| 1 | /* |
| 2 | * Lagarith lossless decoder |
| 3 | * Copyright (c) 2009 Nathan Caldwell <saintdev (at) gmail.com> |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | /** |
| 23 | * @file |
| 24 | * Lagarith lossless decoder |
| 25 | * @author Nathan Caldwell |
| 26 | */ |
| 27 | |
| 28 | #include <inttypes.h> |
| 29 | |
| 30 | #include "avcodec.h" |
| 31 | #include "get_bits.h" |
| 32 | #include "mathops.h" |
| 33 | #include "huffyuvdsp.h" |
| 34 | #include "lagarithrac.h" |
| 35 | #include "thread.h" |
| 36 | |
| 37 | enum LagarithFrameType { |
| 38 | FRAME_RAW = 1, /**< uncompressed */ |
| 39 | FRAME_U_RGB24 = 2, /**< unaligned RGB24 */ |
| 40 | FRAME_ARITH_YUY2 = 3, /**< arithmetic coded YUY2 */ |
| 41 | FRAME_ARITH_RGB24 = 4, /**< arithmetic coded RGB24 */ |
| 42 | FRAME_SOLID_GRAY = 5, /**< solid grayscale color frame */ |
| 43 | FRAME_SOLID_COLOR = 6, /**< solid non-grayscale color frame */ |
| 44 | FRAME_OLD_ARITH_RGB = 7, /**< obsolete arithmetic coded RGB (no longer encoded by upstream since version 1.1.0) */ |
| 45 | FRAME_ARITH_RGBA = 8, /**< arithmetic coded RGBA */ |
| 46 | FRAME_SOLID_RGBA = 9, /**< solid RGBA color frame */ |
| 47 | FRAME_ARITH_YV12 = 10, /**< arithmetic coded YV12 */ |
| 48 | FRAME_REDUCED_RES = 11, /**< reduced resolution YV12 frame */ |
| 49 | }; |
| 50 | |
| 51 | typedef struct LagarithContext { |
| 52 | AVCodecContext *avctx; |
| 53 | HuffYUVDSPContext hdsp; |
| 54 | int zeros; /**< number of consecutive zero bytes encountered */ |
| 55 | int zeros_rem; /**< number of zero bytes remaining to output */ |
| 56 | uint8_t *rgb_planes; |
| 57 | int rgb_planes_allocated; |
| 58 | int rgb_stride; |
| 59 | } LagarithContext; |
| 60 | |
| 61 | /** |
| 62 | * Compute the 52bit mantissa of 1/(double)denom. |
| 63 | * This crazy format uses floats in an entropy coder and we have to match x86 |
| 64 | * rounding exactly, thus ordinary floats aren't portable enough. |
| 65 | * @param denom denominator |
| 66 | * @return 52bit mantissa |
| 67 | * @see softfloat_mul |
| 68 | */ |
| 69 | static uint64_t softfloat_reciprocal(uint32_t denom) |
| 70 | { |
| 71 | int shift = av_log2(denom - 1) + 1; |
| 72 | uint64_t ret = (1ULL << 52) / denom; |
| 73 | uint64_t err = (1ULL << 52) - ret * denom; |
| 74 | ret <<= shift; |
| 75 | err <<= shift; |
| 76 | err += denom / 2; |
| 77 | return ret + err / denom; |
| 78 | } |
| 79 | |
| 80 | /** |
| 81 | * (uint32_t)(x*f), where f has the given mantissa, and exponent 0 |
| 82 | * Used in combination with softfloat_reciprocal computes x/(double)denom. |
| 83 | * @param x 32bit integer factor |
| 84 | * @param mantissa mantissa of f with exponent 0 |
| 85 | * @return 32bit integer value (x*f) |
| 86 | * @see softfloat_reciprocal |
| 87 | */ |
| 88 | static uint32_t softfloat_mul(uint32_t x, uint64_t mantissa) |
| 89 | { |
| 90 | uint64_t l = x * (mantissa & 0xffffffff); |
| 91 | uint64_t h = x * (mantissa >> 32); |
| 92 | h += l >> 32; |
| 93 | l &= 0xffffffff; |
| 94 | l += 1 << av_log2(h >> 21); |
| 95 | h += l >> 32; |
| 96 | return h >> 20; |
| 97 | } |
| 98 | |
| 99 | static uint8_t lag_calc_zero_run(int8_t x) |
| 100 | { |
| 101 | return (x << 1) ^ (x >> 7); |
| 102 | } |
| 103 | |
| 104 | static int lag_decode_prob(GetBitContext *gb, uint32_t *value) |
| 105 | { |
| 106 | static const uint8_t series[] = { 1, 2, 3, 5, 8, 13, 21 }; |
| 107 | int i; |
| 108 | int bit = 0; |
| 109 | int bits = 0; |
| 110 | int prevbit = 0; |
| 111 | unsigned val; |
| 112 | |
| 113 | for (i = 0; i < 7; i++) { |
| 114 | if (prevbit && bit) |
| 115 | break; |
| 116 | prevbit = bit; |
| 117 | bit = get_bits1(gb); |
| 118 | if (bit && !prevbit) |
| 119 | bits += series[i]; |
| 120 | } |
| 121 | bits--; |
| 122 | if (bits < 0 || bits > 31) { |
| 123 | *value = 0; |
| 124 | return -1; |
| 125 | } else if (bits == 0) { |
| 126 | *value = 0; |
| 127 | return 0; |
| 128 | } |
| 129 | |
| 130 | val = get_bits_long(gb, bits); |
| 131 | val |= 1U << bits; |
| 132 | |
| 133 | *value = val - 1; |
| 134 | |
| 135 | return 0; |
| 136 | } |
| 137 | |
| 138 | static int lag_read_prob_header(lag_rac *rac, GetBitContext *gb) |
| 139 | { |
| 140 | int i, j, scale_factor; |
| 141 | unsigned prob, cumulative_target; |
| 142 | unsigned cumul_prob = 0; |
| 143 | unsigned scaled_cumul_prob = 0; |
| 144 | |
| 145 | rac->prob[0] = 0; |
| 146 | rac->prob[257] = UINT_MAX; |
| 147 | /* Read probabilities from bitstream */ |
| 148 | for (i = 1; i < 257; i++) { |
| 149 | if (lag_decode_prob(gb, &rac->prob[i]) < 0) { |
| 150 | av_log(rac->avctx, AV_LOG_ERROR, "Invalid probability encountered.\n"); |
| 151 | return -1; |
| 152 | } |
| 153 | if ((uint64_t)cumul_prob + rac->prob[i] > UINT_MAX) { |
| 154 | av_log(rac->avctx, AV_LOG_ERROR, "Integer overflow encountered in cumulative probability calculation.\n"); |
| 155 | return -1; |
| 156 | } |
| 157 | cumul_prob += rac->prob[i]; |
| 158 | if (!rac->prob[i]) { |
| 159 | if (lag_decode_prob(gb, &prob)) { |
| 160 | av_log(rac->avctx, AV_LOG_ERROR, "Invalid probability run encountered.\n"); |
| 161 | return -1; |
| 162 | } |
| 163 | if (prob > 256 - i) |
| 164 | prob = 256 - i; |
| 165 | for (j = 0; j < prob; j++) |
| 166 | rac->prob[++i] = 0; |
| 167 | } |
| 168 | } |
| 169 | |
| 170 | if (!cumul_prob) { |
| 171 | av_log(rac->avctx, AV_LOG_ERROR, "All probabilities are 0!\n"); |
| 172 | return -1; |
| 173 | } |
| 174 | |
| 175 | /* Scale probabilities so cumulative probability is an even power of 2. */ |
| 176 | scale_factor = av_log2(cumul_prob); |
| 177 | |
| 178 | if (cumul_prob & (cumul_prob - 1)) { |
| 179 | uint64_t mul = softfloat_reciprocal(cumul_prob); |
| 180 | for (i = 1; i <= 128; i++) { |
| 181 | rac->prob[i] = softfloat_mul(rac->prob[i], mul); |
| 182 | scaled_cumul_prob += rac->prob[i]; |
| 183 | } |
| 184 | if (scaled_cumul_prob <= 0) { |
| 185 | av_log(rac->avctx, AV_LOG_ERROR, "Scaled probabilities invalid\n"); |
| 186 | return AVERROR_INVALIDDATA; |
| 187 | } |
| 188 | for (; i < 257; i++) { |
| 189 | rac->prob[i] = softfloat_mul(rac->prob[i], mul); |
| 190 | scaled_cumul_prob += rac->prob[i]; |
| 191 | } |
| 192 | |
| 193 | scale_factor++; |
| 194 | cumulative_target = 1 << scale_factor; |
| 195 | |
| 196 | if (scaled_cumul_prob > cumulative_target) { |
| 197 | av_log(rac->avctx, AV_LOG_ERROR, |
| 198 | "Scaled probabilities are larger than target!\n"); |
| 199 | return -1; |
| 200 | } |
| 201 | |
| 202 | scaled_cumul_prob = cumulative_target - scaled_cumul_prob; |
| 203 | |
| 204 | for (i = 1; scaled_cumul_prob; i = (i & 0x7f) + 1) { |
| 205 | if (rac->prob[i]) { |
| 206 | rac->prob[i]++; |
| 207 | scaled_cumul_prob--; |
| 208 | } |
| 209 | /* Comment from reference source: |
| 210 | * if (b & 0x80 == 0) { // order of operations is 'wrong'; it has been left this way |
| 211 | * // since the compression change is negligible and fixing it |
| 212 | * // breaks backwards compatibility |
| 213 | * b =- (signed int)b; |
| 214 | * b &= 0xFF; |
| 215 | * } else { |
| 216 | * b++; |
| 217 | * b &= 0x7f; |
| 218 | * } |
| 219 | */ |
| 220 | } |
| 221 | } |
| 222 | |
| 223 | rac->scale = scale_factor; |
| 224 | |
| 225 | /* Fill probability array with cumulative probability for each symbol. */ |
| 226 | for (i = 1; i < 257; i++) |
| 227 | rac->prob[i] += rac->prob[i - 1]; |
| 228 | |
| 229 | return 0; |
| 230 | } |
| 231 | |
| 232 | static void add_lag_median_prediction(uint8_t *dst, uint8_t *src1, |
| 233 | uint8_t *diff, int w, int *left, |
| 234 | int *left_top) |
| 235 | { |
| 236 | /* This is almost identical to add_hfyu_median_pred in huffyuvdsp.h. |
| 237 | * However the &0xFF on the gradient predictor yealds incorrect output |
| 238 | * for lagarith. |
| 239 | */ |
| 240 | int i; |
| 241 | uint8_t l, lt; |
| 242 | |
| 243 | l = *left; |
| 244 | lt = *left_top; |
| 245 | |
| 246 | for (i = 0; i < w; i++) { |
| 247 | l = mid_pred(l, src1[i], l + src1[i] - lt) + diff[i]; |
| 248 | lt = src1[i]; |
| 249 | dst[i] = l; |
| 250 | } |
| 251 | |
| 252 | *left = l; |
| 253 | *left_top = lt; |
| 254 | } |
| 255 | |
| 256 | static void lag_pred_line(LagarithContext *l, uint8_t *buf, |
| 257 | int width, int stride, int line) |
| 258 | { |
| 259 | int L, TL; |
| 260 | |
| 261 | if (!line) { |
| 262 | /* Left prediction only for first line */ |
| 263 | L = l->hdsp.add_hfyu_left_pred(buf, buf, width, 0); |
| 264 | } else { |
| 265 | /* Left pixel is actually prev_row[width] */ |
| 266 | L = buf[width - stride - 1]; |
| 267 | |
| 268 | if (line == 1) { |
| 269 | /* Second line, left predict first pixel, the rest of the line is median predicted |
| 270 | * NOTE: In the case of RGB this pixel is top predicted */ |
| 271 | TL = l->avctx->pix_fmt == AV_PIX_FMT_YUV420P ? buf[-stride] : L; |
| 272 | } else { |
| 273 | /* Top left is 2 rows back, last pixel */ |
| 274 | TL = buf[width - (2 * stride) - 1]; |
| 275 | } |
| 276 | |
| 277 | add_lag_median_prediction(buf, buf - stride, buf, |
| 278 | width, &L, &TL); |
| 279 | } |
| 280 | } |
| 281 | |
| 282 | static void lag_pred_line_yuy2(LagarithContext *l, uint8_t *buf, |
| 283 | int width, int stride, int line, |
| 284 | int is_luma) |
| 285 | { |
| 286 | int L, TL; |
| 287 | |
| 288 | if (!line) { |
| 289 | L= buf[0]; |
| 290 | if (is_luma) |
| 291 | buf[0] = 0; |
| 292 | l->hdsp.add_hfyu_left_pred(buf, buf, width, 0); |
| 293 | if (is_luma) |
| 294 | buf[0] = L; |
| 295 | return; |
| 296 | } |
| 297 | if (line == 1) { |
| 298 | const int HEAD = is_luma ? 4 : 2; |
| 299 | int i; |
| 300 | |
| 301 | L = buf[width - stride - 1]; |
| 302 | TL = buf[HEAD - stride - 1]; |
| 303 | for (i = 0; i < HEAD; i++) { |
| 304 | L += buf[i]; |
| 305 | buf[i] = L; |
| 306 | } |
| 307 | for (; i < width; i++) { |
| 308 | L = mid_pred(L & 0xFF, buf[i - stride], (L + buf[i - stride] - TL) & 0xFF) + buf[i]; |
| 309 | TL = buf[i - stride]; |
| 310 | buf[i] = L; |
| 311 | } |
| 312 | } else { |
| 313 | TL = buf[width - (2 * stride) - 1]; |
| 314 | L = buf[width - stride - 1]; |
| 315 | l->hdsp.add_hfyu_median_pred(buf, buf - stride, buf, width, &L, &TL); |
| 316 | } |
| 317 | } |
| 318 | |
| 319 | static int lag_decode_line(LagarithContext *l, lag_rac *rac, |
| 320 | uint8_t *dst, int width, int stride, |
| 321 | int esc_count) |
| 322 | { |
| 323 | int i = 0; |
| 324 | int ret = 0; |
| 325 | |
| 326 | if (!esc_count) |
| 327 | esc_count = -1; |
| 328 | |
| 329 | /* Output any zeros remaining from the previous run */ |
| 330 | handle_zeros: |
| 331 | if (l->zeros_rem) { |
| 332 | int count = FFMIN(l->zeros_rem, width - i); |
| 333 | memset(dst + i, 0, count); |
| 334 | i += count; |
| 335 | l->zeros_rem -= count; |
| 336 | } |
| 337 | |
| 338 | while (i < width) { |
| 339 | dst[i] = lag_get_rac(rac); |
| 340 | ret++; |
| 341 | |
| 342 | if (dst[i]) |
| 343 | l->zeros = 0; |
| 344 | else |
| 345 | l->zeros++; |
| 346 | |
| 347 | i++; |
| 348 | if (l->zeros == esc_count) { |
| 349 | int index = lag_get_rac(rac); |
| 350 | ret++; |
| 351 | |
| 352 | l->zeros = 0; |
| 353 | |
| 354 | l->zeros_rem = lag_calc_zero_run(index); |
| 355 | goto handle_zeros; |
| 356 | } |
| 357 | } |
| 358 | return ret; |
| 359 | } |
| 360 | |
| 361 | static int lag_decode_zero_run_line(LagarithContext *l, uint8_t *dst, |
| 362 | const uint8_t *src, const uint8_t *src_end, |
| 363 | int width, int esc_count) |
| 364 | { |
| 365 | int i = 0; |
| 366 | int count; |
| 367 | uint8_t zero_run = 0; |
| 368 | const uint8_t *src_start = src; |
| 369 | uint8_t mask1 = -(esc_count < 2); |
| 370 | uint8_t mask2 = -(esc_count < 3); |
| 371 | uint8_t *end = dst + (width - 2); |
| 372 | |
| 373 | avpriv_request_sample(l->avctx, "zero_run_line"); |
| 374 | |
| 375 | memset(dst, 0, width); |
| 376 | |
| 377 | output_zeros: |
| 378 | if (l->zeros_rem) { |
| 379 | count = FFMIN(l->zeros_rem, width - i); |
| 380 | if (end - dst < count) { |
| 381 | av_log(l->avctx, AV_LOG_ERROR, "Too many zeros remaining.\n"); |
| 382 | return AVERROR_INVALIDDATA; |
| 383 | } |
| 384 | |
| 385 | memset(dst, 0, count); |
| 386 | l->zeros_rem -= count; |
| 387 | dst += count; |
| 388 | } |
| 389 | |
| 390 | while (dst < end) { |
| 391 | i = 0; |
| 392 | while (!zero_run && dst + i < end) { |
| 393 | i++; |
| 394 | if (i+2 >= src_end - src) |
| 395 | return AVERROR_INVALIDDATA; |
| 396 | zero_run = |
| 397 | !(src[i] | (src[i + 1] & mask1) | (src[i + 2] & mask2)); |
| 398 | } |
| 399 | if (zero_run) { |
| 400 | zero_run = 0; |
| 401 | i += esc_count; |
| 402 | memcpy(dst, src, i); |
| 403 | dst += i; |
| 404 | l->zeros_rem = lag_calc_zero_run(src[i]); |
| 405 | |
| 406 | src += i + 1; |
| 407 | goto output_zeros; |
| 408 | } else { |
| 409 | memcpy(dst, src, i); |
| 410 | src += i; |
| 411 | dst += i; |
| 412 | } |
| 413 | } |
| 414 | return src - src_start; |
| 415 | } |
| 416 | |
| 417 | |
| 418 | |
| 419 | static int lag_decode_arith_plane(LagarithContext *l, uint8_t *dst, |
| 420 | int width, int height, int stride, |
| 421 | const uint8_t *src, int src_size) |
| 422 | { |
| 423 | int i = 0; |
| 424 | int read = 0; |
| 425 | uint32_t length; |
| 426 | uint32_t offset = 1; |
| 427 | int esc_count; |
| 428 | GetBitContext gb; |
| 429 | lag_rac rac; |
| 430 | const uint8_t *src_end = src + src_size; |
| 431 | int ret; |
| 432 | |
| 433 | rac.avctx = l->avctx; |
| 434 | l->zeros = 0; |
| 435 | |
| 436 | if(src_size < 2) |
| 437 | return AVERROR_INVALIDDATA; |
| 438 | |
| 439 | esc_count = src[0]; |
| 440 | if (esc_count < 4) { |
| 441 | length = width * height; |
| 442 | if(src_size < 5) |
| 443 | return AVERROR_INVALIDDATA; |
| 444 | if (esc_count && AV_RL32(src + 1) < length) { |
| 445 | length = AV_RL32(src + 1); |
| 446 | offset += 4; |
| 447 | } |
| 448 | |
| 449 | if ((ret = init_get_bits8(&gb, src + offset, src_size - offset)) < 0) |
| 450 | return ret; |
| 451 | |
| 452 | if (lag_read_prob_header(&rac, &gb) < 0) |
| 453 | return -1; |
| 454 | |
| 455 | ff_lag_rac_init(&rac, &gb, length - stride); |
| 456 | |
| 457 | for (i = 0; i < height; i++) |
| 458 | read += lag_decode_line(l, &rac, dst + (i * stride), width, |
| 459 | stride, esc_count); |
| 460 | |
| 461 | if (read > length) |
| 462 | av_log(l->avctx, AV_LOG_WARNING, |
| 463 | "Output more bytes than length (%d of %"PRIu32")\n", read, |
| 464 | length); |
| 465 | } else if (esc_count < 8) { |
| 466 | esc_count -= 4; |
| 467 | src ++; |
| 468 | src_size --; |
| 469 | if (esc_count > 0) { |
| 470 | /* Zero run coding only, no range coding. */ |
| 471 | for (i = 0; i < height; i++) { |
| 472 | int res = lag_decode_zero_run_line(l, dst + (i * stride), src, |
| 473 | src_end, width, esc_count); |
| 474 | if (res < 0) |
| 475 | return res; |
| 476 | src += res; |
| 477 | } |
| 478 | } else { |
| 479 | if (src_size < width * height) |
| 480 | return AVERROR_INVALIDDATA; // buffer not big enough |
| 481 | /* Plane is stored uncompressed */ |
| 482 | for (i = 0; i < height; i++) { |
| 483 | memcpy(dst + (i * stride), src, width); |
| 484 | src += width; |
| 485 | } |
| 486 | } |
| 487 | } else if (esc_count == 0xff) { |
| 488 | /* Plane is a solid run of given value */ |
| 489 | for (i = 0; i < height; i++) |
| 490 | memset(dst + i * stride, src[1], width); |
| 491 | /* Do not apply prediction. |
| 492 | Note: memset to 0 above, setting first value to src[1] |
| 493 | and applying prediction gives the same result. */ |
| 494 | return 0; |
| 495 | } else { |
| 496 | av_log(l->avctx, AV_LOG_ERROR, |
| 497 | "Invalid zero run escape code! (%#x)\n", esc_count); |
| 498 | return -1; |
| 499 | } |
| 500 | |
| 501 | if (l->avctx->pix_fmt != AV_PIX_FMT_YUV422P) { |
| 502 | for (i = 0; i < height; i++) { |
| 503 | lag_pred_line(l, dst, width, stride, i); |
| 504 | dst += stride; |
| 505 | } |
| 506 | } else { |
| 507 | for (i = 0; i < height; i++) { |
| 508 | lag_pred_line_yuy2(l, dst, width, stride, i, |
| 509 | width == l->avctx->width); |
| 510 | dst += stride; |
| 511 | } |
| 512 | } |
| 513 | |
| 514 | return 0; |
| 515 | } |
| 516 | |
| 517 | /** |
| 518 | * Decode a frame. |
| 519 | * @param avctx codec context |
| 520 | * @param data output AVFrame |
| 521 | * @param data_size size of output data or 0 if no picture is returned |
| 522 | * @param avpkt input packet |
| 523 | * @return number of consumed bytes on success or negative if decode fails |
| 524 | */ |
| 525 | static int lag_decode_frame(AVCodecContext *avctx, |
| 526 | void *data, int *got_frame, AVPacket *avpkt) |
| 527 | { |
| 528 | const uint8_t *buf = avpkt->data; |
| 529 | unsigned int buf_size = avpkt->size; |
| 530 | LagarithContext *l = avctx->priv_data; |
| 531 | ThreadFrame frame = { .f = data }; |
| 532 | AVFrame *const p = data; |
| 533 | uint8_t frametype = 0; |
| 534 | uint32_t offset_gu = 0, offset_bv = 0, offset_ry = 9; |
| 535 | uint32_t offs[4]; |
| 536 | uint8_t *srcs[4], *dst; |
| 537 | int i, j, planes = 3; |
| 538 | int ret; |
| 539 | |
| 540 | p->key_frame = 1; |
| 541 | |
| 542 | frametype = buf[0]; |
| 543 | |
| 544 | offset_gu = AV_RL32(buf + 1); |
| 545 | offset_bv = AV_RL32(buf + 5); |
| 546 | |
| 547 | switch (frametype) { |
| 548 | case FRAME_SOLID_RGBA: |
| 549 | avctx->pix_fmt = AV_PIX_FMT_RGB32; |
| 550 | case FRAME_SOLID_GRAY: |
| 551 | if (frametype == FRAME_SOLID_GRAY) |
| 552 | if (avctx->bits_per_coded_sample == 24) { |
| 553 | avctx->pix_fmt = AV_PIX_FMT_RGB24; |
| 554 | } else { |
| 555 | avctx->pix_fmt = AV_PIX_FMT_0RGB32; |
| 556 | planes = 4; |
| 557 | } |
| 558 | |
| 559 | if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) |
| 560 | return ret; |
| 561 | |
| 562 | dst = p->data[0]; |
| 563 | if (frametype == FRAME_SOLID_RGBA) { |
| 564 | for (j = 0; j < avctx->height; j++) { |
| 565 | for (i = 0; i < avctx->width; i++) |
| 566 | AV_WN32(dst + i * 4, offset_gu); |
| 567 | dst += p->linesize[0]; |
| 568 | } |
| 569 | } else { |
| 570 | for (j = 0; j < avctx->height; j++) { |
| 571 | memset(dst, buf[1], avctx->width * planes); |
| 572 | dst += p->linesize[0]; |
| 573 | } |
| 574 | } |
| 575 | break; |
| 576 | case FRAME_SOLID_COLOR: |
| 577 | if (avctx->bits_per_coded_sample == 24) { |
| 578 | avctx->pix_fmt = AV_PIX_FMT_RGB24; |
| 579 | } else { |
| 580 | avctx->pix_fmt = AV_PIX_FMT_RGB32; |
| 581 | offset_gu |= 0xFFU << 24; |
| 582 | } |
| 583 | |
| 584 | if ((ret = ff_thread_get_buffer(avctx, &frame,0)) < 0) |
| 585 | return ret; |
| 586 | |
| 587 | dst = p->data[0]; |
| 588 | for (j = 0; j < avctx->height; j++) { |
| 589 | for (i = 0; i < avctx->width; i++) |
| 590 | if (avctx->bits_per_coded_sample == 24) { |
| 591 | AV_WB24(dst + i * 3, offset_gu); |
| 592 | } else { |
| 593 | AV_WN32(dst + i * 4, offset_gu); |
| 594 | } |
| 595 | dst += p->linesize[0]; |
| 596 | } |
| 597 | break; |
| 598 | case FRAME_ARITH_RGBA: |
| 599 | avctx->pix_fmt = AV_PIX_FMT_RGB32; |
| 600 | planes = 4; |
| 601 | offset_ry += 4; |
| 602 | offs[3] = AV_RL32(buf + 9); |
| 603 | case FRAME_ARITH_RGB24: |
| 604 | case FRAME_U_RGB24: |
| 605 | if (frametype == FRAME_ARITH_RGB24 || frametype == FRAME_U_RGB24) |
| 606 | avctx->pix_fmt = AV_PIX_FMT_RGB24; |
| 607 | |
| 608 | if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) |
| 609 | return ret; |
| 610 | |
| 611 | offs[0] = offset_bv; |
| 612 | offs[1] = offset_gu; |
| 613 | offs[2] = offset_ry; |
| 614 | |
| 615 | l->rgb_stride = FFALIGN(avctx->width, 16); |
| 616 | av_fast_malloc(&l->rgb_planes, &l->rgb_planes_allocated, |
| 617 | l->rgb_stride * avctx->height * planes + 1); |
| 618 | if (!l->rgb_planes) { |
| 619 | av_log(avctx, AV_LOG_ERROR, "cannot allocate temporary buffer\n"); |
| 620 | return AVERROR(ENOMEM); |
| 621 | } |
| 622 | for (i = 0; i < planes; i++) |
| 623 | srcs[i] = l->rgb_planes + (i + 1) * l->rgb_stride * avctx->height - l->rgb_stride; |
| 624 | for (i = 0; i < planes; i++) |
| 625 | if (buf_size <= offs[i]) { |
| 626 | av_log(avctx, AV_LOG_ERROR, |
| 627 | "Invalid frame offsets\n"); |
| 628 | return AVERROR_INVALIDDATA; |
| 629 | } |
| 630 | |
| 631 | for (i = 0; i < planes; i++) |
| 632 | lag_decode_arith_plane(l, srcs[i], |
| 633 | avctx->width, avctx->height, |
| 634 | -l->rgb_stride, buf + offs[i], |
| 635 | buf_size - offs[i]); |
| 636 | dst = p->data[0]; |
| 637 | for (i = 0; i < planes; i++) |
| 638 | srcs[i] = l->rgb_planes + i * l->rgb_stride * avctx->height; |
| 639 | for (j = 0; j < avctx->height; j++) { |
| 640 | for (i = 0; i < avctx->width; i++) { |
| 641 | uint8_t r, g, b, a; |
| 642 | r = srcs[0][i]; |
| 643 | g = srcs[1][i]; |
| 644 | b = srcs[2][i]; |
| 645 | r += g; |
| 646 | b += g; |
| 647 | if (frametype == FRAME_ARITH_RGBA) { |
| 648 | a = srcs[3][i]; |
| 649 | AV_WN32(dst + i * 4, MKBETAG(a, r, g, b)); |
| 650 | } else { |
| 651 | dst[i * 3 + 0] = r; |
| 652 | dst[i * 3 + 1] = g; |
| 653 | dst[i * 3 + 2] = b; |
| 654 | } |
| 655 | } |
| 656 | dst += p->linesize[0]; |
| 657 | for (i = 0; i < planes; i++) |
| 658 | srcs[i] += l->rgb_stride; |
| 659 | } |
| 660 | break; |
| 661 | case FRAME_ARITH_YUY2: |
| 662 | avctx->pix_fmt = AV_PIX_FMT_YUV422P; |
| 663 | |
| 664 | if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) |
| 665 | return ret; |
| 666 | |
| 667 | if (offset_ry >= buf_size || |
| 668 | offset_gu >= buf_size || |
| 669 | offset_bv >= buf_size) { |
| 670 | av_log(avctx, AV_LOG_ERROR, |
| 671 | "Invalid frame offsets\n"); |
| 672 | return AVERROR_INVALIDDATA; |
| 673 | } |
| 674 | |
| 675 | lag_decode_arith_plane(l, p->data[0], avctx->width, avctx->height, |
| 676 | p->linesize[0], buf + offset_ry, |
| 677 | buf_size - offset_ry); |
| 678 | lag_decode_arith_plane(l, p->data[1], (avctx->width + 1) / 2, |
| 679 | avctx->height, p->linesize[1], |
| 680 | buf + offset_gu, buf_size - offset_gu); |
| 681 | lag_decode_arith_plane(l, p->data[2], (avctx->width + 1) / 2, |
| 682 | avctx->height, p->linesize[2], |
| 683 | buf + offset_bv, buf_size - offset_bv); |
| 684 | break; |
| 685 | case FRAME_ARITH_YV12: |
| 686 | avctx->pix_fmt = AV_PIX_FMT_YUV420P; |
| 687 | |
| 688 | if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) |
| 689 | return ret; |
| 690 | if (buf_size <= offset_ry || buf_size <= offset_gu || buf_size <= offset_bv) { |
| 691 | return AVERROR_INVALIDDATA; |
| 692 | } |
| 693 | |
| 694 | if (offset_ry >= buf_size || |
| 695 | offset_gu >= buf_size || |
| 696 | offset_bv >= buf_size) { |
| 697 | av_log(avctx, AV_LOG_ERROR, |
| 698 | "Invalid frame offsets\n"); |
| 699 | return AVERROR_INVALIDDATA; |
| 700 | } |
| 701 | |
| 702 | lag_decode_arith_plane(l, p->data[0], avctx->width, avctx->height, |
| 703 | p->linesize[0], buf + offset_ry, |
| 704 | buf_size - offset_ry); |
| 705 | lag_decode_arith_plane(l, p->data[2], (avctx->width + 1) / 2, |
| 706 | (avctx->height + 1) / 2, p->linesize[2], |
| 707 | buf + offset_gu, buf_size - offset_gu); |
| 708 | lag_decode_arith_plane(l, p->data[1], (avctx->width + 1) / 2, |
| 709 | (avctx->height + 1) / 2, p->linesize[1], |
| 710 | buf + offset_bv, buf_size - offset_bv); |
| 711 | break; |
| 712 | default: |
| 713 | av_log(avctx, AV_LOG_ERROR, |
| 714 | "Unsupported Lagarith frame type: %#"PRIx8"\n", frametype); |
| 715 | return AVERROR_PATCHWELCOME; |
| 716 | } |
| 717 | |
| 718 | *got_frame = 1; |
| 719 | |
| 720 | return buf_size; |
| 721 | } |
| 722 | |
| 723 | static av_cold int lag_decode_init(AVCodecContext *avctx) |
| 724 | { |
| 725 | LagarithContext *l = avctx->priv_data; |
| 726 | l->avctx = avctx; |
| 727 | |
| 728 | ff_huffyuvdsp_init(&l->hdsp); |
| 729 | |
| 730 | return 0; |
| 731 | } |
| 732 | |
| 733 | static av_cold int lag_decode_end(AVCodecContext *avctx) |
| 734 | { |
| 735 | LagarithContext *l = avctx->priv_data; |
| 736 | |
| 737 | av_freep(&l->rgb_planes); |
| 738 | |
| 739 | return 0; |
| 740 | } |
| 741 | |
| 742 | AVCodec ff_lagarith_decoder = { |
| 743 | .name = "lagarith", |
| 744 | .long_name = NULL_IF_CONFIG_SMALL("Lagarith lossless"), |
| 745 | .type = AVMEDIA_TYPE_VIDEO, |
| 746 | .id = AV_CODEC_ID_LAGARITH, |
| 747 | .priv_data_size = sizeof(LagarithContext), |
| 748 | .init = lag_decode_init, |
| 749 | .close = lag_decode_end, |
| 750 | .decode = lag_decode_frame, |
| 751 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS, |
| 752 | }; |