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 (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  51 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
  52 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
  53 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  54 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  55 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  56
  57 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  58                    long width, long height,
  59                    long lumStride, long chromStride, long dstStride);
  60 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  61                    long width, long height,
  62                    long lumStride, long chromStride, long dstStride);
  63 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  64                       long width, long height,
  65                       long lumStride, long chromStride, long dstStride);
  66 void (*yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  67                       long width, long height,
  68                       long lumStride, long chromStride, long dstStride);
  69 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  70                    long width, long height,
  71                    long lumStride, long chromStride, long srcStride);
  72 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  73                     long width, long height,
  74                     long lumStride, long chromStride, long srcStride);
  75 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
  76                  long srcStride, long dstStride);
  77 void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
  78                         long width, long height, long src1Stride,
  79                         long src2Stride, long dstStride);
  80 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
  81                     uint8_t *dst1, uint8_t *dst2,
  82                     long width, long height,
  83                     long srcStride1, long srcStride2,
  84                     long dstStride1, long dstStride2);
  85 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
  86                      uint8_t *dst,
  87                      long width, long height,
  88                      long srcStride1, long srcStride2,
  89                      long srcStride3, long dstStride);
  90
  91 #if ARCH_X86 && CONFIG_GPL
  92 DECLARE_ASM_CONST(8, uint64_t, mmx_null)     = 0x0000000000000000ULL;
  93 DECLARE_ASM_CONST(8, uint64_t, mmx_one)      = 0xFFFFFFFFFFFFFFFFULL;
  94 DECLARE_ASM_CONST(8, uint64_t, mask32b)      = 0x000000FF000000FFULL;
  95 DECLARE_ASM_CONST(8, uint64_t, mask32g)      = 0x0000FF000000FF00ULL;
  96 DECLARE_ASM_CONST(8, uint64_t, mask32r)      = 0x00FF000000FF0000ULL;
  97 DECLARE_ASM_CONST(8, uint64_t, mask32a)      = 0xFF000000FF000000ULL;
  98 DECLARE_ASM_CONST(8, uint64_t, mask32)       = 0x00FFFFFF00FFFFFFULL;
  99 DECLARE_ASM_CONST(8, uint64_t, mask3216br)   = 0x00F800F800F800F8ULL;
 100 DECLARE_ASM_CONST(8, uint64_t, mask3216g)    = 0x0000FC000000FC00ULL;
 101 DECLARE_ASM_CONST(8, uint64_t, mask3215g)    = 0x0000F8000000F800ULL;
 102 DECLARE_ASM_CONST(8, uint64_t, mul3216)      = 0x2000000420000004ULL;
 103 DECLARE_ASM_CONST(8, uint64_t, mul3215)      = 0x2000000820000008ULL;
 104 DECLARE_ASM_CONST(8, uint64_t, mask24b)      = 0x00FF0000FF0000FFULL;
 105 DECLARE_ASM_CONST(8, uint64_t, mask24g)      = 0xFF0000FF0000FF00ULL;
 106 DECLARE_ASM_CONST(8, uint64_t, mask24r)      = 0x0000FF0000FF0000ULL;
 107 DECLARE_ASM_CONST(8, uint64_t, mask24l)      = 0x0000000000FFFFFFULL;
 108 DECLARE_ASM_CONST(8, uint64_t, mask24h)      = 0x0000FFFFFF000000ULL;
 109 DECLARE_ASM_CONST(8, uint64_t, mask24hh)     = 0xffff000000000000ULL;
 110 DECLARE_ASM_CONST(8, uint64_t, mask24hhh)    = 0xffffffff00000000ULL;
 111 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh)   = 0xffffffffffff0000ULL;
 112 DECLARE_ASM_CONST(8, uint64_t, mask15b)      = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
 113 DECLARE_ASM_CONST(8, uint64_t, mask15rg)     = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
 114 DECLARE_ASM_CONST(8, uint64_t, mask15s)      = 0xFFE0FFE0FFE0FFE0ULL;
 115 DECLARE_ASM_CONST(8, uint64_t, mask15g)      = 0x03E003E003E003E0ULL;
 116 DECLARE_ASM_CONST(8, uint64_t, mask15r)      = 0x7C007C007C007C00ULL;
 117 #define mask16b mask15b
 118 DECLARE_ASM_CONST(8, uint64_t, mask16g)      = 0x07E007E007E007E0ULL;
 119 DECLARE_ASM_CONST(8, uint64_t, mask16r)      = 0xF800F800F800F800ULL;
 120 DECLARE_ASM_CONST(8, uint64_t, red_16mask)   = 0x0000f8000000f800ULL;
 121 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
 122 DECLARE_ASM_CONST(8, uint64_t, blue_16mask)  = 0x0000001f0000001fULL;
 123 DECLARE_ASM_CONST(8, uint64_t, red_15mask)   = 0x00007c0000007c00ULL;
 124 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
 125 DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;
 126 #endif /* ARCH_X86 */
 127
 128 #define RGB2YUV_SHIFT 8
 129 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
 130 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
 131 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 132 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
 133 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
 134 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
 135 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
 136 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 137 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
 138
 139 //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
 140 //plain C versions
 141 #undef HAVE_MMX
 142 #undef HAVE_MMX2
 143 #undef HAVE_AMD3DNOW
 144 #undef HAVE_SSE2
 145 #define HAVE_MMX 0
 146 #define HAVE_MMX2 0
 147 #define HAVE_AMD3DNOW 0
 148 #define HAVE_SSE2 0
 149 #define RENAME(a) a ## _C
 150 #include "rgb2rgb_template.c"
 151
 152 #if ARCH_X86 && CONFIG_GPL
 153
 154 //MMX versions
 155 #undef RENAME
 156 #undef HAVE_MMX
 157 #define HAVE_MMX 1
 158 #define RENAME(a) a ## _MMX
 159 #include "rgb2rgb_template.c"
 160
 161 //MMX2 versions
 162 #undef RENAME
 163 #undef HAVE_MMX2
 164 #define HAVE_MMX2 1
 165 #define RENAME(a) a ## _MMX2
 166 #include "rgb2rgb_template.c"
 167
 168 //3DNOW versions
 169 #undef RENAME
 170 #undef HAVE_MMX2
 171 #undef HAVE_AMD3DNOW
 172 #define HAVE_MMX2 0
 173 #define HAVE_AMD3DNOW 1
 174 #define RENAME(a) a ## _3DNOW
 175 #include "rgb2rgb_template.c"
 176
 177 #endif //ARCH_X86 || ARCH_X86_64
 178
 179 /*
 180  RGB15->RGB16 original by Strepto/Astral
 181  ported to gcc & bugfixed : A'rpi
 182  MMX2, 3DNOW optimization by Nick Kurshev
 183  32-bit C version, and and&add trick by Michael Niedermayer
 184 */
 185
 186 void sws_rgb2rgb_init(int flags){
 187 #if (HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX)  && CONFIG_GPL
 188     if (flags & SWS_CPU_CAPS_MMX2)
 189         rgb2rgb_init_MMX2();
 190     else if (flags & SWS_CPU_CAPS_3DNOW)
 191         rgb2rgb_init_3DNOW();
 192     else if (flags & SWS_CPU_CAPS_MMX)
 193         rgb2rgb_init_MMX();
 194     else
 195 #endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
 196         rgb2rgb_init_C();
 197 }
 198
 199 /**
 200  * Convert the palette to the same packet 32-bit format as the palette
 201  */
 202 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 203 {
 204     long i;
 205
 206     for (i=0; i<num_pixels; i++)
 207         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
 208 }
 209
 210 /**
 211  * Palette format: ABCD -> dst format: ABC
 212  */
 213 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 214 {
 215     long i;
 216
 217     for (i=0; i<num_pixels; i++)
 218     {
 219         //FIXME slow?
 220         dst[0]= palette[src[i]*4+0];
 221         dst[1]= palette[src[i]*4+1];
 222         dst[2]= palette[src[i]*4+2];
 223         dst+= 3;
 224     }
 225 }
 226
 227 /**
 228  * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
 229  */
 230 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 231 {
 232     long i;
 233     for (i=0; i<num_pixels; i++)
 234         ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
 235 }
 236 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 237 {
 238     long i;
 239     for (i=0; i<num_pixels; i++)
 240         ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
 241 }
 242
 243 /**
 244  * Palette is assumed to contain BGR15, see rgb32to15 to convert the palette.
 245  */
 246 void palette8torgb15(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 palette8tobgr15(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 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
 260 {
 261     long i;
 262     long num_pixels = src_size >> 2;
 263     for (i=0; i<num_pixels; i++)
 264     {
 265         #ifdef WORDS_BIGENDIAN
 266             /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
 267             dst[3*i + 0] = src[4*i + 1];
 268             dst[3*i + 1] = src[4*i + 2];
 269             dst[3*i + 2] = src[4*i + 3];
 270         #else
 271             dst[3*i + 0] = src[4*i + 2];
 272             dst[3*i + 1] = src[4*i + 1];
 273             dst[3*i + 2] = src[4*i + 0];
 274         #endif
 275     }
 276 }
 277
 278 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
 279 {
 280     long i;
 281     for (i=0; 3*i<src_size; i++)
 282     {
 283         #ifdef WORDS_BIGENDIAN
 284             /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
 285             dst[4*i + 0] = 255;
 286             dst[4*i + 1] = src[3*i + 0];
 287             dst[4*i + 2] = src[3*i + 1];
 288             dst[4*i + 3] = src[3*i + 2];
 289         #else
 290             dst[4*i + 0] = src[3*i + 2];
 291             dst[4*i + 1] = src[3*i + 1];
 292             dst[4*i + 2] = src[3*i + 0];
 293             dst[4*i + 3] = 255;
 294         #endif
 295     }
 296 }
 297
 298 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 299 {
 300     const uint16_t *end;
 301     uint8_t *d = dst;
 302     const uint16_t *s = (const uint16_t *)src;
 303     end = s + src_size/2;
 304     while (s < end)
 305     {
 306         register uint16_t bgr;
 307         bgr = *s++;
 308         #ifdef WORDS_BIGENDIAN
 309             *d++ = 255;
 310             *d++ = (bgr&0x1F)<<3;
 311             *d++ = (bgr&0x7E0)>>3;
 312             *d++ = (bgr&0xF800)>>8;
 313         #else
 314             *d++ = (bgr&0xF800)>>8;
 315             *d++ = (bgr&0x7E0)>>3;
 316             *d++ = (bgr&0x1F)<<3;
 317             *d++ = 255;
 318         #endif
 319     }
 320 }
 321
 322 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
 323 {
 324     const uint16_t *end;
 325     uint8_t *d = dst;
 326     const uint16_t *s = (const uint16_t *)src;
 327     end = s + src_size/2;
 328     while (s < end)
 329     {
 330         register uint16_t bgr;
 331         bgr = *s++;
 332         *d++ = (bgr&0xF800)>>8;
 333         *d++ = (bgr&0x7E0)>>3;
 334         *d++ = (bgr&0x1F)<<3;
 335     }
 336 }
 337
 338 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 339 {
 340     long i;
 341     long num_pixels = src_size >> 1;
 342
 343     for (i=0; i<num_pixels; i++)
 344     {
 345         unsigned rgb = ((const uint16_t*)src)[i];
 346         ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
 347     }
 348 }
 349
 350 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 351 {
 352     long i;
 353     long num_pixels = src_size >> 1;
 354
 355     for (i=0; i<num_pixels; i++)
 356     {
 357         unsigned rgb = ((const uint16_t*)src)[i];
 358         ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
 359     }
 360 }
 361
 362 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 363 {
 364     const uint16_t *end;
 365     uint8_t *d = dst;
 366     const uint16_t *s = (const uint16_t *)src;
 367     end = s + src_size/2;
 368     while (s < end)
 369     {
 370         register uint16_t bgr;
 371         bgr = *s++;
 372         #ifdef WORDS_BIGENDIAN
 373             *d++ = 255;
 374             *d++ = (bgr&0x1F)<<3;
 375             *d++ = (bgr&0x3E0)>>2;
 376             *d++ = (bgr&0x7C00)>>7;
 377         #else
 378             *d++ = (bgr&0x7C00)>>7;
 379             *d++ = (bgr&0x3E0)>>2;
 380             *d++ = (bgr&0x1F)<<3;
 381             *d++ = 255;
 382         #endif
 383     }
 384 }
 385
 386 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
 387 {
 388     const uint16_t *end;
 389     uint8_t *d = dst;
 390     const uint16_t *s = (const uint16_t *)src;
 391     end = s + src_size/2;
 392     while (s < end)
 393     {
 394         register uint16_t bgr;
 395         bgr = *s++;
 396         *d++ = (bgr&0x7C00)>>7;
 397         *d++ = (bgr&0x3E0)>>2;
 398         *d++ = (bgr&0x1F)<<3;
 399     }
 400 }
 401
 402 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 403 {
 404     long i;
 405     long num_pixels = src_size >> 1;
 406
 407     for (i=0; i<num_pixels; i++)
 408     {
 409         unsigned rgb = ((const uint16_t*)src)[i];
 410         ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
 411     }
 412 }
 413
 414 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 415 {
 416     long i;
 417     long num_pixels = src_size >> 1;
 418
 419     for (i=0; i<num_pixels; i++)
 420     {
 421         unsigned br;
 422         unsigned rgb = ((const uint16_t*)src)[i];
 423         br = rgb&0x7c1F;
 424         ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
 425     }
 426 }
 427
 428 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
 429 {
 430     long i;
 431     long num_pixels = src_size;
 432     for (i=0; i<num_pixels; i++)
 433     {
 434         unsigned b,g,r;
 435         register uint8_t rgb;
 436         rgb = src[i];
 437         r = (rgb&0x07);
 438         g = (rgb&0x38)>>3;
 439         b = (rgb&0xC0)>>6;
 440         dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
 441     }
 442 }