| 1 | /* |
| 2 | * Mirillis FIC decoder |
| 3 | * |
| 4 | * Copyright (c) 2014 Konstantin Shishkov |
| 5 | * Copyright (c) 2014 Derek Buitenhuis |
| 6 | * |
| 7 | * This file is part of FFmpeg. |
| 8 | * |
| 9 | * FFmpeg is free software; you can redistribute it and/or |
| 10 | * modify it under the terms of the GNU Lesser General Public |
| 11 | * License as published by the Free Software Foundation; either |
| 12 | * version 2.1 of the License, or (at your option) any later version. |
| 13 | * |
| 14 | * FFmpeg is distributed in the hope that it will be useful, |
| 15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 17 | * Lesser General Public License for more details. |
| 18 | * |
| 19 | * You should have received a copy of the GNU Lesser General Public |
| 20 | * License along with FFmpeg; if not, write to the Free Software |
| 21 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 22 | */ |
| 23 | |
| 24 | #include "libavutil/common.h" |
| 25 | #include "avcodec.h" |
| 26 | #include "internal.h" |
| 27 | #include "get_bits.h" |
| 28 | #include "golomb.h" |
| 29 | |
| 30 | typedef struct FICThreadContext { |
| 31 | DECLARE_ALIGNED(16, int16_t, block)[64]; |
| 32 | uint8_t *src; |
| 33 | int slice_h; |
| 34 | int src_size; |
| 35 | int y_off; |
| 36 | } FICThreadContext; |
| 37 | |
| 38 | typedef struct FICContext { |
| 39 | AVCodecContext *avctx; |
| 40 | AVFrame *frame; |
| 41 | AVFrame *final_frame; |
| 42 | |
| 43 | FICThreadContext *slice_data; |
| 44 | int slice_data_size; |
| 45 | |
| 46 | const uint8_t *qmat; |
| 47 | |
| 48 | enum AVPictureType cur_frame_type; |
| 49 | |
| 50 | int aligned_width, aligned_height; |
| 51 | int num_slices, slice_h; |
| 52 | |
| 53 | uint8_t cursor_buf[4096]; |
| 54 | } FICContext; |
| 55 | |
| 56 | static const uint8_t fic_qmat_hq[64] = { |
| 57 | 1, 2, 2, 2, 3, 3, 3, 4, |
| 58 | 2, 2, 2, 3, 3, 3, 4, 4, |
| 59 | 2, 2, 3, 3, 3, 4, 4, 4, |
| 60 | 2, 2, 3, 3, 3, 4, 4, 5, |
| 61 | 2, 3, 3, 3, 4, 4, 5, 6, |
| 62 | 3, 3, 3, 4, 4, 5, 6, 7, |
| 63 | 3, 3, 3, 4, 4, 5, 7, 7, |
| 64 | 3, 3, 4, 4, 5, 7, 7, 7, |
| 65 | }; |
| 66 | |
| 67 | static const uint8_t fic_qmat_lq[64] = { |
| 68 | 1, 5, 6, 7, 8, 9, 9, 11, |
| 69 | 5, 5, 7, 8, 9, 9, 11, 12, |
| 70 | 6, 7, 8, 9, 9, 11, 11, 12, |
| 71 | 7, 7, 8, 9, 9, 11, 12, 13, |
| 72 | 7, 8, 9, 9, 10, 11, 13, 16, |
| 73 | 8, 9, 9, 10, 11, 13, 16, 19, |
| 74 | 8, 9, 9, 11, 12, 15, 18, 23, |
| 75 | 9, 9, 11, 12, 15, 18, 23, 27 |
| 76 | }; |
| 77 | |
| 78 | static const uint8_t fic_header[7] = { 0, 0, 1, 'F', 'I', 'C', 'V' }; |
| 79 | |
| 80 | #define FIC_HEADER_SIZE 27 |
| 81 | |
| 82 | static av_always_inline void fic_idct(int16_t *blk, int step, int shift, int rnd) |
| 83 | { |
| 84 | const int t0 = 27246 * blk[3 * step] + 18405 * blk[5 * step]; |
| 85 | const int t1 = 27246 * blk[5 * step] - 18405 * blk[3 * step]; |
| 86 | const int t2 = 6393 * blk[7 * step] + 32139 * blk[1 * step]; |
| 87 | const int t3 = 6393 * blk[1 * step] - 32139 * blk[7 * step]; |
| 88 | const int t4 = 5793 * (t2 + t0 + 0x800 >> 12); |
| 89 | const int t5 = 5793 * (t3 + t1 + 0x800 >> 12); |
| 90 | const int t6 = t2 - t0; |
| 91 | const int t7 = t3 - t1; |
| 92 | const int t8 = 17734 * blk[2 * step] - 42813 * blk[6 * step]; |
| 93 | const int t9 = 17734 * blk[6 * step] + 42814 * blk[2 * step]; |
| 94 | const int tA = (blk[0 * step] - blk[4 * step] << 15) + rnd; |
| 95 | const int tB = (blk[0 * step] + blk[4 * step] << 15) + rnd; |
| 96 | blk[0 * step] = ( t4 + t9 + tB) >> shift; |
| 97 | blk[1 * step] = ( t6 + t7 + t8 + tA) >> shift; |
| 98 | blk[2 * step] = ( t6 - t7 - t8 + tA) >> shift; |
| 99 | blk[3 * step] = ( t5 - t9 + tB) >> shift; |
| 100 | blk[4 * step] = ( -t5 - t9 + tB) >> shift; |
| 101 | blk[5 * step] = (-(t6 - t7) - t8 + tA) >> shift; |
| 102 | blk[6 * step] = (-(t6 + t7) + t8 + tA) >> shift; |
| 103 | blk[7 * step] = ( -t4 + t9 + tB) >> shift; |
| 104 | } |
| 105 | |
| 106 | static void fic_idct_put(uint8_t *dst, int stride, int16_t *block) |
| 107 | { |
| 108 | int i, j; |
| 109 | int16_t *ptr; |
| 110 | |
| 111 | ptr = block; |
| 112 | fic_idct(ptr++, 8, 13, (1 << 12) + (1 << 17)); |
| 113 | for (i = 1; i < 8; i++) { |
| 114 | fic_idct(ptr, 8, 13, 1 << 12); |
| 115 | ptr++; |
| 116 | } |
| 117 | |
| 118 | ptr = block; |
| 119 | for (i = 0; i < 8; i++) { |
| 120 | fic_idct(ptr, 1, 20, 0); |
| 121 | ptr += 8; |
| 122 | } |
| 123 | |
| 124 | ptr = block; |
| 125 | for (j = 0; j < 8; j++) { |
| 126 | for (i = 0; i < 8; i++) |
| 127 | dst[i] = av_clip_uint8(ptr[i]); |
| 128 | dst += stride; |
| 129 | ptr += 8; |
| 130 | } |
| 131 | } |
| 132 | static int fic_decode_block(FICContext *ctx, GetBitContext *gb, |
| 133 | uint8_t *dst, int stride, int16_t *block) |
| 134 | { |
| 135 | int i, num_coeff; |
| 136 | |
| 137 | /* Is it a skip block? */ |
| 138 | if (get_bits1(gb)) { |
| 139 | /* This is a P-frame. */ |
| 140 | ctx->frame->key_frame = 0; |
| 141 | ctx->frame->pict_type = AV_PICTURE_TYPE_P; |
| 142 | |
| 143 | return 0; |
| 144 | } |
| 145 | |
| 146 | memset(block, 0, sizeof(*block) * 64); |
| 147 | |
| 148 | num_coeff = get_bits(gb, 7); |
| 149 | if (num_coeff > 64) |
| 150 | return AVERROR_INVALIDDATA; |
| 151 | |
| 152 | for (i = 0; i < num_coeff; i++) |
| 153 | block[ff_zigzag_direct[i]] = get_se_golomb(gb) * |
| 154 | ctx->qmat[ff_zigzag_direct[i]]; |
| 155 | |
| 156 | fic_idct_put(dst, stride, block); |
| 157 | |
| 158 | return 0; |
| 159 | } |
| 160 | |
| 161 | static int fic_decode_slice(AVCodecContext *avctx, void *tdata) |
| 162 | { |
| 163 | FICContext *ctx = avctx->priv_data; |
| 164 | FICThreadContext *tctx = tdata; |
| 165 | GetBitContext gb; |
| 166 | uint8_t *src = tctx->src; |
| 167 | int slice_h = tctx->slice_h; |
| 168 | int src_size = tctx->src_size; |
| 169 | int y_off = tctx->y_off; |
| 170 | int x, y, p; |
| 171 | |
| 172 | init_get_bits(&gb, src, src_size * 8); |
| 173 | |
| 174 | for (p = 0; p < 3; p++) { |
| 175 | int stride = ctx->frame->linesize[p]; |
| 176 | uint8_t* dst = ctx->frame->data[p] + (y_off >> !!p) * stride; |
| 177 | |
| 178 | for (y = 0; y < (slice_h >> !!p); y += 8) { |
| 179 | for (x = 0; x < (ctx->aligned_width >> !!p); x += 8) { |
| 180 | int ret; |
| 181 | |
| 182 | if ((ret = fic_decode_block(ctx, &gb, dst + x, stride, tctx->block)) != 0) |
| 183 | return ret; |
| 184 | } |
| 185 | |
| 186 | dst += 8 * stride; |
| 187 | } |
| 188 | } |
| 189 | |
| 190 | return 0; |
| 191 | } |
| 192 | |
| 193 | static av_always_inline void fic_alpha_blend(uint8_t *dst, uint8_t *src, |
| 194 | int size, uint8_t *alpha) |
| 195 | { |
| 196 | int i; |
| 197 | |
| 198 | for (i = 0; i < size; i++) |
| 199 | dst[i] += ((src[i] - dst[i]) * alpha[i]) >> 8; |
| 200 | } |
| 201 | |
| 202 | static void fic_draw_cursor(AVCodecContext *avctx, int cur_x, int cur_y) |
| 203 | { |
| 204 | FICContext *ctx = avctx->priv_data; |
| 205 | uint8_t *ptr = ctx->cursor_buf; |
| 206 | uint8_t *dstptr[3]; |
| 207 | uint8_t planes[4][1024]; |
| 208 | uint8_t chroma[3][256]; |
| 209 | int i, j, p; |
| 210 | |
| 211 | /* Convert to YUVA444. */ |
| 212 | for (i = 0; i < 1024; i++) { |
| 213 | planes[0][i] = (( 25 * ptr[0] + 129 * ptr[1] + 66 * ptr[2]) / 255) + 16; |
| 214 | planes[1][i] = ((-38 * ptr[0] + 112 * ptr[1] + -74 * ptr[2]) / 255) + 128; |
| 215 | planes[2][i] = ((-18 * ptr[0] + 112 * ptr[1] + -94 * ptr[2]) / 255) + 128; |
| 216 | planes[3][i] = ptr[3]; |
| 217 | |
| 218 | ptr += 4; |
| 219 | } |
| 220 | |
| 221 | /* Subsample chroma. */ |
| 222 | for (i = 0; i < 32; i += 2) |
| 223 | for (j = 0; j < 32; j += 2) |
| 224 | for (p = 0; p < 3; p++) |
| 225 | chroma[p][16 * (i / 2) + j / 2] = (planes[p + 1][32 * i + j ] + |
| 226 | planes[p + 1][32 * i + j + 1] + |
| 227 | planes[p + 1][32 * (i + 1) + j ] + |
| 228 | planes[p + 1][32 * (i + 1) + j + 1]) / 4; |
| 229 | |
| 230 | /* Seek to x/y pos of cursor. */ |
| 231 | for (i = 0; i < 3; i++) |
| 232 | dstptr[i] = ctx->final_frame->data[i] + |
| 233 | (ctx->final_frame->linesize[i] * (cur_y >> !!i)) + |
| 234 | (cur_x >> !!i) + !!i; |
| 235 | |
| 236 | /* Copy. */ |
| 237 | for (i = 0; i < FFMIN(32, avctx->height - cur_y) - 1; i += 2) { |
| 238 | int lsize = FFMIN(32, avctx->width - cur_x); |
| 239 | int csize = lsize / 2; |
| 240 | |
| 241 | fic_alpha_blend(dstptr[0], |
| 242 | planes[0] + i * 32, lsize, planes[3] + i * 32); |
| 243 | fic_alpha_blend(dstptr[0] + ctx->final_frame->linesize[0], |
| 244 | planes[0] + (i + 1) * 32, lsize, planes[3] + (i + 1) * 32); |
| 245 | fic_alpha_blend(dstptr[1], |
| 246 | chroma[0] + (i / 2) * 16, csize, chroma[2] + (i / 2) * 16); |
| 247 | fic_alpha_blend(dstptr[2], |
| 248 | chroma[1] + (i / 2) * 16, csize, chroma[2] + (i / 2) * 16); |
| 249 | |
| 250 | dstptr[0] += ctx->final_frame->linesize[0] * 2; |
| 251 | dstptr[1] += ctx->final_frame->linesize[1]; |
| 252 | dstptr[2] += ctx->final_frame->linesize[2]; |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | static int fic_decode_frame(AVCodecContext *avctx, void *data, |
| 257 | int *got_frame, AVPacket *avpkt) |
| 258 | { |
| 259 | FICContext *ctx = avctx->priv_data; |
| 260 | uint8_t *src = avpkt->data; |
| 261 | int ret; |
| 262 | int slice, nslices; |
| 263 | int msize; |
| 264 | int tsize; |
| 265 | int cur_x, cur_y; |
| 266 | int skip_cursor = 0; |
| 267 | uint8_t *sdata; |
| 268 | |
| 269 | if ((ret = ff_reget_buffer(avctx, ctx->frame)) < 0) { |
| 270 | av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n"); |
| 271 | return ret; |
| 272 | } |
| 273 | |
| 274 | /* Header + at least one slice (4) */ |
| 275 | if (avpkt->size < FIC_HEADER_SIZE + 4) { |
| 276 | av_log(avctx, AV_LOG_ERROR, "Frame data is too small.\n"); |
| 277 | return AVERROR_INVALIDDATA; |
| 278 | } |
| 279 | |
| 280 | /* Check for header. */ |
| 281 | if (memcmp(src, fic_header, 7)) |
| 282 | av_log(avctx, AV_LOG_WARNING, "Invalid FIC Header.\n"); |
| 283 | |
| 284 | /* Is it a skip frame? */ |
| 285 | if (src[17]) { |
| 286 | if (!ctx->final_frame) { |
| 287 | av_log(avctx, AV_LOG_WARNING, "Initial frame is skipped\n"); |
| 288 | return AVERROR_INVALIDDATA; |
| 289 | } |
| 290 | goto skip; |
| 291 | } |
| 292 | |
| 293 | nslices = src[13]; |
| 294 | if (!nslices) { |
| 295 | av_log(avctx, AV_LOG_ERROR, "Zero slices found.\n"); |
| 296 | return AVERROR_INVALIDDATA; |
| 297 | } |
| 298 | |
| 299 | /* High or Low Quality Matrix? */ |
| 300 | ctx->qmat = src[23] ? fic_qmat_hq : fic_qmat_lq; |
| 301 | |
| 302 | /* Skip cursor data. */ |
| 303 | tsize = AV_RB24(src + 24); |
| 304 | if (tsize > avpkt->size - FIC_HEADER_SIZE) { |
| 305 | av_log(avctx, AV_LOG_ERROR, |
| 306 | "Packet is too small to contain cursor (%d vs %d bytes).\n", |
| 307 | tsize, avpkt->size - FIC_HEADER_SIZE); |
| 308 | return AVERROR_INVALIDDATA; |
| 309 | } |
| 310 | |
| 311 | if (tsize < 32) { |
| 312 | av_log(avctx, AV_LOG_WARNING, |
| 313 | "Cursor data too small. Skipping cursor.\n"); |
| 314 | skip_cursor = 1; |
| 315 | } |
| 316 | |
| 317 | /* Cursor position. */ |
| 318 | cur_x = AV_RL16(src + 33); |
| 319 | cur_y = AV_RL16(src + 35); |
| 320 | if (cur_x > avctx->width || cur_y > avctx->height) { |
| 321 | av_log(avctx, AV_LOG_WARNING, |
| 322 | "Invalid cursor position: (%d,%d). Skipping cusor.\n", |
| 323 | cur_x, cur_y); |
| 324 | skip_cursor = 1; |
| 325 | } |
| 326 | |
| 327 | if (AV_RL16(src + 37) != 32 || AV_RL16(src + 39) != 32) { |
| 328 | av_log(avctx, AV_LOG_WARNING, |
| 329 | "Invalid cursor size. Skipping cursor.\n"); |
| 330 | skip_cursor = 1; |
| 331 | } |
| 332 | |
| 333 | /* Slice height for all but the last slice. */ |
| 334 | ctx->slice_h = 16 * (ctx->aligned_height >> 4) / nslices; |
| 335 | if (ctx->slice_h % 16) |
| 336 | ctx->slice_h = FFALIGN(ctx->slice_h - 16, 16); |
| 337 | |
| 338 | /* First slice offset and remaining data. */ |
| 339 | sdata = src + tsize + FIC_HEADER_SIZE + 4 * nslices; |
| 340 | msize = avpkt->size - nslices * 4 - tsize - FIC_HEADER_SIZE; |
| 341 | |
| 342 | if (msize <= 0) { |
| 343 | av_log(avctx, AV_LOG_ERROR, "Not enough frame data to decode.\n"); |
| 344 | return AVERROR_INVALIDDATA; |
| 345 | } |
| 346 | |
| 347 | /* |
| 348 | * Set the frametype to I initially. It will be set to P if the frame |
| 349 | * has any dependencies (skip blocks). There will be a race condition |
| 350 | * inside the slice decode function to set these, but we do not care. |
| 351 | * since they will only ever be set to 0/P. |
| 352 | */ |
| 353 | ctx->frame->key_frame = 1; |
| 354 | ctx->frame->pict_type = AV_PICTURE_TYPE_I; |
| 355 | |
| 356 | /* Allocate slice data. */ |
| 357 | av_fast_malloc(&ctx->slice_data, &ctx->slice_data_size, |
| 358 | nslices * sizeof(ctx->slice_data[0])); |
| 359 | if (!ctx->slice_data_size) { |
| 360 | av_log(avctx, AV_LOG_ERROR, "Could not allocate slice data.\n"); |
| 361 | return AVERROR(ENOMEM); |
| 362 | } |
| 363 | memset(ctx->slice_data, 0, nslices * sizeof(ctx->slice_data[0])); |
| 364 | |
| 365 | for (slice = 0; slice < nslices; slice++) { |
| 366 | unsigned slice_off = AV_RB32(src + tsize + FIC_HEADER_SIZE + slice * 4); |
| 367 | unsigned slice_size; |
| 368 | int y_off = ctx->slice_h * slice; |
| 369 | int slice_h = ctx->slice_h; |
| 370 | |
| 371 | /* |
| 372 | * Either read the slice size, or consume all data left. |
| 373 | * Also, special case the last slight height. |
| 374 | */ |
| 375 | if (slice == nslices - 1) { |
| 376 | slice_size = msize; |
| 377 | slice_h = FFALIGN(avctx->height - ctx->slice_h * (nslices - 1), 16); |
| 378 | } else { |
| 379 | slice_size = AV_RB32(src + tsize + FIC_HEADER_SIZE + slice * 4 + 4); |
| 380 | } |
| 381 | |
| 382 | if (slice_size < slice_off || slice_size > msize) |
| 383 | continue; |
| 384 | |
| 385 | slice_size -= slice_off; |
| 386 | |
| 387 | ctx->slice_data[slice].src = sdata + slice_off; |
| 388 | ctx->slice_data[slice].src_size = slice_size; |
| 389 | ctx->slice_data[slice].slice_h = slice_h; |
| 390 | ctx->slice_data[slice].y_off = y_off; |
| 391 | } |
| 392 | |
| 393 | if ((ret = avctx->execute(avctx, fic_decode_slice, ctx->slice_data, |
| 394 | NULL, nslices, sizeof(ctx->slice_data[0]))) < 0) |
| 395 | return ret; |
| 396 | |
| 397 | av_frame_free(&ctx->final_frame); |
| 398 | ctx->final_frame = av_frame_clone(ctx->frame); |
| 399 | if (!ctx->final_frame) { |
| 400 | av_log(avctx, AV_LOG_ERROR, "Could not clone frame buffer.\n"); |
| 401 | return AVERROR(ENOMEM); |
| 402 | } |
| 403 | |
| 404 | /* Make sure we use a user-supplied buffer. */ |
| 405 | if ((ret = ff_reget_buffer(avctx, ctx->final_frame)) < 0) { |
| 406 | av_log(avctx, AV_LOG_ERROR, "Could not make frame writable.\n"); |
| 407 | return ret; |
| 408 | } |
| 409 | |
| 410 | /* Draw cursor. */ |
| 411 | if (!skip_cursor) { |
| 412 | memcpy(ctx->cursor_buf, src + 59, 32 * 32 * 4); |
| 413 | fic_draw_cursor(avctx, cur_x, cur_y); |
| 414 | } |
| 415 | |
| 416 | skip: |
| 417 | *got_frame = 1; |
| 418 | if ((ret = av_frame_ref(data, ctx->final_frame)) < 0) |
| 419 | return ret; |
| 420 | |
| 421 | return avpkt->size; |
| 422 | } |
| 423 | |
| 424 | static av_cold int fic_decode_close(AVCodecContext *avctx) |
| 425 | { |
| 426 | FICContext *ctx = avctx->priv_data; |
| 427 | |
| 428 | av_freep(&ctx->slice_data); |
| 429 | av_frame_free(&ctx->final_frame); |
| 430 | av_frame_free(&ctx->frame); |
| 431 | |
| 432 | return 0; |
| 433 | } |
| 434 | |
| 435 | static av_cold int fic_decode_init(AVCodecContext *avctx) |
| 436 | { |
| 437 | FICContext *ctx = avctx->priv_data; |
| 438 | |
| 439 | /* Initialize various context values */ |
| 440 | ctx->avctx = avctx; |
| 441 | ctx->aligned_width = FFALIGN(avctx->width, 16); |
| 442 | ctx->aligned_height = FFALIGN(avctx->height, 16); |
| 443 | |
| 444 | avctx->pix_fmt = AV_PIX_FMT_YUV420P; |
| 445 | avctx->bits_per_raw_sample = 8; |
| 446 | |
| 447 | ctx->frame = av_frame_alloc(); |
| 448 | if (!ctx->frame) |
| 449 | return AVERROR(ENOMEM); |
| 450 | |
| 451 | return 0; |
| 452 | } |
| 453 | |
| 454 | AVCodec ff_fic_decoder = { |
| 455 | .name = "fic", |
| 456 | .long_name = NULL_IF_CONFIG_SMALL("Mirillis FIC"), |
| 457 | .type = AVMEDIA_TYPE_VIDEO, |
| 458 | .id = AV_CODEC_ID_FIC, |
| 459 | .priv_data_size = sizeof(FICContext), |
| 460 | .init = fic_decode_init, |
| 461 | .decode = fic_decode_frame, |
| 462 | .close = fic_decode_close, |
| 463 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS, |
| 464 | }; |