blob: 84b440f741de2a7c6f226e63e8f74028de5d63a1
1 | /* |
2 | * DirectDraw Surface image decoder |
3 | * Copyright (C) 2015 Vittorio Giovara <vittorio.giovara@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 | * DDS decoder |
25 | * |
26 | * https://msdn.microsoft.com/en-us/library/bb943982%28v=vs.85%29.aspx |
27 | */ |
28 | |
29 | #include <stdint.h> |
30 | |
31 | #include "libavutil/libm.h" |
32 | #include "libavutil/imgutils.h" |
33 | |
34 | #include "avcodec.h" |
35 | #include "bytestream.h" |
36 | #include "internal.h" |
37 | #include "texturedsp.h" |
38 | #include "thread.h" |
39 | |
40 | #define DDPF_FOURCC (1 << 2) |
41 | #define DDPF_PALETTE (1 << 5) |
42 | #define DDPF_NORMALMAP (1 << 31) |
43 | |
44 | enum DDSPostProc { |
45 | DDS_NONE = 0, |
46 | DDS_ALPHA_EXP, |
47 | DDS_NORMAL_MAP, |
48 | DDS_RAW_YCOCG, |
49 | DDS_SWAP_ALPHA, |
50 | DDS_SWIZZLE_A2XY, |
51 | DDS_SWIZZLE_RBXG, |
52 | DDS_SWIZZLE_RGXB, |
53 | DDS_SWIZZLE_RXBG, |
54 | DDS_SWIZZLE_RXGB, |
55 | DDS_SWIZZLE_XGBR, |
56 | DDS_SWIZZLE_XRBG, |
57 | DDS_SWIZZLE_XGXR, |
58 | }; |
59 | |
60 | enum DDSDXGIFormat { |
61 | DXGI_FORMAT_R16G16B16A16_TYPELESS = 9, |
62 | DXGI_FORMAT_R16G16B16A16_FLOAT = 10, |
63 | DXGI_FORMAT_R16G16B16A16_UNORM = 11, |
64 | DXGI_FORMAT_R16G16B16A16_UINT = 12, |
65 | DXGI_FORMAT_R16G16B16A16_SNORM = 13, |
66 | DXGI_FORMAT_R16G16B16A16_SINT = 14, |
67 | |
68 | DXGI_FORMAT_R8G8B8A8_TYPELESS = 27, |
69 | DXGI_FORMAT_R8G8B8A8_UNORM = 28, |
70 | DXGI_FORMAT_R8G8B8A8_UNORM_SRGB = 29, |
71 | DXGI_FORMAT_R8G8B8A8_UINT = 30, |
72 | DXGI_FORMAT_R8G8B8A8_SNORM = 31, |
73 | DXGI_FORMAT_R8G8B8A8_SINT = 32, |
74 | |
75 | DXGI_FORMAT_BC1_TYPELESS = 70, |
76 | DXGI_FORMAT_BC1_UNORM = 71, |
77 | DXGI_FORMAT_BC1_UNORM_SRGB = 72, |
78 | DXGI_FORMAT_BC2_TYPELESS = 73, |
79 | DXGI_FORMAT_BC2_UNORM = 74, |
80 | DXGI_FORMAT_BC2_UNORM_SRGB = 75, |
81 | DXGI_FORMAT_BC3_TYPELESS = 76, |
82 | DXGI_FORMAT_BC3_UNORM = 77, |
83 | DXGI_FORMAT_BC3_UNORM_SRGB = 78, |
84 | DXGI_FORMAT_BC4_TYPELESS = 79, |
85 | DXGI_FORMAT_BC4_UNORM = 80, |
86 | DXGI_FORMAT_BC4_SNORM = 81, |
87 | DXGI_FORMAT_BC5_TYPELESS = 82, |
88 | DXGI_FORMAT_BC5_UNORM = 83, |
89 | DXGI_FORMAT_BC5_SNORM = 84, |
90 | DXGI_FORMAT_B5G6R5_UNORM = 85, |
91 | DXGI_FORMAT_B8G8R8A8_UNORM = 87, |
92 | DXGI_FORMAT_B8G8R8X8_UNORM = 88, |
93 | DXGI_FORMAT_B8G8R8A8_TYPELESS = 90, |
94 | DXGI_FORMAT_B8G8R8A8_UNORM_SRGB = 91, |
95 | DXGI_FORMAT_B8G8R8X8_TYPELESS = 92, |
96 | DXGI_FORMAT_B8G8R8X8_UNORM_SRGB = 93, |
97 | }; |
98 | |
99 | typedef struct DDSContext { |
100 | TextureDSPContext texdsp; |
101 | GetByteContext gbc; |
102 | |
103 | int compressed; |
104 | int paletted; |
105 | int bpp; |
106 | enum DDSPostProc postproc; |
107 | |
108 | const uint8_t *tex_data; // Compressed texture |
109 | int tex_ratio; // Compression ratio |
110 | int slice_count; // Number of slices for threaded operations |
111 | |
112 | /* Pointer to the selected compress or decompress function. */ |
113 | int (*tex_funct)(uint8_t *dst, ptrdiff_t stride, const uint8_t *block); |
114 | } DDSContext; |
115 | |
116 | static int parse_pixel_format(AVCodecContext *avctx) |
117 | { |
118 | DDSContext *ctx = avctx->priv_data; |
119 | GetByteContext *gbc = &ctx->gbc; |
120 | uint32_t flags, fourcc, gimp_tag; |
121 | enum DDSDXGIFormat dxgi; |
122 | int size, bpp, r, g, b, a; |
123 | int alpha_exponent, ycocg_classic, ycocg_scaled, normal_map, array; |
124 | |
125 | /* Alternative DDS implementations use reserved1 as custom header. */ |
126 | bytestream2_skip(gbc, 4 * 3); |
127 | gimp_tag = bytestream2_get_le32(gbc); |
128 | alpha_exponent = gimp_tag == MKTAG('A', 'E', 'X', 'P'); |
129 | ycocg_classic = gimp_tag == MKTAG('Y', 'C', 'G', '1'); |
130 | ycocg_scaled = gimp_tag == MKTAG('Y', 'C', 'G', '2'); |
131 | bytestream2_skip(gbc, 4 * 7); |
132 | |
133 | /* Now the real DDPF starts. */ |
134 | size = bytestream2_get_le32(gbc); |
135 | if (size != 32) { |
136 | av_log(avctx, AV_LOG_ERROR, "Invalid pixel format header %d.\n", size); |
137 | return AVERROR_INVALIDDATA; |
138 | } |
139 | flags = bytestream2_get_le32(gbc); |
140 | ctx->compressed = flags & DDPF_FOURCC; |
141 | ctx->paletted = flags & DDPF_PALETTE; |
142 | normal_map = flags & DDPF_NORMALMAP; |
143 | fourcc = bytestream2_get_le32(gbc); |
144 | |
145 | if (ctx->compressed && ctx->paletted) { |
146 | av_log(avctx, AV_LOG_WARNING, |
147 | "Disabling invalid palette flag for compressed dds.\n"); |
148 | ctx->paletted = 0; |
149 | } |
150 | |
151 | bpp = ctx->bpp = bytestream2_get_le32(gbc); // rgbbitcount |
152 | r = bytestream2_get_le32(gbc); // rbitmask |
153 | g = bytestream2_get_le32(gbc); // gbitmask |
154 | b = bytestream2_get_le32(gbc); // bbitmask |
155 | a = bytestream2_get_le32(gbc); // abitmask |
156 | |
157 | bytestream2_skip(gbc, 4); // caps |
158 | bytestream2_skip(gbc, 4); // caps2 |
159 | bytestream2_skip(gbc, 4); // caps3 |
160 | bytestream2_skip(gbc, 4); // caps4 |
161 | bytestream2_skip(gbc, 4); // reserved2 |
162 | |
163 | av_log(avctx, AV_LOG_VERBOSE, "fourcc %s bpp %d " |
164 | "r 0x%x g 0x%x b 0x%x a 0x%x\n", av_fourcc2str(fourcc), bpp, r, g, b, a); |
165 | if (gimp_tag) |
166 | av_log(avctx, AV_LOG_VERBOSE, "and GIMP-DDS tag %s\n", av_fourcc2str(gimp_tag)); |
167 | |
168 | if (ctx->compressed) |
169 | avctx->pix_fmt = AV_PIX_FMT_RGBA; |
170 | |
171 | if (ctx->compressed) { |
172 | switch (fourcc) { |
173 | case MKTAG('D', 'X', 'T', '1'): |
174 | ctx->tex_ratio = 8; |
175 | ctx->tex_funct = ctx->texdsp.dxt1a_block; |
176 | break; |
177 | case MKTAG('D', 'X', 'T', '2'): |
178 | ctx->tex_ratio = 16; |
179 | ctx->tex_funct = ctx->texdsp.dxt2_block; |
180 | break; |
181 | case MKTAG('D', 'X', 'T', '3'): |
182 | ctx->tex_ratio = 16; |
183 | ctx->tex_funct = ctx->texdsp.dxt3_block; |
184 | break; |
185 | case MKTAG('D', 'X', 'T', '4'): |
186 | ctx->tex_ratio = 16; |
187 | ctx->tex_funct = ctx->texdsp.dxt4_block; |
188 | break; |
189 | case MKTAG('D', 'X', 'T', '5'): |
190 | ctx->tex_ratio = 16; |
191 | if (ycocg_scaled) |
192 | ctx->tex_funct = ctx->texdsp.dxt5ys_block; |
193 | else if (ycocg_classic) |
194 | ctx->tex_funct = ctx->texdsp.dxt5y_block; |
195 | else |
196 | ctx->tex_funct = ctx->texdsp.dxt5_block; |
197 | break; |
198 | case MKTAG('R', 'X', 'G', 'B'): |
199 | ctx->tex_ratio = 16; |
200 | ctx->tex_funct = ctx->texdsp.dxt5_block; |
201 | /* This format may be considered as a normal map, |
202 | * but it is handled differently in a separate postproc. */ |
203 | ctx->postproc = DDS_SWIZZLE_RXGB; |
204 | normal_map = 0; |
205 | break; |
206 | case MKTAG('A', 'T', 'I', '1'): |
207 | case MKTAG('B', 'C', '4', 'U'): |
208 | ctx->tex_ratio = 8; |
209 | ctx->tex_funct = ctx->texdsp.rgtc1u_block; |
210 | break; |
211 | case MKTAG('B', 'C', '4', 'S'): |
212 | ctx->tex_ratio = 8; |
213 | ctx->tex_funct = ctx->texdsp.rgtc1s_block; |
214 | break; |
215 | case MKTAG('A', 'T', 'I', '2'): |
216 | /* RGT2 variant with swapped R and G (3Dc)*/ |
217 | ctx->tex_ratio = 16; |
218 | ctx->tex_funct = ctx->texdsp.dxn3dc_block; |
219 | break; |
220 | case MKTAG('B', 'C', '5', 'U'): |
221 | ctx->tex_ratio = 16; |
222 | ctx->tex_funct = ctx->texdsp.rgtc2u_block; |
223 | break; |
224 | case MKTAG('B', 'C', '5', 'S'): |
225 | ctx->tex_ratio = 16; |
226 | ctx->tex_funct = ctx->texdsp.rgtc2s_block; |
227 | break; |
228 | case MKTAG('U', 'Y', 'V', 'Y'): |
229 | ctx->compressed = 0; |
230 | avctx->pix_fmt = AV_PIX_FMT_UYVY422; |
231 | break; |
232 | case MKTAG('Y', 'U', 'Y', '2'): |
233 | ctx->compressed = 0; |
234 | avctx->pix_fmt = AV_PIX_FMT_YUYV422; |
235 | break; |
236 | case MKTAG('P', '8', ' ', ' '): |
237 | /* ATI Palette8, same as normal palette */ |
238 | ctx->compressed = 0; |
239 | ctx->paletted = 1; |
240 | avctx->pix_fmt = AV_PIX_FMT_PAL8; |
241 | break; |
242 | case MKTAG('G', '1', ' ', ' '): |
243 | ctx->compressed = 0; |
244 | avctx->pix_fmt = AV_PIX_FMT_MONOBLACK; |
245 | break; |
246 | case MKTAG('D', 'X', '1', '0'): |
247 | /* DirectX 10 extra header */ |
248 | dxgi = bytestream2_get_le32(gbc); |
249 | bytestream2_skip(gbc, 4); // resourceDimension |
250 | bytestream2_skip(gbc, 4); // miscFlag |
251 | array = bytestream2_get_le32(gbc); |
252 | bytestream2_skip(gbc, 4); // miscFlag2 |
253 | |
254 | if (array != 0) |
255 | av_log(avctx, AV_LOG_VERBOSE, |
256 | "Found array of size %d (ignored).\n", array); |
257 | |
258 | /* Only BC[1-5] are actually compressed. */ |
259 | ctx->compressed = (dxgi >= 70) && (dxgi <= 84); |
260 | |
261 | av_log(avctx, AV_LOG_VERBOSE, "DXGI format %d.\n", dxgi); |
262 | switch (dxgi) { |
263 | /* RGB types. */ |
264 | case DXGI_FORMAT_R16G16B16A16_TYPELESS: |
265 | case DXGI_FORMAT_R16G16B16A16_FLOAT: |
266 | case DXGI_FORMAT_R16G16B16A16_UNORM: |
267 | case DXGI_FORMAT_R16G16B16A16_UINT: |
268 | case DXGI_FORMAT_R16G16B16A16_SNORM: |
269 | case DXGI_FORMAT_R16G16B16A16_SINT: |
270 | avctx->pix_fmt = AV_PIX_FMT_BGRA64; |
271 | break; |
272 | case DXGI_FORMAT_R8G8B8A8_UNORM_SRGB: |
273 | avctx->colorspace = AVCOL_SPC_RGB; |
274 | case DXGI_FORMAT_R8G8B8A8_TYPELESS: |
275 | case DXGI_FORMAT_R8G8B8A8_UNORM: |
276 | case DXGI_FORMAT_R8G8B8A8_UINT: |
277 | case DXGI_FORMAT_R8G8B8A8_SNORM: |
278 | case DXGI_FORMAT_R8G8B8A8_SINT: |
279 | avctx->pix_fmt = AV_PIX_FMT_BGRA; |
280 | break; |
281 | case DXGI_FORMAT_B8G8R8A8_UNORM_SRGB: |
282 | avctx->colorspace = AVCOL_SPC_RGB; |
283 | case DXGI_FORMAT_B8G8R8A8_TYPELESS: |
284 | case DXGI_FORMAT_B8G8R8A8_UNORM: |
285 | avctx->pix_fmt = AV_PIX_FMT_RGBA; |
286 | break; |
287 | case DXGI_FORMAT_B8G8R8X8_UNORM_SRGB: |
288 | avctx->colorspace = AVCOL_SPC_RGB; |
289 | case DXGI_FORMAT_B8G8R8X8_TYPELESS: |
290 | case DXGI_FORMAT_B8G8R8X8_UNORM: |
291 | avctx->pix_fmt = AV_PIX_FMT_RGBA; // opaque |
292 | break; |
293 | case DXGI_FORMAT_B5G6R5_UNORM: |
294 | avctx->pix_fmt = AV_PIX_FMT_RGB565LE; |
295 | break; |
296 | /* Texture types. */ |
297 | case DXGI_FORMAT_BC1_UNORM_SRGB: |
298 | avctx->colorspace = AVCOL_SPC_RGB; |
299 | case DXGI_FORMAT_BC1_TYPELESS: |
300 | case DXGI_FORMAT_BC1_UNORM: |
301 | ctx->tex_ratio = 8; |
302 | ctx->tex_funct = ctx->texdsp.dxt1a_block; |
303 | break; |
304 | case DXGI_FORMAT_BC2_UNORM_SRGB: |
305 | avctx->colorspace = AVCOL_SPC_RGB; |
306 | case DXGI_FORMAT_BC2_TYPELESS: |
307 | case DXGI_FORMAT_BC2_UNORM: |
308 | ctx->tex_ratio = 16; |
309 | ctx->tex_funct = ctx->texdsp.dxt3_block; |
310 | break; |
311 | case DXGI_FORMAT_BC3_UNORM_SRGB: |
312 | avctx->colorspace = AVCOL_SPC_RGB; |
313 | case DXGI_FORMAT_BC3_TYPELESS: |
314 | case DXGI_FORMAT_BC3_UNORM: |
315 | ctx->tex_ratio = 16; |
316 | ctx->tex_funct = ctx->texdsp.dxt5_block; |
317 | break; |
318 | case DXGI_FORMAT_BC4_TYPELESS: |
319 | case DXGI_FORMAT_BC4_UNORM: |
320 | ctx->tex_ratio = 8; |
321 | ctx->tex_funct = ctx->texdsp.rgtc1u_block; |
322 | break; |
323 | case DXGI_FORMAT_BC4_SNORM: |
324 | ctx->tex_ratio = 8; |
325 | ctx->tex_funct = ctx->texdsp.rgtc1s_block; |
326 | break; |
327 | case DXGI_FORMAT_BC5_TYPELESS: |
328 | case DXGI_FORMAT_BC5_UNORM: |
329 | ctx->tex_ratio = 16; |
330 | ctx->tex_funct = ctx->texdsp.rgtc2u_block; |
331 | break; |
332 | case DXGI_FORMAT_BC5_SNORM: |
333 | ctx->tex_ratio = 16; |
334 | ctx->tex_funct = ctx->texdsp.rgtc2s_block; |
335 | break; |
336 | default: |
337 | av_log(avctx, AV_LOG_ERROR, |
338 | "Unsupported DXGI format %d.\n", dxgi); |
339 | return AVERROR_INVALIDDATA; |
340 | } |
341 | break; |
342 | default: |
343 | av_log(avctx, AV_LOG_ERROR, "Unsupported %s fourcc.\n", av_fourcc2str(fourcc)); |
344 | return AVERROR_INVALIDDATA; |
345 | } |
346 | } else if (ctx->paletted) { |
347 | if (bpp == 8) { |
348 | avctx->pix_fmt = AV_PIX_FMT_PAL8; |
349 | } else { |
350 | av_log(avctx, AV_LOG_ERROR, "Unsupported palette bpp %d.\n", bpp); |
351 | return AVERROR_INVALIDDATA; |
352 | } |
353 | } else { |
354 | /* 4 bpp */ |
355 | if (bpp == 4 && r == 0 && g == 0 && b == 0 && a == 0) |
356 | avctx->pix_fmt = AV_PIX_FMT_PAL8; |
357 | /* 8 bpp */ |
358 | else if (bpp == 8 && r == 0xff && g == 0 && b == 0 && a == 0) |
359 | avctx->pix_fmt = AV_PIX_FMT_GRAY8; |
360 | else if (bpp == 8 && r == 0 && g == 0 && b == 0 && a == 0xff) |
361 | avctx->pix_fmt = AV_PIX_FMT_GRAY8; |
362 | /* 16 bpp */ |
363 | else if (bpp == 16 && r == 0xff && g == 0 && b == 0 && a == 0xff00) |
364 | avctx->pix_fmt = AV_PIX_FMT_YA8; |
365 | else if (bpp == 16 && r == 0xff00 && g == 0 && b == 0 && a == 0xff) { |
366 | avctx->pix_fmt = AV_PIX_FMT_YA8; |
367 | ctx->postproc = DDS_SWAP_ALPHA; |
368 | } |
369 | else if (bpp == 16 && r == 0xffff && g == 0 && b == 0 && a == 0) |
370 | avctx->pix_fmt = AV_PIX_FMT_GRAY16LE; |
371 | else if (bpp == 16 && r == 0x7c00 && g == 0x3e0 && b == 0x1f && a == 0) |
372 | avctx->pix_fmt = AV_PIX_FMT_RGB555LE; |
373 | else if (bpp == 16 && r == 0x7c00 && g == 0x3e0 && b == 0x1f && a == 0x8000) |
374 | avctx->pix_fmt = AV_PIX_FMT_RGB555LE; // alpha ignored |
375 | else if (bpp == 16 && r == 0xf800 && g == 0x7e0 && b == 0x1f && a == 0) |
376 | avctx->pix_fmt = AV_PIX_FMT_RGB565LE; |
377 | /* 24 bpp */ |
378 | else if (bpp == 24 && r == 0xff0000 && g == 0xff00 && b == 0xff && a == 0) |
379 | avctx->pix_fmt = AV_PIX_FMT_BGR24; |
380 | /* 32 bpp */ |
381 | else if (bpp == 32 && r == 0xff0000 && g == 0xff00 && b == 0xff && a == 0) |
382 | avctx->pix_fmt = AV_PIX_FMT_BGR0; // opaque |
383 | else if (bpp == 32 && r == 0xff && g == 0xff00 && b == 0xff0000 && a == 0) |
384 | avctx->pix_fmt = AV_PIX_FMT_RGB0; // opaque |
385 | else if (bpp == 32 && r == 0xff0000 && g == 0xff00 && b == 0xff && a == 0xff000000) |
386 | avctx->pix_fmt = AV_PIX_FMT_BGRA; |
387 | else if (bpp == 32 && r == 0xff && g == 0xff00 && b == 0xff0000 && a == 0xff000000) |
388 | avctx->pix_fmt = AV_PIX_FMT_RGBA; |
389 | /* give up */ |
390 | else { |
391 | av_log(avctx, AV_LOG_ERROR, "Unknown pixel format " |
392 | "[bpp %d r 0x%x g 0x%x b 0x%x a 0x%x].\n", bpp, r, g, b, a); |
393 | return AVERROR_INVALIDDATA; |
394 | } |
395 | } |
396 | |
397 | /* Set any remaining post-proc that should happen before frame is ready. */ |
398 | if (alpha_exponent) |
399 | ctx->postproc = DDS_ALPHA_EXP; |
400 | else if (normal_map) |
401 | ctx->postproc = DDS_NORMAL_MAP; |
402 | else if (ycocg_classic && !ctx->compressed) |
403 | ctx->postproc = DDS_RAW_YCOCG; |
404 | |
405 | /* ATI/NVidia variants sometimes add swizzling in bpp. */ |
406 | switch (bpp) { |
407 | case MKTAG('A', '2', 'X', 'Y'): |
408 | ctx->postproc = DDS_SWIZZLE_A2XY; |
409 | break; |
410 | case MKTAG('x', 'G', 'B', 'R'): |
411 | ctx->postproc = DDS_SWIZZLE_XGBR; |
412 | break; |
413 | case MKTAG('x', 'R', 'B', 'G'): |
414 | ctx->postproc = DDS_SWIZZLE_XRBG; |
415 | break; |
416 | case MKTAG('R', 'B', 'x', 'G'): |
417 | ctx->postproc = DDS_SWIZZLE_RBXG; |
418 | break; |
419 | case MKTAG('R', 'G', 'x', 'B'): |
420 | ctx->postproc = DDS_SWIZZLE_RGXB; |
421 | break; |
422 | case MKTAG('R', 'x', 'B', 'G'): |
423 | ctx->postproc = DDS_SWIZZLE_RXBG; |
424 | break; |
425 | case MKTAG('x', 'G', 'x', 'R'): |
426 | ctx->postproc = DDS_SWIZZLE_XGXR; |
427 | break; |
428 | case MKTAG('A', '2', 'D', '5'): |
429 | ctx->postproc = DDS_NORMAL_MAP; |
430 | break; |
431 | } |
432 | |
433 | return 0; |
434 | } |
435 | |
436 | static int decompress_texture_thread(AVCodecContext *avctx, void *arg, |
437 | int slice, int thread_nb) |
438 | { |
439 | DDSContext *ctx = avctx->priv_data; |
440 | AVFrame *frame = arg; |
441 | const uint8_t *d = ctx->tex_data; |
442 | int w_block = avctx->coded_width / TEXTURE_BLOCK_W; |
443 | int h_block = avctx->coded_height / TEXTURE_BLOCK_H; |
444 | int x, y; |
445 | int start_slice, end_slice; |
446 | int base_blocks_per_slice = h_block / ctx->slice_count; |
447 | int remainder_blocks = h_block % ctx->slice_count; |
448 | |
449 | /* When the frame height (in blocks) doesn't divide evenly between the |
450 | * number of slices, spread the remaining blocks evenly between the first |
451 | * operations */ |
452 | start_slice = slice * base_blocks_per_slice; |
453 | /* Add any extra blocks (one per slice) that have been added before this slice */ |
454 | start_slice += FFMIN(slice, remainder_blocks); |
455 | |
456 | end_slice = start_slice + base_blocks_per_slice; |
457 | /* Add an extra block if there are still remainder blocks to be accounted for */ |
458 | if (slice < remainder_blocks) |
459 | end_slice++; |
460 | |
461 | for (y = start_slice; y < end_slice; y++) { |
462 | uint8_t *p = frame->data[0] + y * frame->linesize[0] * TEXTURE_BLOCK_H; |
463 | int off = y * w_block; |
464 | for (x = 0; x < w_block; x++) { |
465 | ctx->tex_funct(p + x * 16, frame->linesize[0], |
466 | d + (off + x) * ctx->tex_ratio); |
467 | } |
468 | } |
469 | |
470 | return 0; |
471 | } |
472 | |
473 | static void do_swizzle(AVFrame *frame, int x, int y) |
474 | { |
475 | int i; |
476 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { |
477 | uint8_t *src = frame->data[0] + i; |
478 | FFSWAP(uint8_t, src[x], src[y]); |
479 | } |
480 | } |
481 | |
482 | static void run_postproc(AVCodecContext *avctx, AVFrame *frame) |
483 | { |
484 | DDSContext *ctx = avctx->priv_data; |
485 | int i, x_off; |
486 | |
487 | switch (ctx->postproc) { |
488 | case DDS_ALPHA_EXP: |
489 | /* Alpha-exponential mode divides each channel by the maximum |
490 | * R, G or B value, and stores the multiplying factor in the |
491 | * alpha channel. */ |
492 | av_log(avctx, AV_LOG_DEBUG, "Post-processing alpha exponent.\n"); |
493 | |
494 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { |
495 | uint8_t *src = frame->data[0] + i; |
496 | int r = src[0]; |
497 | int g = src[1]; |
498 | int b = src[2]; |
499 | int a = src[3]; |
500 | |
501 | src[0] = r * a / 255; |
502 | src[1] = g * a / 255; |
503 | src[2] = b * a / 255; |
504 | src[3] = 255; |
505 | } |
506 | break; |
507 | case DDS_NORMAL_MAP: |
508 | /* Normal maps work in the XYZ color space and they encode |
509 | * X in R or in A, depending on the texture type, Y in G and |
510 | * derive Z with a square root of the distance. |
511 | * |
512 | * http://www.realtimecollisiondetection.net/blog/?p=28 */ |
513 | av_log(avctx, AV_LOG_DEBUG, "Post-processing normal map.\n"); |
514 | |
515 | x_off = ctx->tex_ratio == 8 ? 0 : 3; |
516 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { |
517 | uint8_t *src = frame->data[0] + i; |
518 | int x = src[x_off]; |
519 | int y = src[1]; |
520 | int z = 127; |
521 | |
522 | int d = (255 * 255 - x * x - y * y) / 2; |
523 | if (d > 0) |
524 | z = lrint(sqrtf(d)); |
525 | |
526 | src[0] = x; |
527 | src[1] = y; |
528 | src[2] = z; |
529 | src[3] = 255; |
530 | } |
531 | break; |
532 | case DDS_RAW_YCOCG: |
533 | /* Data is Y-Co-Cg-A and not RGBA, but they are represented |
534 | * with the same masks in the DDPF header. */ |
535 | av_log(avctx, AV_LOG_DEBUG, "Post-processing raw YCoCg.\n"); |
536 | |
537 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { |
538 | uint8_t *src = frame->data[0] + i; |
539 | int a = src[0]; |
540 | int cg = src[1] - 128; |
541 | int co = src[2] - 128; |
542 | int y = src[3]; |
543 | |
544 | src[0] = av_clip_uint8(y + co - cg); |
545 | src[1] = av_clip_uint8(y + cg); |
546 | src[2] = av_clip_uint8(y - co - cg); |
547 | src[3] = a; |
548 | } |
549 | break; |
550 | case DDS_SWAP_ALPHA: |
551 | /* Alpha and Luma are stored swapped. */ |
552 | av_log(avctx, AV_LOG_DEBUG, "Post-processing swapped Luma/Alpha.\n"); |
553 | |
554 | for (i = 0; i < frame->linesize[0] * frame->height; i += 2) { |
555 | uint8_t *src = frame->data[0] + i; |
556 | FFSWAP(uint8_t, src[0], src[1]); |
557 | } |
558 | break; |
559 | case DDS_SWIZZLE_A2XY: |
560 | /* Swap R and G, often used to restore a standard RGTC2. */ |
561 | av_log(avctx, AV_LOG_DEBUG, "Post-processing A2XY swizzle.\n"); |
562 | do_swizzle(frame, 0, 1); |
563 | break; |
564 | case DDS_SWIZZLE_RBXG: |
565 | /* Swap G and A, then B and new A (G). */ |
566 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RBXG swizzle.\n"); |
567 | do_swizzle(frame, 1, 3); |
568 | do_swizzle(frame, 2, 3); |
569 | break; |
570 | case DDS_SWIZZLE_RGXB: |
571 | /* Swap B and A. */ |
572 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RGXB swizzle.\n"); |
573 | do_swizzle(frame, 2, 3); |
574 | break; |
575 | case DDS_SWIZZLE_RXBG: |
576 | /* Swap G and A. */ |
577 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RXBG swizzle.\n"); |
578 | do_swizzle(frame, 1, 3); |
579 | break; |
580 | case DDS_SWIZZLE_RXGB: |
581 | /* Swap R and A (misleading name). */ |
582 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RXGB swizzle.\n"); |
583 | do_swizzle(frame, 0, 3); |
584 | break; |
585 | case DDS_SWIZZLE_XGBR: |
586 | /* Swap B and A, then R and new A (B). */ |
587 | av_log(avctx, AV_LOG_DEBUG, "Post-processing XGBR swizzle.\n"); |
588 | do_swizzle(frame, 2, 3); |
589 | do_swizzle(frame, 0, 3); |
590 | break; |
591 | case DDS_SWIZZLE_XGXR: |
592 | /* Swap G and A, then R and new A (G), then new R (G) and new G (A). |
593 | * This variant does not store any B component. */ |
594 | av_log(avctx, AV_LOG_DEBUG, "Post-processing XGXR swizzle.\n"); |
595 | do_swizzle(frame, 1, 3); |
596 | do_swizzle(frame, 0, 3); |
597 | do_swizzle(frame, 0, 1); |
598 | break; |
599 | case DDS_SWIZZLE_XRBG: |
600 | /* Swap G and A, then R and new A (G). */ |
601 | av_log(avctx, AV_LOG_DEBUG, "Post-processing XRBG swizzle.\n"); |
602 | do_swizzle(frame, 1, 3); |
603 | do_swizzle(frame, 0, 3); |
604 | break; |
605 | } |
606 | } |
607 | |
608 | static int dds_decode(AVCodecContext *avctx, void *data, |
609 | int *got_frame, AVPacket *avpkt) |
610 | { |
611 | DDSContext *ctx = avctx->priv_data; |
612 | GetByteContext *gbc = &ctx->gbc; |
613 | AVFrame *frame = data; |
614 | int mipmap; |
615 | int ret; |
616 | |
617 | ff_texturedsp_init(&ctx->texdsp); |
618 | bytestream2_init(gbc, avpkt->data, avpkt->size); |
619 | |
620 | if (bytestream2_get_bytes_left(gbc) < 128) { |
621 | av_log(avctx, AV_LOG_ERROR, "Frame is too small (%d).\n", |
622 | bytestream2_get_bytes_left(gbc)); |
623 | return AVERROR_INVALIDDATA; |
624 | } |
625 | |
626 | if (bytestream2_get_le32(gbc) != MKTAG('D', 'D', 'S', ' ') || |
627 | bytestream2_get_le32(gbc) != 124) { // header size |
628 | av_log(avctx, AV_LOG_ERROR, "Invalid DDS header.\n"); |
629 | return AVERROR_INVALIDDATA; |
630 | } |
631 | |
632 | bytestream2_skip(gbc, 4); // flags |
633 | |
634 | avctx->height = bytestream2_get_le32(gbc); |
635 | avctx->width = bytestream2_get_le32(gbc); |
636 | ret = av_image_check_size(avctx->width, avctx->height, 0, avctx); |
637 | if (ret < 0) { |
638 | av_log(avctx, AV_LOG_ERROR, "Invalid image size %dx%d.\n", |
639 | avctx->width, avctx->height); |
640 | return ret; |
641 | } |
642 | |
643 | /* Since codec is based on 4x4 blocks, size is aligned to 4. */ |
644 | avctx->coded_width = FFALIGN(avctx->width, TEXTURE_BLOCK_W); |
645 | avctx->coded_height = FFALIGN(avctx->height, TEXTURE_BLOCK_H); |
646 | |
647 | bytestream2_skip(gbc, 4); // pitch |
648 | bytestream2_skip(gbc, 4); // depth |
649 | mipmap = bytestream2_get_le32(gbc); |
650 | if (mipmap != 0) |
651 | av_log(avctx, AV_LOG_VERBOSE, "Found %d mipmaps (ignored).\n", mipmap); |
652 | |
653 | /* Extract pixel format information, considering additional elements |
654 | * in reserved1 and reserved2. */ |
655 | ret = parse_pixel_format(avctx); |
656 | if (ret < 0) |
657 | return ret; |
658 | |
659 | ret = ff_get_buffer(avctx, frame, 0); |
660 | if (ret < 0) |
661 | return ret; |
662 | |
663 | if (ctx->compressed) { |
664 | int size = (avctx->coded_height / TEXTURE_BLOCK_H) * |
665 | (avctx->coded_width / TEXTURE_BLOCK_W) * ctx->tex_ratio; |
666 | ctx->slice_count = av_clip(avctx->thread_count, 1, |
667 | avctx->coded_height / TEXTURE_BLOCK_H); |
668 | |
669 | if (bytestream2_get_bytes_left(gbc) < size) { |
670 | av_log(avctx, AV_LOG_ERROR, |
671 | "Compressed Buffer is too small (%d < %d).\n", |
672 | bytestream2_get_bytes_left(gbc), size); |
673 | return AVERROR_INVALIDDATA; |
674 | } |
675 | |
676 | /* Use the decompress function on the texture, one block per thread. */ |
677 | ctx->tex_data = gbc->buffer; |
678 | avctx->execute2(avctx, decompress_texture_thread, frame, NULL, ctx->slice_count); |
679 | } else if (!ctx->paletted && ctx->bpp == 4 && avctx->pix_fmt == AV_PIX_FMT_PAL8) { |
680 | uint8_t *dst = frame->data[0]; |
681 | int x, y, i; |
682 | |
683 | /* Use the first 64 bytes as palette, then copy the rest. */ |
684 | bytestream2_get_buffer(gbc, frame->data[1], 16 * 4); |
685 | for (i = 0; i < 16; i++) { |
686 | AV_WN32(frame->data[1] + i*4, |
687 | (frame->data[1][2+i*4]<<0)+ |
688 | (frame->data[1][1+i*4]<<8)+ |
689 | (frame->data[1][0+i*4]<<16)+ |
690 | (frame->data[1][3+i*4]<<24) |
691 | ); |
692 | } |
693 | frame->palette_has_changed = 1; |
694 | |
695 | if (bytestream2_get_bytes_left(gbc) < frame->height * frame->width / 2) { |
696 | av_log(avctx, AV_LOG_ERROR, "Buffer is too small (%d < %d).\n", |
697 | bytestream2_get_bytes_left(gbc), frame->height * frame->width / 2); |
698 | return AVERROR_INVALIDDATA; |
699 | } |
700 | |
701 | for (y = 0; y < frame->height; y++) { |
702 | for (x = 0; x < frame->width; x += 2) { |
703 | uint8_t val = bytestream2_get_byte(gbc); |
704 | dst[x ] = val & 0xF; |
705 | dst[x + 1] = val >> 4; |
706 | } |
707 | dst += frame->linesize[0]; |
708 | } |
709 | } else { |
710 | int linesize = av_image_get_linesize(avctx->pix_fmt, frame->width, 0); |
711 | |
712 | if (ctx->paletted) { |
713 | int i; |
714 | /* Use the first 1024 bytes as palette, then copy the rest. */ |
715 | bytestream2_get_buffer(gbc, frame->data[1], 256 * 4); |
716 | for (i = 0; i < 256; i++) |
717 | AV_WN32(frame->data[1] + i*4, |
718 | (frame->data[1][2+i*4]<<0)+ |
719 | (frame->data[1][1+i*4]<<8)+ |
720 | (frame->data[1][0+i*4]<<16)+ |
721 | (frame->data[1][3+i*4]<<24) |
722 | ); |
723 | |
724 | frame->palette_has_changed = 1; |
725 | } |
726 | |
727 | if (bytestream2_get_bytes_left(gbc) < frame->height * linesize) { |
728 | av_log(avctx, AV_LOG_ERROR, "Buffer is too small (%d < %d).\n", |
729 | bytestream2_get_bytes_left(gbc), frame->height * linesize); |
730 | return AVERROR_INVALIDDATA; |
731 | } |
732 | |
733 | av_image_copy_plane(frame->data[0], frame->linesize[0], |
734 | gbc->buffer, linesize, |
735 | linesize, frame->height); |
736 | } |
737 | |
738 | /* Run any post processing here if needed. */ |
739 | if (ctx->postproc != DDS_NONE) |
740 | run_postproc(avctx, frame); |
741 | |
742 | /* Frame is ready to be output. */ |
743 | frame->pict_type = AV_PICTURE_TYPE_I; |
744 | frame->key_frame = 1; |
745 | *got_frame = 1; |
746 | |
747 | return avpkt->size; |
748 | } |
749 | |
750 | AVCodec ff_dds_decoder = { |
751 | .name = "dds", |
752 | .long_name = NULL_IF_CONFIG_SMALL("DirectDraw Surface image decoder"), |
753 | .type = AVMEDIA_TYPE_VIDEO, |
754 | .id = AV_CODEC_ID_DDS, |
755 | .decode = dds_decode, |
756 | .priv_data_size = sizeof(DDSContext), |
757 | .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_SLICE_THREADS, |
758 | .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
759 | }; |
760 |