Commit | Line | Data |
---|---|---|
5c018ee1 VG |
1 | /* |
2 | * DirectDraw Surface image decoder | |
3 | * Copyright (C) 2015 Vittorio Giovara <vittorio.giovara@gmail.com> | |
4 | * | |
5 | * This file is part of Libav. | |
6 | * | |
7 | * Libav 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 | * Libav 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 Libav; 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/imgutils.h" | |
32 | ||
33 | #include "avcodec.h" | |
34 | #include "bytestream.h" | |
35 | #include "internal.h" | |
36 | #include "texturedsp.h" | |
37 | #include "thread.h" | |
38 | ||
39 | #define DDPF_FOURCC (1 << 2) | |
40 | #define DDPF_PALETTE (1 << 5) | |
41 | #define DDPF_NORMALMAP (1 << 31) | |
42 | ||
43 | enum DDSPostProc { | |
44 | DDS_NONE = 0, | |
45 | DDS_ALPHA_EXP, | |
46 | DDS_NORMAL_MAP, | |
47 | DDS_RAW_YCOCG, | |
48 | DDS_SWAP_ALPHA, | |
49 | DDS_SWIZZLE_A2XY, | |
50 | DDS_SWIZZLE_RBXG, | |
51 | DDS_SWIZZLE_RGXB, | |
52 | DDS_SWIZZLE_RXBG, | |
53 | DDS_SWIZZLE_RXGB, | |
54 | DDS_SWIZZLE_XGBR, | |
55 | DDS_SWIZZLE_XRBG, | |
56 | DDS_SWIZZLE_XGXR, | |
57 | } DDSPostProc; | |
58 | ||
59 | enum DDSDXGIFormat { | |
60 | DXGI_FORMAT_R16G16B16A16_TYPELESS = 9, | |
61 | DXGI_FORMAT_R16G16B16A16_FLOAT = 10, | |
62 | DXGI_FORMAT_R16G16B16A16_UNORM = 11, | |
63 | DXGI_FORMAT_R16G16B16A16_UINT = 12, | |
64 | DXGI_FORMAT_R16G16B16A16_SNORM = 13, | |
65 | DXGI_FORMAT_R16G16B16A16_SINT = 14, | |
66 | ||
67 | DXGI_FORMAT_R8G8B8A8_TYPELESS = 27, | |
68 | DXGI_FORMAT_R8G8B8A8_UNORM = 28, | |
69 | DXGI_FORMAT_R8G8B8A8_UNORM_SRGB = 29, | |
70 | DXGI_FORMAT_R8G8B8A8_UINT = 30, | |
71 | DXGI_FORMAT_R8G8B8A8_SNORM = 31, | |
72 | DXGI_FORMAT_R8G8B8A8_SINT = 32, | |
73 | ||
74 | DXGI_FORMAT_BC1_TYPELESS = 70, | |
75 | DXGI_FORMAT_BC1_UNORM = 71, | |
76 | DXGI_FORMAT_BC1_UNORM_SRGB = 72, | |
77 | DXGI_FORMAT_BC2_TYPELESS = 73, | |
78 | DXGI_FORMAT_BC2_UNORM = 74, | |
79 | DXGI_FORMAT_BC2_UNORM_SRGB = 75, | |
80 | DXGI_FORMAT_BC3_TYPELESS = 76, | |
81 | DXGI_FORMAT_BC3_UNORM = 77, | |
82 | DXGI_FORMAT_BC3_UNORM_SRGB = 78, | |
83 | DXGI_FORMAT_BC4_TYPELESS = 79, | |
84 | DXGI_FORMAT_BC4_UNORM = 80, | |
85 | DXGI_FORMAT_BC4_SNORM = 81, | |
86 | DXGI_FORMAT_BC5_TYPELESS = 82, | |
87 | DXGI_FORMAT_BC5_UNORM = 83, | |
88 | DXGI_FORMAT_BC5_SNORM = 84, | |
89 | DXGI_FORMAT_B5G6R5_UNORM = 85, | |
90 | DXGI_FORMAT_B8G8R8A8_UNORM = 87, | |
91 | DXGI_FORMAT_B8G8R8X8_UNORM = 88, | |
92 | DXGI_FORMAT_B8G8R8A8_TYPELESS = 90, | |
93 | DXGI_FORMAT_B8G8R8A8_UNORM_SRGB = 91, | |
94 | DXGI_FORMAT_B8G8R8X8_TYPELESS = 92, | |
95 | DXGI_FORMAT_B8G8R8X8_UNORM_SRGB = 93, | |
96 | } DDSDXGIFormat; | |
97 | ||
98 | typedef struct DDSContext { | |
99 | TextureDSPContext texdsp; | |
100 | GetByteContext gbc; | |
101 | ||
102 | int compressed; | |
103 | int paletted; | |
104 | enum DDSPostProc postproc; | |
105 | ||
106 | const uint8_t *tex_data; // Compressed texture | |
107 | int tex_ratio; // Compression ratio | |
108 | ||
109 | /* Pointer to the selected compress or decompress function. */ | |
110 | int (*tex_funct)(uint8_t *dst, ptrdiff_t stride, const uint8_t *block); | |
111 | } DDSContext; | |
112 | ||
113 | static int parse_pixel_format(AVCodecContext *avctx) | |
114 | { | |
115 | DDSContext *ctx = avctx->priv_data; | |
116 | GetByteContext *gbc = &ctx->gbc; | |
117 | char buf[32]; | |
118 | uint32_t flags, fourcc, gimp_tag; | |
119 | enum DDSDXGIFormat dxgi; | |
120 | int size, bpp, r, g, b, a; | |
121 | int alpha_exponent, ycocg_classic, ycocg_scaled, normal_map, array; | |
122 | ||
123 | /* Alternative DDS implementations use reserved1 as custom header. */ | |
124 | bytestream2_skip(gbc, 4 * 3); | |
125 | gimp_tag = bytestream2_get_le32(gbc); | |
126 | alpha_exponent = gimp_tag == MKTAG('A', 'E', 'X', 'P'); | |
127 | ycocg_classic = gimp_tag == MKTAG('Y', 'C', 'G', '1'); | |
128 | ycocg_scaled = gimp_tag == MKTAG('Y', 'C', 'G', '2'); | |
129 | bytestream2_skip(gbc, 4 * 7); | |
130 | ||
131 | /* Now the real DDPF starts. */ | |
132 | size = bytestream2_get_le32(gbc); | |
133 | if (size != 32) { | |
134 | av_log(avctx, AV_LOG_ERROR, "Invalid pixel format header %d.\n", size); | |
135 | return AVERROR_INVALIDDATA; | |
136 | } | |
137 | flags = bytestream2_get_le32(gbc); | |
138 | ctx->compressed = flags & DDPF_FOURCC; | |
139 | ctx->paletted = flags & DDPF_PALETTE; | |
140 | normal_map = flags & DDPF_NORMALMAP; | |
141 | fourcc = bytestream2_get_le32(gbc); | |
142 | ||
143 | bpp = bytestream2_get_le32(gbc); // rgbbitcount | |
144 | r = bytestream2_get_le32(gbc); // rbitmask | |
145 | g = bytestream2_get_le32(gbc); // gbitmask | |
146 | b = bytestream2_get_le32(gbc); // bbitmask | |
147 | a = bytestream2_get_le32(gbc); // abitmask | |
148 | ||
149 | bytestream2_skip(gbc, 4); // caps | |
150 | bytestream2_skip(gbc, 4); // caps2 | |
151 | bytestream2_skip(gbc, 4); // caps3 | |
152 | bytestream2_skip(gbc, 4); // caps4 | |
153 | bytestream2_skip(gbc, 4); // reserved2 | |
154 | ||
155 | av_get_codec_tag_string(buf, sizeof(buf), fourcc); | |
156 | av_log(avctx, AV_LOG_VERBOSE, "fourcc %s bpp %d " | |
157 | "r 0x%x g 0x%x b 0x%x a 0x%x\n", buf, bpp, r, g, b, a); | |
158 | if (gimp_tag) { | |
159 | av_get_codec_tag_string(buf, sizeof(buf), gimp_tag); | |
160 | av_log(avctx, AV_LOG_VERBOSE, "and GIMP-DDS tag %s\n", buf); | |
161 | } | |
162 | ||
163 | if (ctx->compressed) | |
164 | avctx->pix_fmt = AV_PIX_FMT_RGBA; | |
165 | ||
166 | if (ctx->compressed) { | |
167 | switch (fourcc) { | |
168 | case MKTAG('D', 'X', 'T', '1'): | |
169 | ctx->tex_ratio = 8; | |
170 | ctx->tex_funct = ctx->texdsp.dxt1a_block; | |
171 | break; | |
172 | case MKTAG('D', 'X', 'T', '2'): | |
173 | ctx->tex_ratio = 16; | |
174 | ctx->tex_funct = ctx->texdsp.dxt2_block; | |
175 | break; | |
176 | case MKTAG('D', 'X', 'T', '3'): | |
177 | ctx->tex_ratio = 16; | |
178 | ctx->tex_funct = ctx->texdsp.dxt3_block; | |
179 | break; | |
180 | case MKTAG('D', 'X', 'T', '4'): | |
181 | ctx->tex_ratio = 16; | |
182 | ctx->tex_funct = ctx->texdsp.dxt4_block; | |
183 | break; | |
184 | case MKTAG('D', 'X', 'T', '5'): | |
185 | ctx->tex_ratio = 16; | |
186 | if (ycocg_scaled) | |
187 | ctx->tex_funct = ctx->texdsp.dxt5ys_block; | |
188 | else if (ycocg_classic) | |
189 | ctx->tex_funct = ctx->texdsp.dxt5y_block; | |
190 | else | |
191 | ctx->tex_funct = ctx->texdsp.dxt5_block; | |
192 | break; | |
193 | case MKTAG('R', 'X', 'G', 'B'): | |
194 | ctx->tex_ratio = 16; | |
195 | ctx->tex_funct = ctx->texdsp.dxt5_block; | |
196 | /* This format may be considered as a normal map, | |
197 | * but it is handled differently in a separate postproc. */ | |
198 | ctx->postproc = DDS_SWIZZLE_RXGB; | |
199 | normal_map = 0; | |
200 | break; | |
201 | case MKTAG('A', 'T', 'I', '1'): | |
202 | case MKTAG('B', 'C', '4', 'U'): | |
203 | ctx->tex_ratio = 8; | |
204 | ctx->tex_funct = ctx->texdsp.rgtc1u_block; | |
205 | break; | |
206 | case MKTAG('B', 'C', '4', 'S'): | |
207 | ctx->tex_ratio = 8; | |
208 | ctx->tex_funct = ctx->texdsp.rgtc1s_block; | |
209 | break; | |
210 | case MKTAG('A', 'T', 'I', '2'): | |
211 | /* RGT2 variant with swapped R and G (3Dc)*/ | |
212 | ctx->tex_ratio = 16; | |
213 | ctx->tex_funct = ctx->texdsp.dxn3dc_block; | |
214 | break; | |
215 | case MKTAG('B', 'C', '5', 'U'): | |
216 | ctx->tex_ratio = 16; | |
217 | ctx->tex_funct = ctx->texdsp.rgtc2u_block; | |
218 | break; | |
219 | case MKTAG('B', 'C', '5', 'S'): | |
220 | ctx->tex_ratio = 16; | |
221 | ctx->tex_funct = ctx->texdsp.rgtc2s_block; | |
222 | break; | |
223 | case MKTAG('U', 'Y', 'V', 'Y'): | |
224 | ctx->compressed = 0; | |
225 | avctx->pix_fmt = AV_PIX_FMT_UYVY422; | |
226 | break; | |
227 | case MKTAG('Y', 'U', 'Y', '2'): | |
228 | ctx->compressed = 0; | |
229 | avctx->pix_fmt = AV_PIX_FMT_YUYV422; | |
230 | break; | |
231 | case MKTAG('P', '8', ' ', ' '): | |
232 | /* ATI Palette8, same as normal palette */ | |
233 | ctx->compressed = 0; | |
234 | ctx->paletted = 1; | |
235 | avctx->pix_fmt = AV_PIX_FMT_PAL8; | |
236 | break; | |
237 | case MKTAG('D', 'X', '1', '0'): | |
238 | /* DirectX 10 extra header */ | |
239 | dxgi = bytestream2_get_le32(gbc); | |
240 | bytestream2_skip(gbc, 4); // resourceDimension | |
241 | bytestream2_skip(gbc, 4); // miscFlag | |
242 | array = bytestream2_get_le32(gbc); | |
243 | bytestream2_skip(gbc, 4); // miscFlag2 | |
244 | ||
245 | if (array != 0) | |
246 | av_log(avctx, AV_LOG_VERBOSE, | |
247 | "Found array of size %d (ignored).\n", array); | |
248 | ||
249 | /* Only BC[1-5] are actually compressed. */ | |
250 | ctx->compressed = (dxgi >= 70) && (dxgi <= 84); | |
251 | ||
252 | av_log(avctx, AV_LOG_VERBOSE, "DXGI format %d.\n", dxgi); | |
253 | switch (dxgi) { | |
254 | /* RGB types. */ | |
255 | case DXGI_FORMAT_R16G16B16A16_TYPELESS: | |
256 | case DXGI_FORMAT_R16G16B16A16_FLOAT: | |
257 | case DXGI_FORMAT_R16G16B16A16_UNORM: | |
258 | case DXGI_FORMAT_R16G16B16A16_UINT: | |
259 | case DXGI_FORMAT_R16G16B16A16_SNORM: | |
260 | case DXGI_FORMAT_R16G16B16A16_SINT: | |
261 | avctx->pix_fmt = AV_PIX_FMT_BGRA64; | |
262 | break; | |
263 | case DXGI_FORMAT_R8G8B8A8_UNORM_SRGB: | |
264 | avctx->colorspace = AVCOL_SPC_RGB; | |
265 | case DXGI_FORMAT_R8G8B8A8_TYPELESS: | |
266 | case DXGI_FORMAT_R8G8B8A8_UNORM: | |
267 | case DXGI_FORMAT_R8G8B8A8_UINT: | |
268 | case DXGI_FORMAT_R8G8B8A8_SNORM: | |
269 | case DXGI_FORMAT_R8G8B8A8_SINT: | |
270 | avctx->pix_fmt = AV_PIX_FMT_BGRA; | |
271 | break; | |
272 | case DXGI_FORMAT_B8G8R8A8_UNORM_SRGB: | |
273 | avctx->colorspace = AVCOL_SPC_RGB; | |
274 | case DXGI_FORMAT_B8G8R8A8_TYPELESS: | |
275 | case DXGI_FORMAT_B8G8R8A8_UNORM: | |
276 | avctx->pix_fmt = AV_PIX_FMT_RGBA; | |
277 | break; | |
278 | case DXGI_FORMAT_B8G8R8X8_UNORM_SRGB: | |
279 | avctx->colorspace = AVCOL_SPC_RGB; | |
280 | case DXGI_FORMAT_B8G8R8X8_TYPELESS: | |
281 | case DXGI_FORMAT_B8G8R8X8_UNORM: | |
282 | avctx->pix_fmt = AV_PIX_FMT_RGBA; // opaque | |
283 | break; | |
284 | case DXGI_FORMAT_B5G6R5_UNORM: | |
285 | avctx->pix_fmt = AV_PIX_FMT_RGB565LE; | |
286 | break; | |
287 | /* Texture types. */ | |
288 | case DXGI_FORMAT_BC1_UNORM_SRGB: | |
289 | avctx->colorspace = AVCOL_SPC_RGB; | |
290 | case DXGI_FORMAT_BC1_TYPELESS: | |
291 | case DXGI_FORMAT_BC1_UNORM: | |
292 | ctx->tex_ratio = 8; | |
293 | ctx->tex_funct = ctx->texdsp.dxt1a_block; | |
294 | break; | |
295 | case DXGI_FORMAT_BC2_UNORM_SRGB: | |
296 | avctx->colorspace = AVCOL_SPC_RGB; | |
297 | case DXGI_FORMAT_BC2_TYPELESS: | |
298 | case DXGI_FORMAT_BC2_UNORM: | |
299 | ctx->tex_ratio = 16; | |
300 | ctx->tex_funct = ctx->texdsp.dxt3_block; | |
301 | break; | |
302 | case DXGI_FORMAT_BC3_UNORM_SRGB: | |
303 | avctx->colorspace = AVCOL_SPC_RGB; | |
304 | case DXGI_FORMAT_BC3_TYPELESS: | |
305 | case DXGI_FORMAT_BC3_UNORM: | |
306 | ctx->tex_ratio = 16; | |
307 | ctx->tex_funct = ctx->texdsp.dxt5_block; | |
308 | break; | |
309 | case DXGI_FORMAT_BC4_TYPELESS: | |
310 | case DXGI_FORMAT_BC4_UNORM: | |
311 | ctx->tex_ratio = 8; | |
312 | ctx->tex_funct = ctx->texdsp.rgtc1u_block; | |
313 | break; | |
314 | case DXGI_FORMAT_BC4_SNORM: | |
315 | ctx->tex_ratio = 8; | |
316 | ctx->tex_funct = ctx->texdsp.rgtc1s_block; | |
317 | break; | |
318 | case DXGI_FORMAT_BC5_TYPELESS: | |
319 | case DXGI_FORMAT_BC5_UNORM: | |
320 | ctx->tex_ratio = 16; | |
321 | ctx->tex_funct = ctx->texdsp.rgtc2u_block; | |
322 | break; | |
323 | case DXGI_FORMAT_BC5_SNORM: | |
324 | ctx->tex_ratio = 16; | |
325 | ctx->tex_funct = ctx->texdsp.rgtc2s_block; | |
326 | break; | |
327 | default: | |
328 | av_log(avctx, AV_LOG_ERROR, | |
329 | "Unsupported DXGI format %d.\n", dxgi); | |
330 | return AVERROR_INVALIDDATA; | |
331 | } | |
332 | break; | |
333 | default: | |
334 | av_log(avctx, AV_LOG_ERROR, "Unsupported %s fourcc.\n", buf); | |
335 | return AVERROR_INVALIDDATA; | |
336 | } | |
337 | } else if (ctx->paletted) { | |
338 | if (bpp == 8) { | |
339 | avctx->pix_fmt = AV_PIX_FMT_PAL8; | |
340 | } else { | |
341 | av_log(avctx, AV_LOG_ERROR, "Unsupported palette bpp %d.\n", bpp); | |
342 | return AVERROR_INVALIDDATA; | |
343 | } | |
344 | } else { | |
345 | /* 8 bpp */ | |
346 | if (bpp == 8 && r == 0xff && g == 0 && b == 0 && a == 0) | |
347 | avctx->pix_fmt = AV_PIX_FMT_GRAY8; | |
348 | /* 16 bpp */ | |
349 | else if (bpp == 16 && r == 0xff && g == 0 && b == 0 && a == 0xff00) | |
350 | avctx->pix_fmt = AV_PIX_FMT_YA8; | |
351 | else if (bpp == 16 && r == 0xffff && g == 0 && b == 0 && a == 0) | |
352 | avctx->pix_fmt = AV_PIX_FMT_GRAY16LE; | |
353 | else if (bpp == 16 && r == 0xf800 && g == 0x7e0 && b == 0x1f && a == 0) | |
354 | avctx->pix_fmt = AV_PIX_FMT_RGB565LE; | |
355 | /* 24 bpp */ | |
356 | else if (bpp == 24 && r == 0xff0000 && g == 0xff00 && b == 0xff && a == 0) | |
357 | avctx->pix_fmt = AV_PIX_FMT_BGR24; | |
358 | /* 32 bpp */ | |
359 | else if (bpp == 32 && r == 0xff0000 && g == 0xff00 && b == 0xff && a == 0) | |
360 | avctx->pix_fmt = AV_PIX_FMT_RGBA; // opaque | |
361 | else if (bpp == 32 && r == 0xff && g == 0xff00 && b == 0xff0000 && a == 0) | |
362 | avctx->pix_fmt = AV_PIX_FMT_BGRA; // opaque | |
363 | else if (bpp == 32 && r == 0xff0000 && g == 0xff00 && b == 0xff && a == 0xff000000) | |
364 | avctx->pix_fmt = AV_PIX_FMT_RGBA; | |
365 | else if (bpp == 32 && r == 0xff && g == 0xff00 && b == 0xff0000 && a == 0xff000000) | |
366 | avctx->pix_fmt = AV_PIX_FMT_BGRA; | |
367 | /* give up */ | |
368 | else { | |
369 | av_log(avctx, AV_LOG_ERROR, "Unknown pixel format " | |
370 | "[bpp %d r 0x%x g 0x%x b 0x%x a 0x%x].\n", bpp, r, g, b, a); | |
371 | return AVERROR_INVALIDDATA; | |
372 | } | |
373 | } | |
374 | ||
375 | /* Set any remaining post-proc that should happen before frame is ready. */ | |
376 | if (alpha_exponent) | |
377 | ctx->postproc = DDS_ALPHA_EXP; | |
378 | else if (normal_map) | |
379 | ctx->postproc = DDS_NORMAL_MAP; | |
380 | else if (ycocg_classic && !ctx->compressed) | |
381 | ctx->postproc = DDS_RAW_YCOCG; | |
382 | else if (avctx->pix_fmt == AV_PIX_FMT_YA8) | |
383 | ctx->postproc = DDS_SWAP_ALPHA; | |
384 | ||
385 | /* ATI/NVidia variants sometimes add swizzling in bpp. */ | |
386 | switch (bpp) { | |
387 | case MKTAG('A', '2', 'X', 'Y'): | |
388 | ctx->postproc = DDS_SWIZZLE_A2XY; | |
389 | break; | |
390 | case MKTAG('x', 'G', 'B', 'R'): | |
391 | ctx->postproc = DDS_SWIZZLE_XGBR; | |
392 | break; | |
393 | case MKTAG('x', 'R', 'B', 'G'): | |
394 | ctx->postproc = DDS_SWIZZLE_XRBG; | |
395 | break; | |
396 | case MKTAG('R', 'B', 'x', 'G'): | |
397 | ctx->postproc = DDS_SWIZZLE_RBXG; | |
398 | break; | |
399 | case MKTAG('R', 'G', 'x', 'B'): | |
400 | ctx->postproc = DDS_SWIZZLE_RGXB; | |
401 | break; | |
402 | case MKTAG('R', 'x', 'B', 'G'): | |
403 | ctx->postproc = DDS_SWIZZLE_RXBG; | |
404 | break; | |
405 | case MKTAG('x', 'G', 'x', 'R'): | |
406 | ctx->postproc = DDS_SWIZZLE_XGXR; | |
407 | break; | |
408 | case MKTAG('A', '2', 'D', '5'): | |
409 | ctx->postproc = DDS_NORMAL_MAP; | |
410 | break; | |
411 | } | |
412 | ||
413 | return 0; | |
414 | } | |
415 | ||
416 | static int decompress_texture_thread(AVCodecContext *avctx, void *arg, | |
417 | int block_nb, int thread_nb) | |
418 | { | |
419 | DDSContext *ctx = avctx->priv_data; | |
420 | AVFrame *frame = arg; | |
421 | int x = (TEXTURE_BLOCK_W * block_nb) % avctx->coded_width; | |
422 | int y = TEXTURE_BLOCK_H * (TEXTURE_BLOCK_W * block_nb / avctx->coded_width); | |
423 | uint8_t *p = frame->data[0] + x * 4 + y * frame->linesize[0]; | |
424 | const uint8_t *d = ctx->tex_data + block_nb * ctx->tex_ratio; | |
425 | ||
426 | ctx->tex_funct(p, frame->linesize[0], d); | |
427 | return 0; | |
428 | } | |
429 | ||
430 | static void do_swizzle(AVFrame *frame, int x, int y) | |
431 | { | |
432 | int i; | |
433 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { | |
434 | uint8_t *src = frame->data[0] + i; | |
435 | FFSWAP(uint8_t, src[x], src[y]); | |
436 | } | |
437 | } | |
438 | ||
439 | static void run_postproc(AVCodecContext *avctx, AVFrame *frame) | |
440 | { | |
441 | DDSContext *ctx = avctx->priv_data; | |
442 | int i, x_off; | |
443 | ||
444 | switch (ctx->postproc) { | |
445 | case DDS_ALPHA_EXP: | |
446 | /* Alpha-exponential mode divides each channel by the maximum | |
447 | * R, G or B value, and stores the multiplying factor in the | |
448 | * alpha channel. */ | |
449 | av_log(avctx, AV_LOG_DEBUG, "Post-processing alpha exponent.\n"); | |
450 | ||
451 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { | |
452 | uint8_t *src = frame->data[0] + i; | |
453 | int r = src[0]; | |
454 | int g = src[1]; | |
455 | int b = src[2]; | |
456 | int a = src[3]; | |
457 | ||
458 | src[0] = r * a / 255; | |
459 | src[1] = g * a / 255; | |
460 | src[2] = b * a / 255; | |
461 | src[3] = 255; | |
462 | } | |
463 | break; | |
464 | case DDS_NORMAL_MAP: | |
465 | /* Normal maps work in the XYZ color space and they encode | |
466 | * X in R or in A, depending on the texture type, Y in G and | |
467 | * derive Z with a square root of the distance. | |
468 | * | |
469 | * http://www.realtimecollisiondetection.net/blog/?p=28 */ | |
470 | av_log(avctx, AV_LOG_DEBUG, "Post-processing normal map.\n"); | |
471 | ||
472 | x_off = ctx->tex_ratio == 8 ? 0 : 3; | |
473 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { | |
474 | uint8_t *src = frame->data[0] + i; | |
475 | int x = src[x_off]; | |
476 | int y = src[1]; | |
477 | int z = 127; | |
478 | ||
479 | int d = (255 * 255 - x * x - y * y) / 2; | |
480 | if (d > 0) | |
481 | z = rint(sqrtf(d)); | |
482 | ||
483 | src[0] = x; | |
484 | src[1] = y; | |
485 | src[2] = z; | |
486 | src[3] = 255; | |
487 | } | |
488 | break; | |
489 | case DDS_RAW_YCOCG: | |
490 | /* Data is Y-Co-Cg-A and not RGBA, but they are represented | |
491 | * with the same masks in the DDPF header. */ | |
492 | av_log(avctx, AV_LOG_DEBUG, "Post-processing raw YCoCg.\n"); | |
493 | ||
494 | for (i = 0; i < frame->linesize[0] * frame->height; i += 4) { | |
495 | uint8_t *src = frame->data[0] + i; | |
496 | int a = src[0]; | |
497 | int cg = src[1] - 128; | |
498 | int co = src[2] - 128; | |
499 | int y = src[3]; | |
500 | ||
501 | src[0] = av_clip_uint8(y + co - cg); | |
502 | src[1] = av_clip_uint8(y + cg); | |
503 | src[2] = av_clip_uint8(y - co - cg); | |
504 | src[3] = a; | |
505 | } | |
506 | break; | |
507 | case DDS_SWAP_ALPHA: | |
508 | /* Alpha and Luma are stored swapped. */ | |
509 | av_log(avctx, AV_LOG_DEBUG, "Post-processing swapped Luma/Alpha.\n"); | |
510 | ||
511 | for (i = 0; i < frame->linesize[0] * frame->height; i += 2) { | |
512 | uint8_t *src = frame->data[0] + i; | |
513 | FFSWAP(uint8_t, src[0], src[1]); | |
514 | } | |
515 | break; | |
516 | case DDS_SWIZZLE_A2XY: | |
517 | /* Swap R and G, often used to restore a standard RGTC2. */ | |
518 | av_log(avctx, AV_LOG_DEBUG, "Post-processing A2XY swizzle.\n"); | |
519 | do_swizzle(frame, 0, 1); | |
520 | break; | |
521 | case DDS_SWIZZLE_RBXG: | |
522 | /* Swap G and A, then B and new A (G). */ | |
523 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RBXG swizzle.\n"); | |
524 | do_swizzle(frame, 1, 3); | |
525 | do_swizzle(frame, 2, 3); | |
526 | break; | |
527 | case DDS_SWIZZLE_RGXB: | |
528 | /* Swap B and A. */ | |
529 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RGXB swizzle.\n"); | |
530 | do_swizzle(frame, 2, 3); | |
531 | break; | |
532 | case DDS_SWIZZLE_RXBG: | |
533 | /* Swap G and A. */ | |
534 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RXBG swizzle.\n"); | |
535 | do_swizzle(frame, 1, 3); | |
536 | break; | |
537 | case DDS_SWIZZLE_RXGB: | |
538 | /* Swap R and A (misleading name). */ | |
539 | av_log(avctx, AV_LOG_DEBUG, "Post-processing RXGB swizzle.\n"); | |
540 | do_swizzle(frame, 0, 3); | |
541 | break; | |
542 | case DDS_SWIZZLE_XGBR: | |
543 | /* Swap B and A, then R and new A (B). */ | |
544 | av_log(avctx, AV_LOG_DEBUG, "Post-processing XGBR swizzle.\n"); | |
545 | do_swizzle(frame, 2, 3); | |
546 | do_swizzle(frame, 0, 3); | |
547 | break; | |
548 | case DDS_SWIZZLE_XGXR: | |
549 | /* Swap G and A, then R and new A (G), then new R (G) and new G (A). | |
550 | * This variant does not store any B component. */ | |
551 | av_log(avctx, AV_LOG_DEBUG, "Post-processing XGXR swizzle.\n"); | |
552 | do_swizzle(frame, 1, 3); | |
553 | do_swizzle(frame, 0, 3); | |
554 | do_swizzle(frame, 0, 1); | |
555 | break; | |
556 | case DDS_SWIZZLE_XRBG: | |
557 | /* Swap G and A, then R and new A (G). */ | |
558 | av_log(avctx, AV_LOG_DEBUG, "Post-processing XRBG swizzle.\n"); | |
559 | do_swizzle(frame, 1, 3); | |
560 | do_swizzle(frame, 0, 3); | |
561 | break; | |
562 | } | |
563 | } | |
564 | ||
565 | static int dds_decode(AVCodecContext *avctx, void *data, | |
566 | int *got_frame, AVPacket *avpkt) | |
567 | { | |
568 | DDSContext *ctx = avctx->priv_data; | |
569 | GetByteContext *gbc = &ctx->gbc; | |
570 | AVFrame *frame = data; | |
571 | int blocks, mipmap; | |
572 | int ret; | |
573 | ||
574 | ff_texturedsp_init(&ctx->texdsp); | |
575 | bytestream2_init(gbc, avpkt->data, avpkt->size); | |
576 | ||
577 | if (bytestream2_get_bytes_left(gbc) < 128) { | |
578 | av_log(avctx, AV_LOG_ERROR, "Frame is too small (%d).", | |
579 | bytestream2_get_bytes_left(gbc)); | |
580 | return AVERROR_INVALIDDATA; | |
581 | } | |
582 | ||
583 | if (bytestream2_get_le32(gbc) != MKTAG('D', 'D', 'S', ' ') || | |
584 | bytestream2_get_le32(gbc) != 124) { // header size | |
585 | av_log(avctx, AV_LOG_ERROR, "Invalid DDS header."); | |
586 | return AVERROR_INVALIDDATA; | |
587 | } | |
588 | ||
589 | bytestream2_skip(gbc, 4); // flags | |
590 | ||
591 | avctx->height = bytestream2_get_le32(gbc); | |
592 | avctx->width = bytestream2_get_le32(gbc); | |
593 | ret = av_image_check_size(avctx->width, avctx->height, 0, avctx); | |
594 | if (ret < 0) { | |
595 | av_log(avctx, AV_LOG_ERROR, "Invalid image size %dx%d.\n", | |
596 | avctx->width, avctx->height); | |
597 | return ret; | |
598 | } | |
599 | ||
600 | /* Since codec is based on 4x4 blocks, size is aligned to 4. */ | |
601 | avctx->coded_width = FFALIGN(avctx->width, TEXTURE_BLOCK_W); | |
602 | avctx->coded_height = FFALIGN(avctx->height, TEXTURE_BLOCK_H); | |
603 | ||
604 | bytestream2_skip(gbc, 4); // pitch | |
605 | bytestream2_skip(gbc, 4); // depth | |
606 | mipmap = bytestream2_get_le32(gbc); | |
607 | if (mipmap != 0) | |
608 | av_log(avctx, AV_LOG_VERBOSE, "Found %d mipmaps (ignored).\n", mipmap); | |
609 | ||
610 | /* Extract pixel format information, considering additional elements | |
611 | * in reserved1 and reserved2. */ | |
612 | ret = parse_pixel_format(avctx); | |
613 | if (ret < 0) | |
614 | return ret; | |
615 | ||
616 | ret = ff_get_buffer(avctx, frame, 0); | |
617 | if (ret < 0) | |
618 | return ret; | |
619 | ||
620 | if (ctx->compressed) { | |
621 | /* Use the decompress function on the texture, one block per thread. */ | |
622 | ctx->tex_data = gbc->buffer; | |
623 | blocks = avctx->coded_width * avctx->coded_height / (TEXTURE_BLOCK_W * TEXTURE_BLOCK_H); | |
624 | avctx->execute2(avctx, decompress_texture_thread, frame, NULL, blocks); | |
625 | } else { | |
626 | int linesize = av_image_get_linesize(avctx->pix_fmt, frame->width, 0); | |
627 | ||
628 | if (ctx->paletted) { | |
4d554843 LB |
629 | int i; |
630 | uint32_t *p = (uint32_t *)frame->data[1]; | |
631 | ||
5c018ee1 | 632 | /* Use the first 1024 bytes as palette, then copy the rest. */ |
4d554843 LB |
633 | for (i = 0; i < 256; i++) |
634 | p[i] = bytestream2_get_le32(gbc); | |
635 | ||
5c018ee1 VG |
636 | frame->palette_has_changed = 1; |
637 | } | |
638 | ||
639 | av_image_copy_plane(frame->data[0], frame->linesize[0], | |
640 | gbc->buffer, linesize, | |
641 | linesize, frame->height); | |
642 | } | |
643 | ||
644 | /* Run any post processing here if needed. */ | |
645 | if (avctx->pix_fmt == AV_PIX_FMT_RGBA || avctx->pix_fmt == AV_PIX_FMT_YA8) | |
646 | run_postproc(avctx, frame); | |
647 | ||
648 | /* Frame is ready to be output. */ | |
649 | frame->pict_type = AV_PICTURE_TYPE_I; | |
650 | frame->key_frame = 1; | |
651 | *got_frame = 1; | |
652 | ||
653 | return avpkt->size; | |
654 | } | |
655 | ||
656 | AVCodec ff_dds_decoder = { | |
657 | .name = "dds", | |
658 | .long_name = NULL_IF_CONFIG_SMALL("DirectDraw Surface image decoder"), | |
659 | .type = AVMEDIA_TYPE_VIDEO, | |
660 | .id = AV_CODEC_ID_DDS, | |
661 | .decode = dds_decode, | |
662 | .priv_data_size = sizeof(DDSContext), | |
663 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS, | |
664 | .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | |
665 | }; |