2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of Libav.
6 * Libav 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.
11 * Libav 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.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
28 #include "swscale_internal.h"
30 #include "libavutil/intreadwrite.h"
31 #include "libavutil/cpu.h"
32 #include "libavutil/avutil.h"
33 #include "libavutil/mathematics.h"
34 #include "libavutil/bswap.h"
35 #include "libavutil/pixdesc.h"
37 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_1)[8][8] = {
38 { 0, 1, 0, 1, 0, 1, 0, 1,},
39 { 1, 0, 1, 0, 1, 0, 1, 0,},
40 { 0, 1, 0, 1, 0, 1, 0, 1,},
41 { 1, 0, 1, 0, 1, 0, 1, 0,},
42 { 0, 1, 0, 1, 0, 1, 0, 1,},
43 { 1, 0, 1, 0, 1, 0, 1, 0,},
44 { 0, 1, 0, 1, 0, 1, 0, 1,},
45 { 1, 0, 1, 0, 1, 0, 1, 0,},
47 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_3)[8][8] = {
48 { 1, 2, 1, 2, 1, 2, 1, 2,},
49 { 3, 0, 3, 0, 3, 0, 3, 0,},
50 { 1, 2, 1, 2, 1, 2, 1, 2,},
51 { 3, 0, 3, 0, 3, 0, 3, 0,},
52 { 1, 2, 1, 2, 1, 2, 1, 2,},
53 { 3, 0, 3, 0, 3, 0, 3, 0,},
54 { 1, 2, 1, 2, 1, 2, 1, 2,},
55 { 3, 0, 3, 0, 3, 0, 3, 0,},
57 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_64)[8][8] = {
58 { 18, 34, 30, 46, 17, 33, 29, 45,},
59 { 50, 2, 62, 14, 49, 1, 61, 13,},
60 { 26, 42, 22, 38, 25, 41, 21, 37,},
61 { 58, 10, 54, 6, 57, 9, 53, 5,},
62 { 16, 32, 28, 44, 19, 35, 31, 47,},
63 { 48, 0, 60, 12, 51, 3, 63, 15,},
64 { 24, 40, 20, 36, 27, 43, 23, 39,},
65 { 56, 8, 52, 4, 59, 11, 55, 7,},
67 extern const uint8_t dither_8x8_128[8][8];
68 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_256)[8][8] = {
69 { 72, 136, 120, 184, 68, 132, 116, 180,},
70 { 200, 8, 248, 56, 196, 4, 244, 52,},
71 { 104, 168, 88, 152, 100, 164, 84, 148,},
72 { 232, 40, 216, 24, 228, 36, 212, 20,},
73 { 64, 128, 102, 176, 76, 140, 124, 188,},
74 { 192, 0, 240, 48, 204, 12, 252, 60,},
75 { 96, 160, 80, 144, 108, 172, 92, 156,},
76 { 224, 32, 208, 16, 236, 44, 220, 28,},
79 #define RGB2YUV_SHIFT 15
80 #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
81 #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
82 #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
83 #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
84 #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
85 #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
86 #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
87 #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
88 #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
90 static void fillPlane(uint8_t *plane, int stride, int width, int height, int y,
94 uint8_t *ptr = plane + stride * y;
95 for (i = 0; i < height; i++) {
96 memset(ptr, val, width);
101 static void copyPlane(const uint8_t *src, int srcStride,
102 int srcSliceY, int srcSliceH, int width,
103 uint8_t *dst, int dstStride)
105 dst += dstStride * srcSliceY;
106 if (dstStride == srcStride && srcStride > 0) {
107 memcpy(dst, src, srcSliceH * dstStride);
110 for (i = 0; i < srcSliceH; i++) {
111 memcpy(dst, src, width);
118 static int planarToNv12Wrapper(SwsContext *c, const uint8_t *src[],
119 int srcStride[], int srcSliceY,
120 int srcSliceH, uint8_t *dstParam[],
123 uint8_t *dst = dstParam[1] + dstStride[1] * srcSliceY / 2;
125 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
126 dstParam[0], dstStride[0]);
128 if (c->dstFormat == PIX_FMT_NV12)
129 interleaveBytes(src[1], src[2], dst, c->srcW / 2, srcSliceH / 2,
130 srcStride[1], srcStride[2], dstStride[0]);
132 interleaveBytes(src[2], src[1], dst, c->srcW / 2, srcSliceH / 2,
133 srcStride[2], srcStride[1], dstStride[0]);
138 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t *src[],
139 int srcStride[], int srcSliceY, int srcSliceH,
140 uint8_t *dstParam[], int dstStride[])
142 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
144 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
145 srcStride[1], dstStride[0]);
150 static int planarToUyvyWrapper(SwsContext *c, const uint8_t *src[],
151 int srcStride[], int srcSliceY, int srcSliceH,
152 uint8_t *dstParam[], int dstStride[])
154 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
156 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
157 srcStride[1], dstStride[0]);
162 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t *src[],
163 int srcStride[], int srcSliceY, int srcSliceH,
164 uint8_t *dstParam[], int dstStride[])
166 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
168 yuv422ptoyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
169 srcStride[1], dstStride[0]);
174 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t *src[],
175 int srcStride[], int srcSliceY, int srcSliceH,
176 uint8_t *dstParam[], int dstStride[])
178 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
180 yuv422ptouyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
181 srcStride[1], dstStride[0]);
186 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t *src[],
187 int srcStride[], int srcSliceY, int srcSliceH,
188 uint8_t *dstParam[], int dstStride[])
190 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
191 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;
192 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;
194 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
195 dstStride[1], srcStride[0]);
198 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
203 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t *src[],
204 int srcStride[], int srcSliceY, int srcSliceH,
205 uint8_t *dstParam[], int dstStride[])
207 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
208 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;
209 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;
211 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
212 dstStride[1], srcStride[0]);
217 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t *src[],
218 int srcStride[], int srcSliceY, int srcSliceH,
219 uint8_t *dstParam[], int dstStride[])
221 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
222 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;
223 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;
225 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
226 dstStride[1], srcStride[0]);
229 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
234 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t *src[],
235 int srcStride[], int srcSliceY, int srcSliceH,
236 uint8_t *dstParam[], int dstStride[])
238 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
239 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;
240 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;
242 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
243 dstStride[1], srcStride[0]);
248 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels,
249 const uint8_t *palette)
252 for (i = 0; i < num_pixels; i++)
253 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | (src[(i << 1) + 1] << 24);
256 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, int num_pixels,
257 const uint8_t *palette)
261 for (i = 0; i < num_pixels; i++)
262 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | src[(i << 1) + 1];
265 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels,
266 const uint8_t *palette)
270 for (i = 0; i < num_pixels; i++) {
272 dst[0] = palette[src[i << 1] * 4 + 0];
273 dst[1] = palette[src[i << 1] * 4 + 1];
274 dst[2] = palette[src[i << 1] * 4 + 2];
279 static int packed_16bpc_bswap(SwsContext *c, const uint8_t *src[],
280 int srcStride[], int srcSliceY, int srcSliceH,
281 uint8_t *dst[], int dstStride[])
284 int srcstr = srcStride[0] >> 1;
285 int dststr = dstStride[0] >> 1;
286 uint16_t *dstPtr = (uint16_t *) dst[0];
287 const uint16_t *srcPtr = (const uint16_t *) src[0];
288 int min_stride = FFMIN(srcstr, dststr);
290 for (i = 0; i < srcSliceH; i++) {
291 for (j = 0; j < min_stride; j++) {
292 dstPtr[j] = av_bswap16(srcPtr[j]);
301 static int palToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],
302 int srcSliceY, int srcSliceH, uint8_t *dst[],
305 const enum PixelFormat srcFormat = c->srcFormat;
306 const enum PixelFormat dstFormat = c->dstFormat;
307 void (*conv)(const uint8_t *src, uint8_t *dst, int num_pixels,
308 const uint8_t *palette) = NULL;
310 uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;
311 const uint8_t *srcPtr = src[0];
313 if (srcFormat == PIX_FMT_Y400A) {
315 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
316 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
317 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
318 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
319 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
320 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
322 } else if (usePal(srcFormat)) {
324 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
325 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
326 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
327 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
328 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
329 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
334 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
335 sws_format_name(srcFormat), sws_format_name(dstFormat));
337 for (i = 0; i < srcSliceH; i++) {
338 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
339 srcPtr += srcStride[0];
340 dstPtr += dstStride[0];
347 #define isRGBA32(x) ( \
348 (x) == PIX_FMT_ARGB \
349 || (x) == PIX_FMT_RGBA \
350 || (x) == PIX_FMT_BGRA \
351 || (x) == PIX_FMT_ABGR \
354 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
355 typedef void (* rgbConvFn) (const uint8_t *, uint8_t *, int);
356 static rgbConvFn findRgbConvFn(SwsContext *c)
358 const enum PixelFormat srcFormat = c->srcFormat;
359 const enum PixelFormat dstFormat = c->dstFormat;
360 const int srcId = c->srcFormatBpp;
361 const int dstId = c->dstFormatBpp;
362 rgbConvFn conv = NULL;
364 #define IS_NOT_NE(bpp, fmt) \
365 (((bpp + 7) >> 3) == 2 && \
366 (!(av_pix_fmt_descriptors[fmt].flags & PIX_FMT_BE) != !HAVE_BIGENDIAN))
368 /* if this is non-native rgb444/555/565, don't handle it here. */
369 if (IS_NOT_NE(srcId, srcFormat) || IS_NOT_NE(dstId, dstFormat))
372 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
374 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
375 if ( CONV_IS(ABGR, RGBA)
376 || CONV_IS(ARGB, BGRA)
377 || CONV_IS(BGRA, ARGB)
378 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
379 else if (CONV_IS(ABGR, ARGB)
380 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
381 else if (CONV_IS(ABGR, BGRA)
382 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
383 else if (CONV_IS(BGRA, RGBA)
384 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
385 else if (CONV_IS(BGRA, ABGR)
386 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
389 if ((isBGRinInt(srcFormat) && isBGRinInt(dstFormat)) ||
390 (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
391 switch (srcId | (dstId << 16)) {
392 case 0x000F000C: conv = rgb12to15; break;
393 case 0x000F0010: conv = rgb16to15; break;
394 case 0x000F0018: conv = rgb24to15; break;
395 case 0x000F0020: conv = rgb32to15; break;
396 case 0x0010000F: conv = rgb15to16; break;
397 case 0x00100018: conv = rgb24to16; break;
398 case 0x00100020: conv = rgb32to16; break;
399 case 0x0018000F: conv = rgb15to24; break;
400 case 0x00180010: conv = rgb16to24; break;
401 case 0x00180020: conv = rgb32to24; break;
402 case 0x0020000F: conv = rgb15to32; break;
403 case 0x00200010: conv = rgb16to32; break;
404 case 0x00200018: conv = rgb24to32; break;
406 } else if ((isBGRinInt(srcFormat) && isRGBinInt(dstFormat)) ||
407 (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
408 switch (srcId | (dstId << 16)) {
409 case 0x000C000C: conv = rgb12tobgr12; break;
410 case 0x000F000F: conv = rgb15tobgr15; break;
411 case 0x000F0010: conv = rgb16tobgr15; break;
412 case 0x000F0018: conv = rgb24tobgr15; break;
413 case 0x000F0020: conv = rgb32tobgr15; break;
414 case 0x0010000F: conv = rgb15tobgr16; break;
415 case 0x00100010: conv = rgb16tobgr16; break;
416 case 0x00100018: conv = rgb24tobgr16; break;
417 case 0x00100020: conv = rgb32tobgr16; break;
418 case 0x0018000F: conv = rgb15tobgr24; break;
419 case 0x00180010: conv = rgb16tobgr24; break;
420 case 0x00180018: conv = rgb24tobgr24; break;
421 case 0x00180020: conv = rgb32tobgr24; break;
422 case 0x0020000F: conv = rgb15tobgr32; break;
423 case 0x00200010: conv = rgb16tobgr32; break;
424 case 0x00200018: conv = rgb24tobgr32; break;
431 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
432 static int rgbToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],
433 int srcSliceY, int srcSliceH, uint8_t *dst[],
437 const enum PixelFormat srcFormat = c->srcFormat;
438 const enum PixelFormat dstFormat = c->dstFormat;
439 const int srcBpp = (c->srcFormatBpp + 7) >> 3;
440 const int dstBpp = (c->dstFormatBpp + 7) >> 3;
441 rgbConvFn conv = findRgbConvFn(c);
444 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
445 sws_format_name(srcFormat), sws_format_name(dstFormat));
447 const uint8_t *srcPtr = src[0];
448 uint8_t *dstPtr = dst[0];
449 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) &&
450 !isRGBA32(dstFormat))
451 srcPtr += ALT32_CORR;
453 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) &&
454 !isRGBA32(srcFormat))
455 dstPtr += ALT32_CORR;
457 if (dstStride[0] * srcBpp == srcStride[0] * dstBpp && srcStride[0] > 0 &&
458 !(srcStride[0] % srcBpp))
459 conv(srcPtr, dstPtr + dstStride[0] * srcSliceY,
460 srcSliceH * srcStride[0]);
463 dstPtr += dstStride[0] * srcSliceY;
465 for (i = 0; i < srcSliceH; i++) {
466 conv(srcPtr, dstPtr, c->srcW * srcBpp);
467 srcPtr += srcStride[0];
468 dstPtr += dstStride[0];
475 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t *src[],
476 int srcStride[], int srcSliceY, int srcSliceH,
477 uint8_t *dst[], int dstStride[])
481 dst[0] + srcSliceY * dstStride[0],
482 dst[1] + (srcSliceY >> 1) * dstStride[1],
483 dst[2] + (srcSliceY >> 1) * dstStride[2],
485 dstStride[0], dstStride[1], srcStride[0]);
487 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
491 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t *src[],
492 int srcStride[], int srcSliceY, int srcSliceH,
493 uint8_t *dst[], int dstStride[])
495 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
496 dst[0], dstStride[0]);
498 planar2x(src[1], dst[1] + dstStride[1] * (srcSliceY >> 1), c->chrSrcW,
499 srcSliceH >> 2, srcStride[1], dstStride[1]);
500 planar2x(src[2], dst[2] + dstStride[2] * (srcSliceY >> 1), c->chrSrcW,
501 srcSliceH >> 2, srcStride[2], dstStride[2]);
503 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
507 /* unscaled copy like stuff (assumes nearly identical formats) */
508 static int packedCopyWrapper(SwsContext *c, const uint8_t *src[],
509 int srcStride[], int srcSliceY, int srcSliceH,
510 uint8_t *dst[], int dstStride[])
512 if (dstStride[0] == srcStride[0] && srcStride[0] > 0)
513 memcpy(dst[0] + dstStride[0] * srcSliceY, src[0], srcSliceH * dstStride[0]);
516 const uint8_t *srcPtr = src[0];
517 uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;
520 /* universal length finder */
521 while (length + c->srcW <= FFABS(dstStride[0]) &&
522 length + c->srcW <= FFABS(srcStride[0]))
526 for (i = 0; i < srcSliceH; i++) {
527 memcpy(dstPtr, srcPtr, length);
528 srcPtr += srcStride[0];
529 dstPtr += dstStride[0];
535 #define clip9(x) av_clip_uintp2(x, 9)
536 #define clip10(x) av_clip_uintp2(x, 10)
537 #define DITHER_COPY(dst, dstStride, wfunc, src, srcStride, rfunc, dithers, shift, clip) \
538 for (i = 0; i < height; i++) { \
539 const uint8_t *dither = dithers[i & 7]; \
540 for (j = 0; j < length - 7; j += 8) { \
541 wfunc(&dst[j + 0], clip((rfunc(&src[j + 0]) + dither[0]) >> shift)); \
542 wfunc(&dst[j + 1], clip((rfunc(&src[j + 1]) + dither[1]) >> shift)); \
543 wfunc(&dst[j + 2], clip((rfunc(&src[j + 2]) + dither[2]) >> shift)); \
544 wfunc(&dst[j + 3], clip((rfunc(&src[j + 3]) + dither[3]) >> shift)); \
545 wfunc(&dst[j + 4], clip((rfunc(&src[j + 4]) + dither[4]) >> shift)); \
546 wfunc(&dst[j + 5], clip((rfunc(&src[j + 5]) + dither[5]) >> shift)); \
547 wfunc(&dst[j + 6], clip((rfunc(&src[j + 6]) + dither[6]) >> shift)); \
548 wfunc(&dst[j + 7], clip((rfunc(&src[j + 7]) + dither[7]) >> shift)); \
550 for (; j < length; j++) \
551 wfunc(&dst[j], (rfunc(&src[j]) + dither[j & 7]) >> shift); \
556 static int planarCopyWrapper(SwsContext *c, const uint8_t *src[],
557 int srcStride[], int srcSliceY, int srcSliceH,
558 uint8_t *dst[], int dstStride[])
561 for (plane = 0; plane < 4; plane++) {
562 int length = (plane == 0 || plane == 3) ? c->srcW : -((-c->srcW ) >> c->chrDstHSubSample);
563 int y = (plane == 0 || plane == 3) ? srcSliceY: -((-srcSliceY) >> c->chrDstVSubSample);
564 int height = (plane == 0 || plane == 3) ? srcSliceH: -((-srcSliceH) >> c->chrDstVSubSample);
565 const uint8_t *srcPtr = src[plane];
566 uint8_t *dstPtr = dst[plane] + dstStride[plane] * y;
570 // ignore palette for GRAY8
571 if (plane == 1 && !dst[2]) continue;
572 if (!src[plane] || (plane == 1 && !src[2])) {
573 if (is16BPS(c->dstFormat))
575 fillPlane(dst[plane], dstStride[plane], length, height, y,
576 (plane == 3) ? 255 : 128);
578 if (is9_OR_10BPS(c->srcFormat)) {
579 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1 + 1;
580 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1 + 1;
581 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
583 if (is16BPS(c->dstFormat)) {
584 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
585 #define COPY9_OR_10TO16(rfunc, wfunc) \
586 for (i = 0; i < height; i++) { \
587 for (j = 0; j < length; j++) { \
588 int srcpx = rfunc(&srcPtr2[j]); \
589 wfunc(&dstPtr2[j], (srcpx << (16 - src_depth)) | (srcpx >> (2 * src_depth - 16))); \
591 dstPtr2 += dstStride[plane] / 2; \
592 srcPtr2 += srcStride[plane] / 2; \
594 if (isBE(c->dstFormat)) {
595 if (isBE(c->srcFormat)) {
596 COPY9_OR_10TO16(AV_RB16, AV_WB16);
598 COPY9_OR_10TO16(AV_RL16, AV_WB16);
601 if (isBE(c->srcFormat)) {
602 COPY9_OR_10TO16(AV_RB16, AV_WL16);
604 COPY9_OR_10TO16(AV_RL16, AV_WL16);
607 } else if (is9_OR_10BPS(c->dstFormat)) {
608 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
609 #define COPY9_OR_10TO9_OR_10(loop) \
610 for (i = 0; i < height; i++) { \
611 for (j = 0; j < length; j++) { \
614 dstPtr2 += dstStride[plane] / 2; \
615 srcPtr2 += srcStride[plane] / 2; \
617 #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
618 if (dst_depth > src_depth) { \
619 COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
620 wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
621 } else if (dst_depth < src_depth) { \
622 DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
623 srcPtr2, srcStride[plane] / 2, rfunc, \
624 dither_8x8_1, 1, clip9); \
626 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
628 if (isBE(c->dstFormat)) {
629 if (isBE(c->srcFormat)) {
630 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
632 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
635 if (isBE(c->srcFormat)) {
636 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
638 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
642 #define W8(a, b) { *(a) = (b); }
643 #define COPY9_OR_10TO8(rfunc) \
644 if (src_depth == 9) { \
645 DITHER_COPY(dstPtr, dstStride[plane], W8, \
646 srcPtr2, srcStride[plane] / 2, rfunc, \
647 dither_8x8_1, 1, av_clip_uint8); \
649 DITHER_COPY(dstPtr, dstStride[plane], W8, \
650 srcPtr2, srcStride[plane] / 2, rfunc, \
651 dither_8x8_3, 2, av_clip_uint8); \
653 if (isBE(c->srcFormat)) {
654 COPY9_OR_10TO8(AV_RB16);
656 COPY9_OR_10TO8(AV_RL16);
659 } else if (is9_OR_10BPS(c->dstFormat)) {
660 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1 + 1;
661 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
663 if (is16BPS(c->srcFormat)) {
664 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
665 #define COPY16TO9_OR_10(rfunc, wfunc) \
666 if (dst_depth == 9) { \
667 DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
668 srcPtr2, srcStride[plane] / 2, rfunc, \
669 dither_8x8_128, 7, clip9); \
671 DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
672 srcPtr2, srcStride[plane] / 2, rfunc, \
673 dither_8x8_64, 6, clip10); \
675 if (isBE(c->dstFormat)) {
676 if (isBE(c->srcFormat)) {
677 COPY16TO9_OR_10(AV_RB16, AV_WB16);
679 COPY16TO9_OR_10(AV_RL16, AV_WB16);
682 if (isBE(c->srcFormat)) {
683 COPY16TO9_OR_10(AV_RB16, AV_WL16);
685 COPY16TO9_OR_10(AV_RL16, AV_WL16);
689 #define COPY8TO9_OR_10(wfunc) \
690 for (i = 0; i < height; i++) { \
691 for (j = 0; j < length; j++) { \
692 const int srcpx = srcPtr[j]; \
693 wfunc(&dstPtr2[j], (srcpx << (dst_depth - 8)) | (srcpx >> (16 - dst_depth))); \
695 dstPtr2 += dstStride[plane] / 2; \
696 srcPtr += srcStride[plane]; \
698 if (isBE(c->dstFormat)) {
699 COPY8TO9_OR_10(AV_WB16);
701 COPY8TO9_OR_10(AV_WL16);
704 } else if (is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
705 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
706 #define COPY16TO8(rfunc) \
707 DITHER_COPY(dstPtr, dstStride[plane], W8, \
708 srcPtr2, srcStride[plane] / 2, rfunc, \
709 dither_8x8_256, 8, av_clip_uint8);
710 if (isBE(c->srcFormat)) {
715 } else if (!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
716 for (i = 0; i < height; i++) {
717 for (j = 0; j < length; j++) {
718 dstPtr[ j << 1 ] = srcPtr[j];
719 dstPtr[(j << 1) + 1] = srcPtr[j];
721 srcPtr += srcStride[plane];
722 dstPtr += dstStride[plane];
724 } else if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat) &&
725 isBE(c->srcFormat) != isBE(c->dstFormat)) {
727 for (i = 0; i < height; i++) {
728 for (j = 0; j < length; j++)
729 ((uint16_t *) dstPtr)[j] = av_bswap16(((const uint16_t *) srcPtr)[j]);
730 srcPtr += srcStride[plane];
731 dstPtr += dstStride[plane];
733 } else if (dstStride[plane] == srcStride[plane] &&
734 srcStride[plane] > 0 && srcStride[plane] == length) {
735 memcpy(dst[plane] + dstStride[plane] * y, src[plane],
736 height * dstStride[plane]);
738 if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
740 else if (!av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1)
741 length >>= 3; // monowhite/black
742 for (i = 0; i < height; i++) {
743 memcpy(dstPtr, srcPtr, length);
744 srcPtr += srcStride[plane];
745 dstPtr += dstStride[plane];
754 #define IS_DIFFERENT_ENDIANESS(src_fmt, dst_fmt, pix_fmt) \
755 ((src_fmt == pix_fmt ## BE && dst_fmt == pix_fmt ## LE) || \
756 (src_fmt == pix_fmt ## LE && dst_fmt == pix_fmt ## BE))
759 void ff_get_unscaled_swscale(SwsContext *c)
761 const enum PixelFormat srcFormat = c->srcFormat;
762 const enum PixelFormat dstFormat = c->dstFormat;
763 const int flags = c->flags;
764 const int dstH = c->dstH;
767 needsDither = isAnyRGB(dstFormat) &&
768 c->dstFormatBpp < 24 &&
769 (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
772 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) &&
773 (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
774 c->swScale = planarToNv12Wrapper;
777 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUV422P ||
778 srcFormat == PIX_FMT_YUVA420P) && isAnyRGB(dstFormat) &&
779 !(flags & SWS_ACCURATE_RND) && !(dstH & 1)) {
780 c->swScale = ff_yuv2rgb_get_func_ptr(c);
783 if (srcFormat == PIX_FMT_YUV410P &&
784 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P) &&
785 !(flags & SWS_BITEXACT)) {
786 c->swScale = yvu9ToYv12Wrapper;
790 if (srcFormat == PIX_FMT_BGR24 &&
791 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P) &&
792 !(flags & SWS_ACCURATE_RND))
793 c->swScale = bgr24ToYv12Wrapper;
795 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
796 if (isAnyRGB(srcFormat) && isAnyRGB(dstFormat) && findRgbConvFn(c)
797 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
798 c->swScale= rgbToRgbWrapper;
800 /* bswap 16 bits per pixel/component packed formats */
801 if (IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR444) ||
802 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR48) ||
803 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR555) ||
804 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR565) ||
805 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_GRAY16) ||
806 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB444) ||
807 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB48) ||
808 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB555) ||
809 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB565))
810 c->swScale = packed_16bpc_bswap;
812 if ((usePal(srcFormat) && (
813 dstFormat == PIX_FMT_RGB32 ||
814 dstFormat == PIX_FMT_RGB32_1 ||
815 dstFormat == PIX_FMT_RGB24 ||
816 dstFormat == PIX_FMT_BGR32 ||
817 dstFormat == PIX_FMT_BGR32_1 ||
818 dstFormat == PIX_FMT_BGR24)))
819 c->swScale = palToRgbWrapper;
821 if (srcFormat == PIX_FMT_YUV422P) {
822 if (dstFormat == PIX_FMT_YUYV422)
823 c->swScale = yuv422pToYuy2Wrapper;
824 else if (dstFormat == PIX_FMT_UYVY422)
825 c->swScale = yuv422pToUyvyWrapper;
828 /* LQ converters if -sws 0 or -sws 4*/
829 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
831 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
832 if (dstFormat == PIX_FMT_YUYV422)
833 c->swScale = planarToYuy2Wrapper;
834 else if (dstFormat == PIX_FMT_UYVY422)
835 c->swScale = planarToUyvyWrapper;
838 if (srcFormat == PIX_FMT_YUYV422 &&
839 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
840 c->swScale = yuyvToYuv420Wrapper;
841 if (srcFormat == PIX_FMT_UYVY422 &&
842 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
843 c->swScale = uyvyToYuv420Wrapper;
844 if (srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
845 c->swScale = yuyvToYuv422Wrapper;
846 if (srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
847 c->swScale = uyvyToYuv422Wrapper;
850 if ( srcFormat == dstFormat ||
851 (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P) ||
852 (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P) ||
853 (isPlanarYUV(srcFormat) && isGray(dstFormat)) ||
854 (isPlanarYUV(dstFormat) && isGray(srcFormat)) ||
855 (isGray(dstFormat) && isGray(srcFormat)) ||
856 (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat) &&
857 c->chrDstHSubSample == c->chrSrcHSubSample &&
858 c->chrDstVSubSample == c->chrSrcVSubSample &&
859 dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21 &&
860 srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
862 if (isPacked(c->srcFormat))
863 c->swScale = packedCopyWrapper;
864 else /* Planar YUV or gray */
865 c->swScale = planarCopyWrapper;
869 ff_bfin_get_unscaled_swscale(c);
871 ff_swscale_get_unscaled_altivec(c);
874 static void reset_ptr(const uint8_t *src[], int format)
876 if (!isALPHA(format))
878 if (!isPlanar(format)) {
879 src[3] = src[2] = NULL;
886 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
887 const int linesizes[4])
889 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
892 for (i = 0; i < 4; i++) {
893 int plane = desc->comp[i].plane;
894 if (!data[plane] || !linesizes[plane])
902 * swscale wrapper, so we don't need to export the SwsContext.
903 * Assumes planar YUV to be in YUV order instead of YVU.
905 int attribute_align_arg sws_scale(struct SwsContext *c,
906 const uint8_t * const srcSlice[],
907 const int srcStride[], int srcSliceY,
908 int srcSliceH, uint8_t *const dst[],
909 const int dstStride[])
912 const uint8_t *src2[4] = { srcSlice[0], srcSlice[1], srcSlice[2], srcSlice[3] };
913 uint8_t *dst2[4] = { dst[0], dst[1], dst[2], dst[3] };
915 // do not mess up sliceDir if we have a "trailing" 0-size slice
919 if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {
920 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
923 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
924 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
928 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
929 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
932 if (c->sliceDir == 0) {
933 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
936 if (usePal(c->srcFormat)) {
937 for (i = 0; i < 256; i++) {
938 int p, r, g, b, y, u, v;
939 if (c->srcFormat == PIX_FMT_PAL8) {
940 p = ((const uint32_t *)(srcSlice[1]))[i];
941 r = (p >> 16) & 0xFF;
944 } else if (c->srcFormat == PIX_FMT_RGB8) {
946 g = ((i >> 2) & 7) * 36;
948 } else if (c->srcFormat == PIX_FMT_BGR8) {
950 g = ((i >> 3) & 7) * 36;
952 } else if (c->srcFormat == PIX_FMT_RGB4_BYTE) {
953 r = ( i >> 3 ) * 255;
954 g = ((i >> 1) & 3) * 85;
956 } else if (c->srcFormat == PIX_FMT_GRAY8 ||
957 c->srcFormat == PIX_FMT_Y400A) {
960 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
961 b = ( i >> 3 ) * 255;
962 g = ((i >> 1) & 3) * 85;
965 y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
966 u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
967 v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
968 c->pal_yuv[i] = y + (u << 8) + (v << 16);
970 switch (c->dstFormat) {
975 c->pal_rgb[i] = r + (g << 8) + (b << 16);
977 case PIX_FMT_BGR32_1:
981 c->pal_rgb[i] = (r + (g << 8) + (b << 16)) << 8;
983 case PIX_FMT_RGB32_1:
987 c->pal_rgb[i] = (b + (g << 8) + (r << 16)) << 8;
994 c->pal_rgb[i] = b + (g << 8) + (r << 16);
999 // copy strides, so they can safely be modified
1000 if (c->sliceDir == 1) {
1001 // slices go from top to bottom
1002 int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],
1004 int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],
1007 reset_ptr(src2, c->srcFormat);
1008 reset_ptr((const uint8_t **) dst2, c->dstFormat);
1010 /* reset slice direction at end of frame */
1011 if (srcSliceY + srcSliceH == c->srcH)
1014 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,
1017 // slices go from bottom to top => we flip the image internally
1018 int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],
1020 int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],
1023 src2[0] += (srcSliceH - 1) * srcStride[0];
1024 if (!usePal(c->srcFormat))
1025 src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];
1026 src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];
1027 src2[3] += (srcSliceH - 1) * srcStride[3];
1028 dst2[0] += ( c->dstH - 1) * dstStride[0];
1029 dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];
1030 dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];
1031 dst2[3] += ( c->dstH - 1) * dstStride[3];
1033 reset_ptr(src2, c->srcFormat);
1034 reset_ptr((const uint8_t **) dst2, c->dstFormat);
1036 /* reset slice direction at end of frame */
1040 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,
1041 srcSliceH, dst2, dstStride2);
1045 /* Convert the palette to the same packed 32-bit format as the palette */
1046 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst,
1047 int num_pixels, const uint8_t *palette)
1051 for (i = 0; i < num_pixels; i++)
1052 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
1055 /* Palette format: ABCD -> dst format: ABC */
1056 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst,
1057 int num_pixels, const uint8_t *palette)
1061 for (i = 0; i < num_pixels; i++) {
1063 dst[0] = palette[src[i] * 4 + 0];
1064 dst[1] = palette[src[i] * 4 + 1];
1065 dst[2] = palette[src[i] * 4 + 2];