blob: 71d478c9fc5b98c0db1a95de8178f307c3a6323d
1 | /* |
2 | * Indeo Video v3 compatible decoder |
3 | * Copyright (c) 2009 - 2011 Maxim Poliakovski |
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 | * This is a decoder for Intel Indeo Video v3. |
25 | * It is based on vector quantization, run-length coding and motion compensation. |
26 | * Known container formats: .avi and .mov |
27 | * Known FOURCCs: 'IV31', 'IV32' |
28 | * |
29 | * @see http://wiki.multimedia.cx/index.php?title=Indeo_3 |
30 | */ |
31 | |
32 | #include "libavutil/imgutils.h" |
33 | #include "libavutil/intreadwrite.h" |
34 | #include "avcodec.h" |
35 | #include "copy_block.h" |
36 | #include "bytestream.h" |
37 | #include "get_bits.h" |
38 | #include "hpeldsp.h" |
39 | #include "internal.h" |
40 | |
41 | #include "indeo3data.h" |
42 | |
43 | /* RLE opcodes. */ |
44 | enum { |
45 | RLE_ESC_F9 = 249, ///< same as RLE_ESC_FA + do the same with next block |
46 | RLE_ESC_FA = 250, ///< INTRA: skip block, INTER: copy data from reference |
47 | RLE_ESC_FB = 251, ///< apply null delta to N blocks / skip N blocks |
48 | RLE_ESC_FC = 252, ///< same as RLE_ESC_FD + do the same with next block |
49 | RLE_ESC_FD = 253, ///< apply null delta to all remaining lines of this block |
50 | RLE_ESC_FE = 254, ///< apply null delta to all lines up to the 3rd line |
51 | RLE_ESC_FF = 255 ///< apply null delta to all lines up to the 2nd line |
52 | }; |
53 | |
54 | |
55 | /* Some constants for parsing frame bitstream flags. */ |
56 | #define BS_8BIT_PEL (1 << 1) ///< 8-bit pixel bitdepth indicator |
57 | #define BS_KEYFRAME (1 << 2) ///< intra frame indicator |
58 | #define BS_MV_Y_HALF (1 << 4) ///< vertical mv halfpel resolution indicator |
59 | #define BS_MV_X_HALF (1 << 5) ///< horizontal mv halfpel resolution indicator |
60 | #define BS_NONREF (1 << 8) ///< nonref (discardable) frame indicator |
61 | #define BS_BUFFER 9 ///< indicates which of two frame buffers should be used |
62 | |
63 | |
64 | typedef struct Plane { |
65 | uint8_t *buffers[2]; |
66 | uint8_t *pixels[2]; ///< pointer to the actual pixel data of the buffers above |
67 | uint32_t width; |
68 | uint32_t height; |
69 | ptrdiff_t pitch; |
70 | } Plane; |
71 | |
72 | #define CELL_STACK_MAX 20 |
73 | |
74 | typedef struct Cell { |
75 | int16_t xpos; ///< cell coordinates in 4x4 blocks |
76 | int16_t ypos; |
77 | int16_t width; ///< cell width in 4x4 blocks |
78 | int16_t height; ///< cell height in 4x4 blocks |
79 | uint8_t tree; ///< tree id: 0- MC tree, 1 - VQ tree |
80 | const int8_t *mv_ptr; ///< ptr to the motion vector if any |
81 | } Cell; |
82 | |
83 | typedef struct Indeo3DecodeContext { |
84 | AVCodecContext *avctx; |
85 | HpelDSPContext hdsp; |
86 | |
87 | GetBitContext gb; |
88 | int need_resync; |
89 | int skip_bits; |
90 | const uint8_t *next_cell_data; |
91 | const uint8_t *last_byte; |
92 | const int8_t *mc_vectors; |
93 | unsigned num_vectors; ///< number of motion vectors in mc_vectors |
94 | |
95 | int16_t width, height; |
96 | uint32_t frame_num; ///< current frame number (zero-based) |
97 | int data_size; ///< size of the frame data in bytes |
98 | uint16_t frame_flags; ///< frame properties |
99 | uint8_t cb_offset; ///< needed for selecting VQ tables |
100 | uint8_t buf_sel; ///< active frame buffer: 0 - primary, 1 -secondary |
101 | const uint8_t *y_data_ptr; |
102 | const uint8_t *v_data_ptr; |
103 | const uint8_t *u_data_ptr; |
104 | int32_t y_data_size; |
105 | int32_t v_data_size; |
106 | int32_t u_data_size; |
107 | const uint8_t *alt_quant; ///< secondary VQ table set for the modes 1 and 4 |
108 | Plane planes[3]; |
109 | } Indeo3DecodeContext; |
110 | |
111 | |
112 | static uint8_t requant_tab[8][128]; |
113 | |
114 | /* |
115 | * Build the static requantization table. |
116 | * This table is used to remap pixel values according to a specific |
117 | * quant index and thus avoid overflows while adding deltas. |
118 | */ |
119 | static av_cold void build_requant_tab(void) |
120 | { |
121 | static const int8_t offsets[8] = { 1, 1, 2, -3, -3, 3, 4, 4 }; |
122 | static const int8_t deltas [8] = { 0, 1, 0, 4, 4, 1, 0, 1 }; |
123 | |
124 | int i, j, step; |
125 | |
126 | for (i = 0; i < 8; i++) { |
127 | step = i + 2; |
128 | for (j = 0; j < 128; j++) |
129 | requant_tab[i][j] = (j + offsets[i]) / step * step + deltas[i]; |
130 | } |
131 | |
132 | /* some last elements calculated above will have values >= 128 */ |
133 | /* pixel values shall never exceed 127 so set them to non-overflowing values */ |
134 | /* according with the quantization step of the respective section */ |
135 | requant_tab[0][127] = 126; |
136 | requant_tab[1][119] = 118; |
137 | requant_tab[1][120] = 118; |
138 | requant_tab[2][126] = 124; |
139 | requant_tab[2][127] = 124; |
140 | requant_tab[6][124] = 120; |
141 | requant_tab[6][125] = 120; |
142 | requant_tab[6][126] = 120; |
143 | requant_tab[6][127] = 120; |
144 | |
145 | /* Patch for compatibility with the Intel's binary decoders */ |
146 | requant_tab[1][7] = 10; |
147 | requant_tab[4][8] = 10; |
148 | } |
149 | |
150 | |
151 | static av_cold void free_frame_buffers(Indeo3DecodeContext *ctx) |
152 | { |
153 | int p; |
154 | |
155 | ctx->width = ctx->height = 0; |
156 | |
157 | for (p = 0; p < 3; p++) { |
158 | av_freep(&ctx->planes[p].buffers[0]); |
159 | av_freep(&ctx->planes[p].buffers[1]); |
160 | ctx->planes[p].pixels[0] = ctx->planes[p].pixels[1] = 0; |
161 | } |
162 | } |
163 | |
164 | |
165 | static av_cold int allocate_frame_buffers(Indeo3DecodeContext *ctx, |
166 | AVCodecContext *avctx, int luma_width, int luma_height) |
167 | { |
168 | int p, chroma_width, chroma_height; |
169 | int luma_size, chroma_size; |
170 | ptrdiff_t luma_pitch, chroma_pitch; |
171 | |
172 | if (luma_width < 16 || luma_width > 640 || |
173 | luma_height < 16 || luma_height > 480 || |
174 | luma_width & 3 || luma_height & 3) { |
175 | av_log(avctx, AV_LOG_ERROR, "Invalid picture dimensions: %d x %d!\n", |
176 | luma_width, luma_height); |
177 | return AVERROR_INVALIDDATA; |
178 | } |
179 | |
180 | ctx->width = luma_width ; |
181 | ctx->height = luma_height; |
182 | |
183 | chroma_width = FFALIGN(luma_width >> 2, 4); |
184 | chroma_height = FFALIGN(luma_height >> 2, 4); |
185 | |
186 | luma_pitch = FFALIGN(luma_width, 16); |
187 | chroma_pitch = FFALIGN(chroma_width, 16); |
188 | |
189 | /* Calculate size of the luminance plane. */ |
190 | /* Add one line more for INTRA prediction. */ |
191 | luma_size = luma_pitch * (luma_height + 1); |
192 | |
193 | /* Calculate size of a chrominance planes. */ |
194 | /* Add one line more for INTRA prediction. */ |
195 | chroma_size = chroma_pitch * (chroma_height + 1); |
196 | |
197 | /* allocate frame buffers */ |
198 | for (p = 0; p < 3; p++) { |
199 | ctx->planes[p].pitch = !p ? luma_pitch : chroma_pitch; |
200 | ctx->planes[p].width = !p ? luma_width : chroma_width; |
201 | ctx->planes[p].height = !p ? luma_height : chroma_height; |
202 | |
203 | ctx->planes[p].buffers[0] = av_malloc(!p ? luma_size : chroma_size); |
204 | ctx->planes[p].buffers[1] = av_malloc(!p ? luma_size : chroma_size); |
205 | |
206 | if (!ctx->planes[p].buffers[0] || !ctx->planes[p].buffers[1]) { |
207 | free_frame_buffers(ctx); |
208 | return AVERROR(ENOMEM); |
209 | } |
210 | |
211 | /* fill the INTRA prediction lines with the middle pixel value = 64 */ |
212 | memset(ctx->planes[p].buffers[0], 0x40, ctx->planes[p].pitch); |
213 | memset(ctx->planes[p].buffers[1], 0x40, ctx->planes[p].pitch); |
214 | |
215 | /* set buffer pointers = buf_ptr + pitch and thus skip the INTRA prediction line */ |
216 | ctx->planes[p].pixels[0] = ctx->planes[p].buffers[0] + ctx->planes[p].pitch; |
217 | ctx->planes[p].pixels[1] = ctx->planes[p].buffers[1] + ctx->planes[p].pitch; |
218 | memset(ctx->planes[p].pixels[0], 0, ctx->planes[p].pitch * ctx->planes[p].height); |
219 | memset(ctx->planes[p].pixels[1], 0, ctx->planes[p].pitch * ctx->planes[p].height); |
220 | } |
221 | |
222 | return 0; |
223 | } |
224 | |
225 | /** |
226 | * Copy pixels of the cell(x + mv_x, y + mv_y) from the previous frame into |
227 | * the cell(x, y) in the current frame. |
228 | * |
229 | * @param ctx pointer to the decoder context |
230 | * @param plane pointer to the plane descriptor |
231 | * @param cell pointer to the cell descriptor |
232 | */ |
233 | static int copy_cell(Indeo3DecodeContext *ctx, Plane *plane, Cell *cell) |
234 | { |
235 | int h, w, mv_x, mv_y, offset, offset_dst; |
236 | uint8_t *src, *dst; |
237 | |
238 | /* setup output and reference pointers */ |
239 | offset_dst = (cell->ypos << 2) * plane->pitch + (cell->xpos << 2); |
240 | dst = plane->pixels[ctx->buf_sel] + offset_dst; |
241 | if(cell->mv_ptr){ |
242 | mv_y = cell->mv_ptr[0]; |
243 | mv_x = cell->mv_ptr[1]; |
244 | }else |
245 | mv_x= mv_y= 0; |
246 | |
247 | /* -1 because there is an extra line on top for prediction */ |
248 | if ((cell->ypos << 2) + mv_y < -1 || (cell->xpos << 2) + mv_x < 0 || |
249 | ((cell->ypos + cell->height) << 2) + mv_y > plane->height || |
250 | ((cell->xpos + cell->width) << 2) + mv_x > plane->width) { |
251 | av_log(ctx->avctx, AV_LOG_ERROR, |
252 | "Motion vectors point out of the frame.\n"); |
253 | return AVERROR_INVALIDDATA; |
254 | } |
255 | |
256 | offset = offset_dst + mv_y * plane->pitch + mv_x; |
257 | src = plane->pixels[ctx->buf_sel ^ 1] + offset; |
258 | |
259 | h = cell->height << 2; |
260 | |
261 | for (w = cell->width; w > 0;) { |
262 | /* copy using 16xH blocks */ |
263 | if (!((cell->xpos << 2) & 15) && w >= 4) { |
264 | for (; w >= 4; src += 16, dst += 16, w -= 4) |
265 | ctx->hdsp.put_pixels_tab[0][0](dst, src, plane->pitch, h); |
266 | } |
267 | |
268 | /* copy using 8xH blocks */ |
269 | if (!((cell->xpos << 2) & 7) && w >= 2) { |
270 | ctx->hdsp.put_pixels_tab[1][0](dst, src, plane->pitch, h); |
271 | w -= 2; |
272 | src += 8; |
273 | dst += 8; |
274 | } else if (w >= 1) { |
275 | ctx->hdsp.put_pixels_tab[2][0](dst, src, plane->pitch, h); |
276 | w--; |
277 | src += 4; |
278 | dst += 4; |
279 | } |
280 | } |
281 | |
282 | return 0; |
283 | } |
284 | |
285 | |
286 | /* Average 4/8 pixels at once without rounding using SWAR */ |
287 | #define AVG_32(dst, src, ref) \ |
288 | AV_WN32A(dst, ((AV_RN32(src) + AV_RN32(ref)) >> 1) & 0x7F7F7F7FUL) |
289 | |
290 | #define AVG_64(dst, src, ref) \ |
291 | AV_WN64A(dst, ((AV_RN64(src) + AV_RN64(ref)) >> 1) & 0x7F7F7F7F7F7F7F7FULL) |
292 | |
293 | |
294 | /* |
295 | * Replicate each even pixel as follows: |
296 | * ABCDEFGH -> AACCEEGG |
297 | */ |
298 | static inline uint64_t replicate64(uint64_t a) { |
299 | #if HAVE_BIGENDIAN |
300 | a &= 0xFF00FF00FF00FF00ULL; |
301 | a |= a >> 8; |
302 | #else |
303 | a &= 0x00FF00FF00FF00FFULL; |
304 | a |= a << 8; |
305 | #endif |
306 | return a; |
307 | } |
308 | |
309 | static inline uint32_t replicate32(uint32_t a) { |
310 | #if HAVE_BIGENDIAN |
311 | a &= 0xFF00FF00UL; |
312 | a |= a >> 8; |
313 | #else |
314 | a &= 0x00FF00FFUL; |
315 | a |= a << 8; |
316 | #endif |
317 | return a; |
318 | } |
319 | |
320 | |
321 | /* Fill n lines with 64-bit pixel value pix */ |
322 | static inline void fill_64(uint8_t *dst, const uint64_t pix, int32_t n, |
323 | int32_t row_offset) |
324 | { |
325 | for (; n > 0; dst += row_offset, n--) |
326 | AV_WN64A(dst, pix); |
327 | } |
328 | |
329 | |
330 | /* Error codes for cell decoding. */ |
331 | enum { |
332 | IV3_NOERR = 0, |
333 | IV3_BAD_RLE = 1, |
334 | IV3_BAD_DATA = 2, |
335 | IV3_BAD_COUNTER = 3, |
336 | IV3_UNSUPPORTED = 4, |
337 | IV3_OUT_OF_DATA = 5 |
338 | }; |
339 | |
340 | |
341 | #define BUFFER_PRECHECK \ |
342 | if (*data_ptr >= last_ptr) \ |
343 | return IV3_OUT_OF_DATA; \ |
344 | |
345 | #define RLE_BLOCK_COPY \ |
346 | if (cell->mv_ptr || !skip_flag) \ |
347 | copy_block4(dst, ref, row_offset, row_offset, 4 << v_zoom) |
348 | |
349 | #define RLE_BLOCK_COPY_8 \ |
350 | pix64 = AV_RN64(ref);\ |
351 | if (is_first_row) {/* special prediction case: top line of a cell */\ |
352 | pix64 = replicate64(pix64);\ |
353 | fill_64(dst + row_offset, pix64, 7, row_offset);\ |
354 | AVG_64(dst, ref, dst + row_offset);\ |
355 | } else \ |
356 | fill_64(dst, pix64, 8, row_offset) |
357 | |
358 | #define RLE_LINES_COPY \ |
359 | copy_block4(dst, ref, row_offset, row_offset, num_lines << v_zoom) |
360 | |
361 | #define RLE_LINES_COPY_M10 \ |
362 | pix64 = AV_RN64(ref);\ |
363 | if (is_top_of_cell) {\ |
364 | pix64 = replicate64(pix64);\ |
365 | fill_64(dst + row_offset, pix64, (num_lines << 1) - 1, row_offset);\ |
366 | AVG_64(dst, ref, dst + row_offset);\ |
367 | } else \ |
368 | fill_64(dst, pix64, num_lines << 1, row_offset) |
369 | |
370 | #define APPLY_DELTA_4 \ |
371 | AV_WN16A(dst + line_offset ,\ |
372 | (AV_RN16(ref ) + delta_tab->deltas[dyad1]) & 0x7F7F);\ |
373 | AV_WN16A(dst + line_offset + 2,\ |
374 | (AV_RN16(ref + 2) + delta_tab->deltas[dyad2]) & 0x7F7F);\ |
375 | if (mode >= 3) {\ |
376 | if (is_top_of_cell && !cell->ypos) {\ |
377 | AV_COPY32U(dst, dst + row_offset);\ |
378 | } else {\ |
379 | AVG_32(dst, ref, dst + row_offset);\ |
380 | }\ |
381 | } |
382 | |
383 | #define APPLY_DELTA_8 \ |
384 | /* apply two 32-bit VQ deltas to next even line */\ |
385 | if (is_top_of_cell) { \ |
386 | AV_WN32A(dst + row_offset , \ |
387 | (replicate32(AV_RN32(ref )) + delta_tab->deltas_m10[dyad1]) & 0x7F7F7F7F);\ |
388 | AV_WN32A(dst + row_offset + 4, \ |
389 | (replicate32(AV_RN32(ref + 4)) + delta_tab->deltas_m10[dyad2]) & 0x7F7F7F7F);\ |
390 | } else { \ |
391 | AV_WN32A(dst + row_offset , \ |
392 | (AV_RN32(ref ) + delta_tab->deltas_m10[dyad1]) & 0x7F7F7F7F);\ |
393 | AV_WN32A(dst + row_offset + 4, \ |
394 | (AV_RN32(ref + 4) + delta_tab->deltas_m10[dyad2]) & 0x7F7F7F7F);\ |
395 | } \ |
396 | /* odd lines are not coded but rather interpolated/replicated */\ |
397 | /* first line of the cell on the top of image? - replicate */\ |
398 | /* otherwise - interpolate */\ |
399 | if (is_top_of_cell && !cell->ypos) {\ |
400 | AV_COPY64U(dst, dst + row_offset);\ |
401 | } else \ |
402 | AVG_64(dst, ref, dst + row_offset); |
403 | |
404 | |
405 | #define APPLY_DELTA_1011_INTER \ |
406 | if (mode == 10) { \ |
407 | AV_WN32A(dst , \ |
408 | (AV_RN32(dst ) + delta_tab->deltas_m10[dyad1]) & 0x7F7F7F7F);\ |
409 | AV_WN32A(dst + 4 , \ |
410 | (AV_RN32(dst + 4 ) + delta_tab->deltas_m10[dyad2]) & 0x7F7F7F7F);\ |
411 | AV_WN32A(dst + row_offset , \ |
412 | (AV_RN32(dst + row_offset ) + delta_tab->deltas_m10[dyad1]) & 0x7F7F7F7F);\ |
413 | AV_WN32A(dst + row_offset + 4, \ |
414 | (AV_RN32(dst + row_offset + 4) + delta_tab->deltas_m10[dyad2]) & 0x7F7F7F7F);\ |
415 | } else { \ |
416 | AV_WN16A(dst , \ |
417 | (AV_RN16(dst ) + delta_tab->deltas[dyad1]) & 0x7F7F);\ |
418 | AV_WN16A(dst + 2 , \ |
419 | (AV_RN16(dst + 2 ) + delta_tab->deltas[dyad2]) & 0x7F7F);\ |
420 | AV_WN16A(dst + row_offset , \ |
421 | (AV_RN16(dst + row_offset ) + delta_tab->deltas[dyad1]) & 0x7F7F);\ |
422 | AV_WN16A(dst + row_offset + 2, \ |
423 | (AV_RN16(dst + row_offset + 2) + delta_tab->deltas[dyad2]) & 0x7F7F);\ |
424 | } |
425 | |
426 | |
427 | static int decode_cell_data(Indeo3DecodeContext *ctx, Cell *cell, |
428 | uint8_t *block, uint8_t *ref_block, |
429 | ptrdiff_t row_offset, int h_zoom, int v_zoom, int mode, |
430 | const vqEntry *delta[2], int swap_quads[2], |
431 | const uint8_t **data_ptr, const uint8_t *last_ptr) |
432 | { |
433 | int x, y, line, num_lines; |
434 | int rle_blocks = 0; |
435 | uint8_t code, *dst, *ref; |
436 | const vqEntry *delta_tab; |
437 | unsigned int dyad1, dyad2; |
438 | uint64_t pix64; |
439 | int skip_flag = 0, is_top_of_cell, is_first_row = 1; |
440 | int blk_row_offset, line_offset; |
441 | |
442 | blk_row_offset = (row_offset << (2 + v_zoom)) - (cell->width << 2); |
443 | line_offset = v_zoom ? row_offset : 0; |
444 | |
445 | if (cell->height & v_zoom || cell->width & h_zoom) |
446 | return IV3_BAD_DATA; |
447 | |
448 | for (y = 0; y < cell->height; is_first_row = 0, y += 1 + v_zoom) { |
449 | for (x = 0; x < cell->width; x += 1 + h_zoom) { |
450 | ref = ref_block; |
451 | dst = block; |
452 | |
453 | if (rle_blocks > 0) { |
454 | if (mode <= 4) { |
455 | RLE_BLOCK_COPY; |
456 | } else if (mode == 10 && !cell->mv_ptr) { |
457 | RLE_BLOCK_COPY_8; |
458 | } |
459 | rle_blocks--; |
460 | } else { |
461 | for (line = 0; line < 4;) { |
462 | num_lines = 1; |
463 | is_top_of_cell = is_first_row && !line; |
464 | |
465 | /* select primary VQ table for odd, secondary for even lines */ |
466 | if (mode <= 4) |
467 | delta_tab = delta[line & 1]; |
468 | else |
469 | delta_tab = delta[1]; |
470 | BUFFER_PRECHECK; |
471 | code = bytestream_get_byte(data_ptr); |
472 | if (code < 248) { |
473 | if (code < delta_tab->num_dyads) { |
474 | BUFFER_PRECHECK; |
475 | dyad1 = bytestream_get_byte(data_ptr); |
476 | dyad2 = code; |
477 | if (dyad1 >= delta_tab->num_dyads || dyad1 >= 248) |
478 | return IV3_BAD_DATA; |
479 | } else { |
480 | /* process QUADS */ |
481 | code -= delta_tab->num_dyads; |
482 | dyad1 = code / delta_tab->quad_exp; |
483 | dyad2 = code % delta_tab->quad_exp; |
484 | if (swap_quads[line & 1]) |
485 | FFSWAP(unsigned int, dyad1, dyad2); |
486 | } |
487 | if (mode <= 4) { |
488 | APPLY_DELTA_4; |
489 | } else if (mode == 10 && !cell->mv_ptr) { |
490 | APPLY_DELTA_8; |
491 | } else { |
492 | APPLY_DELTA_1011_INTER; |
493 | } |
494 | } else { |
495 | /* process RLE codes */ |
496 | switch (code) { |
497 | case RLE_ESC_FC: |
498 | skip_flag = 0; |
499 | rle_blocks = 1; |
500 | code = 253; |
501 | /* FALLTHROUGH */ |
502 | case RLE_ESC_FF: |
503 | case RLE_ESC_FE: |
504 | case RLE_ESC_FD: |
505 | num_lines = 257 - code - line; |
506 | if (num_lines <= 0) |
507 | return IV3_BAD_RLE; |
508 | if (mode <= 4) { |
509 | RLE_LINES_COPY; |
510 | } else if (mode == 10 && !cell->mv_ptr) { |
511 | RLE_LINES_COPY_M10; |
512 | } |
513 | break; |
514 | case RLE_ESC_FB: |
515 | BUFFER_PRECHECK; |
516 | code = bytestream_get_byte(data_ptr); |
517 | rle_blocks = (code & 0x1F) - 1; /* set block counter */ |
518 | if (code >= 64 || rle_blocks < 0) |
519 | return IV3_BAD_COUNTER; |
520 | skip_flag = code & 0x20; |
521 | num_lines = 4 - line; /* enforce next block processing */ |
522 | if (mode >= 10 || (cell->mv_ptr || !skip_flag)) { |
523 | if (mode <= 4) { |
524 | RLE_LINES_COPY; |
525 | } else if (mode == 10 && !cell->mv_ptr) { |
526 | RLE_LINES_COPY_M10; |
527 | } |
528 | } |
529 | break; |
530 | case RLE_ESC_F9: |
531 | skip_flag = 1; |
532 | rle_blocks = 1; |
533 | /* FALLTHROUGH */ |
534 | case RLE_ESC_FA: |
535 | if (line) |
536 | return IV3_BAD_RLE; |
537 | num_lines = 4; /* enforce next block processing */ |
538 | if (cell->mv_ptr) { |
539 | if (mode <= 4) { |
540 | RLE_LINES_COPY; |
541 | } else if (mode == 10 && !cell->mv_ptr) { |
542 | RLE_LINES_COPY_M10; |
543 | } |
544 | } |
545 | break; |
546 | default: |
547 | return IV3_UNSUPPORTED; |
548 | } |
549 | } |
550 | |
551 | line += num_lines; |
552 | ref += row_offset * (num_lines << v_zoom); |
553 | dst += row_offset * (num_lines << v_zoom); |
554 | } |
555 | } |
556 | |
557 | /* move to next horizontal block */ |
558 | block += 4 << h_zoom; |
559 | ref_block += 4 << h_zoom; |
560 | } |
561 | |
562 | /* move to next line of blocks */ |
563 | ref_block += blk_row_offset; |
564 | block += blk_row_offset; |
565 | } |
566 | return IV3_NOERR; |
567 | } |
568 | |
569 | |
570 | /** |
571 | * Decode a vector-quantized cell. |
572 | * It consists of several routines, each of which handles one or more "modes" |
573 | * with which a cell can be encoded. |
574 | * |
575 | * @param ctx pointer to the decoder context |
576 | * @param avctx ptr to the AVCodecContext |
577 | * @param plane pointer to the plane descriptor |
578 | * @param cell pointer to the cell descriptor |
579 | * @param data_ptr pointer to the compressed data |
580 | * @param last_ptr pointer to the last byte to catch reads past end of buffer |
581 | * @return number of consumed bytes or negative number in case of error |
582 | */ |
583 | static int decode_cell(Indeo3DecodeContext *ctx, AVCodecContext *avctx, |
584 | Plane *plane, Cell *cell, const uint8_t *data_ptr, |
585 | const uint8_t *last_ptr) |
586 | { |
587 | int x, mv_x, mv_y, mode, vq_index, prim_indx, second_indx; |
588 | int zoom_fac; |
589 | int offset, error = 0, swap_quads[2]; |
590 | uint8_t code, *block, *ref_block = 0; |
591 | const vqEntry *delta[2]; |
592 | const uint8_t *data_start = data_ptr; |
593 | |
594 | /* get coding mode and VQ table index from the VQ descriptor byte */ |
595 | code = *data_ptr++; |
596 | mode = code >> 4; |
597 | vq_index = code & 0xF; |
598 | |
599 | /* setup output and reference pointers */ |
600 | offset = (cell->ypos << 2) * plane->pitch + (cell->xpos << 2); |
601 | block = plane->pixels[ctx->buf_sel] + offset; |
602 | |
603 | if (!cell->mv_ptr) { |
604 | /* use previous line as reference for INTRA cells */ |
605 | ref_block = block - plane->pitch; |
606 | } else if (mode >= 10) { |
607 | /* for mode 10 and 11 INTER first copy the predicted cell into the current one */ |
608 | /* so we don't need to do data copying for each RLE code later */ |
609 | int ret = copy_cell(ctx, plane, cell); |
610 | if (ret < 0) |
611 | return ret; |
612 | } else { |
613 | /* set the pointer to the reference pixels for modes 0-4 INTER */ |
614 | mv_y = cell->mv_ptr[0]; |
615 | mv_x = cell->mv_ptr[1]; |
616 | |
617 | /* -1 because there is an extra line on top for prediction */ |
618 | if ((cell->ypos << 2) + mv_y < -1 || (cell->xpos << 2) + mv_x < 0 || |
619 | ((cell->ypos + cell->height) << 2) + mv_y > plane->height || |
620 | ((cell->xpos + cell->width) << 2) + mv_x > plane->width) { |
621 | av_log(ctx->avctx, AV_LOG_ERROR, |
622 | "Motion vectors point out of the frame.\n"); |
623 | return AVERROR_INVALIDDATA; |
624 | } |
625 | |
626 | offset += mv_y * plane->pitch + mv_x; |
627 | ref_block = plane->pixels[ctx->buf_sel ^ 1] + offset; |
628 | } |
629 | |
630 | /* select VQ tables as follows: */ |
631 | /* modes 0 and 3 use only the primary table for all lines in a block */ |
632 | /* while modes 1 and 4 switch between primary and secondary tables on alternate lines */ |
633 | if (mode == 1 || mode == 4) { |
634 | code = ctx->alt_quant[vq_index]; |
635 | prim_indx = (code >> 4) + ctx->cb_offset; |
636 | second_indx = (code & 0xF) + ctx->cb_offset; |
637 | } else { |
638 | vq_index += ctx->cb_offset; |
639 | prim_indx = second_indx = vq_index; |
640 | } |
641 | |
642 | if (prim_indx >= 24 || second_indx >= 24) { |
643 | av_log(avctx, AV_LOG_ERROR, "Invalid VQ table indexes! Primary: %d, secondary: %d!\n", |
644 | prim_indx, second_indx); |
645 | return AVERROR_INVALIDDATA; |
646 | } |
647 | |
648 | delta[0] = &vq_tab[second_indx]; |
649 | delta[1] = &vq_tab[prim_indx]; |
650 | swap_quads[0] = second_indx >= 16; |
651 | swap_quads[1] = prim_indx >= 16; |
652 | |
653 | /* requantize the prediction if VQ index of this cell differs from VQ index */ |
654 | /* of the predicted cell in order to avoid overflows. */ |
655 | if (vq_index >= 8 && ref_block) { |
656 | for (x = 0; x < cell->width << 2; x++) |
657 | ref_block[x] = requant_tab[vq_index & 7][ref_block[x] & 127]; |
658 | } |
659 | |
660 | error = IV3_NOERR; |
661 | |
662 | switch (mode) { |
663 | case 0: /*------------------ MODES 0 & 1 (4x4 block processing) --------------------*/ |
664 | case 1: |
665 | case 3: /*------------------ MODES 3 & 4 (4x8 block processing) --------------------*/ |
666 | case 4: |
667 | if (mode >= 3 && cell->mv_ptr) { |
668 | av_log(avctx, AV_LOG_ERROR, "Attempt to apply Mode 3/4 to an INTER cell!\n"); |
669 | return AVERROR_INVALIDDATA; |
670 | } |
671 | |
672 | zoom_fac = mode >= 3; |
673 | error = decode_cell_data(ctx, cell, block, ref_block, plane->pitch, |
674 | 0, zoom_fac, mode, delta, swap_quads, |
675 | &data_ptr, last_ptr); |
676 | break; |
677 | case 10: /*-------------------- MODE 10 (8x8 block processing) ---------------------*/ |
678 | case 11: /*----------------- MODE 11 (4x8 INTER block processing) ------------------*/ |
679 | if (mode == 10 && !cell->mv_ptr) { /* MODE 10 INTRA processing */ |
680 | error = decode_cell_data(ctx, cell, block, ref_block, plane->pitch, |
681 | 1, 1, mode, delta, swap_quads, |
682 | &data_ptr, last_ptr); |
683 | } else { /* mode 10 and 11 INTER processing */ |
684 | if (mode == 11 && !cell->mv_ptr) { |
685 | av_log(avctx, AV_LOG_ERROR, "Attempt to use Mode 11 for an INTRA cell!\n"); |
686 | return AVERROR_INVALIDDATA; |
687 | } |
688 | |
689 | zoom_fac = mode == 10; |
690 | error = decode_cell_data(ctx, cell, block, ref_block, plane->pitch, |
691 | zoom_fac, 1, mode, delta, swap_quads, |
692 | &data_ptr, last_ptr); |
693 | } |
694 | break; |
695 | default: |
696 | av_log(avctx, AV_LOG_ERROR, "Unsupported coding mode: %d\n", mode); |
697 | return AVERROR_INVALIDDATA; |
698 | }//switch mode |
699 | |
700 | switch (error) { |
701 | case IV3_BAD_RLE: |
702 | av_log(avctx, AV_LOG_ERROR, "Mode %d: RLE code %X is not allowed at the current line\n", |
703 | mode, data_ptr[-1]); |
704 | return AVERROR_INVALIDDATA; |
705 | case IV3_BAD_DATA: |
706 | av_log(avctx, AV_LOG_ERROR, "Mode %d: invalid VQ data\n", mode); |
707 | return AVERROR_INVALIDDATA; |
708 | case IV3_BAD_COUNTER: |
709 | av_log(avctx, AV_LOG_ERROR, "Mode %d: RLE-FB invalid counter: %d\n", mode, code); |
710 | return AVERROR_INVALIDDATA; |
711 | case IV3_UNSUPPORTED: |
712 | av_log(avctx, AV_LOG_ERROR, "Mode %d: unsupported RLE code: %X\n", mode, data_ptr[-1]); |
713 | return AVERROR_INVALIDDATA; |
714 | case IV3_OUT_OF_DATA: |
715 | av_log(avctx, AV_LOG_ERROR, "Mode %d: attempt to read past end of buffer\n", mode); |
716 | return AVERROR_INVALIDDATA; |
717 | } |
718 | |
719 | return data_ptr - data_start; /* report number of bytes consumed from the input buffer */ |
720 | } |
721 | |
722 | |
723 | /* Binary tree codes. */ |
724 | enum { |
725 | H_SPLIT = 0, |
726 | V_SPLIT = 1, |
727 | INTRA_NULL = 2, |
728 | INTER_DATA = 3 |
729 | }; |
730 | |
731 | |
732 | #define SPLIT_CELL(size, new_size) (new_size) = ((size) > 2) ? ((((size) + 2) >> 2) << 1) : 1 |
733 | |
734 | #define UPDATE_BITPOS(n) \ |
735 | ctx->skip_bits += (n); \ |
736 | ctx->need_resync = 1 |
737 | |
738 | #define RESYNC_BITSTREAM \ |
739 | if (ctx->need_resync && !(get_bits_count(&ctx->gb) & 7)) { \ |
740 | skip_bits_long(&ctx->gb, ctx->skip_bits); \ |
741 | ctx->skip_bits = 0; \ |
742 | ctx->need_resync = 0; \ |
743 | } |
744 | |
745 | #define CHECK_CELL \ |
746 | if (curr_cell.xpos + curr_cell.width > (plane->width >> 2) || \ |
747 | curr_cell.ypos + curr_cell.height > (plane->height >> 2)) { \ |
748 | av_log(avctx, AV_LOG_ERROR, "Invalid cell: x=%d, y=%d, w=%d, h=%d\n", \ |
749 | curr_cell.xpos, curr_cell.ypos, curr_cell.width, curr_cell.height); \ |
750 | return AVERROR_INVALIDDATA; \ |
751 | } |
752 | |
753 | |
754 | static int parse_bintree(Indeo3DecodeContext *ctx, AVCodecContext *avctx, |
755 | Plane *plane, int code, Cell *ref_cell, |
756 | const int depth, const int strip_width) |
757 | { |
758 | Cell curr_cell; |
759 | int bytes_used, ret; |
760 | |
761 | if (depth <= 0) { |
762 | av_log(avctx, AV_LOG_ERROR, "Stack overflow (corrupted binary tree)!\n"); |
763 | return AVERROR_INVALIDDATA; // unwind recursion |
764 | } |
765 | |
766 | curr_cell = *ref_cell; // clone parent cell |
767 | if (code == H_SPLIT) { |
768 | SPLIT_CELL(ref_cell->height, curr_cell.height); |
769 | ref_cell->ypos += curr_cell.height; |
770 | ref_cell->height -= curr_cell.height; |
771 | if (ref_cell->height <= 0 || curr_cell.height <= 0) |
772 | return AVERROR_INVALIDDATA; |
773 | } else if (code == V_SPLIT) { |
774 | if (curr_cell.width > strip_width) { |
775 | /* split strip */ |
776 | curr_cell.width = (curr_cell.width <= (strip_width << 1) ? 1 : 2) * strip_width; |
777 | } else |
778 | SPLIT_CELL(ref_cell->width, curr_cell.width); |
779 | ref_cell->xpos += curr_cell.width; |
780 | ref_cell->width -= curr_cell.width; |
781 | if (ref_cell->width <= 0 || curr_cell.width <= 0) |
782 | return AVERROR_INVALIDDATA; |
783 | } |
784 | |
785 | while (get_bits_left(&ctx->gb) >= 2) { /* loop until return */ |
786 | RESYNC_BITSTREAM; |
787 | switch (code = get_bits(&ctx->gb, 2)) { |
788 | case H_SPLIT: |
789 | case V_SPLIT: |
790 | if (parse_bintree(ctx, avctx, plane, code, &curr_cell, depth - 1, strip_width)) |
791 | return AVERROR_INVALIDDATA; |
792 | break; |
793 | case INTRA_NULL: |
794 | if (!curr_cell.tree) { /* MC tree INTRA code */ |
795 | curr_cell.mv_ptr = 0; /* mark the current strip as INTRA */ |
796 | curr_cell.tree = 1; /* enter the VQ tree */ |
797 | } else { /* VQ tree NULL code */ |
798 | RESYNC_BITSTREAM; |
799 | code = get_bits(&ctx->gb, 2); |
800 | if (code >= 2) { |
801 | av_log(avctx, AV_LOG_ERROR, "Invalid VQ_NULL code: %d\n", code); |
802 | return AVERROR_INVALIDDATA; |
803 | } |
804 | if (code == 1) |
805 | av_log(avctx, AV_LOG_ERROR, "SkipCell procedure not implemented yet!\n"); |
806 | |
807 | CHECK_CELL |
808 | if (!curr_cell.mv_ptr) |
809 | return AVERROR_INVALIDDATA; |
810 | |
811 | ret = copy_cell(ctx, plane, &curr_cell); |
812 | return ret; |
813 | } |
814 | break; |
815 | case INTER_DATA: |
816 | if (!curr_cell.tree) { /* MC tree INTER code */ |
817 | unsigned mv_idx; |
818 | /* get motion vector index and setup the pointer to the mv set */ |
819 | if (!ctx->need_resync) |
820 | ctx->next_cell_data = &ctx->gb.buffer[(get_bits_count(&ctx->gb) + 7) >> 3]; |
821 | if (ctx->next_cell_data >= ctx->last_byte) { |
822 | av_log(avctx, AV_LOG_ERROR, "motion vector out of array\n"); |
823 | return AVERROR_INVALIDDATA; |
824 | } |
825 | mv_idx = *(ctx->next_cell_data++); |
826 | if (mv_idx >= ctx->num_vectors) { |
827 | av_log(avctx, AV_LOG_ERROR, "motion vector index out of range\n"); |
828 | return AVERROR_INVALIDDATA; |
829 | } |
830 | curr_cell.mv_ptr = &ctx->mc_vectors[mv_idx << 1]; |
831 | curr_cell.tree = 1; /* enter the VQ tree */ |
832 | UPDATE_BITPOS(8); |
833 | } else { /* VQ tree DATA code */ |
834 | if (!ctx->need_resync) |
835 | ctx->next_cell_data = &ctx->gb.buffer[(get_bits_count(&ctx->gb) + 7) >> 3]; |
836 | |
837 | CHECK_CELL |
838 | bytes_used = decode_cell(ctx, avctx, plane, &curr_cell, |
839 | ctx->next_cell_data, ctx->last_byte); |
840 | if (bytes_used < 0) |
841 | return AVERROR_INVALIDDATA; |
842 | |
843 | UPDATE_BITPOS(bytes_used << 3); |
844 | ctx->next_cell_data += bytes_used; |
845 | return 0; |
846 | } |
847 | break; |
848 | } |
849 | }//while |
850 | |
851 | return AVERROR_INVALIDDATA; |
852 | } |
853 | |
854 | |
855 | static int decode_plane(Indeo3DecodeContext *ctx, AVCodecContext *avctx, |
856 | Plane *plane, const uint8_t *data, int32_t data_size, |
857 | int32_t strip_width) |
858 | { |
859 | Cell curr_cell; |
860 | unsigned num_vectors; |
861 | |
862 | /* each plane data starts with mc_vector_count field, */ |
863 | /* an optional array of motion vectors followed by the vq data */ |
864 | num_vectors = bytestream_get_le32(&data); data_size -= 4; |
865 | if (num_vectors > 256) { |
866 | av_log(ctx->avctx, AV_LOG_ERROR, |
867 | "Read invalid number of motion vectors %d\n", num_vectors); |
868 | return AVERROR_INVALIDDATA; |
869 | } |
870 | if (num_vectors * 2 > data_size) |
871 | return AVERROR_INVALIDDATA; |
872 | |
873 | ctx->num_vectors = num_vectors; |
874 | ctx->mc_vectors = num_vectors ? data : 0; |
875 | |
876 | /* init the bitreader */ |
877 | init_get_bits(&ctx->gb, &data[num_vectors * 2], (data_size - num_vectors * 2) << 3); |
878 | ctx->skip_bits = 0; |
879 | ctx->need_resync = 0; |
880 | |
881 | ctx->last_byte = data + data_size; |
882 | |
883 | /* initialize the 1st cell and set its dimensions to whole plane */ |
884 | curr_cell.xpos = curr_cell.ypos = 0; |
885 | curr_cell.width = plane->width >> 2; |
886 | curr_cell.height = plane->height >> 2; |
887 | curr_cell.tree = 0; // we are in the MC tree now |
888 | curr_cell.mv_ptr = 0; // no motion vector = INTRA cell |
889 | |
890 | return parse_bintree(ctx, avctx, plane, INTRA_NULL, &curr_cell, CELL_STACK_MAX, strip_width); |
891 | } |
892 | |
893 | |
894 | #define OS_HDR_ID MKBETAG('F', 'R', 'M', 'H') |
895 | |
896 | static int decode_frame_headers(Indeo3DecodeContext *ctx, AVCodecContext *avctx, |
897 | const uint8_t *buf, int buf_size) |
898 | { |
899 | GetByteContext gb; |
900 | const uint8_t *bs_hdr; |
901 | uint32_t frame_num, word2, check_sum, data_size; |
902 | int y_offset, u_offset, v_offset; |
903 | uint32_t starts[3], ends[3]; |
904 | uint16_t height, width; |
905 | int i, j; |
906 | |
907 | bytestream2_init(&gb, buf, buf_size); |
908 | |
909 | /* parse and check the OS header */ |
910 | frame_num = bytestream2_get_le32(&gb); |
911 | word2 = bytestream2_get_le32(&gb); |
912 | check_sum = bytestream2_get_le32(&gb); |
913 | data_size = bytestream2_get_le32(&gb); |
914 | |
915 | if ((frame_num ^ word2 ^ data_size ^ OS_HDR_ID) != check_sum) { |
916 | av_log(avctx, AV_LOG_ERROR, "OS header checksum mismatch!\n"); |
917 | return AVERROR_INVALIDDATA; |
918 | } |
919 | |
920 | /* parse the bitstream header */ |
921 | bs_hdr = gb.buffer; |
922 | |
923 | if (bytestream2_get_le16(&gb) != 32) { |
924 | av_log(avctx, AV_LOG_ERROR, "Unsupported codec version!\n"); |
925 | return AVERROR_INVALIDDATA; |
926 | } |
927 | |
928 | ctx->frame_num = frame_num; |
929 | ctx->frame_flags = bytestream2_get_le16(&gb); |
930 | ctx->data_size = (bytestream2_get_le32(&gb) + 7) >> 3; |
931 | ctx->cb_offset = bytestream2_get_byte(&gb); |
932 | |
933 | if (ctx->data_size == 16) |
934 | return 4; |
935 | ctx->data_size = FFMIN(ctx->data_size, buf_size - 16); |
936 | |
937 | bytestream2_skip(&gb, 3); // skip reserved byte and checksum |
938 | |
939 | /* check frame dimensions */ |
940 | height = bytestream2_get_le16(&gb); |
941 | width = bytestream2_get_le16(&gb); |
942 | if (av_image_check_size(width, height, 0, avctx)) |
943 | return AVERROR_INVALIDDATA; |
944 | |
945 | if (width != ctx->width || height != ctx->height) { |
946 | int res; |
947 | |
948 | ff_dlog(avctx, "Frame dimensions changed!\n"); |
949 | |
950 | if (width < 16 || width > 640 || |
951 | height < 16 || height > 480 || |
952 | width & 3 || height & 3) { |
953 | av_log(avctx, AV_LOG_ERROR, |
954 | "Invalid picture dimensions: %d x %d!\n", width, height); |
955 | return AVERROR_INVALIDDATA; |
956 | } |
957 | free_frame_buffers(ctx); |
958 | if ((res = allocate_frame_buffers(ctx, avctx, width, height)) < 0) |
959 | return res; |
960 | if ((res = ff_set_dimensions(avctx, width, height)) < 0) |
961 | return res; |
962 | } |
963 | |
964 | y_offset = bytestream2_get_le32(&gb); |
965 | v_offset = bytestream2_get_le32(&gb); |
966 | u_offset = bytestream2_get_le32(&gb); |
967 | bytestream2_skip(&gb, 4); |
968 | |
969 | /* unfortunately there is no common order of planes in the buffer */ |
970 | /* so we use that sorting algo for determining planes data sizes */ |
971 | starts[0] = y_offset; |
972 | starts[1] = v_offset; |
973 | starts[2] = u_offset; |
974 | |
975 | for (j = 0; j < 3; j++) { |
976 | ends[j] = ctx->data_size; |
977 | for (i = 2; i >= 0; i--) |
978 | if (starts[i] < ends[j] && starts[i] > starts[j]) |
979 | ends[j] = starts[i]; |
980 | } |
981 | |
982 | ctx->y_data_size = ends[0] - starts[0]; |
983 | ctx->v_data_size = ends[1] - starts[1]; |
984 | ctx->u_data_size = ends[2] - starts[2]; |
985 | if (FFMIN3(y_offset, v_offset, u_offset) < 0 || |
986 | FFMAX3(y_offset, v_offset, u_offset) >= ctx->data_size - 16 || |
987 | FFMIN3(y_offset, v_offset, u_offset) < gb.buffer - bs_hdr + 16 || |
988 | FFMIN3(ctx->y_data_size, ctx->v_data_size, ctx->u_data_size) <= 0) { |
989 | av_log(avctx, AV_LOG_ERROR, "One of the y/u/v offsets is invalid\n"); |
990 | return AVERROR_INVALIDDATA; |
991 | } |
992 | |
993 | ctx->y_data_ptr = bs_hdr + y_offset; |
994 | ctx->v_data_ptr = bs_hdr + v_offset; |
995 | ctx->u_data_ptr = bs_hdr + u_offset; |
996 | ctx->alt_quant = gb.buffer; |
997 | |
998 | if (ctx->data_size == 16) { |
999 | av_log(avctx, AV_LOG_DEBUG, "Sync frame encountered!\n"); |
1000 | return 16; |
1001 | } |
1002 | |
1003 | if (ctx->frame_flags & BS_8BIT_PEL) { |
1004 | avpriv_request_sample(avctx, "8-bit pixel format"); |
1005 | return AVERROR_PATCHWELCOME; |
1006 | } |
1007 | |
1008 | if (ctx->frame_flags & BS_MV_X_HALF || ctx->frame_flags & BS_MV_Y_HALF) { |
1009 | avpriv_request_sample(avctx, "Halfpel motion vectors"); |
1010 | return AVERROR_PATCHWELCOME; |
1011 | } |
1012 | |
1013 | return 0; |
1014 | } |
1015 | |
1016 | |
1017 | /** |
1018 | * Convert and output the current plane. |
1019 | * All pixel values will be upsampled by shifting right by one bit. |
1020 | * |
1021 | * @param[in] plane pointer to the descriptor of the plane being processed |
1022 | * @param[in] buf_sel indicates which frame buffer the input data stored in |
1023 | * @param[out] dst pointer to the buffer receiving converted pixels |
1024 | * @param[in] dst_pitch pitch for moving to the next y line |
1025 | * @param[in] dst_height output plane height |
1026 | */ |
1027 | static void output_plane(const Plane *plane, int buf_sel, uint8_t *dst, |
1028 | ptrdiff_t dst_pitch, int dst_height) |
1029 | { |
1030 | int x,y; |
1031 | const uint8_t *src = plane->pixels[buf_sel]; |
1032 | ptrdiff_t pitch = plane->pitch; |
1033 | |
1034 | dst_height = FFMIN(dst_height, plane->height); |
1035 | for (y = 0; y < dst_height; y++) { |
1036 | /* convert four pixels at once using SWAR */ |
1037 | for (x = 0; x < plane->width >> 2; x++) { |
1038 | AV_WN32A(dst, (AV_RN32A(src) & 0x7F7F7F7F) << 1); |
1039 | src += 4; |
1040 | dst += 4; |
1041 | } |
1042 | |
1043 | for (x <<= 2; x < plane->width; x++) |
1044 | *dst++ = *src++ << 1; |
1045 | |
1046 | src += pitch - plane->width; |
1047 | dst += dst_pitch - plane->width; |
1048 | } |
1049 | } |
1050 | |
1051 | |
1052 | static av_cold int decode_init(AVCodecContext *avctx) |
1053 | { |
1054 | Indeo3DecodeContext *ctx = avctx->priv_data; |
1055 | |
1056 | ctx->avctx = avctx; |
1057 | avctx->pix_fmt = AV_PIX_FMT_YUV410P; |
1058 | |
1059 | build_requant_tab(); |
1060 | |
1061 | ff_hpeldsp_init(&ctx->hdsp, avctx->flags); |
1062 | |
1063 | return allocate_frame_buffers(ctx, avctx, avctx->width, avctx->height); |
1064 | } |
1065 | |
1066 | |
1067 | static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, |
1068 | AVPacket *avpkt) |
1069 | { |
1070 | Indeo3DecodeContext *ctx = avctx->priv_data; |
1071 | const uint8_t *buf = avpkt->data; |
1072 | int buf_size = avpkt->size; |
1073 | AVFrame *frame = data; |
1074 | int res; |
1075 | |
1076 | res = decode_frame_headers(ctx, avctx, buf, buf_size); |
1077 | if (res < 0) |
1078 | return res; |
1079 | |
1080 | /* skip sync(null) frames */ |
1081 | if (res) { |
1082 | // we have processed 16 bytes but no data was decoded |
1083 | *got_frame = 0; |
1084 | return buf_size; |
1085 | } |
1086 | |
1087 | /* skip droppable INTER frames if requested */ |
1088 | if (ctx->frame_flags & BS_NONREF && |
1089 | (avctx->skip_frame >= AVDISCARD_NONREF)) |
1090 | return 0; |
1091 | |
1092 | /* skip INTER frames if requested */ |
1093 | if (!(ctx->frame_flags & BS_KEYFRAME) && avctx->skip_frame >= AVDISCARD_NONKEY) |
1094 | return 0; |
1095 | |
1096 | /* use BS_BUFFER flag for buffer switching */ |
1097 | ctx->buf_sel = (ctx->frame_flags >> BS_BUFFER) & 1; |
1098 | |
1099 | if ((res = ff_get_buffer(avctx, frame, 0)) < 0) |
1100 | return res; |
1101 | |
1102 | /* decode luma plane */ |
1103 | if ((res = decode_plane(ctx, avctx, ctx->planes, ctx->y_data_ptr, ctx->y_data_size, 40))) |
1104 | return res; |
1105 | |
1106 | /* decode chroma planes */ |
1107 | if ((res = decode_plane(ctx, avctx, &ctx->planes[1], ctx->u_data_ptr, ctx->u_data_size, 10))) |
1108 | return res; |
1109 | |
1110 | if ((res = decode_plane(ctx, avctx, &ctx->planes[2], ctx->v_data_ptr, ctx->v_data_size, 10))) |
1111 | return res; |
1112 | |
1113 | output_plane(&ctx->planes[0], ctx->buf_sel, |
1114 | frame->data[0], frame->linesize[0], |
1115 | avctx->height); |
1116 | output_plane(&ctx->planes[1], ctx->buf_sel, |
1117 | frame->data[1], frame->linesize[1], |
1118 | (avctx->height + 3) >> 2); |
1119 | output_plane(&ctx->planes[2], ctx->buf_sel, |
1120 | frame->data[2], frame->linesize[2], |
1121 | (avctx->height + 3) >> 2); |
1122 | |
1123 | *got_frame = 1; |
1124 | |
1125 | return buf_size; |
1126 | } |
1127 | |
1128 | |
1129 | static av_cold int decode_close(AVCodecContext *avctx) |
1130 | { |
1131 | free_frame_buffers(avctx->priv_data); |
1132 | |
1133 | return 0; |
1134 | } |
1135 | |
1136 | AVCodec ff_indeo3_decoder = { |
1137 | .name = "indeo3", |
1138 | .long_name = NULL_IF_CONFIG_SMALL("Intel Indeo 3"), |
1139 | .type = AVMEDIA_TYPE_VIDEO, |
1140 | .id = AV_CODEC_ID_INDEO3, |
1141 | .priv_data_size = sizeof(Indeo3DecodeContext), |
1142 | .init = decode_init, |
1143 | .close = decode_close, |
1144 | .decode = decode_frame, |
1145 | .capabilities = AV_CODEC_CAP_DR1, |
1146 | }; |
1147 |