_fdct = fdct_ifast; |
0 | 1280 |
88 | 1281 use_permuted_idct = 1; |
34 | 1282 |
2 | 1283 #ifdef HAVE_MMX |
0 | 1284 dsputil_init_mmx(); |
1285 #endif | |
62 | 1286 #ifdef ARCH_ARMV4L |
1287 dsputil_init_armv4l(); | |
1288 #endif | |
88 | 1289 #ifdef HAVE_MLIB |
1290 dsputil_init_mlib(); | |
1291 use_permuted_idct = 0; | |
1292 #endif | |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
209
diff
changeset
|
1293 #ifdef ARCH_ALPHA |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
209
diff
changeset
|
1294 dsputil_init_alpha(); |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
209
diff
changeset
|
1295 use_permuted_idct = 0; |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
209
diff
changeset
|
1296 #endif |
88 | 1297 |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1298 #ifdef SIMPLE_IDCT |
480 | 1299 if (ff_idct == NULL) { |
1300 ff_idct_put = simple_idct_put; | |
1301 ff_idct_add = simple_idct_add; | |
1302 use_permuted_idct=0; | |
1303 } else { | |
1304 ff_idct_put = gen_idct_put; | |
1305 ff_idct_add = gen_idct_add; | |
1306 } | |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1307 #endif |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
1308 |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1309 if(use_permuted_idct) |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1310 #ifdef SIMPLE_IDCT |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1311 for(i=0; i<64; i++) permutation[i]= simple_mmx_permutation[i]; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1312 #else |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1313 for(i=0; i<64; i++) permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2); |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1314 #endif |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1315 else |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1316 for(i=0; i<64; i++) permutation[i]=i; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
1317 |
220 | 1318 for(i=0; i<64; i++) inv_zigzag_direct16[zigzag_direct[i]]= i+1; |
1319 for(i=0; i<64; i++) zigzag_direct_noperm[i]= zigzag_direct[i]; | |
1320 | |
88 | 1321 if (use_permuted_idct) { |
1322 /* permute for IDCT */ | |
1323 for(i=0;i<64;i++) { | |
1324 j = zigzag_direct[i]; | |
1325 zigzag_direct[i] = block_permute_op(j); | |
1326 j = ff_alternate_horizontal_scan[i]; | |
1327 ff_alternate_horizontal_scan[i] = block_permute_op(j); | |
1328 j = ff_alternate_vertical_scan[i]; | |
1329 ff_alternate_vertical_scan[i] = block_permute_op(j); | |
1330 } | |
1331 block_permute(default_intra_matrix); | |
1332 block_permute(default_non_intra_matrix); | |
312 | 1333 block_permute(ff_mpeg4_default_intra_matrix); |
1334 block_permute(ff_mpeg4_default_non_intra_matrix); | |
88 | 1335 } |
200 | 1336 |
1337 build_zigzag_end(); | |
0 | 1338 } |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1339 |
403
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1340 /* remove any non bit exact operation (testing purpose) */ |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1341 void avcodec_set_bit_exact(void) |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1342 { |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1343 #ifdef HAVE_MMX |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1344 dsputil_set_bit_exact_mmx(); |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1345 #endif |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1346 } |
2c3e25f4c496
removed unused stuff - added dsputil_set_bit_exact() support for easier testing
glantau
parents:
398
diff
changeset
|
1347 |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1348 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:
220
diff
changeset
|
1349 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:
220
diff
changeset
|
1350 AVCodecContext *avctx) |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1351 { |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1352 int quad, diff, x, y; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1353 UINT8 *orig, *coded; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1354 UINT32 *sq = squareTbl + 256; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1355 |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1356 quad = 0; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1357 diff = 0; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1358 |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1359 /* Luminance */ |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1360 orig = orig_image[0]; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1361 coded = coded_image[0]; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1362 |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1363 for (y=0;y<avctx->height;y++) { |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1364 for (x=0;x<avctx->width;x++) { |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1365 diff = *(orig + x) - *(coded + x); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1366 quad += sq[diff]; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1367 } |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1368 orig += orig_linesize[0]; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1369 coded += coded_linesize; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1370 } |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1371 |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1372 avctx->psnr_y = (float) quad / (float) (avctx->width * avctx->height); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1373 |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1374 if (avctx->psnr_y) { |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1375 avctx->psnr_y = (float) (255 * 255) / avctx->psnr_y; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1376 avctx->psnr_y = 10 * (float) log10 (avctx->psnr_y); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1377 } else |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1378 avctx->psnr_y = 99.99; |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1379 } |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
220
diff
changeset
|
1380 |