libswscale/rgb2rgb.c

   1 /*
   2  * software RGB to RGB converter
   3  * pluralize by software PAL8 to RGB converter
   4  *              software YUV to YUV converter
   5  *              software YUV to RGB converter
   6  * Written by Nick Kurshev.
   7  * palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)
   8  *
   9  * This file is part of FFmpeg.
  10  *
  11  * FFmpeg is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License as published by
  13  * the Free Software Foundation; either version 2 of the License, or
  14  * (at your option) any later version.
  15  *
  16  * FFmpeg is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19  * GNU General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU General Public License
  22  * along with FFmpeg; if not, write to the Free Software
  23  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  24  *
  25  * The C code (not assembly, MMX, ...) of this file can be used
  26  * under the LGPL license.
  27  */
  28 #include <inttypes.h>
  29 #include "config.h"
  30 #include "libavutil/x86_cpu.h"
  31 #include "libavutil/bswap.h"
  32 #include "rgb2rgb.h"
  33 #include "swscale.h"
  34 #include "swscale_internal.h"
  35
  36 #define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
  37
  38 void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  39 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  40 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  41 void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  42 void (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
  43 void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
  44 void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
  45 void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  46 void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
  47 void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
  48 void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  49 void (*rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size);
  50 //void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  51 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  52 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
  53 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
  54 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  55 //void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  56 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  57 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  58
  59 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  60                    long width, long height,
  61                    long lumStride, long chromStride, long dstStride);
  62 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  63                    long width, long height,
  64                    long lumStride, long chromStride, long dstStride);
  65 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  66                       long width, long height,
  67                       long lumStride, long chromStride, long dstStride);
  68 void (*yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  69                       long width, long height,
  70                       long lumStride, long chromStride, long dstStride);
  71 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  72                    long width, long height,
  73                    long lumStride, long chromStride, long srcStride);
  74 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  75                     long width, long height,
  76                     long lumStride, long chromStride, long srcStride);
  77 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
  78                  long srcStride, long dstStride);
  79 void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
  80                         long width, long height, long src1Stride,
  81                         long src2Stride, long dstStride);
  82 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
  83                     uint8_t *dst1, uint8_t *dst2,
  84                     long width, long height,
  85                     long srcStride1, long srcStride2,
  86                     long dstStride1, long dstStride2);
  87 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
  88                      uint8_t *dst,
  89                      long width, long height,
  90                      long srcStride1, long srcStride2,
  91                      long srcStride3, long dstStride);
  92
  93 #if defined(ARCH_X86) && defined(CONFIG_GPL)
  94 DECLARE_ASM_CONST(8, uint64_t, mmx_null)     = 0x0000000000000000ULL;
  95 DECLARE_ASM_CONST(8, uint64_t, mmx_one)      = 0xFFFFFFFFFFFFFFFFULL;
  96 DECLARE_ASM_CONST(8, uint64_t, mask32b)      = 0x000000FF000000FFULL;
  97 DECLARE_ASM_CONST(8, uint64_t, mask32g)      = 0x0000FF000000FF00ULL;
  98 DECLARE_ASM_CONST(8, uint64_t, mask32r)      = 0x00FF000000FF0000ULL;
  99 DECLARE_ASM_CONST(8, uint64_t, mask32)       = 0x00FFFFFF00FFFFFFULL;
 100 DECLARE_ASM_CONST(8, uint64_t, mask3216br)   = 0x00F800F800F800F8ULL;
 101 DECLARE_ASM_CONST(8, uint64_t, mask3216g)    = 0x0000FC000000FC00ULL;
 102 DECLARE_ASM_CONST(8, uint64_t, mask3215g)    = 0x0000F8000000F800ULL;
 103 DECLARE_ASM_CONST(8, uint64_t, mul3216)      = 0x2000000420000004ULL;
 104 DECLARE_ASM_CONST(8, uint64_t, mul3215)      = 0x2000000820000008ULL;
 105 DECLARE_ASM_CONST(8, uint64_t, mask24b)      = 0x00FF0000FF0000FFULL;
 106 DECLARE_ASM_CONST(8, uint64_t, mask24g)      = 0xFF0000FF0000FF00ULL;
 107 DECLARE_ASM_CONST(8, uint64_t, mask24r)      = 0x0000FF0000FF0000ULL;
 108 DECLARE_ASM_CONST(8, uint64_t, mask24l)      = 0x0000000000FFFFFFULL;
 109 DECLARE_ASM_CONST(8, uint64_t, mask24h)      = 0x0000FFFFFF000000ULL;
 110 DECLARE_ASM_CONST(8, uint64_t, mask24hh)     = 0xffff000000000000ULL;
 111 DECLARE_ASM_CONST(8, uint64_t, mask24hhh)    = 0xffffffff00000000ULL;
 112 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh)   = 0xffffffffffff0000ULL;
 113 DECLARE_ASM_CONST(8, uint64_t, mask15b)      = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
 114 DECLARE_ASM_CONST(8, uint64_t, mask15rg)     = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
 115 DECLARE_ASM_CONST(8, uint64_t, mask15s)      = 0xFFE0FFE0FFE0FFE0ULL;
 116 DECLARE_ASM_CONST(8, uint64_t, mask15g)      = 0x03E003E003E003E0ULL;
 117 DECLARE_ASM_CONST(8, uint64_t, mask15r)      = 0x7C007C007C007C00ULL;
 118 #define mask16b mask15b
 119 DECLARE_ASM_CONST(8, uint64_t, mask16g)      = 0x07E007E007E007E0ULL;
 120 DECLARE_ASM_CONST(8, uint64_t, mask16r)      = 0xF800F800F800F800ULL;
 121 DECLARE_ASM_CONST(8, uint64_t, red_16mask)   = 0x0000f8000000f800ULL;
 122 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
 123 DECLARE_ASM_CONST(8, uint64_t, blue_16mask)  = 0x0000001f0000001fULL;
 124 DECLARE_ASM_CONST(8, uint64_t, red_15mask)   = 0x00007c0000007c00ULL;
 125 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
 126 DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;
 127
 128 #if 0
 129 static volatile uint64_t __attribute__((aligned(8))) b5Dither;
 130 static volatile uint64_t __attribute__((aligned(8))) g5Dither;
 131 static volatile uint64_t __attribute__((aligned(8))) g6Dither;
 132 static volatile uint64_t __attribute__((aligned(8))) r5Dither;
 133
 134 static uint64_t __attribute__((aligned(8))) dither4[2]={
 135     0x0103010301030103LL,
 136     0x0200020002000200LL,};
 137
 138 static uint64_t __attribute__((aligned(8))) dither8[2]={
 139     0x0602060206020602LL,
 140     0x0004000400040004LL,};
 141 #endif
 142 #endif /* defined(ARCH_X86) */
 143
 144 #define RGB2YUV_SHIFT 8
 145 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
 146 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
 147 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 148 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
 149 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
 150 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
 151 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
 152 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 153 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
 154
 155 //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
 156 //plain C versions
 157 #undef HAVE_MMX
 158 #undef HAVE_MMX2
 159 #undef HAVE_3DNOW
 160 #undef HAVE_SSE2
 161 #define RENAME(a) a ## _C
 162 #include "rgb2rgb_template.c"
 163
 164 #if defined(ARCH_X86) && defined(CONFIG_GPL)
 165
 166 //MMX versions
 167 #undef RENAME
 168 #define HAVE_MMX
 169 #undef HAVE_MMX2
 170 #undef HAVE_3DNOW
 171 #undef HAVE_SSE2
 172 #define RENAME(a) a ## _MMX
 173 #include "rgb2rgb_template.c"
 174
 175 //MMX2 versions
 176 #undef RENAME
 177 #define HAVE_MMX
 178 #define HAVE_MMX2
 179 #undef HAVE_3DNOW
 180 #undef HAVE_SSE2
 181 #define RENAME(a) a ## _MMX2
 182 #include "rgb2rgb_template.c"
 183
 184 //3DNOW versions
 185 #undef RENAME
 186 #define HAVE_MMX
 187 #undef HAVE_MMX2
 188 #define HAVE_3DNOW
 189 #undef HAVE_SSE2
 190 #define RENAME(a) a ## _3DNOW
 191 #include "rgb2rgb_template.c"
 192
 193 #endif //ARCH_X86 || ARCH_X86_64
 194
 195 /*
 196  RGB15->RGB16 original by Strepto/Astral
 197  ported to gcc & bugfixed : A'rpi
 198  MMX2, 3DNOW optimization by Nick Kurshev
 199  32-bit C version, and and&add trick by Michael Niedermayer
 200 */
 201
 202 void sws_rgb2rgb_init(int flags){
 203 #if (defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX))  && defined(CONFIG_GPL)
 204     if (flags & SWS_CPU_CAPS_MMX2)
 205         rgb2rgb_init_MMX2();
 206     else if (flags & SWS_CPU_CAPS_3DNOW)
 207         rgb2rgb_init_3DNOW();
 208     else if (flags & SWS_CPU_CAPS_MMX)
 209         rgb2rgb_init_MMX();
 210     else
 211 #endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
 212         rgb2rgb_init_C();
 213 }
 214
 215 /**
 216  * Convert the palette to the same packet 32-bit format as the palette
 217  */
 218 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 219 {
 220     long i;
 221
 222     for (i=0; i<num_pixels; i++)
 223         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
 224 }
 225
 226 /**
 227  * Palette format: ABCD -> dst format: ABC
 228  */
 229 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 230 {
 231     long i;
 232
 233     for (i=0; i<num_pixels; i++)
 234     {
 235         //FIXME slow?
 236         dst[0]= palette[src[i]*4+0];
 237         dst[1]= palette[src[i]*4+1];
 238         dst[2]= palette[src[i]*4+2];
 239         dst+= 3;
 240     }
 241 }
 242
 243 /**
 244  * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
 245  */
 246 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 247 {
 248     long i;
 249     for (i=0; i<num_pixels; i++)
 250         ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
 251 }
 252 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 253 {
 254     long i;
 255     for (i=0; i<num_pixels; i++)
 256         ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
 257 }
 258
 259 /**
 260  * Palette is assumed to contain BGR15, see rgb32to15 to convert the palette.
 261  */
 262 void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 263 {
 264     long i;
 265     for (i=0; i<num_pixels; i++)
 266         ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
 267 }
 268 void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 269 {
 270     long i;
 271     for (i=0; i<num_pixels; i++)
 272         ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
 273 }
 274
 275 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
 276 {
 277     long i;
 278     long num_pixels = src_size >> 2;
 279     for (i=0; i<num_pixels; i++)
 280     {
 281         #ifdef WORDS_BIGENDIAN
 282             /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
 283             dst[3*i + 0] = src[4*i + 1];
 284             dst[3*i + 1] = src[4*i + 2];
 285             dst[3*i + 2] = src[4*i + 3];
 286         #else
 287             dst[3*i + 0] = src[4*i + 2];
 288             dst[3*i + 1] = src[4*i + 1];
 289             dst[3*i + 2] = src[4*i + 0];
 290         #endif
 291     }
 292 }
 293
 294 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
 295 {
 296     long i;
 297     for (i=0; 3*i<src_size; i++)
 298     {
 299         #ifdef WORDS_BIGENDIAN
 300             /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
 301             dst[4*i + 0] = 0;
 302             dst[4*i + 1] = src[3*i + 0];
 303             dst[4*i + 2] = src[3*i + 1];
 304             dst[4*i + 3] = src[3*i + 2];
 305         #else
 306             dst[4*i + 0] = src[3*i + 2];
 307             dst[4*i + 1] = src[3*i + 1];
 308             dst[4*i + 2] = src[3*i + 0];
 309             dst[4*i + 3] = 0;
 310         #endif
 311     }
 312 }
 313
 314 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 315 {
 316     const uint16_t *end;
 317     uint8_t *d = dst;
 318     const uint16_t *s = (const uint16_t *)src;
 319     end = s + src_size/2;
 320     while (s < end)
 321     {
 322         register uint16_t bgr;
 323         bgr = *s++;
 324         #ifdef WORDS_BIGENDIAN
 325             *d++ = 0;
 326             *d++ = (bgr&0x1F)<<3;
 327             *d++ = (bgr&0x7E0)>>3;
 328             *d++ = (bgr&0xF800)>>8;
 329         #else
 330             *d++ = (bgr&0xF800)>>8;
 331             *d++ = (bgr&0x7E0)>>3;
 332             *d++ = (bgr&0x1F)<<3;
 333             *d++ = 0;
 334         #endif
 335     }
 336 }
 337
 338 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
 339 {
 340     const uint16_t *end;
 341     uint8_t *d = dst;
 342     const uint16_t *s = (const uint16_t *)src;
 343     end = s + src_size/2;
 344     while (s < end)
 345     {
 346         register uint16_t bgr;
 347         bgr = *s++;
 348         *d++ = (bgr&0xF800)>>8;
 349         *d++ = (bgr&0x7E0)>>3;
 350         *d++ = (bgr&0x1F)<<3;
 351     }
 352 }
 353
 354 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 355 {
 356     long i;
 357     long num_pixels = src_size >> 1;
 358
 359     for (i=0; i<num_pixels; i++)
 360     {
 361         unsigned rgb = ((const uint16_t*)src)[i];
 362         ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
 363     }
 364 }
 365
 366 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 367 {
 368     long i;
 369     long num_pixels = src_size >> 1;
 370
 371     for (i=0; i<num_pixels; i++)
 372     {
 373         unsigned rgb = ((const uint16_t*)src)[i];
 374         ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
 375     }
 376 }
 377
 378 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 379 {
 380     const uint16_t *end;
 381     uint8_t *d = dst;
 382     const uint16_t *s = (const uint16_t *)src;
 383     end = s + src_size/2;
 384     while (s < end)
 385     {
 386         register uint16_t bgr;
 387         bgr = *s++;
 388         #ifdef WORDS_BIGENDIAN
 389             *d++ = 0;
 390             *d++ = (bgr&0x1F)<<3;
 391             *d++ = (bgr&0x3E0)>>2;
 392             *d++ = (bgr&0x7C00)>>7;
 393         #else
 394             *d++ = (bgr&0x7C00)>>7;
 395             *d++ = (bgr&0x3E0)>>2;
 396             *d++ = (bgr&0x1F)<<3;
 397             *d++ = 0;
 398         #endif
 399     }
 400 }
 401
 402 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
 403 {
 404     const uint16_t *end;
 405     uint8_t *d = dst;
 406     const uint16_t *s = (const uint16_t *)src;
 407     end = s + src_size/2;
 408     while (s < end)
 409     {
 410         register uint16_t bgr;
 411         bgr = *s++;
 412         *d++ = (bgr&0x7C00)>>7;
 413         *d++ = (bgr&0x3E0)>>2;
 414         *d++ = (bgr&0x1F)<<3;
 415     }
 416 }
 417
 418 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 419 {
 420     long i;
 421     long num_pixels = src_size >> 1;
 422
 423     for (i=0; i<num_pixels; i++)
 424     {
 425         unsigned rgb = ((const uint16_t*)src)[i];
 426         ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
 427     }
 428 }
 429
 430 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 431 {
 432     long i;
 433     long num_pixels = src_size >> 1;
 434
 435     for (i=0; i<num_pixels; i++)
 436     {
 437         unsigned br;
 438         unsigned rgb = ((const uint16_t*)src)[i];
 439         br = rgb&0x7c1F;
 440         ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
 441     }
 442 }
 443
 444 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
 445 {
 446     long i;
 447     long num_pixels = src_size;
 448     for (i=0; i<num_pixels; i++)
 449     {
 450         unsigned b,g,r;
 451         register uint8_t rgb;
 452         rgb = src[i];
 453         r = (rgb&0x07);
 454         g = (rgb&0x38)>>3;
 455         b = (rgb&0xC0)>>6;
 456         dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
 457     }
 458 }