Mercurial > libavcodec.hg
annotate dsputil.c @ 200:6ab301aaa652 libavcodec
(commit by michael)
dequantizers skip trailing zeros
msmpeg4 non-intra decoder has its dequantizer "build in" now
| author | arpi_esp |
|---|---|
| date | Sun, 13 Jan 2002 04:59:37 +0000 |
| parents | 9e0e56869d05 |
| children | c0d8ecae7ac5 |
| rev | line source |
|---|---|
| 0 | 1 /* |
| 2 * DSP utils | |
| 3 * Copyright (c) 2000, 2001 Gerard Lantau. | |
| 4 * | |
| 5 * This program is free software; you can redistribute it and/or modify | |
| 6 * it under the terms of the GNU General Public License as published by | |
| 7 * the Free Software Foundation; either version 2 of the License, or | |
| 8 * (at your option) any later version. | |
| 9 * | |
| 10 * This program 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 | |
| 13 * GNU General Public License for more details. | |
| 14 * | |
| 15 * You should have received a copy of the GNU General Public License | |
| 16 * along with this program; if not, write to the Free Software | |
| 17 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | |
| 18 */ | |
| 19 #include <stdlib.h> | |
| 20 #include <stdio.h> | |
| 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" |
| 0 | 24 |
|
19
82d4c9be9873
MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
arpi_esp
parents:
6
diff
changeset
|
25 void (*ff_idct)(DCTELEM *block); |
| 0 | 26 void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size); |
| 27 void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); | |
| 28 void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); | |
| 29 | |
| 30 op_pixels_abs_func pix_abs16x16; | |
| 31 op_pixels_abs_func pix_abs16x16_x2; | |
| 32 op_pixels_abs_func pix_abs16x16_y2; | |
| 33 op_pixels_abs_func pix_abs16x16_xy2; | |
| 34 | |
| 50 | 35 UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; |
| 0 | 36 UINT32 squareTbl[512]; |
| 37 | |
| 34 | 38 extern UINT16 default_intra_matrix[64]; |
| 39 extern UINT16 default_non_intra_matrix[64]; | |
| 40 | |
| 41 UINT8 zigzag_direct[64] = { | |
| 42 0, 1, 8, 16, 9, 2, 3, 10, | |
| 43 17, 24, 32, 25, 18, 11, 4, 5, | |
| 44 12, 19, 26, 33, 40, 48, 41, 34, | |
| 45 27, 20, 13, 6, 7, 14, 21, 28, | |
| 46 35, 42, 49, 56, 57, 50, 43, 36, | |
| 47 29, 22, 15, 23, 30, 37, 44, 51, | |
| 48 58, 59, 52, 45, 38, 31, 39, 46, | |
| 49 53, 60, 61, 54, 47, 55, 62, 63 | |
| 50 }; | |
| 51 | |
| 52 UINT8 ff_alternate_horizontal_scan[64] = { | |
| 53 0, 1, 2, 3, 8, 9, 16, 17, | |
| 54 10, 11, 4, 5, 6, 7, 15, 14, | |
| 55 13, 12, 19, 18, 24, 25, 32, 33, | |
| 56 26, 27, 20, 21, 22, 23, 28, 29, | |
| 57 30, 31, 34, 35, 40, 41, 48, 49, | |
| 58 42, 43, 36, 37, 38, 39, 44, 45, | |
| 59 46, 47, 50, 51, 56, 57, 58, 59, | |
| 60 52, 53, 54, 55, 60, 61, 62, 63, | |
| 61 }; | |
| 62 | |
| 63 UINT8 ff_alternate_vertical_scan[64] = { | |
| 64 0, 8, 16, 24, 1, 9, 2, 10, | |
| 65 17, 25, 32, 40, 48, 56, 57, 49, | |
| 66 41, 33, 26, 18, 3, 11, 4, 12, | |
| 67 19, 27, 34, 42, 50, 58, 35, 43, | |
| 68 51, 59, 20, 28, 5, 13, 6, 14, | |
| 69 21, 29, 36, 44, 52, 60, 37, 45, | |
| 70 53, 61, 22, 30, 7, 15, 23, 31, | |
| 71 38, 46, 54, 62, 39, 47, 55, 63, | |
| 72 }; | |
| 73 | |
|
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
74 |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
75 static UINT8 simple_mmx_permutation[64]={ |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
76 0x00, 0x08, 0x01, 0x09, 0x04, 0x0C, 0x05, 0x0D, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
77 0x10, 0x18, 0x11, 0x19, 0x14, 0x1C, 0x15, 0x1D, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
78 0x02, 0x0A, 0x03, 0x0B, 0x06, 0x0E, 0x07, 0x0F, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
79 0x12, 0x1A, 0x13, 0x1B, 0x16, 0x1E, 0x17, 0x1F, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
80 0x20, 0x28, 0x21, 0x29, 0x24, 0x2C, 0x25, 0x2D, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
81 0x30, 0x38, 0x31, 0x39, 0x34, 0x3C, 0x35, 0x3D, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
82 0x22, 0x2A, 0x23, 0x2B, 0x26, 0x2E, 0x27, 0x2F, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
83 0x32, 0x3A, 0x33, 0x3B, 0x36, 0x3E, 0x37, 0x3F, |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
84 }; |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
85 |
| 200 | 86 /* used to skip zeros at the end */ |
| 87 UINT8 zigzag_end[64]; | |
| 88 | |
|
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
89 UINT8 permutation[64]; |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
90 //UINT8 invPermutation[64]; |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
91 |
| 200 | 92 static void build_zigzag_end() |
| 93 { | |
| 94 int lastIndex; | |
| 95 int lastIndexAfterPerm=0; | |
| 96 for(lastIndex=0; lastIndex<64; lastIndex++) | |
| 97 { | |
| 98 if(zigzag_direct[lastIndex] > lastIndexAfterPerm) | |
| 99 lastIndexAfterPerm= zigzag_direct[lastIndex]; | |
| 100 zigzag_end[lastIndex]= lastIndexAfterPerm + 1; | |
| 101 } | |
| 102 } | |
| 103 | |
| 0 | 104 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size) |
| 105 { | |
| 106 DCTELEM *p; | |
| 107 const UINT8 *pix; | |
| 108 int i; | |
| 109 | |
| 110 /* read the pixels */ | |
| 111 p = block; | |
| 112 pix = pixels; | |
| 113 for(i=0;i<8;i++) { | |
| 114 p[0] = pix[0]; | |
| 115 p[1] = pix[1]; | |
| 116 p[2] = pix[2]; | |
| 117 p[3] = pix[3]; | |
| 118 p[4] = pix[4]; | |
| 119 p[5] = pix[5]; | |
| 120 p[6] = pix[6]; | |
| 121 p[7] = pix[7]; | |
| 122 pix += line_size; | |
| 123 p += 8; | |
| 124 } | |
| 125 } | |
| 126 | |
| 127 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size) | |
| 128 { | |
| 129 const DCTELEM *p; | |
| 130 UINT8 *pix; | |
| 131 int i; | |
| 132 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
| 133 | |
| 134 /* read the pixels */ | |
| 135 p = block; | |
| 136 pix = pixels; | |
| 137 for(i=0;i<8;i++) { | |
| 138 pix[0] = cm[p[0]]; | |
| 139 pix[1] = cm[p[1]]; | |
| 140 pix[2] = cm[p[2]]; | |
| 141 pix[3] = cm[p[3]]; | |
| 142 pix[4] = cm[p[4]]; | |
| 143 pix[5] = cm[p[5]]; | |
| 144 pix[6] = cm[p[6]]; | |
| 145 pix[7] = cm[p[7]]; | |
| 146 pix += line_size; | |
| 147 p += 8; | |
| 148 } | |
| 149 } | |
| 150 | |
| 151 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size) | |
| 152 { | |
| 153 const DCTELEM *p; | |
| 154 UINT8 *pix; | |
| 155 int i; | |
| 156 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
| 157 | |
| 158 /* read the pixels */ | |
| 159 p = block; | |
| 160 pix = pixels; | |
| 161 for(i=0;i<8;i++) { | |
| 162 pix[0] = cm[pix[0] + p[0]]; | |
| 163 pix[1] = cm[pix[1] + p[1]]; | |
| 164 pix[2] = cm[pix[2] + p[2]]; | |
| 165 pix[3] = cm[pix[3] + p[3]]; | |
| 166 pix[4] = cm[pix[4] + p[4]]; | |
| 167 pix[5] = cm[pix[5] + p[5]]; | |
| 168 pix[6] = cm[pix[6] + p[6]]; | |
| 169 pix[7] = cm[pix[7] + p[7]]; | |
| 170 pix += line_size; | |
| 171 p += 8; | |
| 172 } | |
| 173 } | |
| 174 | |
| 175 #define PIXOP(BTYPE, OPNAME, OP, INCR) \ | |
| 176 \ | |
| 177 static void OPNAME ## _pixels(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
| 178 { \ | |
| 179 BTYPE *p; \ | |
| 180 const UINT8 *pix; \ | |
| 181 \ | |
| 182 p = block; \ | |
| 183 pix = pixels; \ | |
| 184 do { \ | |
| 185 OP(p[0], pix[0]); \ | |
| 186 OP(p[1], pix[1]); \ | |
| 187 OP(p[2], pix[2]); \ | |
| 188 OP(p[3], pix[3]); \ | |
| 189 OP(p[4], pix[4]); \ | |
| 190 OP(p[5], pix[5]); \ | |
| 191 OP(p[6], pix[6]); \ | |
| 192 OP(p[7], pix[7]); \ | |
| 193 pix += line_size; \ | |
| 194 p += INCR; \ | |
| 195 } while (--h);; \ | |
| 196 } \ | |
| 197 \ | |
| 198 static void OPNAME ## _pixels_x2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
| 199 { \ | |
| 200 BTYPE *p; \ | |
| 201 const UINT8 *pix; \ | |
| 202 \ | |
| 203 p = block; \ | |
| 204 pix = pixels; \ | |
| 205 do { \ | |
| 206 OP(p[0], avg2(pix[0], pix[1])); \ | |
| 207 OP(p[1], avg2(pix[1], pix[2])); \ | |
| 208 OP(p[2], avg2(pix[2], pix[3])); \ | |
| 209 OP(p[3], avg2(pix[3], pix[4])); \ | |
| 210 OP(p[4], avg2(pix[4], pix[5])); \ | |
| 211 OP(p[5], avg2(pix[5], pix[6])); \ | |
| 212 OP(p[6], avg2(pix[6], pix[7])); \ | |
| 213 OP(p[7], avg2(pix[7], pix[8])); \ | |
| 214 pix += line_size; \ | |
| 215 p += INCR; \ | |
| 216 } while (--h); \ | |
| 217 } \ | |
| 218 \ | |
| 219 static void OPNAME ## _pixels_y2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
| 220 { \ | |
| 221 BTYPE *p; \ | |
| 222 const UINT8 *pix; \ | |
| 223 const UINT8 *pix1; \ | |
| 224 \ | |
| 225 p = block; \ | |
| 226 pix = pixels; \ | |
| 227 pix1 = pixels + line_size; \ | |
| 228 do { \ | |
| 229 OP(p[0], avg2(pix[0], pix1[0])); \ | |
| 230 OP(p[1], avg2(pix[1], pix1[1])); \ | |
| 231 OP(p[2], avg2(pix[2], pix1[2])); \ | |
| 232 OP(p[3], avg2(pix[3], pix1[3])); \ | |
| 233 OP(p[4], avg2(pix[4], pix1[4])); \ | |
| 234 OP(p[5], avg2(pix[5], pix1[5])); \ | |
| 235 OP(p[6], avg2(pix[6], pix1[6])); \ | |
| 236 OP(p[7], avg2(pix[7], pix1[7])); \ | |
| 237 pix += line_size; \ | |
| 238 pix1 += line_size; \ | |
| 239 p += INCR; \ | |
| 240 } while(--h); \ | |
| 241 } \ | |
| 242 \ | |
| 243 static void OPNAME ## _pixels_xy2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \ | |
| 244 { \ | |
| 245 BTYPE *p; \ | |
| 246 const UINT8 *pix; \ | |
| 247 const UINT8 *pix1; \ | |
| 248 \ | |
| 249 p = block; \ | |
| 250 pix = pixels; \ | |
| 251 pix1 = pixels + line_size; \ | |
| 252 do { \ | |
| 253 OP(p[0], avg4(pix[0], pix[1], pix1[0], pix1[1])); \ | |
| 254 OP(p[1], avg4(pix[1], pix[2], pix1[1], pix1[2])); \ | |
| 255 OP(p[2], avg4(pix[2], pix[3], pix1[2], pix1[3])); \ | |
| 256 OP(p[3], avg4(pix[3], pix[4], pix1[3], pix1[4])); \ | |
| 257 OP(p[4], avg4(pix[4], pix[5], pix1[4], pix1[5])); \ | |
| 258 OP(p[5], avg4(pix[5], pix[6], pix1[5], pix1[6])); \ | |
| 259 OP(p[6], avg4(pix[6], pix[7], pix1[6], pix1[7])); \ | |
| 260 OP(p[7], avg4(pix[7], pix[8], pix1[7], pix1[8])); \ | |
| 261 pix += line_size; \ | |
| 262 pix1 += line_size; \ | |
| 263 p += INCR; \ | |
| 264 } while(--h); \ | |
| 265 } \ | |
| 266 \ | |
| 267 void (*OPNAME ## _pixels_tab[4])(BTYPE *block, const UINT8 *pixels, int line_size, int h) = { \ | |
| 268 OPNAME ## _pixels, \ | |
| 269 OPNAME ## _pixels_x2, \ | |
| 270 OPNAME ## _pixels_y2, \ | |
| 271 OPNAME ## _pixels_xy2, \ | |
| 272 }; | |
| 273 | |
| 274 | |
| 275 /* rounding primitives */ | |
| 276 #define avg2(a,b) ((a+b+1)>>1) | |
| 277 #define avg4(a,b,c,d) ((a+b+c+d+2)>>2) | |
| 278 | |
| 279 #define op_put(a, b) a = b | |
| 280 #define op_avg(a, b) a = avg2(a, b) | |
| 281 #define op_sub(a, b) a -= b | |
| 282 | |
| 283 PIXOP(UINT8, put, op_put, line_size) | |
| 284 PIXOP(UINT8, avg, op_avg, line_size) | |
| 285 | |
| 286 PIXOP(DCTELEM, sub, op_sub, 8) | |
| 287 | |
| 288 /* not rounding primitives */ | |
| 289 #undef avg2 | |
| 290 #undef avg4 | |
| 291 #define avg2(a,b) ((a+b)>>1) | |
| 292 #define avg4(a,b,c,d) ((a+b+c+d+1)>>2) | |
| 293 | |
| 294 PIXOP(UINT8, put_no_rnd, op_put, line_size) | |
| 295 PIXOP(UINT8, avg_no_rnd, op_avg, line_size) | |
| 296 | |
| 297 /* motion estimation */ | |
| 298 | |
| 299 #undef avg2 | |
| 300 #undef avg4 | |
| 301 #define avg2(a,b) ((a+b+1)>>1) | |
| 302 #define avg4(a,b,c,d) ((a+b+c+d+2)>>2) | |
| 303 | |
| 304 int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h) | |
| 305 { | |
| 306 int s, i; | |
| 307 | |
| 308 s = 0; | |
| 309 for(i=0;i<h;i++) { | |
| 310 s += abs(pix1[0] - pix2[0]); | |
| 311 s += abs(pix1[1] - pix2[1]); | |
| 312 s += abs(pix1[2] - pix2[2]); | |
| 313 s += abs(pix1[3] - pix2[3]); | |
| 314 s += abs(pix1[4] - pix2[4]); | |
| 315 s += abs(pix1[5] - pix2[5]); | |
| 316 s += abs(pix1[6] - pix2[6]); | |
| 317 s += abs(pix1[7] - pix2[7]); | |
| 318 s += abs(pix1[8] - pix2[8]); | |
| 319 s += abs(pix1[9] - pix2[9]); | |
| 320 s += abs(pix1[10] - pix2[10]); | |
| 321 s += abs(pix1[11] - pix2[11]); | |
| 322 s += abs(pix1[12] - pix2[12]); | |
| 323 s += abs(pix1[13] - pix2[13]); | |
| 324 s += abs(pix1[14] - pix2[14]); | |
| 325 s += abs(pix1[15] - pix2[15]); | |
| 326 pix1 += line_size; | |
| 327 pix2 += line_size; | |
| 328 } | |
| 329 return s; | |
| 330 } | |
| 331 | |
| 332 int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h) | |
| 333 { | |
| 334 int s, i; | |
| 335 | |
| 336 s = 0; | |
| 337 for(i=0;i<h;i++) { | |
| 338 s += abs(pix1[0] - avg2(pix2[0], pix2[1])); | |
| 339 s += abs(pix1[1] - avg2(pix2[1], pix2[2])); | |
| 340 s += abs(pix1[2] - avg2(pix2[2], pix2[3])); | |
| 341 s += abs(pix1[3] - avg2(pix2[3], pix2[4])); | |
| 342 s += abs(pix1[4] - avg2(pix2[4], pix2[5])); | |
| 343 s += abs(pix1[5] - avg2(pix2[5], pix2[6])); | |
| 344 s += abs(pix1[6] - avg2(pix2[6], pix2[7])); | |
| 345 s += abs(pix1[7] - avg2(pix2[7], pix2[8])); | |
| 346 s += abs(pix1[8] - avg2(pix2[8], pix2[9])); | |
| 347 s += abs(pix1[9] - avg2(pix2[9], pix2[10])); | |
| 348 s += abs(pix1[10] - avg2(pix2[10], pix2[11])); | |
| 349 s += abs(pix1[11] - avg2(pix2[11], pix2[12])); | |
| 350 s += abs(pix1[12] - avg2(pix2[12], pix2[13])); | |
| 351 s += abs(pix1[13] - avg2(pix2[13], pix2[14])); | |
| 352 s += abs(pix1[14] - avg2(pix2[14], pix2[15])); | |
| 353 s += abs(pix1[15] - avg2(pix2[15], pix2[16])); | |
| 354 pix1 += line_size; | |
| 355 pix2 += line_size; | |
| 356 } | |
| 357 return s; | |
| 358 } | |
| 359 | |
| 360 int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h) | |
| 361 { | |
| 362 int s, i; | |
| 363 UINT8 *pix3 = pix2 + line_size; | |
| 364 | |
| 365 s = 0; | |
| 366 for(i=0;i<h;i++) { | |
| 367 s += abs(pix1[0] - avg2(pix2[0], pix3[0])); | |
| 368 s += abs(pix1[1] - avg2(pix2[1], pix3[1])); | |
| 369 s += abs(pix1[2] - avg2(pix2[2], pix3[2])); | |
| 370 s += abs(pix1[3] - avg2(pix2[3], pix3[3])); | |
| 371 s += abs(pix1[4] - avg2(pix2[4], pix3[4])); | |
| 372 s += abs(pix1[5] - avg2(pix2[5], pix3[5])); | |
| 373 s += abs(pix1[6] - avg2(pix2[6], pix3[6])); | |
| 374 s += abs(pix1[7] - avg2(pix2[7], pix3[7])); | |
| 375 s += abs(pix1[8] - avg2(pix2[8], pix3[8])); | |
| 376 s += abs(pix1[9] - avg2(pix2[9], pix3[9])); | |
| 377 s += abs(pix1[10] - avg2(pix2[10], pix3[10])); | |
| 378 s += abs(pix1[11] - avg2(pix2[11], pix3[11])); | |
| 379 s += abs(pix1[12] - avg2(pix2[12], pix3[12])); | |
| 380 s += abs(pix1[13] - avg2(pix2[13], pix3[13])); | |
| 381 s += abs(pix1[14] - avg2(pix2[14], pix3[14])); | |
| 382 s += abs(pix1[15] - avg2(pix2[15], pix3[15])); | |
| 383 pix1 += line_size; | |
| 384 pix2 += line_size; | |
| 385 pix3 += line_size; | |
| 386 } | |
| 387 return s; | |
| 388 } | |
| 389 | |
| 390 int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h) | |
| 391 { | |
| 392 int s, i; | |
| 393 UINT8 *pix3 = pix2 + line_size; | |
| 394 | |
| 395 s = 0; | |
| 396 for(i=0;i<h;i++) { | |
| 397 s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1])); | |
| 398 s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2])); | |
| 399 s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3])); | |
| 400 s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4])); | |
| 401 s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5])); | |
| 402 s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6])); | |
| 403 s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7])); | |
| 404 s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8])); | |
| 405 s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9])); | |
| 406 s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10])); | |
| 407 s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11])); | |
| 408 s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12])); | |
| 409 s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13])); | |
| 410 s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14])); | |
| 411 s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15])); | |
| 412 s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16])); | |
| 413 pix1 += line_size; | |
| 414 pix2 += line_size; | |
| 415 pix3 += line_size; | |
| 416 } | |
| 417 return s; | |
| 418 } | |
| 419 | |
| 34 | 420 /* permute block according so that it corresponds to the MMX idct |
| 421 order */ | |
|
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
422 #ifdef SIMPLE_IDCT |
|
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
423 /* 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
|
424 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
|
425 { |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
426 int i; |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
427 INT16 temp[64]; |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
428 |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
429 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
|
430 |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
431 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
|
432 } |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
433 #else |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
434 |
| 34 | 435 void block_permute(INT16 *block) |
| 436 { | |
| 437 int tmp1, tmp2, tmp3, tmp4, tmp5, tmp6; | |
| 438 int i; | |
| 439 | |
| 440 for(i=0;i<8;i++) { | |
| 441 tmp1 = block[1]; | |
| 442 tmp2 = block[2]; | |
| 443 tmp3 = block[3]; | |
| 444 tmp4 = block[4]; | |
| 445 tmp5 = block[5]; | |
| 446 tmp6 = block[6]; | |
| 447 block[1] = tmp2; | |
| 448 block[2] = tmp4; | |
| 449 block[3] = tmp6; | |
| 450 block[4] = tmp1; | |
| 451 block[5] = tmp3; | |
| 452 block[6] = tmp5; | |
| 453 block += 8; | |
| 454 } | |
| 455 } | |
|
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
456 #endif |
| 34 | 457 |
| 0 | 458 void dsputil_init(void) |
| 459 { | |
| 34 | 460 int i, j; |
| 88 | 461 int use_permuted_idct; |
| 0 | 462 |
| 463 for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i; | |
| 464 for(i=0;i<MAX_NEG_CROP;i++) { | |
| 465 cropTbl[i] = 0; | |
| 466 cropTbl[i + MAX_NEG_CROP + 256] = 255; | |
| 467 } | |
| 468 | |
| 469 for(i=0;i<512;i++) { | |
| 470 squareTbl[i] = (i - 256) * (i - 256); | |
| 471 } | |
| 472 | |
|
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
473 #ifdef SIMPLE_IDCT |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
474 ff_idct = simple_idct; |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
475 #else |
|
19
82d4c9be9873
MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
arpi_esp
parents:
6
diff
changeset
|
476 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
|
477 #endif |
| 0 | 478 get_pixels = get_pixels_c; |
| 479 put_pixels_clamped = put_pixels_clamped_c; | |
| 480 add_pixels_clamped = add_pixels_clamped_c; | |
| 481 | |
| 482 pix_abs16x16 = pix_abs16x16_c; | |
| 483 pix_abs16x16_x2 = pix_abs16x16_x2_c; | |
| 484 pix_abs16x16_y2 = pix_abs16x16_y2_c; | |
| 485 pix_abs16x16_xy2 = pix_abs16x16_xy2_c; | |
| 486 av_fdct = jpeg_fdct_ifast; | |
| 487 | |
| 88 | 488 use_permuted_idct = 1; |
| 34 | 489 |
| 2 | 490 #ifdef HAVE_MMX |
| 0 | 491 dsputil_init_mmx(); |
| 492 #endif | |
| 62 | 493 #ifdef ARCH_ARMV4L |
| 494 dsputil_init_armv4l(); | |
| 495 #endif | |
| 88 | 496 #ifdef HAVE_MLIB |
| 497 dsputil_init_mlib(); | |
| 498 use_permuted_idct = 0; | |
| 499 #endif | |
| 500 | |
|
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
501 #ifdef SIMPLE_IDCT |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
502 if(ff_idct == simple_idct) use_permuted_idct=0; |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
503 #endif |
|
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
504 |
|
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
505 if(use_permuted_idct) |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
506 #ifdef SIMPLE_IDCT |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
507 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
|
508 #else |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
509 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
|
510 #endif |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
511 else |
|
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
512 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
|
513 |
| 88 | 514 if (use_permuted_idct) { |
| 515 /* permute for IDCT */ | |
| 516 for(i=0;i<64;i++) { | |
| 517 j = zigzag_direct[i]; | |
| 518 zigzag_direct[i] = block_permute_op(j); | |
| 519 j = ff_alternate_horizontal_scan[i]; | |
| 520 ff_alternate_horizontal_scan[i] = block_permute_op(j); | |
| 521 j = ff_alternate_vertical_scan[i]; | |
| 522 ff_alternate_vertical_scan[i] = block_permute_op(j); | |
| 523 } | |
| 524 block_permute(default_intra_matrix); | |
| 525 block_permute(default_non_intra_matrix); | |
| 526 } | |
| 200 | 527 |
| 528 build_zigzag_end(); | |
| 0 | 529 } |
