Commit | Line | Data |
---|---|---|
2ba45a60 DM |
1 | /* |
2 | * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at> | |
3 | * | |
4 | * This file is part of FFmpeg. | |
5 | * | |
6 | * FFmpeg is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU Lesser General Public | |
8 | * License as published by the Free Software Foundation; either | |
9 | * version 2.1 of the License, or (at your option) any later version. | |
10 | * | |
11 | * FFmpeg is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * Lesser General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU Lesser General Public | |
17 | * License along with FFmpeg; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
19 | */ | |
20 | ||
21 | #include <inttypes.h> | |
22 | #include <math.h> | |
23 | #include <stdio.h> | |
24 | #include <string.h> | |
25 | ||
26 | #include "libavutil/avassert.h" | |
27 | #include "libavutil/avutil.h" | |
28 | #include "libavutil/bswap.h" | |
29 | #include "libavutil/cpu.h" | |
f6fa7814 | 30 | #include "libavutil/imgutils.h" |
2ba45a60 DM |
31 | #include "libavutil/intreadwrite.h" |
32 | #include "libavutil/mathematics.h" | |
33 | #include "libavutil/pixdesc.h" | |
34 | #include "config.h" | |
35 | #include "rgb2rgb.h" | |
36 | #include "swscale_internal.h" | |
37 | #include "swscale.h" | |
38 | ||
39 | DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_128)[9][8] = { | |
40 | { 36, 68, 60, 92, 34, 66, 58, 90, }, | |
41 | { 100, 4, 124, 28, 98, 2, 122, 26, }, | |
42 | { 52, 84, 44, 76, 50, 82, 42, 74, }, | |
43 | { 116, 20, 108, 12, 114, 18, 106, 10, }, | |
44 | { 32, 64, 56, 88, 38, 70, 62, 94, }, | |
45 | { 96, 0, 120, 24, 102, 6, 126, 30, }, | |
46 | { 48, 80, 40, 72, 54, 86, 46, 78, }, | |
47 | { 112, 16, 104, 8, 118, 22, 110, 14, }, | |
48 | { 36, 68, 60, 92, 34, 66, 58, 90, }, | |
49 | }; | |
50 | ||
51 | DECLARE_ALIGNED(8, static const uint8_t, sws_pb_64)[8] = { | |
52 | 64, 64, 64, 64, 64, 64, 64, 64 | |
53 | }; | |
54 | ||
55 | static av_always_inline void fillPlane(uint8_t *plane, int stride, int width, | |
56 | int height, int y, uint8_t val) | |
57 | { | |
58 | int i; | |
59 | uint8_t *ptr = plane + stride * y; | |
60 | for (i = 0; i < height; i++) { | |
61 | memset(ptr, val, width); | |
62 | ptr += stride; | |
63 | } | |
64 | } | |
65 | ||
66 | static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW, | |
67 | const uint8_t *_src, const int16_t *filter, | |
68 | const int32_t *filterPos, int filterSize) | |
69 | { | |
70 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat); | |
71 | int i; | |
72 | int32_t *dst = (int32_t *) _dst; | |
73 | const uint16_t *src = (const uint16_t *) _src; | |
74 | int bits = desc->comp[0].depth_minus1; | |
75 | int sh = bits - 4; | |
76 | ||
77 | if((isAnyRGB(c->srcFormat) || c->srcFormat==AV_PIX_FMT_PAL8) && desc->comp[0].depth_minus1<15) | |
78 | sh= 9; | |
79 | ||
80 | for (i = 0; i < dstW; i++) { | |
81 | int j; | |
82 | int srcPos = filterPos[i]; | |
83 | int val = 0; | |
84 | ||
85 | for (j = 0; j < filterSize; j++) { | |
86 | val += src[srcPos + j] * filter[filterSize * i + j]; | |
87 | } | |
88 | // filter=14 bit, input=16 bit, output=30 bit, >> 11 makes 19 bit | |
89 | dst[i] = FFMIN(val >> sh, (1 << 19) - 1); | |
90 | } | |
91 | } | |
92 | ||
93 | static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW, | |
94 | const uint8_t *_src, const int16_t *filter, | |
95 | const int32_t *filterPos, int filterSize) | |
96 | { | |
97 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat); | |
98 | int i; | |
99 | const uint16_t *src = (const uint16_t *) _src; | |
100 | int sh = desc->comp[0].depth_minus1; | |
101 | ||
102 | if(sh<15) | |
103 | sh= isAnyRGB(c->srcFormat) || c->srcFormat==AV_PIX_FMT_PAL8 ? 13 : desc->comp[0].depth_minus1; | |
104 | ||
105 | for (i = 0; i < dstW; i++) { | |
106 | int j; | |
107 | int srcPos = filterPos[i]; | |
108 | int val = 0; | |
109 | ||
110 | for (j = 0; j < filterSize; j++) { | |
111 | val += src[srcPos + j] * filter[filterSize * i + j]; | |
112 | } | |
113 | // filter=14 bit, input=16 bit, output=30 bit, >> 15 makes 15 bit | |
114 | dst[i] = FFMIN(val >> sh, (1 << 15) - 1); | |
115 | } | |
116 | } | |
117 | ||
118 | // bilinear / bicubic scaling | |
119 | static void hScale8To15_c(SwsContext *c, int16_t *dst, int dstW, | |
120 | const uint8_t *src, const int16_t *filter, | |
121 | const int32_t *filterPos, int filterSize) | |
122 | { | |
123 | int i; | |
124 | for (i = 0; i < dstW; i++) { | |
125 | int j; | |
126 | int srcPos = filterPos[i]; | |
127 | int val = 0; | |
128 | for (j = 0; j < filterSize; j++) { | |
129 | val += ((int)src[srcPos + j]) * filter[filterSize * i + j]; | |
130 | } | |
131 | dst[i] = FFMIN(val >> 7, (1 << 15) - 1); // the cubic equation does overflow ... | |
132 | } | |
133 | } | |
134 | ||
135 | static void hScale8To19_c(SwsContext *c, int16_t *_dst, int dstW, | |
136 | const uint8_t *src, const int16_t *filter, | |
137 | const int32_t *filterPos, int filterSize) | |
138 | { | |
139 | int i; | |
140 | int32_t *dst = (int32_t *) _dst; | |
141 | for (i = 0; i < dstW; i++) { | |
142 | int j; | |
143 | int srcPos = filterPos[i]; | |
144 | int val = 0; | |
145 | for (j = 0; j < filterSize; j++) { | |
146 | val += ((int)src[srcPos + j]) * filter[filterSize * i + j]; | |
147 | } | |
148 | dst[i] = FFMIN(val >> 3, (1 << 19) - 1); // the cubic equation does overflow ... | |
149 | } | |
150 | } | |
151 | ||
152 | // FIXME all pal and rgb srcFormats could do this conversion as well | |
153 | // FIXME all scalers more complex than bilinear could do half of this transform | |
154 | static void chrRangeToJpeg_c(int16_t *dstU, int16_t *dstV, int width) | |
155 | { | |
156 | int i; | |
157 | for (i = 0; i < width; i++) { | |
158 | dstU[i] = (FFMIN(dstU[i], 30775) * 4663 - 9289992) >> 12; // -264 | |
159 | dstV[i] = (FFMIN(dstV[i], 30775) * 4663 - 9289992) >> 12; // -264 | |
160 | } | |
161 | } | |
162 | ||
163 | static void chrRangeFromJpeg_c(int16_t *dstU, int16_t *dstV, int width) | |
164 | { | |
165 | int i; | |
166 | for (i = 0; i < width; i++) { | |
167 | dstU[i] = (dstU[i] * 1799 + 4081085) >> 11; // 1469 | |
168 | dstV[i] = (dstV[i] * 1799 + 4081085) >> 11; // 1469 | |
169 | } | |
170 | } | |
171 | ||
172 | static void lumRangeToJpeg_c(int16_t *dst, int width) | |
173 | { | |
174 | int i; | |
175 | for (i = 0; i < width; i++) | |
176 | dst[i] = (FFMIN(dst[i], 30189) * 19077 - 39057361) >> 14; | |
177 | } | |
178 | ||
179 | static void lumRangeFromJpeg_c(int16_t *dst, int width) | |
180 | { | |
181 | int i; | |
182 | for (i = 0; i < width; i++) | |
183 | dst[i] = (dst[i] * 14071 + 33561947) >> 14; | |
184 | } | |
185 | ||
186 | static void chrRangeToJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width) | |
187 | { | |
188 | int i; | |
189 | int32_t *dstU = (int32_t *) _dstU; | |
190 | int32_t *dstV = (int32_t *) _dstV; | |
191 | for (i = 0; i < width; i++) { | |
192 | dstU[i] = (FFMIN(dstU[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264 | |
193 | dstV[i] = (FFMIN(dstV[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264 | |
194 | } | |
195 | } | |
196 | ||
197 | static void chrRangeFromJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width) | |
198 | { | |
199 | int i; | |
200 | int32_t *dstU = (int32_t *) _dstU; | |
201 | int32_t *dstV = (int32_t *) _dstV; | |
202 | for (i = 0; i < width; i++) { | |
203 | dstU[i] = (dstU[i] * 1799 + (4081085 << 4)) >> 11; // 1469 | |
204 | dstV[i] = (dstV[i] * 1799 + (4081085 << 4)) >> 11; // 1469 | |
205 | } | |
206 | } | |
207 | ||
208 | static void lumRangeToJpeg16_c(int16_t *_dst, int width) | |
209 | { | |
210 | int i; | |
211 | int32_t *dst = (int32_t *) _dst; | |
212 | for (i = 0; i < width; i++) { | |
213 | dst[i] = ((int)(FFMIN(dst[i], 30189 << 4) * 4769U - (39057361 << 2))) >> 12; | |
214 | } | |
215 | } | |
216 | ||
217 | static void lumRangeFromJpeg16_c(int16_t *_dst, int width) | |
218 | { | |
219 | int i; | |
220 | int32_t *dst = (int32_t *) _dst; | |
221 | for (i = 0; i < width; i++) | |
222 | dst[i] = (dst[i]*(14071/4) + (33561947<<4)/4)>>12; | |
223 | } | |
224 | ||
225 | // *** horizontal scale Y line to temp buffer | |
226 | static av_always_inline void hyscale(SwsContext *c, int16_t *dst, int dstWidth, | |
227 | const uint8_t *src_in[4], | |
228 | int srcW, int xInc, | |
229 | const int16_t *hLumFilter, | |
230 | const int32_t *hLumFilterPos, | |
231 | int hLumFilterSize, | |
232 | uint8_t *formatConvBuffer, | |
233 | uint32_t *pal, int isAlpha) | |
234 | { | |
235 | void (*toYV12)(uint8_t *, const uint8_t *, const uint8_t *, const uint8_t *, int, uint32_t *) = | |
236 | isAlpha ? c->alpToYV12 : c->lumToYV12; | |
237 | void (*convertRange)(int16_t *, int) = isAlpha ? NULL : c->lumConvertRange; | |
238 | const uint8_t *src = src_in[isAlpha ? 3 : 0]; | |
239 | ||
240 | if (toYV12) { | |
241 | toYV12(formatConvBuffer, src, src_in[1], src_in[2], srcW, pal); | |
242 | src = formatConvBuffer; | |
243 | } else if (c->readLumPlanar && !isAlpha) { | |
244 | c->readLumPlanar(formatConvBuffer, src_in, srcW, c->input_rgb2yuv_table); | |
245 | src = formatConvBuffer; | |
246 | } else if (c->readAlpPlanar && isAlpha) { | |
247 | c->readAlpPlanar(formatConvBuffer, src_in, srcW, NULL); | |
248 | src = formatConvBuffer; | |
249 | } | |
250 | ||
251 | if (!c->hyscale_fast) { | |
252 | c->hyScale(c, dst, dstWidth, src, hLumFilter, | |
253 | hLumFilterPos, hLumFilterSize); | |
254 | } else { // fast bilinear upscale / crap downscale | |
255 | c->hyscale_fast(c, dst, dstWidth, src, srcW, xInc); | |
256 | } | |
257 | ||
258 | if (convertRange) | |
259 | convertRange(dst, dstWidth); | |
260 | } | |
261 | ||
262 | static av_always_inline void hcscale(SwsContext *c, int16_t *dst1, | |
263 | int16_t *dst2, int dstWidth, | |
264 | const uint8_t *src_in[4], | |
265 | int srcW, int xInc, | |
266 | const int16_t *hChrFilter, | |
267 | const int32_t *hChrFilterPos, | |
268 | int hChrFilterSize, | |
269 | uint8_t *formatConvBuffer, uint32_t *pal) | |
270 | { | |
271 | const uint8_t *src1 = src_in[1], *src2 = src_in[2]; | |
272 | if (c->chrToYV12) { | |
273 | uint8_t *buf2 = formatConvBuffer + | |
274 | FFALIGN(srcW*2+78, 16); | |
275 | c->chrToYV12(formatConvBuffer, buf2, src_in[0], src1, src2, srcW, pal); | |
276 | src1= formatConvBuffer; | |
277 | src2= buf2; | |
278 | } else if (c->readChrPlanar) { | |
279 | uint8_t *buf2 = formatConvBuffer + | |
280 | FFALIGN(srcW*2+78, 16); | |
281 | c->readChrPlanar(formatConvBuffer, buf2, src_in, srcW, c->input_rgb2yuv_table); | |
282 | src1 = formatConvBuffer; | |
283 | src2 = buf2; | |
284 | } | |
285 | ||
286 | if (!c->hcscale_fast) { | |
287 | c->hcScale(c, dst1, dstWidth, src1, hChrFilter, hChrFilterPos, hChrFilterSize); | |
288 | c->hcScale(c, dst2, dstWidth, src2, hChrFilter, hChrFilterPos, hChrFilterSize); | |
289 | } else { // fast bilinear upscale / crap downscale | |
290 | c->hcscale_fast(c, dst1, dst2, dstWidth, src1, src2, srcW, xInc); | |
291 | } | |
292 | ||
293 | if (c->chrConvertRange) | |
294 | c->chrConvertRange(dst1, dst2, dstWidth); | |
295 | } | |
296 | ||
297 | #define DEBUG_SWSCALE_BUFFERS 0 | |
298 | #define DEBUG_BUFFERS(...) \ | |
299 | if (DEBUG_SWSCALE_BUFFERS) \ | |
300 | av_log(c, AV_LOG_DEBUG, __VA_ARGS__) | |
301 | ||
302 | static int swscale(SwsContext *c, const uint8_t *src[], | |
303 | int srcStride[], int srcSliceY, | |
304 | int srcSliceH, uint8_t *dst[], int dstStride[]) | |
305 | { | |
306 | /* load a few things into local vars to make the code more readable? | |
307 | * and faster */ | |
308 | const int srcW = c->srcW; | |
309 | const int dstW = c->dstW; | |
310 | const int dstH = c->dstH; | |
311 | const int chrDstW = c->chrDstW; | |
312 | const int chrSrcW = c->chrSrcW; | |
313 | const int lumXInc = c->lumXInc; | |
314 | const int chrXInc = c->chrXInc; | |
315 | const enum AVPixelFormat dstFormat = c->dstFormat; | |
316 | const int flags = c->flags; | |
317 | int32_t *vLumFilterPos = c->vLumFilterPos; | |
318 | int32_t *vChrFilterPos = c->vChrFilterPos; | |
319 | int32_t *hLumFilterPos = c->hLumFilterPos; | |
320 | int32_t *hChrFilterPos = c->hChrFilterPos; | |
321 | int16_t *hLumFilter = c->hLumFilter; | |
322 | int16_t *hChrFilter = c->hChrFilter; | |
323 | int32_t *lumMmxFilter = c->lumMmxFilter; | |
324 | int32_t *chrMmxFilter = c->chrMmxFilter; | |
325 | const int vLumFilterSize = c->vLumFilterSize; | |
326 | const int vChrFilterSize = c->vChrFilterSize; | |
327 | const int hLumFilterSize = c->hLumFilterSize; | |
328 | const int hChrFilterSize = c->hChrFilterSize; | |
329 | int16_t **lumPixBuf = c->lumPixBuf; | |
330 | int16_t **chrUPixBuf = c->chrUPixBuf; | |
331 | int16_t **chrVPixBuf = c->chrVPixBuf; | |
332 | int16_t **alpPixBuf = c->alpPixBuf; | |
333 | const int vLumBufSize = c->vLumBufSize; | |
334 | const int vChrBufSize = c->vChrBufSize; | |
335 | uint8_t *formatConvBuffer = c->formatConvBuffer; | |
336 | uint32_t *pal = c->pal_yuv; | |
337 | yuv2planar1_fn yuv2plane1 = c->yuv2plane1; | |
338 | yuv2planarX_fn yuv2planeX = c->yuv2planeX; | |
339 | yuv2interleavedX_fn yuv2nv12cX = c->yuv2nv12cX; | |
340 | yuv2packed1_fn yuv2packed1 = c->yuv2packed1; | |
341 | yuv2packed2_fn yuv2packed2 = c->yuv2packed2; | |
342 | yuv2packedX_fn yuv2packedX = c->yuv2packedX; | |
343 | yuv2anyX_fn yuv2anyX = c->yuv2anyX; | |
344 | const int chrSrcSliceY = srcSliceY >> c->chrSrcVSubSample; | |
345 | const int chrSrcSliceH = FF_CEIL_RSHIFT(srcSliceH, c->chrSrcVSubSample); | |
346 | int should_dither = is9_OR_10BPS(c->srcFormat) || | |
347 | is16BPS(c->srcFormat); | |
348 | int lastDstY; | |
349 | ||
350 | /* vars which will change and which we need to store back in the context */ | |
351 | int dstY = c->dstY; | |
352 | int lumBufIndex = c->lumBufIndex; | |
353 | int chrBufIndex = c->chrBufIndex; | |
354 | int lastInLumBuf = c->lastInLumBuf; | |
355 | int lastInChrBuf = c->lastInChrBuf; | |
356 | ||
357 | if (!usePal(c->srcFormat)) { | |
358 | pal = c->input_rgb2yuv_table; | |
359 | } | |
360 | ||
361 | if (isPacked(c->srcFormat)) { | |
362 | src[0] = | |
363 | src[1] = | |
364 | src[2] = | |
365 | src[3] = src[0]; | |
366 | srcStride[0] = | |
367 | srcStride[1] = | |
368 | srcStride[2] = | |
369 | srcStride[3] = srcStride[0]; | |
370 | } | |
371 | srcStride[1] <<= c->vChrDrop; | |
372 | srcStride[2] <<= c->vChrDrop; | |
373 | ||
374 | DEBUG_BUFFERS("swscale() %p[%d] %p[%d] %p[%d] %p[%d] -> %p[%d] %p[%d] %p[%d] %p[%d]\n", | |
375 | src[0], srcStride[0], src[1], srcStride[1], | |
376 | src[2], srcStride[2], src[3], srcStride[3], | |
377 | dst[0], dstStride[0], dst[1], dstStride[1], | |
378 | dst[2], dstStride[2], dst[3], dstStride[3]); | |
379 | DEBUG_BUFFERS("srcSliceY: %d srcSliceH: %d dstY: %d dstH: %d\n", | |
380 | srcSliceY, srcSliceH, dstY, dstH); | |
381 | DEBUG_BUFFERS("vLumFilterSize: %d vLumBufSize: %d vChrFilterSize: %d vChrBufSize: %d\n", | |
382 | vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize); | |
383 | ||
384 | if (dstStride[0]&15 || dstStride[1]&15 || | |
385 | dstStride[2]&15 || dstStride[3]&15) { | |
386 | static int warnedAlready = 0; // FIXME maybe move this into the context | |
387 | if (flags & SWS_PRINT_INFO && !warnedAlready) { | |
388 | av_log(c, AV_LOG_WARNING, | |
389 | "Warning: dstStride is not aligned!\n" | |
390 | " ->cannot do aligned memory accesses anymore\n"); | |
391 | warnedAlready = 1; | |
392 | } | |
393 | } | |
394 | ||
395 | if ( (uintptr_t)dst[0]&15 || (uintptr_t)dst[1]&15 || (uintptr_t)dst[2]&15 | |
396 | || (uintptr_t)src[0]&15 || (uintptr_t)src[1]&15 || (uintptr_t)src[2]&15 | |
397 | || dstStride[0]&15 || dstStride[1]&15 || dstStride[2]&15 || dstStride[3]&15 | |
398 | || srcStride[0]&15 || srcStride[1]&15 || srcStride[2]&15 || srcStride[3]&15 | |
399 | ) { | |
400 | static int warnedAlready=0; | |
401 | int cpu_flags = av_get_cpu_flags(); | |
402 | if (HAVE_MMXEXT && (cpu_flags & AV_CPU_FLAG_SSE2) && !warnedAlready){ | |
403 | av_log(c, AV_LOG_WARNING, "Warning: data is not aligned! This can lead to a speedloss\n"); | |
404 | warnedAlready=1; | |
405 | } | |
406 | } | |
407 | ||
408 | /* Note the user might start scaling the picture in the middle so this | |
409 | * will not get executed. This is not really intended but works | |
410 | * currently, so people might do it. */ | |
411 | if (srcSliceY == 0) { | |
412 | lumBufIndex = -1; | |
413 | chrBufIndex = -1; | |
414 | dstY = 0; | |
415 | lastInLumBuf = -1; | |
416 | lastInChrBuf = -1; | |
417 | } | |
418 | ||
419 | if (!should_dither) { | |
420 | c->chrDither8 = c->lumDither8 = sws_pb_64; | |
421 | } | |
422 | lastDstY = dstY; | |
423 | ||
424 | for (; dstY < dstH; dstY++) { | |
425 | const int chrDstY = dstY >> c->chrDstVSubSample; | |
426 | uint8_t *dest[4] = { | |
427 | dst[0] + dstStride[0] * dstY, | |
428 | dst[1] + dstStride[1] * chrDstY, | |
429 | dst[2] + dstStride[2] * chrDstY, | |
430 | (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? dst[3] + dstStride[3] * dstY : NULL, | |
431 | }; | |
432 | int use_mmx_vfilter= c->use_mmx_vfilter; | |
433 | ||
434 | // First line needed as input | |
435 | const int firstLumSrcY = FFMAX(1 - vLumFilterSize, vLumFilterPos[dstY]); | |
436 | const int firstLumSrcY2 = FFMAX(1 - vLumFilterSize, vLumFilterPos[FFMIN(dstY | ((1 << c->chrDstVSubSample) - 1), dstH - 1)]); | |
437 | // First line needed as input | |
438 | const int firstChrSrcY = FFMAX(1 - vChrFilterSize, vChrFilterPos[chrDstY]); | |
439 | ||
440 | // Last line needed as input | |
441 | int lastLumSrcY = FFMIN(c->srcH, firstLumSrcY + vLumFilterSize) - 1; | |
442 | int lastLumSrcY2 = FFMIN(c->srcH, firstLumSrcY2 + vLumFilterSize) - 1; | |
443 | int lastChrSrcY = FFMIN(c->chrSrcH, firstChrSrcY + vChrFilterSize) - 1; | |
444 | int enough_lines; | |
445 | ||
446 | // handle holes (FAST_BILINEAR & weird filters) | |
447 | if (firstLumSrcY > lastInLumBuf) | |
448 | lastInLumBuf = firstLumSrcY - 1; | |
449 | if (firstChrSrcY > lastInChrBuf) | |
450 | lastInChrBuf = firstChrSrcY - 1; | |
451 | av_assert0(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1); | |
452 | av_assert0(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1); | |
453 | ||
454 | DEBUG_BUFFERS("dstY: %d\n", dstY); | |
455 | DEBUG_BUFFERS("\tfirstLumSrcY: %d lastLumSrcY: %d lastInLumBuf: %d\n", | |
456 | firstLumSrcY, lastLumSrcY, lastInLumBuf); | |
457 | DEBUG_BUFFERS("\tfirstChrSrcY: %d lastChrSrcY: %d lastInChrBuf: %d\n", | |
458 | firstChrSrcY, lastChrSrcY, lastInChrBuf); | |
459 | ||
460 | // Do we have enough lines in this slice to output the dstY line | |
461 | enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH && | |
462 | lastChrSrcY < FF_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample); | |
463 | ||
464 | if (!enough_lines) { | |
465 | lastLumSrcY = srcSliceY + srcSliceH - 1; | |
466 | lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1; | |
467 | DEBUG_BUFFERS("buffering slice: lastLumSrcY %d lastChrSrcY %d\n", | |
468 | lastLumSrcY, lastChrSrcY); | |
469 | } | |
470 | ||
471 | // Do horizontal scaling | |
472 | while (lastInLumBuf < lastLumSrcY) { | |
473 | const uint8_t *src1[4] = { | |
474 | src[0] + (lastInLumBuf + 1 - srcSliceY) * srcStride[0], | |
475 | src[1] + (lastInLumBuf + 1 - srcSliceY) * srcStride[1], | |
476 | src[2] + (lastInLumBuf + 1 - srcSliceY) * srcStride[2], | |
477 | src[3] + (lastInLumBuf + 1 - srcSliceY) * srcStride[3], | |
478 | }; | |
479 | lumBufIndex++; | |
480 | av_assert0(lumBufIndex < 2 * vLumBufSize); | |
481 | av_assert0(lastInLumBuf + 1 - srcSliceY < srcSliceH); | |
482 | av_assert0(lastInLumBuf + 1 - srcSliceY >= 0); | |
483 | hyscale(c, lumPixBuf[lumBufIndex], dstW, src1, srcW, lumXInc, | |
484 | hLumFilter, hLumFilterPos, hLumFilterSize, | |
485 | formatConvBuffer, pal, 0); | |
486 | if (CONFIG_SWSCALE_ALPHA && alpPixBuf) | |
487 | hyscale(c, alpPixBuf[lumBufIndex], dstW, src1, srcW, | |
488 | lumXInc, hLumFilter, hLumFilterPos, hLumFilterSize, | |
489 | formatConvBuffer, pal, 1); | |
490 | lastInLumBuf++; | |
491 | DEBUG_BUFFERS("\t\tlumBufIndex %d: lastInLumBuf: %d\n", | |
492 | lumBufIndex, lastInLumBuf); | |
493 | } | |
494 | while (lastInChrBuf < lastChrSrcY) { | |
495 | const uint8_t *src1[4] = { | |
496 | src[0] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[0], | |
497 | src[1] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[1], | |
498 | src[2] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[2], | |
499 | src[3] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[3], | |
500 | }; | |
501 | chrBufIndex++; | |
502 | av_assert0(chrBufIndex < 2 * vChrBufSize); | |
503 | av_assert0(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH)); | |
504 | av_assert0(lastInChrBuf + 1 - chrSrcSliceY >= 0); | |
505 | // FIXME replace parameters through context struct (some at least) | |
506 | ||
507 | if (c->needs_hcscale) | |
508 | hcscale(c, chrUPixBuf[chrBufIndex], chrVPixBuf[chrBufIndex], | |
509 | chrDstW, src1, chrSrcW, chrXInc, | |
510 | hChrFilter, hChrFilterPos, hChrFilterSize, | |
511 | formatConvBuffer, pal); | |
512 | lastInChrBuf++; | |
513 | DEBUG_BUFFERS("\t\tchrBufIndex %d: lastInChrBuf: %d\n", | |
514 | chrBufIndex, lastInChrBuf); | |
515 | } | |
516 | // wrap buf index around to stay inside the ring buffer | |
517 | if (lumBufIndex >= vLumBufSize) | |
518 | lumBufIndex -= vLumBufSize; | |
519 | if (chrBufIndex >= vChrBufSize) | |
520 | chrBufIndex -= vChrBufSize; | |
521 | if (!enough_lines) | |
522 | break; // we can't output a dstY line so let's try with the next slice | |
523 | ||
524 | #if HAVE_MMX_INLINE | |
525 | updateMMXDitherTables(c, dstY, lumBufIndex, chrBufIndex, | |
526 | lastInLumBuf, lastInChrBuf); | |
527 | #endif | |
528 | if (should_dither) { | |
529 | c->chrDither8 = ff_dither_8x8_128[chrDstY & 7]; | |
530 | c->lumDither8 = ff_dither_8x8_128[dstY & 7]; | |
531 | } | |
532 | if (dstY >= dstH - 2) { | |
533 | /* hmm looks like we can't use MMX here without overwriting | |
534 | * this array's tail */ | |
535 | ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX, | |
536 | &yuv2packed1, &yuv2packed2, &yuv2packedX, &yuv2anyX); | |
537 | use_mmx_vfilter= 0; | |
538 | } | |
539 | ||
540 | { | |
541 | const int16_t **lumSrcPtr = (const int16_t **)(void*) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize; | |
542 | const int16_t **chrUSrcPtr = (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; | |
543 | const int16_t **chrVSrcPtr = (const int16_t **)(void*) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; | |
544 | const int16_t **alpSrcPtr = (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? | |
545 | (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL; | |
546 | int16_t *vLumFilter = c->vLumFilter; | |
547 | int16_t *vChrFilter = c->vChrFilter; | |
548 | ||
549 | if (isPlanarYUV(dstFormat) || | |
550 | (isGray(dstFormat) && !isALPHA(dstFormat))) { // YV12 like | |
551 | const int chrSkipMask = (1 << c->chrDstVSubSample) - 1; | |
552 | ||
553 | vLumFilter += dstY * vLumFilterSize; | |
554 | vChrFilter += chrDstY * vChrFilterSize; | |
555 | ||
556 | // av_assert0(use_mmx_vfilter != ( | |
557 | // yuv2planeX == yuv2planeX_10BE_c | |
558 | // || yuv2planeX == yuv2planeX_10LE_c | |
559 | // || yuv2planeX == yuv2planeX_9BE_c | |
560 | // || yuv2planeX == yuv2planeX_9LE_c | |
561 | // || yuv2planeX == yuv2planeX_16BE_c | |
562 | // || yuv2planeX == yuv2planeX_16LE_c | |
563 | // || yuv2planeX == yuv2planeX_8_c) || !ARCH_X86); | |
564 | ||
565 | if(use_mmx_vfilter){ | |
566 | vLumFilter= (int16_t *)c->lumMmxFilter; | |
567 | vChrFilter= (int16_t *)c->chrMmxFilter; | |
568 | } | |
569 | ||
570 | if (vLumFilterSize == 1) { | |
571 | yuv2plane1(lumSrcPtr[0], dest[0], dstW, c->lumDither8, 0); | |
572 | } else { | |
573 | yuv2planeX(vLumFilter, vLumFilterSize, | |
574 | lumSrcPtr, dest[0], | |
575 | dstW, c->lumDither8, 0); | |
576 | } | |
577 | ||
578 | if (!((dstY & chrSkipMask) || isGray(dstFormat))) { | |
579 | if (yuv2nv12cX) { | |
580 | yuv2nv12cX(c, vChrFilter, | |
581 | vChrFilterSize, chrUSrcPtr, chrVSrcPtr, | |
582 | dest[1], chrDstW); | |
583 | } else if (vChrFilterSize == 1) { | |
584 | yuv2plane1(chrUSrcPtr[0], dest[1], chrDstW, c->chrDither8, 0); | |
585 | yuv2plane1(chrVSrcPtr[0], dest[2], chrDstW, c->chrDither8, 3); | |
586 | } else { | |
587 | yuv2planeX(vChrFilter, | |
588 | vChrFilterSize, chrUSrcPtr, dest[1], | |
589 | chrDstW, c->chrDither8, 0); | |
590 | yuv2planeX(vChrFilter, | |
591 | vChrFilterSize, chrVSrcPtr, dest[2], | |
592 | chrDstW, c->chrDither8, use_mmx_vfilter ? (c->uv_offx2 >> 1) : 3); | |
593 | } | |
594 | } | |
595 | ||
596 | if (CONFIG_SWSCALE_ALPHA && alpPixBuf) { | |
597 | if(use_mmx_vfilter){ | |
598 | vLumFilter= (int16_t *)c->alpMmxFilter; | |
599 | } | |
600 | if (vLumFilterSize == 1) { | |
601 | yuv2plane1(alpSrcPtr[0], dest[3], dstW, | |
602 | c->lumDither8, 0); | |
603 | } else { | |
604 | yuv2planeX(vLumFilter, | |
605 | vLumFilterSize, alpSrcPtr, dest[3], | |
606 | dstW, c->lumDither8, 0); | |
607 | } | |
608 | } | |
609 | } else if (yuv2packedX) { | |
610 | av_assert1(lumSrcPtr + vLumFilterSize - 1 < (const int16_t **)lumPixBuf + vLumBufSize * 2); | |
611 | av_assert1(chrUSrcPtr + vChrFilterSize - 1 < (const int16_t **)chrUPixBuf + vChrBufSize * 2); | |
612 | if (c->yuv2packed1 && vLumFilterSize == 1 && | |
613 | vChrFilterSize <= 2) { // unscaled RGB | |
614 | int chrAlpha = vChrFilterSize == 1 ? 0 : vChrFilter[2 * dstY + 1]; | |
615 | yuv2packed1(c, *lumSrcPtr, chrUSrcPtr, chrVSrcPtr, | |
616 | alpPixBuf ? *alpSrcPtr : NULL, | |
617 | dest[0], dstW, chrAlpha, dstY); | |
618 | } else if (c->yuv2packed2 && vLumFilterSize == 2 && | |
619 | vChrFilterSize == 2) { // bilinear upscale RGB | |
620 | int lumAlpha = vLumFilter[2 * dstY + 1]; | |
621 | int chrAlpha = vChrFilter[2 * dstY + 1]; | |
622 | lumMmxFilter[2] = | |
623 | lumMmxFilter[3] = vLumFilter[2 * dstY] * 0x10001; | |
624 | chrMmxFilter[2] = | |
625 | chrMmxFilter[3] = vChrFilter[2 * chrDstY] * 0x10001; | |
626 | yuv2packed2(c, lumSrcPtr, chrUSrcPtr, chrVSrcPtr, | |
627 | alpPixBuf ? alpSrcPtr : NULL, | |
628 | dest[0], dstW, lumAlpha, chrAlpha, dstY); | |
629 | } else { // general RGB | |
630 | yuv2packedX(c, vLumFilter + dstY * vLumFilterSize, | |
631 | lumSrcPtr, vLumFilterSize, | |
632 | vChrFilter + dstY * vChrFilterSize, | |
633 | chrUSrcPtr, chrVSrcPtr, vChrFilterSize, | |
634 | alpSrcPtr, dest[0], dstW, dstY); | |
635 | } | |
636 | } else { | |
637 | av_assert1(!yuv2packed1 && !yuv2packed2); | |
638 | yuv2anyX(c, vLumFilter + dstY * vLumFilterSize, | |
639 | lumSrcPtr, vLumFilterSize, | |
640 | vChrFilter + dstY * vChrFilterSize, | |
641 | chrUSrcPtr, chrVSrcPtr, vChrFilterSize, | |
642 | alpSrcPtr, dest, dstW, dstY); | |
643 | } | |
644 | } | |
645 | } | |
646 | if (isPlanar(dstFormat) && isALPHA(dstFormat) && !alpPixBuf) { | |
647 | int length = dstW; | |
648 | int height = dstY - lastDstY; | |
649 | ||
650 | if (is16BPS(dstFormat) || isNBPS(dstFormat)) { | |
651 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat); | |
652 | fillPlane16(dst[3], dstStride[3], length, height, lastDstY, | |
653 | 1, desc->comp[3].depth_minus1, | |
654 | isBE(dstFormat)); | |
655 | } else | |
656 | fillPlane(dst[3], dstStride[3], length, height, lastDstY, 255); | |
657 | } | |
658 | ||
659 | #if HAVE_MMXEXT_INLINE | |
660 | if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT) | |
661 | __asm__ volatile ("sfence" ::: "memory"); | |
662 | #endif | |
663 | emms_c(); | |
664 | ||
665 | /* store changed local vars back in the context */ | |
666 | c->dstY = dstY; | |
667 | c->lumBufIndex = lumBufIndex; | |
668 | c->chrBufIndex = chrBufIndex; | |
669 | c->lastInLumBuf = lastInLumBuf; | |
670 | c->lastInChrBuf = lastInChrBuf; | |
671 | ||
672 | return dstY - lastDstY; | |
673 | } | |
674 | ||
675 | av_cold void ff_sws_init_range_convert(SwsContext *c) | |
676 | { | |
677 | c->lumConvertRange = NULL; | |
678 | c->chrConvertRange = NULL; | |
679 | if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) { | |
680 | if (c->dstBpc <= 14) { | |
681 | if (c->srcRange) { | |
682 | c->lumConvertRange = lumRangeFromJpeg_c; | |
683 | c->chrConvertRange = chrRangeFromJpeg_c; | |
684 | } else { | |
685 | c->lumConvertRange = lumRangeToJpeg_c; | |
686 | c->chrConvertRange = chrRangeToJpeg_c; | |
687 | } | |
688 | } else { | |
689 | if (c->srcRange) { | |
690 | c->lumConvertRange = lumRangeFromJpeg16_c; | |
691 | c->chrConvertRange = chrRangeFromJpeg16_c; | |
692 | } else { | |
693 | c->lumConvertRange = lumRangeToJpeg16_c; | |
694 | c->chrConvertRange = chrRangeToJpeg16_c; | |
695 | } | |
696 | } | |
697 | } | |
698 | } | |
699 | ||
700 | static av_cold void sws_init_swscale(SwsContext *c) | |
701 | { | |
702 | enum AVPixelFormat srcFormat = c->srcFormat; | |
703 | ||
704 | ff_sws_init_output_funcs(c, &c->yuv2plane1, &c->yuv2planeX, | |
705 | &c->yuv2nv12cX, &c->yuv2packed1, | |
706 | &c->yuv2packed2, &c->yuv2packedX, &c->yuv2anyX); | |
707 | ||
708 | ff_sws_init_input_funcs(c); | |
709 | ||
710 | ||
711 | if (c->srcBpc == 8) { | |
712 | if (c->dstBpc <= 14) { | |
713 | c->hyScale = c->hcScale = hScale8To15_c; | |
714 | if (c->flags & SWS_FAST_BILINEAR) { | |
715 | c->hyscale_fast = ff_hyscale_fast_c; | |
716 | c->hcscale_fast = ff_hcscale_fast_c; | |
717 | } | |
718 | } else { | |
719 | c->hyScale = c->hcScale = hScale8To19_c; | |
720 | } | |
721 | } else { | |
722 | c->hyScale = c->hcScale = c->dstBpc > 14 ? hScale16To19_c | |
723 | : hScale16To15_c; | |
724 | } | |
725 | ||
726 | ff_sws_init_range_convert(c); | |
727 | ||
728 | if (!(isGray(srcFormat) || isGray(c->dstFormat) || | |
729 | srcFormat == AV_PIX_FMT_MONOBLACK || srcFormat == AV_PIX_FMT_MONOWHITE)) | |
730 | c->needs_hcscale = 1; | |
731 | } | |
732 | ||
733 | SwsFunc ff_getSwsFunc(SwsContext *c) | |
734 | { | |
735 | sws_init_swscale(c); | |
736 | ||
737 | if (ARCH_PPC) | |
738 | ff_sws_init_swscale_ppc(c); | |
739 | if (ARCH_X86) | |
740 | ff_sws_init_swscale_x86(c); | |
741 | ||
742 | return swscale; | |
743 | } | |
744 | ||
745 | static void reset_ptr(const uint8_t *src[], int format) | |
746 | { | |
747 | if (!isALPHA(format)) | |
748 | src[3] = NULL; | |
749 | if (!isPlanar(format)) { | |
750 | src[3] = src[2] = NULL; | |
751 | ||
752 | if (!usePal(format)) | |
753 | src[1] = NULL; | |
754 | } | |
755 | } | |
756 | ||
757 | static int check_image_pointers(const uint8_t * const data[4], enum AVPixelFormat pix_fmt, | |
758 | const int linesizes[4]) | |
759 | { | |
760 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); | |
761 | int i; | |
762 | ||
763 | for (i = 0; i < 4; i++) { | |
764 | int plane = desc->comp[i].plane; | |
765 | if (!data[plane] || !linesizes[plane]) | |
766 | return 0; | |
767 | } | |
768 | ||
769 | return 1; | |
770 | } | |
771 | ||
772 | static void xyz12Torgb48(struct SwsContext *c, uint16_t *dst, | |
773 | const uint16_t *src, int stride, int h) | |
774 | { | |
775 | int xp,yp; | |
776 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat); | |
777 | ||
778 | for (yp=0; yp<h; yp++) { | |
779 | for (xp=0; xp+2<stride; xp+=3) { | |
780 | int x, y, z, r, g, b; | |
781 | ||
782 | if (desc->flags & AV_PIX_FMT_FLAG_BE) { | |
783 | x = AV_RB16(src + xp + 0); | |
784 | y = AV_RB16(src + xp + 1); | |
785 | z = AV_RB16(src + xp + 2); | |
786 | } else { | |
787 | x = AV_RL16(src + xp + 0); | |
788 | y = AV_RL16(src + xp + 1); | |
789 | z = AV_RL16(src + xp + 2); | |
790 | } | |
791 | ||
792 | x = c->xyzgamma[x>>4]; | |
793 | y = c->xyzgamma[y>>4]; | |
794 | z = c->xyzgamma[z>>4]; | |
795 | ||
796 | // convert from XYZlinear to sRGBlinear | |
797 | r = c->xyz2rgb_matrix[0][0] * x + | |
798 | c->xyz2rgb_matrix[0][1] * y + | |
799 | c->xyz2rgb_matrix[0][2] * z >> 12; | |
800 | g = c->xyz2rgb_matrix[1][0] * x + | |
801 | c->xyz2rgb_matrix[1][1] * y + | |
802 | c->xyz2rgb_matrix[1][2] * z >> 12; | |
803 | b = c->xyz2rgb_matrix[2][0] * x + | |
804 | c->xyz2rgb_matrix[2][1] * y + | |
805 | c->xyz2rgb_matrix[2][2] * z >> 12; | |
806 | ||
807 | // limit values to 12-bit depth | |
f6fa7814 DM |
808 | r = av_clip(r, 0, 4095); |
809 | g = av_clip(g, 0, 4095); | |
810 | b = av_clip(b, 0, 4095); | |
2ba45a60 DM |
811 | |
812 | // convert from sRGBlinear to RGB and scale from 12bit to 16bit | |
813 | if (desc->flags & AV_PIX_FMT_FLAG_BE) { | |
814 | AV_WB16(dst + xp + 0, c->rgbgamma[r] << 4); | |
815 | AV_WB16(dst + xp + 1, c->rgbgamma[g] << 4); | |
816 | AV_WB16(dst + xp + 2, c->rgbgamma[b] << 4); | |
817 | } else { | |
818 | AV_WL16(dst + xp + 0, c->rgbgamma[r] << 4); | |
819 | AV_WL16(dst + xp + 1, c->rgbgamma[g] << 4); | |
820 | AV_WL16(dst + xp + 2, c->rgbgamma[b] << 4); | |
821 | } | |
822 | } | |
823 | src += stride; | |
824 | dst += stride; | |
825 | } | |
826 | } | |
827 | ||
828 | static void rgb48Toxyz12(struct SwsContext *c, uint16_t *dst, | |
829 | const uint16_t *src, int stride, int h) | |
830 | { | |
831 | int xp,yp; | |
832 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->dstFormat); | |
833 | ||
834 | for (yp=0; yp<h; yp++) { | |
835 | for (xp=0; xp+2<stride; xp+=3) { | |
836 | int x, y, z, r, g, b; | |
837 | ||
838 | if (desc->flags & AV_PIX_FMT_FLAG_BE) { | |
839 | r = AV_RB16(src + xp + 0); | |
840 | g = AV_RB16(src + xp + 1); | |
841 | b = AV_RB16(src + xp + 2); | |
842 | } else { | |
843 | r = AV_RL16(src + xp + 0); | |
844 | g = AV_RL16(src + xp + 1); | |
845 | b = AV_RL16(src + xp + 2); | |
846 | } | |
847 | ||
848 | r = c->rgbgammainv[r>>4]; | |
849 | g = c->rgbgammainv[g>>4]; | |
850 | b = c->rgbgammainv[b>>4]; | |
851 | ||
852 | // convert from sRGBlinear to XYZlinear | |
853 | x = c->rgb2xyz_matrix[0][0] * r + | |
854 | c->rgb2xyz_matrix[0][1] * g + | |
855 | c->rgb2xyz_matrix[0][2] * b >> 12; | |
856 | y = c->rgb2xyz_matrix[1][0] * r + | |
857 | c->rgb2xyz_matrix[1][1] * g + | |
858 | c->rgb2xyz_matrix[1][2] * b >> 12; | |
859 | z = c->rgb2xyz_matrix[2][0] * r + | |
860 | c->rgb2xyz_matrix[2][1] * g + | |
861 | c->rgb2xyz_matrix[2][2] * b >> 12; | |
862 | ||
863 | // limit values to 12-bit depth | |
f6fa7814 DM |
864 | x = av_clip(x, 0, 4095); |
865 | y = av_clip(y, 0, 4095); | |
866 | z = av_clip(z, 0, 4095); | |
2ba45a60 DM |
867 | |
868 | // convert from XYZlinear to X'Y'Z' and scale from 12bit to 16bit | |
869 | if (desc->flags & AV_PIX_FMT_FLAG_BE) { | |
870 | AV_WB16(dst + xp + 0, c->xyzgammainv[x] << 4); | |
871 | AV_WB16(dst + xp + 1, c->xyzgammainv[y] << 4); | |
872 | AV_WB16(dst + xp + 2, c->xyzgammainv[z] << 4); | |
873 | } else { | |
874 | AV_WL16(dst + xp + 0, c->xyzgammainv[x] << 4); | |
875 | AV_WL16(dst + xp + 1, c->xyzgammainv[y] << 4); | |
876 | AV_WL16(dst + xp + 2, c->xyzgammainv[z] << 4); | |
877 | } | |
878 | } | |
879 | src += stride; | |
880 | dst += stride; | |
881 | } | |
882 | } | |
883 | ||
884 | /** | |
885 | * swscale wrapper, so we don't need to export the SwsContext. | |
886 | * Assumes planar YUV to be in YUV order instead of YVU. | |
887 | */ | |
888 | int attribute_align_arg sws_scale(struct SwsContext *c, | |
889 | const uint8_t * const srcSlice[], | |
890 | const int srcStride[], int srcSliceY, | |
891 | int srcSliceH, uint8_t *const dst[], | |
892 | const int dstStride[]) | |
893 | { | |
894 | int i, ret; | |
895 | const uint8_t *src2[4]; | |
896 | uint8_t *dst2[4]; | |
897 | uint8_t *rgb0_tmp = NULL; | |
898 | ||
899 | if (!srcStride || !dstStride || !dst || !srcSlice) { | |
900 | av_log(c, AV_LOG_ERROR, "One of the input parameters to sws_scale() is NULL, please check the calling code\n"); | |
901 | return 0; | |
902 | } | |
f6fa7814 DM |
903 | if (c->cascaded_context[0] && srcSliceY == 0 && srcSliceH == c->cascaded_context[0]->srcH) { |
904 | ret = sws_scale(c->cascaded_context[0], | |
905 | srcSlice, srcStride, srcSliceY, srcSliceH, | |
906 | c->cascaded_tmp, c->cascaded_tmpStride); | |
907 | if (ret < 0) | |
908 | return ret; | |
909 | ret = sws_scale(c->cascaded_context[1], | |
910 | (const uint8_t * const * )c->cascaded_tmp, c->cascaded_tmpStride, 0, c->cascaded_context[0]->dstH, | |
911 | dst, dstStride); | |
912 | return ret; | |
913 | } | |
914 | ||
2ba45a60 DM |
915 | memcpy(src2, srcSlice, sizeof(src2)); |
916 | memcpy(dst2, dst, sizeof(dst2)); | |
917 | ||
918 | // do not mess up sliceDir if we have a "trailing" 0-size slice | |
919 | if (srcSliceH == 0) | |
920 | return 0; | |
921 | ||
922 | if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) { | |
923 | av_log(c, AV_LOG_ERROR, "bad src image pointers\n"); | |
924 | return 0; | |
925 | } | |
926 | if (!check_image_pointers((const uint8_t* const*)dst, c->dstFormat, dstStride)) { | |
927 | av_log(c, AV_LOG_ERROR, "bad dst image pointers\n"); | |
928 | return 0; | |
929 | } | |
930 | ||
931 | if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) { | |
932 | av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n"); | |
933 | return 0; | |
934 | } | |
935 | if (c->sliceDir == 0) { | |
936 | if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1; | |
937 | } | |
938 | ||
939 | if (usePal(c->srcFormat)) { | |
940 | for (i = 0; i < 256; i++) { | |
941 | int r, g, b, y, u, v, a = 0xff; | |
942 | if (c->srcFormat == AV_PIX_FMT_PAL8) { | |
943 | uint32_t p = ((const uint32_t *)(srcSlice[1]))[i]; | |
944 | a = (p >> 24) & 0xFF; | |
945 | r = (p >> 16) & 0xFF; | |
946 | g = (p >> 8) & 0xFF; | |
947 | b = p & 0xFF; | |
948 | } else if (c->srcFormat == AV_PIX_FMT_RGB8) { | |
949 | r = ( i >> 5 ) * 36; | |
950 | g = ((i >> 2) & 7) * 36; | |
951 | b = ( i & 3) * 85; | |
952 | } else if (c->srcFormat == AV_PIX_FMT_BGR8) { | |
953 | b = ( i >> 6 ) * 85; | |
954 | g = ((i >> 3) & 7) * 36; | |
955 | r = ( i & 7) * 36; | |
956 | } else if (c->srcFormat == AV_PIX_FMT_RGB4_BYTE) { | |
957 | r = ( i >> 3 ) * 255; | |
958 | g = ((i >> 1) & 3) * 85; | |
959 | b = ( i & 1) * 255; | |
960 | } else if (c->srcFormat == AV_PIX_FMT_GRAY8 || c->srcFormat == AV_PIX_FMT_GRAY8A) { | |
961 | r = g = b = i; | |
962 | } else { | |
963 | av_assert1(c->srcFormat == AV_PIX_FMT_BGR4_BYTE); | |
964 | b = ( i >> 3 ) * 255; | |
965 | g = ((i >> 1) & 3) * 85; | |
966 | r = ( i & 1) * 255; | |
967 | } | |
968 | #define RGB2YUV_SHIFT 15 | |
969 | #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
970 | #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
971 | #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
972 | #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
973 | #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
974 | #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
975 | #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
976 | #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
977 | #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)) | |
978 | ||
979 | y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT); | |
980 | u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT); | |
981 | v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT); | |
982 | c->pal_yuv[i]= y + (u<<8) + (v<<16) + ((unsigned)a<<24); | |
983 | ||
984 | switch (c->dstFormat) { | |
985 | case AV_PIX_FMT_BGR32: | |
986 | #if !HAVE_BIGENDIAN | |
987 | case AV_PIX_FMT_RGB24: | |
988 | #endif | |
989 | c->pal_rgb[i]= r + (g<<8) + (b<<16) + ((unsigned)a<<24); | |
990 | break; | |
991 | case AV_PIX_FMT_BGR32_1: | |
992 | #if HAVE_BIGENDIAN | |
993 | case AV_PIX_FMT_BGR24: | |
994 | #endif | |
995 | c->pal_rgb[i]= a + (r<<8) + (g<<16) + ((unsigned)b<<24); | |
996 | break; | |
997 | case AV_PIX_FMT_RGB32_1: | |
998 | #if HAVE_BIGENDIAN | |
999 | case AV_PIX_FMT_RGB24: | |
1000 | #endif | |
1001 | c->pal_rgb[i]= a + (b<<8) + (g<<16) + ((unsigned)r<<24); | |
1002 | break; | |
1003 | case AV_PIX_FMT_RGB32: | |
1004 | #if !HAVE_BIGENDIAN | |
1005 | case AV_PIX_FMT_BGR24: | |
1006 | #endif | |
1007 | default: | |
1008 | c->pal_rgb[i]= b + (g<<8) + (r<<16) + ((unsigned)a<<24); | |
1009 | } | |
1010 | } | |
1011 | } | |
1012 | ||
1013 | if (c->src0Alpha && !c->dst0Alpha && isALPHA(c->dstFormat)) { | |
1014 | uint8_t *base; | |
1015 | int x,y; | |
1016 | rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32); | |
1017 | if (!rgb0_tmp) | |
1018 | return AVERROR(ENOMEM); | |
1019 | ||
1020 | base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp; | |
1021 | for (y=0; y<srcSliceH; y++){ | |
1022 | memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*c->srcW); | |
1023 | for (x=c->src0Alpha-1; x<4*c->srcW; x+=4) { | |
1024 | base[ srcStride[0]*y + x] = 0xFF; | |
1025 | } | |
1026 | } | |
1027 | src2[0] = base; | |
1028 | } | |
1029 | ||
1030 | if (c->srcXYZ && !(c->dstXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) { | |
1031 | uint8_t *base; | |
1032 | rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32); | |
1033 | if (!rgb0_tmp) | |
1034 | return AVERROR(ENOMEM); | |
1035 | ||
1036 | base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp; | |
1037 | ||
1038 | xyz12Torgb48(c, (uint16_t*)base, (const uint16_t*)src2[0], srcStride[0]/2, srcSliceH); | |
1039 | src2[0] = base; | |
1040 | } | |
1041 | ||
1042 | if (!srcSliceY && (c->flags & SWS_BITEXACT) && c->dither == SWS_DITHER_ED && c->dither_error[0]) | |
1043 | for (i = 0; i < 4; i++) | |
1044 | memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (c->dstW+2)); | |
1045 | ||
1046 | ||
1047 | // copy strides, so they can safely be modified | |
1048 | if (c->sliceDir == 1) { | |
1049 | // slices go from top to bottom | |
1050 | int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2], | |
1051 | srcStride[3] }; | |
1052 | int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2], | |
1053 | dstStride[3] }; | |
1054 | ||
1055 | reset_ptr(src2, c->srcFormat); | |
1056 | reset_ptr((void*)dst2, c->dstFormat); | |
1057 | ||
1058 | /* reset slice direction at end of frame */ | |
1059 | if (srcSliceY + srcSliceH == c->srcH) | |
1060 | c->sliceDir = 0; | |
1061 | ||
1062 | ret = c->swscale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, | |
1063 | dstStride2); | |
1064 | } else { | |
1065 | // slices go from bottom to top => we flip the image internally | |
1066 | int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2], | |
1067 | -srcStride[3] }; | |
1068 | int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2], | |
1069 | -dstStride[3] }; | |
1070 | ||
1071 | src2[0] += (srcSliceH - 1) * srcStride[0]; | |
1072 | if (!usePal(c->srcFormat)) | |
1073 | src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1]; | |
1074 | src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2]; | |
1075 | src2[3] += (srcSliceH - 1) * srcStride[3]; | |
1076 | dst2[0] += ( c->dstH - 1) * dstStride[0]; | |
1077 | dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1]; | |
1078 | dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2]; | |
1079 | dst2[3] += ( c->dstH - 1) * dstStride[3]; | |
1080 | ||
1081 | reset_ptr(src2, c->srcFormat); | |
1082 | reset_ptr((void*)dst2, c->dstFormat); | |
1083 | ||
1084 | /* reset slice direction at end of frame */ | |
1085 | if (!srcSliceY) | |
1086 | c->sliceDir = 0; | |
1087 | ||
1088 | ret = c->swscale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, | |
1089 | srcSliceH, dst2, dstStride2); | |
1090 | } | |
1091 | ||
1092 | ||
1093 | if (c->dstXYZ && !(c->srcXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) { | |
1094 | /* replace on the same data */ | |
1095 | rgb48Toxyz12(c, (uint16_t*)dst2[0], (const uint16_t*)dst2[0], dstStride[0]/2, ret); | |
1096 | } | |
1097 | ||
1098 | av_free(rgb0_tmp); | |
1099 | return ret; | |
1100 | } | |
1101 |