swscale.c 100 KB
Newer Older
1
/*
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public 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
 * the C code (not assembly, mmx, ...) of this file can be used
 * under the LGPL license too
22
 */
23

24
/*
25
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
26
  supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
27
  {BGR,RGB}{1,4,8,15,16} support dithering
28

29 30 31 32 33 34
  unscaled special converters (YV12=I420=IYUV, Y800=Y8)
  YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
  x -> x
  YUV9 -> YV12
  YUV9/YV12 -> Y800
  Y800 -> YUV9/YV12
Michael Niedermayer's avatar
Michael Niedermayer committed
35 36
  BGR24 -> BGR32 & RGB24 -> RGB32
  BGR32 -> BGR24 & RGB32 -> RGB24
Michael Niedermayer's avatar
Michael Niedermayer committed
37
  BGR15 -> BGR16
Michael Niedermayer's avatar
Michael Niedermayer committed
38 39
*/

40
/*
Diego Biurrun's avatar
typos  
Diego Biurrun committed
41
tested special converters (most are tested actually, but I did not write it down ...)
42
 YV12 -> BGR16
Michael Niedermayer's avatar
Michael Niedermayer committed
43
 YV12 -> YV12
Michael Niedermayer's avatar
Michael Niedermayer committed
44
 BGR15 -> BGR16
45
 BGR16 -> BGR16
46
 YVU9 -> YV12
Michael Niedermayer's avatar
Michael Niedermayer committed
47 48

untested special converters
Diego Biurrun's avatar
Diego Biurrun committed
49
  YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be ok)
50 51
  YV12/I420 -> YV12/I420
  YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
Michael Niedermayer's avatar
Michael Niedermayer committed
52 53
  BGR24 -> BGR32 & RGB24 -> RGB32
  BGR32 -> BGR24 & RGB32 -> RGB24
Michael Niedermayer's avatar
Michael Niedermayer committed
54
  BGR24 -> YV12
55 56
*/

57
#define _SVID_SOURCE //needed for MAP_ANONYMOUS
58
#include <inttypes.h>
59
#include <string.h>
60
#include <math.h>
61
#include <stdio.h>
Bohdan Horst's avatar
Bohdan Horst committed
62
#include <unistd.h>
63
#include "config.h"
Michael Niedermayer's avatar
Michael Niedermayer committed
64
#include <assert.h>
65 66
#ifdef HAVE_SYS_MMAN_H
#include <sys/mman.h>
67 68 69
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
#define MAP_ANONYMOUS MAP_ANON
#endif
70
#endif
71
#include "swscale.h"
72
#include "swscale_internal.h"
73
#include "rgb2rgb.h"
74 75
#include "libavutil/x86_cpu.h"
#include "libavutil/bswap.h"
Arpi's avatar
Arpi committed
76

77 78 79 80 81
unsigned swscale_version(void)
{
    return LIBSWSCALE_VERSION_INT;
}

Michael Niedermayer's avatar
Michael Niedermayer committed
82
#undef MOVNTQ
83
#undef PAVGB
84

85
//#undef HAVE_MMX2
Michael Niedermayer's avatar
101  
Michael Niedermayer committed
86
//#define HAVE_3DNOW
87
//#undef HAVE_MMX
88
//#undef ARCH_X86
Michael Niedermayer's avatar
101++  
Michael Niedermayer committed
89
//#define WORDS_BIGENDIAN
90
#define DITHER1XBPP
91

Michael Niedermayer's avatar
Michael Niedermayer committed
92 93
#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit

94
#define RET 0xC3 //near return opcode for X86
95

96 97 98 99 100
#ifdef M_PI
#define PI M_PI
#else
#define PI 3.14159265358979323846
#endif
101

102 103
#define isSupportedIn(x)    (       \
           (x)==PIX_FMT_YUV420P     \
104
        || (x)==PIX_FMT_YUVA420P    \
105 106 107
        || (x)==PIX_FMT_YUYV422     \
        || (x)==PIX_FMT_UYVY422     \
        || (x)==PIX_FMT_RGB32       \
108
        || (x)==PIX_FMT_RGB32_1     \
109 110 111 112
        || (x)==PIX_FMT_BGR24       \
        || (x)==PIX_FMT_BGR565      \
        || (x)==PIX_FMT_BGR555      \
        || (x)==PIX_FMT_BGR32       \
113
        || (x)==PIX_FMT_BGR32_1     \
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
        || (x)==PIX_FMT_RGB24       \
        || (x)==PIX_FMT_RGB565      \
        || (x)==PIX_FMT_RGB555      \
        || (x)==PIX_FMT_GRAY8       \
        || (x)==PIX_FMT_YUV410P     \
        || (x)==PIX_FMT_GRAY16BE    \
        || (x)==PIX_FMT_GRAY16LE    \
        || (x)==PIX_FMT_YUV444P     \
        || (x)==PIX_FMT_YUV422P     \
        || (x)==PIX_FMT_YUV411P     \
        || (x)==PIX_FMT_PAL8        \
        || (x)==PIX_FMT_BGR8        \
        || (x)==PIX_FMT_RGB8        \
        || (x)==PIX_FMT_BGR4_BYTE   \
        || (x)==PIX_FMT_RGB4_BYTE   \
Andreas Öman's avatar
Andreas Öman committed
129
        || (x)==PIX_FMT_YUV440P     \
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
    )
#define isSupportedOut(x)   (       \
           (x)==PIX_FMT_YUV420P     \
        || (x)==PIX_FMT_YUYV422     \
        || (x)==PIX_FMT_UYVY422     \
        || (x)==PIX_FMT_YUV444P     \
        || (x)==PIX_FMT_YUV422P     \
        || (x)==PIX_FMT_YUV411P     \
        || isRGB(x)                 \
        || isBGR(x)                 \
        || (x)==PIX_FMT_NV12        \
        || (x)==PIX_FMT_NV21        \
        || (x)==PIX_FMT_GRAY16BE    \
        || (x)==PIX_FMT_GRAY16LE    \
        || (x)==PIX_FMT_GRAY8       \
        || (x)==PIX_FMT_YUV410P     \
    )
#define isPacked(x)         (       \
           (x)==PIX_FMT_PAL8        \
        || (x)==PIX_FMT_YUYV422     \
        || (x)==PIX_FMT_UYVY422     \
        || isRGB(x)                 \
        || isBGR(x)                 \
    )
154

155
#define RGB2YUV_SHIFT 15
156 157 158 159 160 161 162 163 164
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
165

166 167
extern const int32_t Inverse_Table_6_9[8][4];

168 169 170 171 172 173 174 175 176 177 178
static const double rgb2yuv_table[8][9]={
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
};

179 180
/*
NOTES
181
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
182

183
TODO
Diego Biurrun's avatar
typo  
Diego Biurrun committed
184
more intelligent misalignment avoidance for the horizontal scaler
185 186
write special vertical cubic upscale version
Optimize C code (yv12 / minmax)
187
add support for packed pixel yuv input & output
188 189
add support for Y8 output
optimize bgr24 & bgr32
190
add BGR4 output support
191
write special BGR->BGR scaler
192
*/
193

194
#if defined(ARCH_X86) && defined (CONFIG_GPL)
195 196 197 198 199 200 201 202
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
203

204 205 206 207
static volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;
static volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;
static volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;
static volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;
Michael Niedermayer's avatar
Michael Niedermayer committed
208

209
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
210 211
        0x0103010301030103LL,
        0x0200020002000200LL,};
Michael Niedermayer's avatar
Michael Niedermayer committed
212

213
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
214 215
        0x0602060206020602LL,
        0x0004000400040004LL,};
216

217 218 219 220 221 222
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
223

224 225 226
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
Michael Niedermayer's avatar
Michael Niedermayer committed
227

Michael Niedermayer's avatar
Michael Niedermayer committed
228
#ifdef FAST_BGR2YV12
229 230 231
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
Michael Niedermayer's avatar
Michael Niedermayer committed
232
#else
233 234 235
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
Diego Biurrun's avatar
Diego Biurrun committed
236
#endif /* FAST_BGR2YV12 */
237 238 239
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
240 241 242 243 244 245 246 247 248 249 250 251 252 253

DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
DECLARE_ALIGNED(8, const uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
DECLARE_ALIGNED(8, const uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;

DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toUV[2][4]) = {
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
};

DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;

254
#endif /* defined(ARCH_X86) */
255 256 257 258

// clipping helper table for C implementations:
static unsigned char clip_table[768];

259
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
260

Michael Niedermayer's avatar
Michael Niedermayer committed
261 262 263 264 265
extern const uint8_t dither_2x2_4[2][8];
extern const uint8_t dither_2x2_8[2][8];
extern const uint8_t dither_8x8_32[8][8];
extern const uint8_t dither_8x8_73[8][8];
extern const uint8_t dither_8x8_220[8][8];
266

267
const char *sws_format_name(enum PixelFormat format)
268
{
269 270 271
    switch (format) {
        case PIX_FMT_YUV420P:
            return "yuv420p";
272 273
        case PIX_FMT_YUVA420P:
            return "yuva420p";
274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
        case PIX_FMT_YUYV422:
            return "yuyv422";
        case PIX_FMT_RGB24:
            return "rgb24";
        case PIX_FMT_BGR24:
            return "bgr24";
        case PIX_FMT_YUV422P:
            return "yuv422p";
        case PIX_FMT_YUV444P:
            return "yuv444p";
        case PIX_FMT_RGB32:
            return "rgb32";
        case PIX_FMT_YUV410P:
            return "yuv410p";
        case PIX_FMT_YUV411P:
            return "yuv411p";
        case PIX_FMT_RGB565:
            return "rgb565";
        case PIX_FMT_RGB555:
            return "rgb555";
Kostya Shishkov's avatar
Kostya Shishkov committed
294 295 296 297
        case PIX_FMT_GRAY16BE:
            return "gray16be";
        case PIX_FMT_GRAY16LE:
            return "gray16le";
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345
        case PIX_FMT_GRAY8:
            return "gray8";
        case PIX_FMT_MONOWHITE:
            return "mono white";
        case PIX_FMT_MONOBLACK:
            return "mono black";
        case PIX_FMT_PAL8:
            return "Palette";
        case PIX_FMT_YUVJ420P:
            return "yuvj420p";
        case PIX_FMT_YUVJ422P:
            return "yuvj422p";
        case PIX_FMT_YUVJ444P:
            return "yuvj444p";
        case PIX_FMT_XVMC_MPEG2_MC:
            return "xvmc_mpeg2_mc";
        case PIX_FMT_XVMC_MPEG2_IDCT:
            return "xvmc_mpeg2_idct";
        case PIX_FMT_UYVY422:
            return "uyvy422";
        case PIX_FMT_UYYVYY411:
            return "uyyvyy411";
        case PIX_FMT_RGB32_1:
            return "rgb32x";
        case PIX_FMT_BGR32_1:
            return "bgr32x";
        case PIX_FMT_BGR32:
            return "bgr32";
        case PIX_FMT_BGR565:
            return "bgr565";
        case PIX_FMT_BGR555:
            return "bgr555";
        case PIX_FMT_BGR8:
            return "bgr8";
        case PIX_FMT_BGR4:
            return "bgr4";
        case PIX_FMT_BGR4_BYTE:
            return "bgr4 byte";
        case PIX_FMT_RGB8:
            return "rgb8";
        case PIX_FMT_RGB4:
            return "rgb4";
        case PIX_FMT_RGB4_BYTE:
            return "rgb4 byte";
        case PIX_FMT_NV12:
            return "nv12";
        case PIX_FMT_NV21:
            return "nv21";
Andreas Öman's avatar
Andreas Öman committed
346 347
        case PIX_FMT_YUV440P:
            return "yuv440p";
348 349 350
        default:
            return "Unknown format";
    }
351 352
}

353
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
354 355
                               int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                               uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
Michael Niedermayer's avatar
Michael Niedermayer committed
356
{
357 358 359 360 361 362 363 364 365 366 367 368
    //FIXME Optimize (just quickly writen not opti..)
    int i;
    for (i=0; i<dstW; i++)
    {
        int val=1<<18;
        int j;
        for (j=0; j<lumFilterSize; j++)
            val += lumSrc[j][i] * lumFilter[j];

        dest[i]= av_clip_uint8(val>>19);
    }

369
    if (uDest)
370 371 372 373 374 375 376 377
        for (i=0; i<chrDstW; i++)
        {
            int u=1<<18;
            int v=1<<18;
            int j;
            for (j=0; j<chrFilterSize; j++)
            {
                u += chrSrc[j][i] * chrFilter[j];
378
                v += chrSrc[j][i + VOFW] * chrFilter[j];
379 380 381 382 383
            }

            uDest[i]= av_clip_uint8(u>>19);
            vDest[i]= av_clip_uint8(v>>19);
        }
Michael Niedermayer's avatar
Michael Niedermayer committed
384 385
}

Ville Syrjälä's avatar
Ville Syrjälä committed
386
static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
387 388
                                int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
Ville Syrjälä's avatar
Ville Syrjälä committed
389
{
390 391 392 393 394 395 396 397 398 399 400 401
    //FIXME Optimize (just quickly writen not opti..)
    int i;
    for (i=0; i<dstW; i++)
    {
        int val=1<<18;
        int j;
        for (j=0; j<lumFilterSize; j++)
            val += lumSrc[j][i] * lumFilter[j];

        dest[i]= av_clip_uint8(val>>19);
    }

402
    if (!uDest)
403 404 405 406 407 408 409 410 411 412 413
        return;

    if (dstFormat == PIX_FMT_NV12)
        for (i=0; i<chrDstW; i++)
        {
            int u=1<<18;
            int v=1<<18;
            int j;
            for (j=0; j<chrFilterSize; j++)
            {
                u += chrSrc[j][i] * chrFilter[j];
414
                v += chrSrc[j][i + VOFW] * chrFilter[j];
415 416 417 418 419 420 421 422 423 424 425 426 427 428
            }

            uDest[2*i]= av_clip_uint8(u>>19);
            uDest[2*i+1]= av_clip_uint8(v>>19);
        }
    else
        for (i=0; i<chrDstW; i++)
        {
            int u=1<<18;
            int v=1<<18;
            int j;
            for (j=0; j<chrFilterSize; j++)
            {
                u += chrSrc[j][i] * chrFilter[j];
429
                v += chrSrc[j][i + VOFW] * chrFilter[j];
430 431 432 433 434
            }

            uDest[2*i]= av_clip_uint8(v>>19);
            uDest[2*i+1]= av_clip_uint8(u>>19);
        }
Ville Syrjälä's avatar
Ville Syrjälä committed
435
}
Michael Niedermayer's avatar
Michael Niedermayer committed
436

437
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type) \
438 439 440 441 442 443
    for (i=0; i<(dstW>>1); i++){\
        int j;\
        int Y1 = 1<<18;\
        int Y2 = 1<<18;\
        int U  = 1<<18;\
        int V  = 1<<18;\
444
        type av_unused *r, *b, *g;\
445 446 447 448 449 450 451 452 453 454
        const int i2= 2*i;\
        \
        for (j=0; j<lumFilterSize; j++)\
        {\
            Y1 += lumSrc[j][i2] * lumFilter[j];\
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
        }\
        for (j=0; j<chrFilterSize; j++)\
        {\
            U += chrSrc[j][i] * chrFilter[j];\
455
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
456 457 458 459 460
        }\
        Y1>>=19;\
        Y2>>=19;\
        U >>=19;\
        V >>=19;\
461 462 463

#define YSCALE_YUV_2_PACKEDX_C(type) \
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type)\
464 465 466 467 468 469 470 471 472 473 474
        if ((Y1|Y2|U|V)&256)\
        {\
            if (Y1>255)   Y1=255; \
            else if (Y1<0)Y1=0;   \
            if (Y2>255)   Y2=255; \
            else if (Y2<0)Y2=0;   \
            if (U>255)    U=255;  \
            else if (U<0) U=0;    \
            if (V>255)    V=255;  \
            else if (V<0) V=0;    \
        }
475

476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
#define YSCALE_YUV_2_PACKEDX_FULL_C \
    for (i=0; i<dstW; i++){\
        int j;\
        int Y = 0;\
        int U = -128<<19;\
        int V = -128<<19;\
        int R,G,B;\
        \
        for (j=0; j<lumFilterSize; j++){\
            Y += lumSrc[j][i     ] * lumFilter[j];\
        }\
        for (j=0; j<chrFilterSize; j++){\
            U += chrSrc[j][i     ] * chrFilter[j];\
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
        }\
        Y >>=10;\
        U >>=10;\
        V >>=10;\

#define YSCALE_YUV_2_RGBX_FULL_C(rnd) \
    YSCALE_YUV_2_PACKEDX_FULL_C\
        Y-= c->oy;\
        Y*= c->cy;\
        Y+= rnd;\
        R= Y + V*c->cvr;\
        G= Y + V*c->cvg + U*c->cug;\
        B= Y +            U*c->cub;\
        if ((R|G|B)&(0xC0000000)){\
            if (R>=(256<<22))   R=(256<<22)-1; \
            else if (R<0)R=0;   \
            if (G>=(256<<22))   G=(256<<22)-1; \
            else if (G<0)G=0;   \
            if (B>=(256<<22))   B=(256<<22)-1; \
            else if (B<0)B=0;   \
        }\


513
#define YSCALE_YUV_2_GRAY16_C \
514 515 516 517 518 519
    for (i=0; i<(dstW>>1); i++){\
        int j;\
        int Y1 = 1<<18;\
        int Y2 = 1<<18;\
        int U  = 1<<18;\
        int V  = 1<<18;\
520
        \
521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
        const int i2= 2*i;\
        \
        for (j=0; j<lumFilterSize; j++)\
        {\
            Y1 += lumSrc[j][i2] * lumFilter[j];\
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
        }\
        Y1>>=11;\
        Y2>>=11;\
        if ((Y1|Y2|U|V)&65536)\
        {\
            if (Y1>65535)   Y1=65535; \
            else if (Y1<0)Y1=0;   \
            if (Y2>65535)   Y2=65535; \
            else if (Y2<0)Y2=0;   \
        }

Michael Niedermayer's avatar
Michael Niedermayer committed
538
#define YSCALE_YUV_2_RGBX_C(type) \
539
    YSCALE_YUV_2_PACKEDX_NOCLIP_C(type)  \
540 541 542 543 544 545 546 547 548 549
    r = (type *)c->table_rV[V];   \
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
    b = (type *)c->table_bU[U];   \

#define YSCALE_YUV_2_PACKED2_C   \
    for (i=0; i<(dstW>>1); i++){ \
        const int i2= 2*i;       \
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
550
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
Michael Niedermayer's avatar
Michael Niedermayer committed
551

552 553 554 555 556 557
#define YSCALE_YUV_2_GRAY16_2_C   \
    for (i=0; i<(dstW>>1); i++){ \
        const int i2= 2*i;       \
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \

Michael Niedermayer's avatar
Michael Niedermayer committed
558
#define YSCALE_YUV_2_RGB2_C(type) \
559 560 561 562 563
    YSCALE_YUV_2_PACKED2_C\
    type *r, *b, *g;\
    r = (type *)c->table_rV[V];\
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
    b = (type *)c->table_bU[U];\
Michael Niedermayer's avatar
Michael Niedermayer committed
564

Michael Niedermayer's avatar
Michael Niedermayer committed
565
#define YSCALE_YUV_2_PACKED1_C \
566 567 568 569 570
    for (i=0; i<(dstW>>1); i++){\
        const int i2= 2*i;\
        int Y1= buf0[i2  ]>>7;\
        int Y2= buf0[i2+1]>>7;\
        int U= (uvbuf1[i     ])>>7;\
571
        int V= (uvbuf1[i+VOFW])>>7;\
Michael Niedermayer's avatar
Michael Niedermayer committed
572

573 574 575 576 577 578
#define YSCALE_YUV_2_GRAY16_1_C \
    for (i=0; i<(dstW>>1); i++){\
        const int i2= 2*i;\
        int Y1= buf0[i2  ]<<1;\
        int Y2= buf0[i2+1]<<1;\

Michael Niedermayer's avatar
Michael Niedermayer committed
579
#define YSCALE_YUV_2_RGB1_C(type) \
580 581 582 583 584
    YSCALE_YUV_2_PACKED1_C\
    type *r, *b, *g;\
    r = (type *)c->table_rV[V];\
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
    b = (type *)c->table_bU[U];\
Michael Niedermayer's avatar
Michael Niedermayer committed
585

Michael Niedermayer's avatar
Michael Niedermayer committed
586
#define YSCALE_YUV_2_PACKED1B_C \
587 588 589 590 591
    for (i=0; i<(dstW>>1); i++){\
        const int i2= 2*i;\
        int Y1= buf0[i2  ]>>7;\
        int Y2= buf0[i2+1]>>7;\
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
592
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
Michael Niedermayer's avatar
Michael Niedermayer committed
593 594

#define YSCALE_YUV_2_RGB1B_C(type) \
595 596 597 598 599
    YSCALE_YUV_2_PACKED1B_C\
    type *r, *b, *g;\
    r = (type *)c->table_rV[V];\
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
    b = (type *)c->table_bU[U];\
Michael Niedermayer's avatar
Michael Niedermayer committed
600

601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651
#define YSCALE_YUV_2_MONOBLACK2_C \
    const uint8_t * const d128=dither_8x8_220[y&7];\
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
    for (i=0; i<dstW-7; i+=8){\
        int acc;\
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
        ((uint8_t*)dest)[0]= acc;\
        dest++;\
    }\


#define YSCALE_YUV_2_MONOBLACKX_C \
    const uint8_t * const d128=dither_8x8_220[y&7];\
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
    int acc=0;\
    for (i=0; i<dstW-1; i+=2){\
        int j;\
        int Y1=1<<18;\
        int Y2=1<<18;\
\
        for (j=0; j<lumFilterSize; j++)\
        {\
            Y1 += lumSrc[j][i] * lumFilter[j];\
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
        }\
        Y1>>=19;\
        Y2>>=19;\
        if ((Y1|Y2)&256)\
        {\
            if (Y1>255)   Y1=255;\
            else if (Y1<0)Y1=0;\
            if (Y2>255)   Y2=255;\
            else if (Y2<0)Y2=0;\
        }\
        acc+= acc + g[Y1+d128[(i+0)&7]];\
        acc+= acc + g[Y2+d128[(i+1)&7]];\
        if ((i&7)==6){\
            ((uint8_t*)dest)[0]= acc;\
            dest++;\
        }\
    }


#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
652 653 654 655
    switch(c->dstFormat)\
    {\
    case PIX_FMT_RGB32:\
    case PIX_FMT_BGR32:\
656 657
    case PIX_FMT_RGB32_1:\
    case PIX_FMT_BGR32_1:\
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749
        func(uint32_t)\
            ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
            ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
        }                \
        break;\
    case PIX_FMT_RGB24:\
        func(uint8_t)\
            ((uint8_t*)dest)[0]= r[Y1];\
            ((uint8_t*)dest)[1]= g[Y1];\
            ((uint8_t*)dest)[2]= b[Y1];\
            ((uint8_t*)dest)[3]= r[Y2];\
            ((uint8_t*)dest)[4]= g[Y2];\
            ((uint8_t*)dest)[5]= b[Y2];\
            dest+=6;\
        }\
        break;\
    case PIX_FMT_BGR24:\
        func(uint8_t)\
            ((uint8_t*)dest)[0]= b[Y1];\
            ((uint8_t*)dest)[1]= g[Y1];\
            ((uint8_t*)dest)[2]= r[Y1];\
            ((uint8_t*)dest)[3]= b[Y2];\
            ((uint8_t*)dest)[4]= g[Y2];\
            ((uint8_t*)dest)[5]= r[Y2];\
            dest+=6;\
        }\
        break;\
    case PIX_FMT_RGB565:\
    case PIX_FMT_BGR565:\
        {\
            const int dr1= dither_2x2_8[y&1    ][0];\
            const int dg1= dither_2x2_4[y&1    ][0];\
            const int db1= dither_2x2_8[(y&1)^1][0];\
            const int dr2= dither_2x2_8[y&1    ][1];\
            const int dg2= dither_2x2_4[y&1    ][1];\
            const int db2= dither_2x2_8[(y&1)^1][1];\
            func(uint16_t)\
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
            }\
        }\
        break;\
    case PIX_FMT_RGB555:\
    case PIX_FMT_BGR555:\
        {\
            const int dr1= dither_2x2_8[y&1    ][0];\
            const int dg1= dither_2x2_8[y&1    ][1];\
            const int db1= dither_2x2_8[(y&1)^1][0];\
            const int dr2= dither_2x2_8[y&1    ][1];\
            const int dg2= dither_2x2_8[y&1    ][0];\
            const int db2= dither_2x2_8[(y&1)^1][1];\
            func(uint16_t)\
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
            }\
        }\
        break;\
    case PIX_FMT_RGB8:\
    case PIX_FMT_BGR8:\
        {\
            const uint8_t * const d64= dither_8x8_73[y&7];\
            const uint8_t * const d32= dither_8x8_32[y&7];\
            func(uint8_t)\
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
            }\
        }\
        break;\
    case PIX_FMT_RGB4:\
    case PIX_FMT_BGR4:\
        {\
            const uint8_t * const d64= dither_8x8_73 [y&7];\
            const uint8_t * const d128=dither_8x8_220[y&7];\
            func(uint8_t)\
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
            }\
        }\
        break;\
    case PIX_FMT_RGB4_BYTE:\
    case PIX_FMT_BGR4_BYTE:\
        {\
            const uint8_t * const d64= dither_8x8_73 [y&7];\
            const uint8_t * const d128=dither_8x8_220[y&7];\
            func(uint8_t)\
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
            }\
        }\
        break;\
    case PIX_FMT_MONOBLACK:\
        {\
750
            func_monoblack\
751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
        }\
        break;\
    case PIX_FMT_YUYV422:\
        func2\
            ((uint8_t*)dest)[2*i2+0]= Y1;\
            ((uint8_t*)dest)[2*i2+1]= U;\
            ((uint8_t*)dest)[2*i2+2]= Y2;\
            ((uint8_t*)dest)[2*i2+3]= V;\
        }                \
        break;\
    case PIX_FMT_UYVY422:\
        func2\
            ((uint8_t*)dest)[2*i2+0]= U;\
            ((uint8_t*)dest)[2*i2+1]= Y1;\
            ((uint8_t*)dest)[2*i2+2]= V;\
            ((uint8_t*)dest)[2*i2+3]= Y2;\
        }                \
        break;\
769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784
    case PIX_FMT_GRAY16BE:\
        func_g16\
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
            ((uint8_t*)dest)[2*i2+1]= Y1;\
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
            ((uint8_t*)dest)[2*i2+3]= Y2;\
        }                \
        break;\
    case PIX_FMT_GRAY16LE:\
        func_g16\
            ((uint8_t*)dest)[2*i2+0]= Y1;\
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
            ((uint8_t*)dest)[2*i2+2]= Y2;\
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
        }                \
        break;\
785
    }\
Michael Niedermayer's avatar
Michael Niedermayer committed
786 787


Michael Niedermayer's avatar
Michael Niedermayer committed
788
static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
789 790
                                  int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                                  uint8_t *dest, int dstW, int y)
Michael Niedermayer's avatar
Michael Niedermayer committed
791
{
792
    int i;
793
    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOBLACKX_C)
Michael Niedermayer's avatar
Michael Niedermayer committed
794 795
}

796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831
static inline void yuv2rgbXinC_full(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                                    uint8_t *dest, int dstW, int y)
{
    int i;
    int step= fmt_depth(c->dstFormat)/8;

    switch(c->dstFormat){
    case PIX_FMT_ARGB:
        dest++;
    case PIX_FMT_RGB24:
    case PIX_FMT_RGBA:
        YSCALE_YUV_2_RGBX_FULL_C(1<<21)
            dest[0]= R>>22;
            dest[1]= G>>22;
            dest[2]= B>>22;
            dest[3]= 0;
            dest+= step;
        }
        break;
    case PIX_FMT_ABGR:
        dest++;
    case PIX_FMT_BGR24:
    case PIX_FMT_BGRA:
        YSCALE_YUV_2_RGBX_FULL_C(1<<21)
            dest[0]= B>>22;
            dest[1]= G>>22;
            dest[2]= R>>22;
            dest[3]= 0;
            dest+= step;
        }
        break;
    default:
        assert(0);
    }
}
Michael Niedermayer's avatar
Michael Niedermayer committed
832

Michael Niedermayer's avatar
Michael Niedermayer committed
833 834
//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
//Plain C versions
835
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) || !defined(CONFIG_GPL)
836 837 838
#define COMPILE_C
#endif

839
#ifdef ARCH_POWERPC
840
#if (defined (HAVE_ALTIVEC) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
841 842 843 844
#define COMPILE_ALTIVEC
#endif //HAVE_ALTIVEC
#endif //ARCH_POWERPC

845
#if defined(ARCH_X86)
846

847
#if ((defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
848 849 850
#define COMPILE_MMX
#endif

851
#if (defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
852 853 854
#define COMPILE_MMX2
#endif

855
#if ((defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
856 857
#define COMPILE_3DNOW
#endif
858
#endif //ARCH_X86 || ARCH_X86_64
859 860 861 862 863 864

#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW

#ifdef COMPILE_C
Michael Niedermayer's avatar
Michael Niedermayer committed
865 866 867
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
868
#undef HAVE_ALTIVEC
Michael Niedermayer's avatar
Michael Niedermayer committed
869 870
#define RENAME(a) a ## _C
#include "swscale_template.c"
871
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
872

873 874 875 876 877 878 879
#ifdef COMPILE_ALTIVEC
#undef RENAME
#define HAVE_ALTIVEC
#define RENAME(a) a ## _altivec
#include "swscale_template.c"
#endif

880
#if defined(ARCH_X86)
Michael Niedermayer's avatar
Michael Niedermayer committed
881

Michael Niedermayer's avatar
Michael Niedermayer committed
882 883 884 885 886 887 888 889 890
//X86 versions
/*
#undef RENAME
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _X86
#include "swscale_template.c"
Michael Niedermayer's avatar
Michael Niedermayer committed
891
*/
Michael Niedermayer's avatar
Michael Niedermayer committed
892
//MMX versions
893
#ifdef COMPILE_MMX
Michael Niedermayer's avatar
Michael Niedermayer committed
894 895 896 897 898 899
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define RENAME(a) a ## _MMX
#include "swscale_template.c"
900
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
901 902

//MMX2 versions
903
#ifdef COMPILE_MMX2
Michael Niedermayer's avatar
Michael Niedermayer committed
904 905 906 907 908 909
#undef RENAME
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
#define RENAME(a) a ## _MMX2
#include "swscale_template.c"
910
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
911 912

//3DNOW versions
913
#ifdef COMPILE_3DNOW
Michael Niedermayer's avatar
Michael Niedermayer committed
914 915 916 917 918 919
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
#define RENAME(a) a ## _3DNow
#include "swscale_template.c"
920
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
921

922
#endif //ARCH_X86 || ARCH_X86_64
Michael Niedermayer's avatar
Michael Niedermayer committed
923

Gabucino's avatar
Gabucino committed
924
// minor note: the HAVE_xyz is messed up after that line so don't use it
925

Michael Niedermayer's avatar
Michael Niedermayer committed
926 927
static double getSplineCoeff(double a, double b, double c, double d, double dist)
{
928 929 930 931 932 933 934
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
    if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;
    else                return getSplineCoeff(        0.0,
                                             b+ 2.0*c + 3.0*d,
                                                    c + 3.0*d,
                                            -b- 3.0*c - 6.0*d,
                                            dist-1.0);
Michael Niedermayer's avatar
Michael Niedermayer committed
935
}
936

937
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
938 939
                             int srcW, int dstW, int filterAlign, int one, int flags,
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
940
{
941 942 943 944 945 946
    int i;
    int filterSize;
    int filter2Size;
    int minFilterSize;
    double *filter=NULL;
    double *filter2=NULL;
Michael Niedermayer's avatar
Michael Niedermayer committed
947
    int ret= -1;
948
#if defined(ARCH_X86)
949
    if (flags & SWS_CPU_CAPS_MMX)
Diego Biurrun's avatar
Diego Biurrun committed
950
        asm volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
951
#endif
952

953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
    // Note the +1 is for the MMXscaler which reads over the end
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));

    if (FFABS(xInc - 0x10000) <10) // unscaled
    {
        int i;
        filterSize= 1;
        filter= av_malloc(dstW*sizeof(double)*filterSize);
        for (i=0; i<dstW*filterSize; i++) filter[i]=0;

        for (i=0; i<dstW; i++)
        {
            filter[i*filterSize]=1;
            (*filterPos)[i]=i;
        }

    }
    else if (flags&SWS_POINT) // lame looking point sampling mode
    {
        int i;
        int xDstInSrc;
        filterSize= 1;
        filter= av_malloc(dstW*sizeof(double)*filterSize);

        xDstInSrc= xInc/2 - 0x8000;
        for (i=0; i<dstW; i++)
        {
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

            (*filterPos)[i]= xx;
            filter[i]= 1.0;
            xDstInSrc+= xInc;
        }
    }
    else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
    {
        int i;
        int xDstInSrc;
        if      (flags&SWS_BICUBIC) filterSize= 4;
        else if (flags&SWS_X      ) filterSize= 4;
        else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA
        filter= av_malloc(dstW*sizeof(double)*filterSize);

        xDstInSrc= xInc/2 - 0x8000;
        for (i=0; i<dstW; i++)
        {
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
            int j;

            (*filterPos)[i]= xx;
                //Bilinear upscale / linear interpolate / Area averaging
                for (j=0; j<filterSize; j++)
                {
                    double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);
                    double coeff= 1.0 - d;
                    if (coeff<0) coeff=0;
                    filter[i*filterSize + j]= coeff;
                    xx++;
                }
            xDstInSrc+= xInc;
        }
    }
    else
    {
        double xDstInSrc;
        double sizeFactor, filterSizeInSrc;
        const double xInc1= (double)xInc / (double)(1<<16);

        if      (flags&SWS_BICUBIC)      sizeFactor=  4.0;
        else if (flags&SWS_X)            sizeFactor=  8.0;
        else if (flags&SWS_AREA)         sizeFactor=  1.0; //downscale only, for upscale it is bilinear
        else if (flags&SWS_GAUSS)        sizeFactor=  8.0;   // infinite ;)
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;
        else if (flags&SWS_SINC)         sizeFactor= 20.0; // infinite ;)
        else if (flags&SWS_SPLINE)       sizeFactor= 20.0;  // infinite ;)
        else if (flags&SWS_BILINEAR)     sizeFactor=  2.0;
        else {
            sizeFactor= 0.0; //GCC warning killer
1031
            assert(0);
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115
        }

        if (xInc1 <= 1.0)       filterSizeInSrc= sizeFactor; // upscale
        else                    filterSizeInSrc= sizeFactor*srcW / (double)dstW;

        filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
        if (filterSize > srcW-2) filterSize=srcW-2;

        filter= av_malloc(dstW*sizeof(double)*filterSize);

        xDstInSrc= xInc1 / 2.0 - 0.5;
        for (i=0; i<dstW; i++)
        {
            int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);
            int j;
            (*filterPos)[i]= xx;
            for (j=0; j<filterSize; j++)
            {
                double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;
                double coeff;
                if (flags & SWS_BICUBIC)
                {
                    double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;
                    double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

                    if (d<1.0)
                        coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;
                    else if (d<2.0)
                        coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;
                    else
                        coeff=0.0;
                }
/*                else if (flags & SWS_X)
                {
                    double p= param ? param*0.01 : 0.3;
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
                    coeff*= pow(2.0, - p*d*d);
                }*/
                else if (flags & SWS_X)
                {
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

                    if (d<1.0)
                        coeff = cos(d*PI);
                    else
                        coeff=-1.0;
                    if (coeff<0.0)      coeff= -pow(-coeff, A);
                    else                coeff=  pow( coeff, A);
                    coeff= coeff*0.5 + 0.5;
                }
                else if (flags & SWS_AREA)
                {
                    double srcPixelSize= 1.0/xInc1;
                    if      (d + srcPixelSize/2 < 0.5) coeff= 1.0;
                    else if (d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;
                    else coeff=0.0;
                }
                else if (flags & SWS_GAUSS)
                {
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
                    coeff = pow(2.0, - p*d*d);
                }
                else if (flags & SWS_SINC)
                {
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
                }
                else if (flags & SWS_LANCZOS)
                {
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
                    coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
                    if (d>p) coeff=0;
                }
                else if (flags & SWS_BILINEAR)
                {
                    coeff= 1.0 - d;
                    if (coeff<0) coeff=0;
                }
                else if (flags & SWS_SPLINE)
                {
                    double p=-2.196152422706632;
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);
                }
                else {
                    coeff= 0.0; //GCC warning killer
1116
                    assert(0);
1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
                }

                filter[i*filterSize + j]= coeff;
                xx++;
            }
            xDstInSrc+= xInc1;
        }
    }

    /* apply src & dst Filter to filter -> filter2
       av_free(filter);
    */
1129
    assert(filterSize>0);
1130 1131 1132
    filter2Size= filterSize;
    if (srcFilter) filter2Size+= srcFilter->length - 1;
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1133
    assert(filter2Size>0);
1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
    filter2= av_malloc(filter2Size*dstW*sizeof(double));

    for (i=0; i<dstW; i++)
    {
        int j;
        SwsVector scaleFilter;
        SwsVector *outVec;

        scaleFilter.coeff= filter + i*filterSize;
        scaleFilter.length= filterSize;

        if (srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);
        else           outVec= &scaleFilter;

1148
        assert(outVec->length == filter2Size);
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
        //FIXME dstFilter

        for (j=0; j<outVec->length; j++)
        {
            filter2[i*filter2Size + j]= outVec->coeff[j];
        }

        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

        if (outVec != &scaleFilter) sws_freeVec(outVec);
    }
1160
    av_freep(&filter);
1161 1162

    /* try to reduce the filter-size (step1 find size and shift left) */
Diego Biurrun's avatar
Diego Biurrun committed
1163
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178
    minFilterSize= 0;
    for (i=dstW-1; i>=0; i--)
    {
        int min= filter2Size;
        int j;
        double cutOff=0.0;

        /* get rid off near zero elements on the left by shifting left */
        for (j=0; j<filter2Size; j++)
        {
            int k;
            cutOff += FFABS(filter2[i*filter2Size]);

            if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;

Diego Biurrun's avatar
Diego Biurrun committed
1179
            /* preserve monotonicity because the core can't handle the filter otherwise */
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

            // Move filter coeffs left
            for (k=1; k<filter2Size; k++)
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
            filter2[i*filter2Size + k - 1]= 0.0;
            (*filterPos)[i]++;
        }

        cutOff=0.0;
        /* count near zeros on the right */
        for (j=filter2Size-1; j>0; j--)
        {
            cutOff += FFABS(filter2[i*filter2Size + j]);

            if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;
            min--;
        }

        if (min>minFilterSize) minFilterSize= min;
    }

    if (flags & SWS_CPU_CAPS_ALTIVEC) {
        // we can handle the special case 4,
        // so we don't want to go to the full 8
        if (minFilterSize < 5)
1206 1207
            filterAlign = 4;

1208 1209 1210 1211 1212 1213
        // we really don't want to waste our time
        // doing useless computation, so fall-back on
        // the scalar C code for very small filter.
        // vectorizing is worth it only if you have
        // decent-sized vector.
        if (minFilterSize < 3)
1214
            filterAlign = 1;
1215 1216 1217 1218 1219 1220 1221 1222
    }

    if (flags & SWS_CPU_CAPS_MMX) {
        // special case for unscaled vertical filtering
        if (minFilterSize == 1 && filterAlign == 2)
            filterAlign= 1;
    }