| 1 | /* |
| 2 | * Simple IDCT |
| 3 | * |
| 4 | * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at> |
| 5 | * |
| 6 | * This file is part of FFmpeg. |
| 7 | * |
| 8 | * FFmpeg is free software; you can redistribute it and/or |
| 9 | * modify it under the terms of the GNU Lesser General Public |
| 10 | * License as published by the Free Software Foundation; either |
| 11 | * version 2.1 of the License, or (at your option) any later version. |
| 12 | * |
| 13 | * FFmpeg is distributed in the hope that it will be useful, |
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 16 | * Lesser General Public License for more details. |
| 17 | * |
| 18 | * You should have received a copy of the GNU Lesser General Public |
| 19 | * License along with FFmpeg; if not, write to the Free Software |
| 20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 21 | */ |
| 22 | |
| 23 | /** |
| 24 | * @file |
| 25 | * simpleidct in C. |
| 26 | */ |
| 27 | |
| 28 | /* |
| 29 | based upon some outcommented c code from mpeg2dec (idct_mmx.c |
| 30 | written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>) |
| 31 | */ |
| 32 | |
| 33 | #include "simple_idct.h" |
| 34 | |
| 35 | #include "bit_depth_template.c" |
| 36 | |
| 37 | #undef W1 |
| 38 | #undef W2 |
| 39 | #undef W3 |
| 40 | #undef W4 |
| 41 | #undef W5 |
| 42 | #undef W6 |
| 43 | #undef W7 |
| 44 | #undef ROW_SHIFT |
| 45 | #undef COL_SHIFT |
| 46 | #undef DC_SHIFT |
| 47 | #undef MUL |
| 48 | #undef MAC |
| 49 | |
| 50 | #if BIT_DEPTH == 8 |
| 51 | |
| 52 | #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 53 | #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 54 | #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 55 | #define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 56 | #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 57 | #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 58 | #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
| 59 | |
| 60 | #define ROW_SHIFT 11 |
| 61 | #define COL_SHIFT 20 |
| 62 | #define DC_SHIFT 3 |
| 63 | |
| 64 | #define MUL(a, b) MUL16(a, b) |
| 65 | #define MAC(a, b, c) MAC16(a, b, c) |
| 66 | |
| 67 | #elif BIT_DEPTH == 10 || BIT_DEPTH == 12 |
| 68 | |
| 69 | #if BIT_DEPTH == 10 |
| 70 | #define W1 (22725*4) // 90901 |
| 71 | #define W2 (21407*4) // 85627 |
| 72 | #define W3 (19265*4) // 77062 |
| 73 | #define W4 (16384*4) // 65535 |
| 74 | #define W5 (12873*4) // 51491 |
| 75 | #define W6 ( 8867*4) // 35468 |
| 76 | #define W7 ( 4520*4) // 18081 |
| 77 | |
| 78 | #define ROW_SHIFT 15 |
| 79 | #define COL_SHIFT 20 |
| 80 | #define DC_SHIFT 1 |
| 81 | #else |
| 82 | #define W1 45451 |
| 83 | #define W2 42813 |
| 84 | #define W3 38531 |
| 85 | #define W4 32767 |
| 86 | #define W5 25746 |
| 87 | #define W6 17734 |
| 88 | #define W7 9041 |
| 89 | |
| 90 | #define ROW_SHIFT 16 |
| 91 | #define COL_SHIFT 17 |
| 92 | #define DC_SHIFT -1 |
| 93 | #endif |
| 94 | |
| 95 | #define MUL(a, b) ((a) * (b)) |
| 96 | #define MAC(a, b, c) ((a) += (b) * (c)) |
| 97 | |
| 98 | #else |
| 99 | |
| 100 | #error "Unsupported bitdepth" |
| 101 | |
| 102 | #endif |
| 103 | |
| 104 | static inline void FUNC(idctRowCondDC)(int16_t *row, int extra_shift) |
| 105 | { |
| 106 | int a0, a1, a2, a3, b0, b1, b2, b3; |
| 107 | |
| 108 | #if HAVE_FAST_64BIT |
| 109 | #define ROW0_MASK (0xffffLL << 48 * HAVE_BIGENDIAN) |
| 110 | if (((((uint64_t *)row)[0] & ~ROW0_MASK) | ((uint64_t *)row)[1]) == 0) { |
| 111 | uint64_t temp; |
| 112 | if (DC_SHIFT - extra_shift >= 0) { |
| 113 | temp = (row[0] * (1 << (DC_SHIFT - extra_shift))) & 0xffff; |
| 114 | } else { |
| 115 | temp = ((row[0] + (1<<(extra_shift - DC_SHIFT-1))) >> (extra_shift - DC_SHIFT)) & 0xffff; |
| 116 | } |
| 117 | temp += temp * (1 << 16); |
| 118 | temp += temp * ((uint64_t) 1 << 32); |
| 119 | ((uint64_t *)row)[0] = temp; |
| 120 | ((uint64_t *)row)[1] = temp; |
| 121 | return; |
| 122 | } |
| 123 | #else |
| 124 | if (!(((uint32_t*)row)[1] | |
| 125 | ((uint32_t*)row)[2] | |
| 126 | ((uint32_t*)row)[3] | |
| 127 | row[1])) { |
| 128 | uint32_t temp; |
| 129 | if (DC_SHIFT - extra_shift >= 0) { |
| 130 | temp = (row[0] * (1 << (DC_SHIFT - extra_shift))) & 0xffff; |
| 131 | } else { |
| 132 | temp = ((row[0] + (1<<(extra_shift - DC_SHIFT-1))) >> (extra_shift - DC_SHIFT)) & 0xffff; |
| 133 | } |
| 134 | temp += temp * (1 << 16); |
| 135 | ((uint32_t*)row)[0]=((uint32_t*)row)[1] = |
| 136 | ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp; |
| 137 | return; |
| 138 | } |
| 139 | #endif |
| 140 | |
| 141 | a0 = (W4 * row[0]) + (1 << (ROW_SHIFT + extra_shift - 1)); |
| 142 | a1 = a0; |
| 143 | a2 = a0; |
| 144 | a3 = a0; |
| 145 | |
| 146 | a0 += W2 * row[2]; |
| 147 | a1 += W6 * row[2]; |
| 148 | a2 -= W6 * row[2]; |
| 149 | a3 -= W2 * row[2]; |
| 150 | |
| 151 | b0 = MUL(W1, row[1]); |
| 152 | MAC(b0, W3, row[3]); |
| 153 | b1 = MUL(W3, row[1]); |
| 154 | MAC(b1, -W7, row[3]); |
| 155 | b2 = MUL(W5, row[1]); |
| 156 | MAC(b2, -W1, row[3]); |
| 157 | b3 = MUL(W7, row[1]); |
| 158 | MAC(b3, -W5, row[3]); |
| 159 | |
| 160 | if (AV_RN64A(row + 4)) { |
| 161 | a0 += W4*row[4] + W6*row[6]; |
| 162 | a1 += - W4*row[4] - W2*row[6]; |
| 163 | a2 += - W4*row[4] + W2*row[6]; |
| 164 | a3 += W4*row[4] - W6*row[6]; |
| 165 | |
| 166 | MAC(b0, W5, row[5]); |
| 167 | MAC(b0, W7, row[7]); |
| 168 | |
| 169 | MAC(b1, -W1, row[5]); |
| 170 | MAC(b1, -W5, row[7]); |
| 171 | |
| 172 | MAC(b2, W7, row[5]); |
| 173 | MAC(b2, W3, row[7]); |
| 174 | |
| 175 | MAC(b3, W3, row[5]); |
| 176 | MAC(b3, -W1, row[7]); |
| 177 | } |
| 178 | |
| 179 | row[0] = (a0 + b0) >> (ROW_SHIFT + extra_shift); |
| 180 | row[7] = (a0 - b0) >> (ROW_SHIFT + extra_shift); |
| 181 | row[1] = (a1 + b1) >> (ROW_SHIFT + extra_shift); |
| 182 | row[6] = (a1 - b1) >> (ROW_SHIFT + extra_shift); |
| 183 | row[2] = (a2 + b2) >> (ROW_SHIFT + extra_shift); |
| 184 | row[5] = (a2 - b2) >> (ROW_SHIFT + extra_shift); |
| 185 | row[3] = (a3 + b3) >> (ROW_SHIFT + extra_shift); |
| 186 | row[4] = (a3 - b3) >> (ROW_SHIFT + extra_shift); |
| 187 | } |
| 188 | |
| 189 | #define IDCT_COLS do { \ |
| 190 | a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); \ |
| 191 | a1 = a0; \ |
| 192 | a2 = a0; \ |
| 193 | a3 = a0; \ |
| 194 | \ |
| 195 | a0 += W2*col[8*2]; \ |
| 196 | a1 += W6*col[8*2]; \ |
| 197 | a2 += -W6*col[8*2]; \ |
| 198 | a3 += -W2*col[8*2]; \ |
| 199 | \ |
| 200 | b0 = MUL(W1, col[8*1]); \ |
| 201 | b1 = MUL(W3, col[8*1]); \ |
| 202 | b2 = MUL(W5, col[8*1]); \ |
| 203 | b3 = MUL(W7, col[8*1]); \ |
| 204 | \ |
| 205 | MAC(b0, W3, col[8*3]); \ |
| 206 | MAC(b1, -W7, col[8*3]); \ |
| 207 | MAC(b2, -W1, col[8*3]); \ |
| 208 | MAC(b3, -W5, col[8*3]); \ |
| 209 | \ |
| 210 | if (col[8*4]) { \ |
| 211 | a0 += W4*col[8*4]; \ |
| 212 | a1 += -W4*col[8*4]; \ |
| 213 | a2 += -W4*col[8*4]; \ |
| 214 | a3 += W4*col[8*4]; \ |
| 215 | } \ |
| 216 | \ |
| 217 | if (col[8*5]) { \ |
| 218 | MAC(b0, W5, col[8*5]); \ |
| 219 | MAC(b1, -W1, col[8*5]); \ |
| 220 | MAC(b2, W7, col[8*5]); \ |
| 221 | MAC(b3, W3, col[8*5]); \ |
| 222 | } \ |
| 223 | \ |
| 224 | if (col[8*6]) { \ |
| 225 | a0 += W6*col[8*6]; \ |
| 226 | a1 += -W2*col[8*6]; \ |
| 227 | a2 += W2*col[8*6]; \ |
| 228 | a3 += -W6*col[8*6]; \ |
| 229 | } \ |
| 230 | \ |
| 231 | if (col[8*7]) { \ |
| 232 | MAC(b0, W7, col[8*7]); \ |
| 233 | MAC(b1, -W5, col[8*7]); \ |
| 234 | MAC(b2, W3, col[8*7]); \ |
| 235 | MAC(b3, -W1, col[8*7]); \ |
| 236 | } \ |
| 237 | } while (0) |
| 238 | |
| 239 | static inline void FUNC(idctSparseColPut)(pixel *dest, int line_size, |
| 240 | int16_t *col) |
| 241 | { |
| 242 | int a0, a1, a2, a3, b0, b1, b2, b3; |
| 243 | |
| 244 | IDCT_COLS; |
| 245 | |
| 246 | dest[0] = av_clip_pixel((a0 + b0) >> COL_SHIFT); |
| 247 | dest += line_size; |
| 248 | dest[0] = av_clip_pixel((a1 + b1) >> COL_SHIFT); |
| 249 | dest += line_size; |
| 250 | dest[0] = av_clip_pixel((a2 + b2) >> COL_SHIFT); |
| 251 | dest += line_size; |
| 252 | dest[0] = av_clip_pixel((a3 + b3) >> COL_SHIFT); |
| 253 | dest += line_size; |
| 254 | dest[0] = av_clip_pixel((a3 - b3) >> COL_SHIFT); |
| 255 | dest += line_size; |
| 256 | dest[0] = av_clip_pixel((a2 - b2) >> COL_SHIFT); |
| 257 | dest += line_size; |
| 258 | dest[0] = av_clip_pixel((a1 - b1) >> COL_SHIFT); |
| 259 | dest += line_size; |
| 260 | dest[0] = av_clip_pixel((a0 - b0) >> COL_SHIFT); |
| 261 | } |
| 262 | |
| 263 | static inline void FUNC(idctSparseColAdd)(pixel *dest, int line_size, |
| 264 | int16_t *col) |
| 265 | { |
| 266 | int a0, a1, a2, a3, b0, b1, b2, b3; |
| 267 | |
| 268 | IDCT_COLS; |
| 269 | |
| 270 | dest[0] = av_clip_pixel(dest[0] + ((a0 + b0) >> COL_SHIFT)); |
| 271 | dest += line_size; |
| 272 | dest[0] = av_clip_pixel(dest[0] + ((a1 + b1) >> COL_SHIFT)); |
| 273 | dest += line_size; |
| 274 | dest[0] = av_clip_pixel(dest[0] + ((a2 + b2) >> COL_SHIFT)); |
| 275 | dest += line_size; |
| 276 | dest[0] = av_clip_pixel(dest[0] + ((a3 + b3) >> COL_SHIFT)); |
| 277 | dest += line_size; |
| 278 | dest[0] = av_clip_pixel(dest[0] + ((a3 - b3) >> COL_SHIFT)); |
| 279 | dest += line_size; |
| 280 | dest[0] = av_clip_pixel(dest[0] + ((a2 - b2) >> COL_SHIFT)); |
| 281 | dest += line_size; |
| 282 | dest[0] = av_clip_pixel(dest[0] + ((a1 - b1) >> COL_SHIFT)); |
| 283 | dest += line_size; |
| 284 | dest[0] = av_clip_pixel(dest[0] + ((a0 - b0) >> COL_SHIFT)); |
| 285 | } |
| 286 | |
| 287 | static inline void FUNC(idctSparseCol)(int16_t *col) |
| 288 | { |
| 289 | int a0, a1, a2, a3, b0, b1, b2, b3; |
| 290 | |
| 291 | IDCT_COLS; |
| 292 | |
| 293 | col[0 ] = ((a0 + b0) >> COL_SHIFT); |
| 294 | col[8 ] = ((a1 + b1) >> COL_SHIFT); |
| 295 | col[16] = ((a2 + b2) >> COL_SHIFT); |
| 296 | col[24] = ((a3 + b3) >> COL_SHIFT); |
| 297 | col[32] = ((a3 - b3) >> COL_SHIFT); |
| 298 | col[40] = ((a2 - b2) >> COL_SHIFT); |
| 299 | col[48] = ((a1 - b1) >> COL_SHIFT); |
| 300 | col[56] = ((a0 - b0) >> COL_SHIFT); |
| 301 | } |
| 302 | |
| 303 | void FUNC(ff_simple_idct_put)(uint8_t *dest_, int line_size, int16_t *block) |
| 304 | { |
| 305 | pixel *dest = (pixel *)dest_; |
| 306 | int i; |
| 307 | |
| 308 | line_size /= sizeof(pixel); |
| 309 | |
| 310 | for (i = 0; i < 8; i++) |
| 311 | FUNC(idctRowCondDC)(block + i*8, 0); |
| 312 | |
| 313 | for (i = 0; i < 8; i++) |
| 314 | FUNC(idctSparseColPut)(dest + i, line_size, block + i); |
| 315 | } |
| 316 | |
| 317 | void FUNC(ff_simple_idct_add)(uint8_t *dest_, int line_size, int16_t *block) |
| 318 | { |
| 319 | pixel *dest = (pixel *)dest_; |
| 320 | int i; |
| 321 | |
| 322 | line_size /= sizeof(pixel); |
| 323 | |
| 324 | for (i = 0; i < 8; i++) |
| 325 | FUNC(idctRowCondDC)(block + i*8, 0); |
| 326 | |
| 327 | for (i = 0; i < 8; i++) |
| 328 | FUNC(idctSparseColAdd)(dest + i, line_size, block + i); |
| 329 | } |
| 330 | |
| 331 | void FUNC(ff_simple_idct)(int16_t *block) |
| 332 | { |
| 333 | int i; |
| 334 | |
| 335 | for (i = 0; i < 8; i++) |
| 336 | FUNC(idctRowCondDC)(block + i*8, 0); |
| 337 | |
| 338 | for (i = 0; i < 8; i++) |
| 339 | FUNC(idctSparseCol)(block + i); |
| 340 | } |