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#ifndef SWSCALE_SWSCALE_INTERNAL_H 22#define SWSCALE_SWSCALE_INTERNAL_H 23 24#include "config.h" 25 26#if HAVE_ALTIVEC_H 27#include <altivec.h> 28#endif 29 30#include "libavutil/avassert.h" 31#include "libavutil/avutil.h" 32#include "libavutil/common.h" 33#include "libavutil/intreadwrite.h" 34#include "libavutil/log.h" 35#include "libavutil/pixfmt.h" 36#include "libavutil/pixdesc.h" 37 38#define STR(s) AV_TOSTRING(s) // AV_STRINGIFY is too long 39 40#define YUVRGB_TABLE_HEADROOM 128 41 42#define MAX_FILTER_SIZE SWS_MAX_FILTER_SIZE 43 44#define DITHER1XBPP 45 46#if HAVE_BIGENDIAN 47#define ALT32_CORR (-1) 48#else 49#define ALT32_CORR 1 50#endif 51 52#if ARCH_X86_64 53# define APCK_PTR2 8 54# define APCK_COEF 16 55# define APCK_SIZE 24 56#else 57# define APCK_PTR2 4 58# define APCK_COEF 8 59# define APCK_SIZE 16 60#endif 61 62struct SwsContext; 63 64typedef enum SwsDither { 65 SWS_DITHER_NONE = 0, 66 SWS_DITHER_AUTO, 67 SWS_DITHER_BAYER, 68 SWS_DITHER_ED, 69 SWS_DITHER_A_DITHER, 70 SWS_DITHER_X_DITHER, 71 NB_SWS_DITHER, 72} SwsDither; 73 74typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t *src[], 75 int srcStride[], int srcSliceY, int srcSliceH, 76 uint8_t *dst[], int dstStride[]); 77 78/** 79 * Write one line of horizontally scaled data to planar output 80 * without any additional vertical scaling (or point-scaling). 81 * 82 * @param src scaled source data, 15bit for 8-10bit output, 83 * 19-bit for 16bit output (in int32_t) 84 * @param dest pointer to the output plane. For >8bit 85 * output, this is in uint16_t 86 * @param dstW width of destination in pixels 87 * @param dither ordered dither array of type int16_t and size 8 88 * @param offset Dither offset 89 */ 90typedef void (*yuv2planar1_fn)(const int16_t *src, uint8_t *dest, int dstW, 91 const uint8_t *dither, int offset); 92 93/** 94 * Write one line of horizontally scaled data to planar output 95 * with multi-point vertical scaling between input pixels. 96 * 97 * @param filter vertical luma/alpha scaling coefficients, 12bit [0,4096] 98 * @param src scaled luma (Y) or alpha (A) source data, 15bit for 8-10bit output, 99 * 19-bit for 16bit output (in int32_t) 100 * @param filterSize number of vertical input lines to scale 101 * @param dest pointer to output plane. For >8bit 102 * output, this is in uint16_t 103 * @param dstW width of destination pixels 104 * @param offset Dither offset 105 */ 106typedef void (*yuv2planarX_fn)(const int16_t *filter, int filterSize, 107 const int16_t **src, uint8_t *dest, int dstW, 108 const uint8_t *dither, int offset); 109 110/** 111 * Write one line of horizontally scaled chroma to interleaved output 112 * with multi-point vertical scaling between input pixels. 113 * 114 * @param c SWS scaling context 115 * @param chrFilter vertical chroma scaling coefficients, 12bit [0,4096] 116 * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output, 117 * 19-bit for 16bit output (in int32_t) 118 * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output, 119 * 19-bit for 16bit output (in int32_t) 120 * @param chrFilterSize number of vertical chroma input lines to scale 121 * @param dest pointer to the output plane. For >8bit 122 * output, this is in uint16_t 123 * @param dstW width of chroma planes 124 */ 125typedef void (*yuv2interleavedX_fn)(struct SwsContext *c, 126 const int16_t *chrFilter, 127 int chrFilterSize, 128 const int16_t **chrUSrc, 129 const int16_t **chrVSrc, 130 uint8_t *dest, int dstW); 131 132/** 133 * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB 134 * output without any additional vertical scaling (or point-scaling). Note 135 * that this function may do chroma scaling, see the "uvalpha" argument. 136 * 137 * @param c SWS scaling context 138 * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output, 139 * 19-bit for 16bit output (in int32_t) 140 * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output, 141 * 19-bit for 16bit output (in int32_t) 142 * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output, 143 * 19-bit for 16bit output (in int32_t) 144 * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output, 145 * 19-bit for 16bit output (in int32_t) 146 * @param dest pointer to the output plane. For 16bit output, this is 147 * uint16_t 148 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels 149 * to write into dest[] 150 * @param uvalpha chroma scaling coefficient for the second line of chroma 151 * pixels, either 2048 or 0. If 0, one chroma input is used 152 * for 2 output pixels (or if the SWS_FLAG_FULL_CHR_INT flag 153 * is set, it generates 1 output pixel). If 2048, two chroma 154 * input pixels should be averaged for 2 output pixels (this 155 * only happens if SWS_FLAG_FULL_CHR_INT is not set) 156 * @param y vertical line number for this output. This does not need 157 * to be used to calculate the offset in the destination, 158 * but can be used to generate comfort noise using dithering 159 * for some output formats. 160 */ 161typedef void (*yuv2packed1_fn)(struct SwsContext *c, const int16_t *lumSrc, 162 const int16_t *chrUSrc[2], 163 const int16_t *chrVSrc[2], 164 const int16_t *alpSrc, uint8_t *dest, 165 int dstW, int uvalpha, int y); 166/** 167 * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB 168 * output by doing bilinear scaling between two input lines. 169 * 170 * @param c SWS scaling context 171 * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output, 172 * 19-bit for 16bit output (in int32_t) 173 * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output, 174 * 19-bit for 16bit output (in int32_t) 175 * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output, 176 * 19-bit for 16bit output (in int32_t) 177 * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output, 178 * 19-bit for 16bit output (in int32_t) 179 * @param dest pointer to the output plane. For 16bit output, this is 180 * uint16_t 181 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels 182 * to write into dest[] 183 * @param yalpha luma/alpha scaling coefficients for the second input line. 184 * The first line's coefficients can be calculated by using 185 * 4096 - yalpha 186 * @param uvalpha chroma scaling coefficient for the second input line. The 187 * first line's coefficients can be calculated by using 188 * 4096 - uvalpha 189 * @param y vertical line number for this output. This does not need 190 * to be used to calculate the offset in the destination, 191 * but can be used to generate comfort noise using dithering 192 * for some output formats. 193 */ 194typedef void (*yuv2packed2_fn)(struct SwsContext *c, const int16_t *lumSrc[2], 195 const int16_t *chrUSrc[2], 196 const int16_t *chrVSrc[2], 197 const int16_t *alpSrc[2], 198 uint8_t *dest, 199 int dstW, int yalpha, int uvalpha, int y); 200/** 201 * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB 202 * output by doing multi-point vertical scaling between input pixels. 203 * 204 * @param c SWS scaling context 205 * @param lumFilter vertical luma/alpha scaling coefficients, 12bit [0,4096] 206 * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output, 207 * 19-bit for 16bit output (in int32_t) 208 * @param lumFilterSize number of vertical luma/alpha input lines to scale 209 * @param chrFilter vertical chroma scaling coefficients, 12bit [0,4096] 210 * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output, 211 * 19-bit for 16bit output (in int32_t) 212 * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output, 213 * 19-bit for 16bit output (in int32_t) 214 * @param chrFilterSize number of vertical chroma input lines to scale 215 * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output, 216 * 19-bit for 16bit output (in int32_t) 217 * @param dest pointer to the output plane. For 16bit output, this is 218 * uint16_t 219 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels 220 * to write into dest[] 221 * @param y vertical line number for this output. This does not need 222 * to be used to calculate the offset in the destination, 223 * but can be used to generate comfort noise using dithering 224 * or some output formats. 225 */ 226typedef void (*yuv2packedX_fn)(struct SwsContext *c, const int16_t *lumFilter, 227 const int16_t **lumSrc, int lumFilterSize, 228 const int16_t *chrFilter, 229 const int16_t **chrUSrc, 230 const int16_t **chrVSrc, int chrFilterSize, 231 const int16_t **alpSrc, uint8_t *dest, 232 int dstW, int y); 233 234/** 235 * Write one line of horizontally scaled Y/U/V/A to YUV/RGB 236 * output by doing multi-point vertical scaling between input pixels. 237 * 238 * @param c SWS scaling context 239 * @param lumFilter vertical luma/alpha scaling coefficients, 12bit [0,4096] 240 * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output, 241 * 19-bit for 16bit output (in int32_t) 242 * @param lumFilterSize number of vertical luma/alpha input lines to scale 243 * @param chrFilter vertical chroma scaling coefficients, 12bit [0,4096] 244 * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output, 245 * 19-bit for 16bit output (in int32_t) 246 * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output, 247 * 19-bit for 16bit output (in int32_t) 248 * @param chrFilterSize number of vertical chroma input lines to scale 249 * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output, 250 * 19-bit for 16bit output (in int32_t) 251 * @param dest pointer to the output planes. For 16bit output, this is 252 * uint16_t 253 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels 254 * to write into dest[] 255 * @param y vertical line number for this output. This does not need 256 * to be used to calculate the offset in the destination, 257 * but can be used to generate comfort noise using dithering 258 * or some output formats. 259 */ 260typedef void (*yuv2anyX_fn)(struct SwsContext *c, const int16_t *lumFilter, 261 const int16_t **lumSrc, int lumFilterSize, 262 const int16_t *chrFilter, 263 const int16_t **chrUSrc, 264 const int16_t **chrVSrc, int chrFilterSize, 265 const int16_t **alpSrc, uint8_t **dest, 266 int dstW, int y); 267 268/* This struct should be aligned on at least a 32-byte boundary. */ 269typedef struct SwsContext { 270 /** 271 * info on struct for av_log 272 */ 273 const AVClass *av_class; 274 275 /** 276 * Note that src, dst, srcStride, dstStride will be copied in the 277 * sws_scale() wrapper so they can be freely modified here. 278 */ 279 SwsFunc swscale; 280 int srcW; ///< Width of source luma/alpha planes. 281 int srcH; ///< Height of source luma/alpha planes. 282 int dstH; ///< Height of destination luma/alpha planes. 283 int chrSrcW; ///< Width of source chroma planes. 284 int chrSrcH; ///< Height of source chroma planes. 285 int chrDstW; ///< Width of destination chroma planes. 286 int chrDstH; ///< Height of destination chroma planes. 287 int lumXInc, chrXInc; 288 int lumYInc, chrYInc; 289 enum AVPixelFormat dstFormat; ///< Destination pixel format. 290 enum AVPixelFormat srcFormat; ///< Source pixel format. 291 int dstFormatBpp; ///< Number of bits per pixel of the destination pixel format. 292 int srcFormatBpp; ///< Number of bits per pixel of the source pixel format. 293 int dstBpc, srcBpc; 294 int chrSrcHSubSample; ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in source image. 295 int chrSrcVSubSample; ///< Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in source image. 296 int chrDstHSubSample; ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in destination image. 297 int chrDstVSubSample; ///< Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in destination image. 298 int vChrDrop; ///< Binary logarithm of extra vertical subsampling factor in source image chroma planes specified by user. 299 int sliceDir; ///< Direction that slices are fed to the scaler (1 = top-to-bottom, -1 = bottom-to-top). 300 double param[2]; ///< Input parameters for scaling algorithms that need them. 301 302 uint32_t pal_yuv[256]; 303 uint32_t pal_rgb[256]; 304 305 /** 306 * @name Scaled horizontal lines ring buffer. 307 * The horizontal scaler keeps just enough scaled lines in a ring buffer 308 * so they may be passed to the vertical scaler. The pointers to the 309 * allocated buffers for each line are duplicated in sequence in the ring 310 * buffer to simplify indexing and avoid wrapping around between lines 311 * inside the vertical scaler code. The wrapping is done before the 312 * vertical scaler is called. 313 */ 314 //@{ 315 int16_t **lumPixBuf; ///< Ring buffer for scaled horizontal luma plane lines to be fed to the vertical scaler. 316 int16_t **chrUPixBuf; ///< Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler. 317 int16_t **chrVPixBuf; ///< Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler. 318 int16_t **alpPixBuf; ///< Ring buffer for scaled horizontal alpha plane lines to be fed to the vertical scaler. 319 int vLumBufSize; ///< Number of vertical luma/alpha lines allocated in the ring buffer. 320 int vChrBufSize; ///< Number of vertical chroma lines allocated in the ring buffer. 321 int lastInLumBuf; ///< Last scaled horizontal luma/alpha line from source in the ring buffer. 322 int lastInChrBuf; ///< Last scaled horizontal chroma line from source in the ring buffer. 323 int lumBufIndex; ///< Index in ring buffer of the last scaled horizontal luma/alpha line from source. 324 int chrBufIndex; ///< Index in ring buffer of the last scaled horizontal chroma line from source. 325 //@} 326 327 uint8_t *formatConvBuffer; 328 329 /** 330 * @name Horizontal and vertical filters. 331 * To better understand the following fields, here is a pseudo-code of 332 * their usage in filtering a horizontal line: 333 * @code 334 * for (i = 0; i < width; i++) { 335 * dst[i] = 0; 336 * for (j = 0; j < filterSize; j++) 337 * dst[i] += src[ filterPos[i] + j ] * filter[ filterSize * i + j ]; 338 * dst[i] >>= FRAC_BITS; // The actual implementation is fixed-point. 339 * } 340 * @endcode 341 */ 342 //@{ 343 int16_t *hLumFilter; ///< Array of horizontal filter coefficients for luma/alpha planes. 344 int16_t *hChrFilter; ///< Array of horizontal filter coefficients for chroma planes. 345 int16_t *vLumFilter; ///< Array of vertical filter coefficients for luma/alpha planes. 346 int16_t *vChrFilter; ///< Array of vertical filter coefficients for chroma planes. 347 int32_t *hLumFilterPos; ///< Array of horizontal filter starting positions for each dst[i] for luma/alpha planes. 348 int32_t *hChrFilterPos; ///< Array of horizontal filter starting positions for each dst[i] for chroma planes. 349 int32_t *vLumFilterPos; ///< Array of vertical filter starting positions for each dst[i] for luma/alpha planes. 350 int32_t *vChrFilterPos; ///< Array of vertical filter starting positions for each dst[i] for chroma planes. 351 int hLumFilterSize; ///< Horizontal filter size for luma/alpha pixels. 352 int hChrFilterSize; ///< Horizontal filter size for chroma pixels. 353 int vLumFilterSize; ///< Vertical filter size for luma/alpha pixels. 354 int vChrFilterSize; ///< Vertical filter size for chroma pixels. 355 //@} 356 357 int lumMmxextFilterCodeSize; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code size for luma/alpha planes. 358 int chrMmxextFilterCodeSize; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code size for chroma planes. 359 uint8_t *lumMmxextFilterCode; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code for luma/alpha planes. 360 uint8_t *chrMmxextFilterCode; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code for chroma planes. 361 362 int canMMXEXTBeUsed; 363 364 int dstY; ///< Last destination vertical line output from last slice. 365 int flags; ///< Flags passed by the user to select scaler algorithm, optimizations, subsampling, etc... 366 void *yuvTable; // pointer to the yuv->rgb table start so it can be freed() 367 // alignment ensures the offset can be added in a single 368 // instruction on e.g. ARM 369 DECLARE_ALIGNED(16, int, table_gV)[256 + 2*YUVRGB_TABLE_HEADROOM]; 370 uint8_t *table_rV[256 + 2*YUVRGB_TABLE_HEADROOM]; 371 uint8_t *table_gU[256 + 2*YUVRGB_TABLE_HEADROOM]; 372 uint8_t *table_bU[256 + 2*YUVRGB_TABLE_HEADROOM]; 373 DECLARE_ALIGNED(16, int32_t, input_rgb2yuv_table)[16+40*4]; // This table can contain both C and SIMD formatted values, the C vales are always at the XY_IDX points 374#define RY_IDX 0 375#define GY_IDX 1 376#define BY_IDX 2 377#define RU_IDX 3 378#define GU_IDX 4 379#define BU_IDX 5 380#define RV_IDX 6 381#define GV_IDX 7 382#define BV_IDX 8 383#define RGB2YUV_SHIFT 15 384 385 int *dither_error[4]; 386 387 //Colorspace stuff 388 int contrast, brightness, saturation; // for sws_getColorspaceDetails 389 int srcColorspaceTable[4]; 390 int dstColorspaceTable[4]; 391 int srcRange; ///< 0 = MPG YUV range, 1 = JPG YUV range (source image). 392 int dstRange; ///< 0 = MPG YUV range, 1 = JPG YUV range (destination image). 393 int src0Alpha; 394 int dst0Alpha; 395 int srcXYZ; 396 int dstXYZ; 397 int src_h_chr_pos; 398 int dst_h_chr_pos; 399 int src_v_chr_pos; 400 int dst_v_chr_pos; 401 int yuv2rgb_y_offset; 402 int yuv2rgb_y_coeff; 403 int yuv2rgb_v2r_coeff; 404 int yuv2rgb_v2g_coeff; 405 int yuv2rgb_u2g_coeff; 406 int yuv2rgb_u2b_coeff; 407 408#define RED_DITHER "0*8" 409#define GREEN_DITHER "1*8" 410#define BLUE_DITHER "2*8" 411#define Y_COEFF "3*8" 412#define VR_COEFF "4*8" 413#define UB_COEFF "5*8" 414#define VG_COEFF "6*8" 415#define UG_COEFF "7*8" 416#define Y_OFFSET "8*8" 417#define U_OFFSET "9*8" 418#define V_OFFSET "10*8" 419#define LUM_MMX_FILTER_OFFSET "11*8" 420#define CHR_MMX_FILTER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE) 421#define DSTW_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2" 422#define ESP_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+8" 423#define VROUNDER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+16" 424#define U_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+24" 425#define V_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+32" 426#define Y_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+40" 427#define ALP_MMX_FILTER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+48" 428#define UV_OFF_PX "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+48" 429#define UV_OFF_BYTE "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+56" 430#define DITHER16 "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+64" 431#define DITHER32 "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+80" 432#define DITHER32_INT (11*8+4*4*MAX_FILTER_SIZE*3+80) // value equal to above, used for checking that the struct hasn't been changed by mistake 433 434 DECLARE_ALIGNED(8, uint64_t, redDither); 435 DECLARE_ALIGNED(8, uint64_t, greenDither); 436 DECLARE_ALIGNED(8, uint64_t, blueDither); 437 438 DECLARE_ALIGNED(8, uint64_t, yCoeff); 439 DECLARE_ALIGNED(8, uint64_t, vrCoeff); 440 DECLARE_ALIGNED(8, uint64_t, ubCoeff); 441 DECLARE_ALIGNED(8, uint64_t, vgCoeff); 442 DECLARE_ALIGNED(8, uint64_t, ugCoeff); 443 DECLARE_ALIGNED(8, uint64_t, yOffset); 444 DECLARE_ALIGNED(8, uint64_t, uOffset); 445 DECLARE_ALIGNED(8, uint64_t, vOffset); 446 int32_t lumMmxFilter[4 * MAX_FILTER_SIZE]; 447 int32_t chrMmxFilter[4 * MAX_FILTER_SIZE]; 448 int dstW; ///< Width of destination luma/alpha planes. 449 DECLARE_ALIGNED(8, uint64_t, esp); 450 DECLARE_ALIGNED(8, uint64_t, vRounder); 451 DECLARE_ALIGNED(8, uint64_t, u_temp); 452 DECLARE_ALIGNED(8, uint64_t, v_temp); 453 DECLARE_ALIGNED(8, uint64_t, y_temp); 454 int32_t alpMmxFilter[4 * MAX_FILTER_SIZE]; 455 // alignment of these values is not necessary, but merely here 456 // to maintain the same offset across x8632 and x86-64. Once we 457 // use proper offset macros in the asm, they can be removed. 458 DECLARE_ALIGNED(8, ptrdiff_t, uv_off); ///< offset (in pixels) between u and v planes 459 DECLARE_ALIGNED(8, ptrdiff_t, uv_offx2); ///< offset (in bytes) between u and v planes 460 DECLARE_ALIGNED(8, uint16_t, dither16)[8]; 461 DECLARE_ALIGNED(8, uint32_t, dither32)[8]; 462 463 const uint8_t *chrDither8, *lumDither8; 464 465#if HAVE_ALTIVEC 466 vector signed short CY; 467 vector signed short CRV; 468 vector signed short CBU; 469 vector signed short CGU; 470 vector signed short CGV; 471 vector signed short OY; 472 vector unsigned short CSHIFT; 473 vector signed short *vYCoeffsBank, *vCCoeffsBank; 474#endif 475 476 int use_mmx_vfilter; 477 478/* pre defined color-spaces gamma */ 479#define XYZ_GAMMA (2.6f) 480#define RGB_GAMMA (2.2f) 481 int16_t *xyzgamma; 482 int16_t *rgbgamma; 483 int16_t *xyzgammainv; 484 int16_t *rgbgammainv; 485 int16_t xyz2rgb_matrix[3][4]; 486 int16_t rgb2xyz_matrix[3][4]; 487 488 /* function pointers for swscale() */ 489 yuv2planar1_fn yuv2plane1; 490 yuv2planarX_fn yuv2planeX; 491 yuv2interleavedX_fn yuv2nv12cX; 492 yuv2packed1_fn yuv2packed1; 493 yuv2packed2_fn yuv2packed2; 494 yuv2packedX_fn yuv2packedX; 495 yuv2anyX_fn yuv2anyX; 496 497 /// Unscaled conversion of luma plane to YV12 for horizontal scaler. 498 void (*lumToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3, 499 int width, uint32_t *pal); 500 /// Unscaled conversion of alpha plane to YV12 for horizontal scaler. 501 void (*alpToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3, 502 int width, uint32_t *pal); 503 /// Unscaled conversion of chroma planes to YV12 for horizontal scaler. 504 void (*chrToYV12)(uint8_t *dstU, uint8_t *dstV, 505 const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, 506 int width, uint32_t *pal); 507 508 /** 509 * Functions to read planar input, such as planar RGB, and convert 510 * internally to Y/UV/A. 511 */ 512 /** @{ */ 513 void (*readLumPlanar)(uint8_t *dst, const uint8_t *src[4], int width, int32_t *rgb2yuv); 514 void (*readChrPlanar)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4], 515 int width, int32_t *rgb2yuv); 516 void (*readAlpPlanar)(uint8_t *dst, const uint8_t *src[4], int width, int32_t *rgb2yuv); 517 /** @} */ 518 519 /** 520 * Scale one horizontal line of input data using a bilinear filter 521 * to produce one line of output data. Compared to SwsContext->hScale(), 522 * please take note of the following caveats when using these: 523 * - Scaling is done using only 7bit instead of 14bit coefficients. 524 * - You can use no more than 5 input pixels to produce 4 output 525 * pixels. Therefore, this filter should not be used for downscaling 526 * by more than ~20% in width (because that equals more than 5/4th 527 * downscaling and thus more than 5 pixels input per 4 pixels output). 528 * - In general, bilinear filters create artifacts during downscaling 529 * (even when <20%), because one output pixel will span more than one 530 * input pixel, and thus some pixels will need edges of both neighbor 531 * pixels to interpolate the output pixel. Since you can use at most 532 * two input pixels per output pixel in bilinear scaling, this is 533 * impossible and thus downscaling by any size will create artifacts. 534 * To enable this type of scaling, set SWS_FLAG_FAST_BILINEAR 535 * in SwsContext->flags. 536 */ 537 /** @{ */ 538 void (*hyscale_fast)(struct SwsContext *c, 539 int16_t *dst, int dstWidth, 540 const uint8_t *src, int srcW, int xInc); 541 void (*hcscale_fast)(struct SwsContext *c, 542 int16_t *dst1, int16_t *dst2, int dstWidth, 543 const uint8_t *src1, const uint8_t *src2, 544 int srcW, int xInc); 545 /** @} */ 546 547 /** 548 * Scale one horizontal line of input data using a filter over the input 549 * lines, to produce one (differently sized) line of output data. 550 * 551 * @param dst pointer to destination buffer for horizontally scaled 552 * data. If the number of bits per component of one 553 * destination pixel (SwsContext->dstBpc) is <= 10, data 554 * will be 15bpc in 16bits (int16_t) width. Else (i.e. 555 * SwsContext->dstBpc == 16), data will be 19bpc in 556 * 32bits (int32_t) width. 557 * @param dstW width of destination image 558 * @param src pointer to source data to be scaled. If the number of 559 * bits per component of a source pixel (SwsContext->srcBpc) 560 * is 8, this is 8bpc in 8bits (uint8_t) width. Else 561 * (i.e. SwsContext->dstBpc > 8), this is native depth 562 * in 16bits (uint16_t) width. In other words, for 9-bit 563 * YUV input, this is 9bpc, for 10-bit YUV input, this is 564 * 10bpc, and for 16-bit RGB or YUV, this is 16bpc. 565 * @param filter filter coefficients to be used per output pixel for 566 * scaling. This contains 14bpp filtering coefficients. 567 * Guaranteed to contain dstW * filterSize entries. 568 * @param filterPos position of the first input pixel to be used for 569 * each output pixel during scaling. Guaranteed to 570 * contain dstW entries. 571 * @param filterSize the number of input coefficients to be used (and 572 * thus the number of input pixels to be used) for 573 * creating a single output pixel. Is aligned to 4 574 * (and input coefficients thus padded with zeroes) 575 * to simplify creating SIMD code. 576 */ 577 /** @{ */ 578 void (*hyScale)(struct SwsContext *c, int16_t *dst, int dstW, 579 const uint8_t *src, const int16_t *filter, 580 const int32_t *filterPos, int filterSize); 581 void (*hcScale)(struct SwsContext *c, int16_t *dst, int dstW, 582 const uint8_t *src, const int16_t *filter, 583 const int32_t *filterPos, int filterSize); 584 /** @} */ 585 586 /// Color range conversion function for luma plane if needed. 587 void (*lumConvertRange)(int16_t *dst, int width); 588 /// Color range conversion function for chroma planes if needed. 589 void (*chrConvertRange)(int16_t *dst1, int16_t *dst2, int width); 590 591 int needs_hcscale; ///< Set if there are chroma planes to be converted. 592 593 SwsDither dither; 594} SwsContext; 595//FIXME check init (where 0) 596 597SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c); 598int ff_yuv2rgb_c_init_tables(SwsContext *c, const int inv_table[4], 599 int fullRange, int brightness, 600 int contrast, int saturation); 601void ff_yuv2rgb_init_tables_ppc(SwsContext *c, const int inv_table[4], 602 int brightness, int contrast, int saturation); 603 604void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex, 605 int lastInLumBuf, int lastInChrBuf); 606 607av_cold void ff_sws_init_range_convert(SwsContext *c); 608 609SwsFunc ff_yuv2rgb_init_x86(SwsContext *c); 610SwsFunc ff_yuv2rgb_init_ppc(SwsContext *c); 611 612#if FF_API_SWS_FORMAT_NAME 613/** 614 * @deprecated Use av_get_pix_fmt_name() instead. 615 */ 616attribute_deprecated 617const char *sws_format_name(enum AVPixelFormat format); 618#endif 619 620static av_always_inline int is16BPS(enum AVPixelFormat pix_fmt) 621{ 622 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 623 av_assert0(desc); 624 return desc->comp[0].depth_minus1 == 15; 625} 626 627static av_always_inline int is9_OR_10BPS(enum AVPixelFormat pix_fmt) 628{ 629 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 630 av_assert0(desc); 631 return desc->comp[0].depth_minus1 >= 8 && desc->comp[0].depth_minus1 <= 13; 632} 633 634#define isNBPS(x) is9_OR_10BPS(x) 635 636static av_always_inline int isBE(enum AVPixelFormat pix_fmt) 637{ 638 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 639 av_assert0(desc); 640 return desc->flags & AV_PIX_FMT_FLAG_BE; 641} 642 643static av_always_inline int isYUV(enum AVPixelFormat pix_fmt) 644{ 645 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 646 av_assert0(desc); 647 return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components >= 2; 648} 649 650static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt) 651{ 652 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 653 av_assert0(desc); 654 return ((desc->flags & AV_PIX_FMT_FLAG_PLANAR) && isYUV(pix_fmt)); 655} 656 657static av_always_inline int isRGB(enum AVPixelFormat pix_fmt) 658{ 659 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 660 av_assert0(desc); 661 return (desc->flags & AV_PIX_FMT_FLAG_RGB); 662} 663 664#if 0 // FIXME 665#define isGray(x) \ 666 (!(av_pix_fmt_desc_get(x)->flags & AV_PIX_FMT_FLAG_PAL) && \ 667 av_pix_fmt_desc_get(x)->nb_components <= 2) 668#else 669#define isGray(x) \ 670 ((x) == AV_PIX_FMT_GRAY8 || \ 671 (x) == AV_PIX_FMT_Y400A || \ 672 (x) == AV_PIX_FMT_GRAY16BE || \ 673 (x) == AV_PIX_FMT_GRAY16LE) 674#endif 675 676#define isRGBinInt(x) \ 677 ( \ 678 (x) == AV_PIX_FMT_RGB48BE || \ 679 (x) == AV_PIX_FMT_RGB48LE || \ 680 (x) == AV_PIX_FMT_RGB32 || \ 681 (x) == AV_PIX_FMT_RGB32_1 || \ 682 (x) == AV_PIX_FMT_RGB24 || \ 683 (x) == AV_PIX_FMT_RGB565BE || \ 684 (x) == AV_PIX_FMT_RGB565LE || \ 685 (x) == AV_PIX_FMT_RGB555BE || \ 686 (x) == AV_PIX_FMT_RGB555LE || \ 687 (x) == AV_PIX_FMT_RGB444BE || \ 688 (x) == AV_PIX_FMT_RGB444LE || \ 689 (x) == AV_PIX_FMT_RGB8 || \ 690 (x) == AV_PIX_FMT_RGB4 || \ 691 (x) == AV_PIX_FMT_RGB4_BYTE || \ 692 (x) == AV_PIX_FMT_RGBA64BE || \ 693 (x) == AV_PIX_FMT_RGBA64LE || \ 694 (x) == AV_PIX_FMT_MONOBLACK || \ 695 (x) == AV_PIX_FMT_MONOWHITE \ 696 ) 697#define isBGRinInt(x) \ 698 ( \ 699 (x) == AV_PIX_FMT_BGR48BE || \ 700 (x) == AV_PIX_FMT_BGR48LE || \ 701 (x) == AV_PIX_FMT_BGR32 || \ 702 (x) == AV_PIX_FMT_BGR32_1 || \ 703 (x) == AV_PIX_FMT_BGR24 || \ 704 (x) == AV_PIX_FMT_BGR565BE || \ 705 (x) == AV_PIX_FMT_BGR565LE || \ 706 (x) == AV_PIX_FMT_BGR555BE || \ 707 (x) == AV_PIX_FMT_BGR555LE || \ 708 (x) == AV_PIX_FMT_BGR444BE || \ 709 (x) == AV_PIX_FMT_BGR444LE || \ 710 (x) == AV_PIX_FMT_BGR8 || \ 711 (x) == AV_PIX_FMT_BGR4 || \ 712 (x) == AV_PIX_FMT_BGR4_BYTE || \ 713 (x) == AV_PIX_FMT_BGRA64BE || \ 714 (x) == AV_PIX_FMT_BGRA64LE || \ 715 (x) == AV_PIX_FMT_MONOBLACK || \ 716 (x) == AV_PIX_FMT_MONOWHITE \ 717 ) 718 719#define isRGBinBytes(x) ( \ 720 (x) == AV_PIX_FMT_RGB48BE \ 721 || (x) == AV_PIX_FMT_RGB48LE \ 722 || (x) == AV_PIX_FMT_RGBA64BE \ 723 || (x) == AV_PIX_FMT_RGBA64LE \ 724 || (x) == AV_PIX_FMT_RGBA \ 725 || (x) == AV_PIX_FMT_ARGB \ 726 || (x) == AV_PIX_FMT_RGB24 \ 727 ) 728#define isBGRinBytes(x) ( \ 729 (x) == AV_PIX_FMT_BGR48BE \ 730 || (x) == AV_PIX_FMT_BGR48LE \ 731 || (x) == AV_PIX_FMT_BGRA64BE \ 732 || (x) == AV_PIX_FMT_BGRA64LE \ 733 || (x) == AV_PIX_FMT_BGRA \ 734 || (x) == AV_PIX_FMT_ABGR \ 735 || (x) == AV_PIX_FMT_BGR24 \ 736 ) 737 738#define isBayer(x) ( \ 739 (x)==AV_PIX_FMT_BAYER_BGGR8 \ 740 || (x)==AV_PIX_FMT_BAYER_BGGR16LE \ 741 || (x)==AV_PIX_FMT_BAYER_BGGR16BE \ 742 || (x)==AV_PIX_FMT_BAYER_RGGB8 \ 743 || (x)==AV_PIX_FMT_BAYER_RGGB16LE \ 744 || (x)==AV_PIX_FMT_BAYER_RGGB16BE \ 745 || (x)==AV_PIX_FMT_BAYER_GBRG8 \ 746 || (x)==AV_PIX_FMT_BAYER_GBRG16LE \ 747 || (x)==AV_PIX_FMT_BAYER_GBRG16BE \ 748 || (x)==AV_PIX_FMT_BAYER_GRBG8 \ 749 || (x)==AV_PIX_FMT_BAYER_GRBG16LE \ 750 || (x)==AV_PIX_FMT_BAYER_GRBG16BE \ 751 ) 752 753#define isAnyRGB(x) \ 754 ( \ 755 isBayer(x) || \ 756 isRGBinInt(x) || \ 757 isBGRinInt(x) || \ 758 isRGB(x) \ 759 ) 760 761static av_always_inline int isALPHA(enum AVPixelFormat pix_fmt) 762{ 763 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 764 av_assert0(desc); 765 if (pix_fmt == AV_PIX_FMT_PAL8) 766 return 1; 767 return desc->flags & AV_PIX_FMT_FLAG_ALPHA; 768} 769 770#if 1 771#define isPacked(x) ( \ 772 (x)==AV_PIX_FMT_PAL8 \ 773 || (x)==AV_PIX_FMT_YUYV422 \ 774 || (x)==AV_PIX_FMT_YVYU422 \ 775 || (x)==AV_PIX_FMT_UYVY422 \ 776 || (x)==AV_PIX_FMT_Y400A \ 777 || isRGBinInt(x) \ 778 || isBGRinInt(x) \ 779 ) 780#else 781static av_always_inline int isPacked(enum AVPixelFormat pix_fmt) 782{ 783 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 784 av_assert0(desc); 785 return ((desc->nb_components >= 2 && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR)) || 786 pix_fmt == AV_PIX_FMT_PAL8); 787} 788 789#endif 790static av_always_inline int isPlanar(enum AVPixelFormat pix_fmt) 791{ 792 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 793 av_assert0(desc); 794 return (desc->nb_components >= 2 && (desc->flags & AV_PIX_FMT_FLAG_PLANAR)); 795} 796 797static av_always_inline int isPackedRGB(enum AVPixelFormat pix_fmt) 798{ 799 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 800 av_assert0(desc); 801 return ((desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)) == AV_PIX_FMT_FLAG_RGB); 802} 803 804static av_always_inline int isPlanarRGB(enum AVPixelFormat pix_fmt) 805{ 806 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 807 av_assert0(desc); 808 return ((desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)) == 809 (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)); 810} 811 812static av_always_inline int usePal(enum AVPixelFormat pix_fmt) 813{ 814 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt); 815 av_assert0(desc); 816 return (desc->flags & AV_PIX_FMT_FLAG_PAL) || (desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL); 817} 818 819extern const uint64_t ff_dither4[2]; 820extern const uint64_t ff_dither8[2]; 821 822extern const uint8_t ff_dither_2x2_4[3][8]; 823extern const uint8_t ff_dither_2x2_8[3][8]; 824extern const uint8_t ff_dither_4x4_16[5][8]; 825extern const uint8_t ff_dither_8x8_32[9][8]; 826extern const uint8_t ff_dither_8x8_73[9][8]; 827extern const uint8_t ff_dither_8x8_128[9][8]; 828extern const uint8_t ff_dither_8x8_220[9][8]; 829 830extern const int32_t ff_yuv2rgb_coeffs[8][4]; 831 832extern const AVClass sws_context_class; 833 834/** 835 * Set c->swscale to an unscaled converter if one exists for the specific 836 * source and destination formats, bit depths, flags, etc. 837 */ 838void ff_get_unscaled_swscale(SwsContext *c); 839void ff_get_unscaled_swscale_ppc(SwsContext *c); 840void ff_get_unscaled_swscale_arm(SwsContext *c); 841 842/** 843 * Return function pointer to fastest main scaler path function depending 844 * on architecture and available optimizations. 845 */ 846SwsFunc ff_getSwsFunc(SwsContext *c); 847 848void ff_sws_init_input_funcs(SwsContext *c); 849void ff_sws_init_output_funcs(SwsContext *c, 850 yuv2planar1_fn *yuv2plane1, 851 yuv2planarX_fn *yuv2planeX, 852 yuv2interleavedX_fn *yuv2nv12cX, 853 yuv2packed1_fn *yuv2packed1, 854 yuv2packed2_fn *yuv2packed2, 855 yuv2packedX_fn *yuv2packedX, 856 yuv2anyX_fn *yuv2anyX); 857void ff_sws_init_swscale_ppc(SwsContext *c); 858void ff_sws_init_swscale_x86(SwsContext *c); 859 860static inline void fillPlane16(uint8_t *plane, int stride, int width, int height, int y, 861 int alpha, int bits, const int big_endian) 862{ 863 int i, j; 864 uint8_t *ptr = plane + stride * y; 865 int v = alpha ? 0xFFFF>>(15-bits) : (1<<bits); 866 for (i = 0; i < height; i++) { 867#define FILL(wfunc) \ 868 for (j = 0; j < width; j++) {\ 869 wfunc(ptr+2*j, v);\ 870 } 871 if (big_endian) { 872 FILL(AV_WB16); 873 } else { 874 FILL(AV_WL16); 875 } 876 ptr += stride; 877 } 878} 879 880#endif /* SWSCALE_SWSCALE_INTERNAL_H */ 881