swscale.c 102 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
/*
Michael Niedermayer's avatar
Michael Niedermayer committed
25
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, 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
/*
41 42
tested special converters (most are tested actually but i didnt write it down ...)
 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
#include <inttypes.h>
58
#include <string.h>
59
#include <math.h>
60
#include <stdio.h>
Bohdan Horst's avatar
Bohdan Horst committed
61
#include <unistd.h>
62
#include "config.h"
Michael Niedermayer's avatar
Michael Niedermayer committed
63
#include <assert.h>
64 65
#ifdef HAVE_SYS_MMAN_H
#include <sys/mman.h>
66 67 68
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
#define MAP_ANONYMOUS MAP_ANON
#endif
69
#endif
70
#include "swscale.h"
71
#include "swscale_internal.h"
72
#include "x86_cpu.h"
73
#include "bswap.h"
74
#include "rgb2rgb.h"
75
#include "libavcodec/opt.h"
Arpi's avatar
Arpi committed
76

Michael Niedermayer's avatar
Michael Niedermayer committed
77
#undef MOVNTQ
78
#undef PAVGB
79

80
//#undef HAVE_MMX2
Michael Niedermayer's avatar
101  
Michael Niedermayer committed
81
//#define HAVE_3DNOW
82
//#undef HAVE_MMX
83
//#undef ARCH_X86
Michael Niedermayer's avatar
101++  
Michael Niedermayer committed
84
//#define WORDS_BIGENDIAN
85
#define DITHER1XBPP
86

Michael Niedermayer's avatar
Michael Niedermayer committed
87 88
#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit

89
#define RET 0xC3 //near return opcode for X86
90

91
#ifdef MP_DEBUG
Michael Niedermayer's avatar
Michael Niedermayer committed
92
#define ASSERT(x) assert(x);
93
#else
94
#define ASSERT(x) ;
95 96 97 98 99 100 101
#endif

#ifdef M_PI
#define PI M_PI
#else
#define PI 3.14159265358979323846
#endif
102

103 104
#define isSupportedIn(x)    (       \
           (x)==PIX_FMT_YUV420P     \
105
        || (x)==PIX_FMT_YUVA420P    \
106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
        || (x)==PIX_FMT_YUYV422     \
        || (x)==PIX_FMT_UYVY422     \
        || (x)==PIX_FMT_RGB32       \
        || (x)==PIX_FMT_BGR24       \
        || (x)==PIX_FMT_BGR565      \
        || (x)==PIX_FMT_BGR555      \
        || (x)==PIX_FMT_BGR32       \
        || (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
128
        || (x)==PIX_FMT_YUV440P     \
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152
    )
#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)                 \
    )
153 154

#define RGB2YUV_SHIFT 16
155 156 157 158 159 160 161 162 163
#define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
#define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
#define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
#define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
#define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
#define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
#define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
#define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
#define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
164

165 166
extern const int32_t Inverse_Table_6_9[8][4];

167 168
/*
NOTES
169
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
170

171
TODO
Diego Biurrun's avatar
typo  
Diego Biurrun committed
172
more intelligent misalignment avoidance for the horizontal scaler
173 174
write special vertical cubic upscale version
Optimize C code (yv12 / minmax)
175
add support for packed pixel yuv input & output
176 177
add support for Y8 output
optimize bgr24 & bgr32
178
add BGR4 output support
179
write special BGR->BGR scaler
180
*/
181

182
#if defined(ARCH_X86) && defined (CONFIG_GPL)
183 184
static uint64_t attribute_used __attribute__((aligned(8))) bF8=       0xF8F8F8F8F8F8F8F8LL;
static uint64_t attribute_used __attribute__((aligned(8))) bFC=       0xFCFCFCFCFCFCFCFCLL;
185
static uint64_t                __attribute__((aligned(8))) w10=       0x0010001000100010LL;
186 187 188 189 190
static uint64_t attribute_used __attribute__((aligned(8))) w02=       0x0002000200020002LL;
static uint64_t attribute_used __attribute__((aligned(8))) bm00001111=0x00000000FFFFFFFFLL;
static uint64_t attribute_used __attribute__((aligned(8))) bm00000111=0x0000000000FFFFFFLL;
static uint64_t attribute_used __attribute__((aligned(8))) bm11111000=0xFFFFFFFFFF000000LL;
static uint64_t attribute_used __attribute__((aligned(8))) bm01010101=0x00FF00FF00FF00FFLL;
191

192 193 194 195
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
196 197

static uint64_t __attribute__((aligned(8))) dither4[2]={
198 199
        0x0103010301030103LL,
        0x0200020002000200LL,};
Michael Niedermayer's avatar
Michael Niedermayer committed
200 201

static uint64_t __attribute__((aligned(8))) dither8[2]={
202 203
        0x0602060206020602LL,
        0x0004000400040004LL,};
204

205
static uint64_t                __attribute__((aligned(8))) b16Mask=   0x001F001F001F001FLL;
206 207
static uint64_t attribute_used __attribute__((aligned(8))) g16Mask=   0x07E007E007E007E0LL;
static uint64_t attribute_used __attribute__((aligned(8))) r16Mask=   0xF800F800F800F800LL;
208
static uint64_t                __attribute__((aligned(8))) b15Mask=   0x001F001F001F001FLL;
209 210
static uint64_t attribute_used __attribute__((aligned(8))) g15Mask=   0x03E003E003E003E0LL;
static uint64_t attribute_used __attribute__((aligned(8))) r15Mask=   0x7C007C007C007C00LL;
211

212 213 214
static uint64_t attribute_used __attribute__((aligned(8))) M24A=      0x00FF0000FF0000FFLL;
static uint64_t attribute_used __attribute__((aligned(8))) M24B=      0xFF0000FF0000FF00LL;
static uint64_t attribute_used __attribute__((aligned(8))) M24C=      0x0000FF0000FF0000LL;
Michael Niedermayer's avatar
Michael Niedermayer committed
215

Michael Niedermayer's avatar
Michael Niedermayer committed
216
#ifdef FAST_BGR2YV12
217 218 219
static const uint64_t bgr2YCoeff   attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
static const uint64_t bgr2UCoeff   attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
static const uint64_t bgr2VCoeff   attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
Michael Niedermayer's avatar
Michael Niedermayer committed
220
#else
221 222 223
static const uint64_t bgr2YCoeff   attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
static const uint64_t bgr2UCoeff   attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
static const uint64_t bgr2VCoeff   attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
Diego Biurrun's avatar
Diego Biurrun committed
224
#endif /* FAST_BGR2YV12 */
225 226 227
static const uint64_t bgr2YOffset  attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8))) = 0x8080808080808080ULL;
static const uint64_t w1111        attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
228
#endif /* defined(ARCH_X86) */
229 230 231 232

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

233
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
234

Michael Niedermayer's avatar
Michael Niedermayer committed
235 236 237 238 239
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];
240

241 242 243 244
static const char * sws_context_to_name(void * ptr) {
    return "swscaler";
}

245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
#define OFFSET(x) offsetof(SwsContext, x)
#define DEFAULT 0
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM

static const AVOption options[] = {
    { "sws_flags", "scaler/cpu flags", OFFSET(flags), FF_OPT_TYPE_FLAGS, DEFAULT, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "fast_bilinear", "fast bilinear", 0, FF_OPT_TYPE_CONST, SWS_FAST_BILINEAR, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "bilinear", "bilinear", 0, FF_OPT_TYPE_CONST, SWS_BILINEAR, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "bicubic", "bicubic", 0, FF_OPT_TYPE_CONST, SWS_BICUBIC, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "experimental", "experimental", 0, FF_OPT_TYPE_CONST, SWS_X, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "neighbor", "nearest neighbor", 0, FF_OPT_TYPE_CONST, SWS_POINT, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "area", "averaging area", 0, FF_OPT_TYPE_CONST, SWS_AREA, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "bicublin", "luma bicubic, chroma bilinear", 0, FF_OPT_TYPE_CONST, SWS_BICUBLIN, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "gauss", "gaussian", 0, FF_OPT_TYPE_CONST, SWS_GAUSS, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "sinc", "sinc", 0, FF_OPT_TYPE_CONST, SWS_SINC, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "lanczos", "lanczos", 0, FF_OPT_TYPE_CONST, SWS_LANCZOS, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "spline", "natural bicubic spline", 0, FF_OPT_TYPE_CONST, SWS_SPLINE, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "print_info", "print info", 0, FF_OPT_TYPE_CONST, SWS_PRINT_INFO, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "accurate_rnd", "accurate rounding", 0, FF_OPT_TYPE_CONST, SWS_ACCURATE_RND, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "mmx", "MMX SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_MMX, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "mmx2", "MMX2 SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_MMX2, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "3dnow", "3DNOW SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_3DNOW, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "altivec", "AltiVec SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_ALTIVEC, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "bfin", "Blackfin SIMD acceleration", 0, FF_OPT_TYPE_CONST, SWS_CPU_CAPS_BFIN, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "full_chroma_int", "full chroma interpolation", 0 , FF_OPT_TYPE_CONST, SWS_FULL_CHR_H_INT, INT_MIN, INT_MAX, VE, "sws_flags" },
    { "full_chroma_inp", "full chroma input", 0 , FF_OPT_TYPE_CONST, SWS_FULL_CHR_H_INP, INT_MIN, INT_MAX, VE, "sws_flags" },
    { NULL }
};

#undef VE
#undef DEFAULT

static AVClass sws_context_class = { "SWScaler", sws_context_to_name, options };
278

279
char *sws_format_name(enum PixelFormat format)
280
{
281 282 283
    switch (format) {
        case PIX_FMT_YUV420P:
            return "yuv420p";
284 285
        case PIX_FMT_YUVA420P:
            return "yuva420p";
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
        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
306 307 308 309
        case PIX_FMT_GRAY16BE:
            return "gray16be";
        case PIX_FMT_GRAY16LE:
            return "gray16le";
310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
        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
358 359
        case PIX_FMT_YUV440P:
            return "yuv440p";
360 361 362
        default:
            return "Unknown format";
    }
363 364
}

365
#if defined(ARCH_X86) && defined (CONFIG_GPL)
Michael Niedermayer's avatar
Michael Niedermayer committed
366 367
void in_asm_used_var_warning_killer()
{
368 369 370 371
    volatile int i= bF8+bFC+w10+
    bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+
    M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101;
    if (i) i=0;
Michael Niedermayer's avatar
Michael Niedermayer committed
372 373
}
#endif
374

375
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
376 377
                               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
378
{
379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
    //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);
    }

    if (uDest != NULL)
        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];
                v += chrSrc[j][i + 2048] * chrFilter[j];
            }

            uDest[i]= av_clip_uint8(u>>19);
            vDest[i]= av_clip_uint8(v>>19);
        }
Michael Niedermayer's avatar
Michael Niedermayer committed
406 407
}

Ville Syrjälä's avatar
Ville Syrjälä committed
408
static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
409 410
                                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
411
{
412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
    //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);
    }

    if (uDest == NULL)
        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];
                v += chrSrc[j][i + 2048] * chrFilter[j];
            }

            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];
                v += chrSrc[j][i + 2048] * chrFilter[j];
            }

            uDest[2*i]= av_clip_uint8(v>>19);
            uDest[2*i+1]= av_clip_uint8(u>>19);
        }
Ville Syrjälä's avatar
Ville Syrjälä committed
457
}
Michael Niedermayer's avatar
Michael Niedermayer committed
458

Michael Niedermayer's avatar
Michael Niedermayer committed
459
#define YSCALE_YUV_2_PACKEDX_C(type) \
460 461 462 463 464 465
    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;\
466
        type av_unused *r, *b, *g;\
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
        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];\
            V += chrSrc[j][i+2048] * chrFilter[j];\
        }\
        Y1>>=19;\
        Y2>>=19;\
        U >>=19;\
        V >>=19;\
        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;    \
        }
494

Michael Niedermayer's avatar
Michael Niedermayer committed
495
#define YSCALE_YUV_2_RGBX_C(type) \
496 497 498 499 500 501 502 503 504 505 506 507
    YSCALE_YUV_2_PACKEDX_C(type)  \
    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;  \
        int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;  \
Michael Niedermayer's avatar
Michael Niedermayer committed
508 509

#define YSCALE_YUV_2_RGB2_C(type) \
510 511 512 513 514
    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
515

Michael Niedermayer's avatar
Michael Niedermayer committed
516
#define YSCALE_YUV_2_PACKED1_C \
517 518 519 520 521 522
    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;\
        int V= (uvbuf1[i+2048])>>7;\
Michael Niedermayer's avatar
Michael Niedermayer committed
523 524

#define YSCALE_YUV_2_RGB1_C(type) \
525 526 527 528 529
    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
530

Michael Niedermayer's avatar
Michael Niedermayer committed
531
#define YSCALE_YUV_2_PACKED1B_C \
532 533 534 535 536 537
    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;\
        int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\
Michael Niedermayer's avatar
Michael Niedermayer committed
538 539

#define YSCALE_YUV_2_RGB1B_C(type) \
540 541 542 543 544
    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
545

Michael Niedermayer's avatar
Michael Niedermayer committed
546
#define YSCALE_YUV_2_ANYRGB_C(func, func2)\
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
    switch(c->dstFormat)\
    {\
    case PIX_FMT_RGB32:\
    case PIX_FMT_BGR32:\
        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:\
        {\
            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++;\
            }\
Michael Niedermayer's avatar
Michael Niedermayer committed
658 659 660 661
\
/*\
((uint8_t*)dest)-= dstW>>4;\
{\
662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682
            int acc=0;\
            int left=0;\
            static int top[1024];\
            static int last_new[1024][1024];\
            static int last_in3[1024][1024];\
            static int drift[1024][1024];\
            int topLeft=0;\
            int shift=0;\
            int count=0;\
            const uint8_t * const d128=dither_8x8_220[y&7];\
            int error_new=0;\
            int error_in3=0;\
            int f=0;\
            \
            for (i=dstW>>1; i<dstW; i++){\
                int in= ((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19);\
                int in2 = (76309 * (in - 16) + 32768) >> 16;\
                int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\
                int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\
                         + (last_new[y][i] - in3)*f/256;\
                int new= old> 128 ? 255 : 0;\
Michael Niedermayer's avatar
Michael Niedermayer committed
683
\
684 685 686 687 688
                error_new+= FFABS(last_new[y][i] - new);\
                error_in3+= FFABS(last_in3[y][i] - in3);\
                f= error_new - error_in3*4;\
                if (f<0) f=0;\
                if (f>256) f=256;\
Michael Niedermayer's avatar
Michael Niedermayer committed
689
\
690 691 692 693
                topLeft= top[i];\
                left= top[i]= old - new;\
                last_new[y][i]= new;\
                last_in3[y][i]= in3;\
Michael Niedermayer's avatar
Michael Niedermayer committed
694
\
695 696 697 698 699 700
                acc+= acc + (new&1);\
                if ((i&7)==6){\
                    ((uint8_t*)dest)[0]= acc;\
                    ((uint8_t*)dest)++;\
                }\
            }\
Michael Niedermayer's avatar
Michael Niedermayer committed
701 702
}\
*/\
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721
        }\
        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;\
    }\
Michael Niedermayer's avatar
Michael Niedermayer committed
722 723


Michael Niedermayer's avatar
Michael Niedermayer committed
724
static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
725 726
                                  int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                                  uint8_t *dest, int dstW, int y)
Michael Niedermayer's avatar
Michael Niedermayer committed
727
{
728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872
    int i;
    switch(c->dstFormat)
    {
    case PIX_FMT_BGR32:
    case PIX_FMT_RGB32:
        YSCALE_YUV_2_RGBX_C(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:
        YSCALE_YUV_2_RGBX_C(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:
        YSCALE_YUV_2_RGBX_C(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];
            YSCALE_YUV_2_RGBX_C(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];
            YSCALE_YUV_2_RGBX_C(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];
            YSCALE_YUV_2_RGBX_C(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];
            YSCALE_YUV_2_RGBX_C(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];
            YSCALE_YUV_2_RGBX_C(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:
        {
            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++;
                }
            }
        }
        break;
    case PIX_FMT_YUYV422:
        YSCALE_YUV_2_PACKEDX_C(void)
            ((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:
        YSCALE_YUV_2_PACKEDX_C(void)
            ((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;
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
873 874 875
}


Michael Niedermayer's avatar
Michael Niedermayer committed
876 877
//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
//Plain C versions
878
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) || !defined(CONFIG_GPL)
879 880 881
#define COMPILE_C
#endif

882
#ifdef ARCH_POWERPC
883
#if (defined (HAVE_ALTIVEC) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
884 885 886 887
#define COMPILE_ALTIVEC
#endif //HAVE_ALTIVEC
#endif //ARCH_POWERPC

888
#if defined(ARCH_X86)
889

890
#if ((defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
891 892 893
#define COMPILE_MMX
#endif

894
#if (defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
895 896 897
#define COMPILE_MMX2
#endif

898
#if ((defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
899 900
#define COMPILE_3DNOW
#endif
901
#endif //ARCH_X86 || ARCH_X86_64
902 903 904 905 906 907

#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW

#ifdef COMPILE_C
Michael Niedermayer's avatar
Michael Niedermayer committed
908 909 910
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
911
#undef HAVE_ALTIVEC
Michael Niedermayer's avatar
Michael Niedermayer committed
912 913
#define RENAME(a) a ## _C
#include "swscale_template.c"
914
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
915

916 917 918 919 920 921 922 923 924
#ifdef ARCH_POWERPC
#ifdef COMPILE_ALTIVEC
#undef RENAME
#define HAVE_ALTIVEC
#define RENAME(a) a ## _altivec
#include "swscale_template.c"
#endif
#endif //ARCH_POWERPC

925
#if defined(ARCH_X86)
Michael Niedermayer's avatar
Michael Niedermayer committed
926

Michael Niedermayer's avatar
Michael Niedermayer committed
927 928 929 930 931 932 933 934 935
//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
936
*/
Michael Niedermayer's avatar
Michael Niedermayer committed
937
//MMX versions
938
#ifdef COMPILE_MMX
Michael Niedermayer's avatar
Michael Niedermayer committed
939 940 941 942 943 944
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define RENAME(a) a ## _MMX
#include "swscale_template.c"
945
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
946 947

//MMX2 versions
948
#ifdef COMPILE_MMX2
Michael Niedermayer's avatar
Michael Niedermayer committed
949 950 951 952 953 954
#undef RENAME
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
#define RENAME(a) a ## _MMX2
#include "swscale_template.c"
955
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
956 957

//3DNOW versions
958
#ifdef COMPILE_3DNOW
Michael Niedermayer's avatar
Michael Niedermayer committed
959 960 961 962 963 964
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
#define RENAME(a) a ## _3DNow
#include "swscale_template.c"
965
#endif
Michael Niedermayer's avatar
Michael Niedermayer committed
966

967
#endif //ARCH_X86 || ARCH_X86_64
Michael Niedermayer's avatar
Michael Niedermayer committed
968

Gabucino's avatar
Gabucino committed
969
// minor note: the HAVE_xyz is messed up after that line so don't use it
970

Michael Niedermayer's avatar
Michael Niedermayer committed
971 972
static double getSplineCoeff(double a, double b, double c, double d, double dist)
{
973 974 975 976 977 978 979
//    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
980
}
981

982
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
983 984
                             int srcW, int dstW, int filterAlign, int one, int flags,
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
985
{
986 987 988 989 990 991
    int i;
    int filterSize;
    int filter2Size;
    int minFilterSize;
    double *filter=NULL;
    double *filter2=NULL;
992
#if defined(ARCH_X86)
993
    if (flags & SWS_CPU_CAPS_MMX)
Diego Biurrun's avatar
Diego Biurrun committed
994
        asm volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
995
#endif
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 1031 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 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 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 1206
    // 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
            ASSERT(0)
        }

        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
                    ASSERT(0)
                }

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

    /* apply src & dst Filter to filter -> filter2
       av_free(filter);
    */
    ASSERT(filterSize>0)
    filter2Size= filterSize;
    if (srcFilter) filter2Size+= srcFilter->length - 1;
    if (dstFilter) filter2Size+= dstFilter->length - 1;
    ASSERT(filter2Size>0)
    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;

        ASSERT(outVec->length == filter2Size)
        //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);
    }
    av_free(filter); filter=NULL;

    /* try to reduce the filter-size (step1 find size and shift left) */
Diego Biurrun's avatar
Diego Biurrun committed
1207
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222
    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
1223
            /* preserve monotonicity because the core can't handle the filter otherwise */
1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
            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)
1250 1251
            filterAlign = 4;

1252 1253 1254 1255 1256 1257
        // 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)
1258
            filterAlign = 1;
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
    }

    if (flags & SWS_CPU_CAPS_MMX) {
        // special case for u