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
  12  * modify it under the terms of the GNU Lesser General Public
  13  * License as published by the Free Software Foundation; either
  14  * version 2.1 of the License, or (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 GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License 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 #include <inttypes.h>
  26 #include "config.h"
  27 #include "libavutil/x86_cpu.h"
  28 #include "libavutil/bswap.h"
  29 #include "rgb2rgb.h"
  30 #include "swscale.h"
  31 #include "swscale_internal.h"
  32
  33 #define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
  34
  35 void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
  36 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  37 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  38 void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  39 void (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
  40 void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
  41 void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
  42 void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  43 void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
  44 void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
  45 void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  46 void (*rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size);
  47 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
  48 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
  49 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
  50 void (*shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, long src_size);
  51 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
  52 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
  53
  54 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
  55                    long width, long height,
  56                    long lumStride, long chromStride, long dstStride);
  57 void (*yv12touyvy)(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 (*yuv422ptoyuy2)(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 (*yuv422ptouyvy)(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 (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
  67                    long width, long height,
  68                    long lumStride, long chromStride, long srcStride);
  69 void (*rgb24toyv12)(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 (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
  73                  long srcStride, long dstStride);
  74 void (*interleaveBytes)(const uint8_t *src1, const uint8_t *src2, uint8_t *dst,
  75                         long width, long height, long src1Stride,
  76                         long src2Stride, long dstStride);
  77 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
  78                     uint8_t *dst1, uint8_t *dst2,
  79                     long width, long height,
  80                     long srcStride1, long srcStride2,
  81                     long dstStride1, long dstStride2);
  82 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
  83                      uint8_t *dst,
  84                      long width, long height,
  85                      long srcStride1, long srcStride2,
  86                      long srcStride3, long dstStride);
  87 void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
  88                      long width, long height,
  89                      long lumStride, long chromStride, long srcStride);
  90 void (*uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
  91                      long width, long height,
  92                      long lumStride, long chromStride, long srcStride);
  93 void (*yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
  94                      long width, long height,
  95                      long lumStride, long chromStride, long srcStride);
  96 void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
  97                      long width, long height,
  98                      long lumStride, long chromStride, long srcStride);
  99
 100
 101 #if ARCH_X86
 102 DECLARE_ASM_CONST(8, uint64_t, mmx_ff)       = 0x00000000000000FFULL;
 103 DECLARE_ASM_CONST(8, uint64_t, mmx_null)     = 0x0000000000000000ULL;
 104 DECLARE_ASM_CONST(8, uint64_t, mmx_one)      = 0xFFFFFFFFFFFFFFFFULL;
 105 DECLARE_ASM_CONST(8, uint64_t, mask32b)      = 0x000000FF000000FFULL;
 106 DECLARE_ASM_CONST(8, uint64_t, mask32g)      = 0x0000FF000000FF00ULL;
 107 DECLARE_ASM_CONST(8, uint64_t, mask32r)      = 0x00FF000000FF0000ULL;
 108 DECLARE_ASM_CONST(8, uint64_t, mask32a)      = 0xFF000000FF000000ULL;
 109 DECLARE_ASM_CONST(8, uint64_t, mask32)       = 0x00FFFFFF00FFFFFFULL;
 110 DECLARE_ASM_CONST(8, uint64_t, mask3216br)   = 0x00F800F800F800F8ULL;
 111 DECLARE_ASM_CONST(8, uint64_t, mask3216g)    = 0x0000FC000000FC00ULL;
 112 DECLARE_ASM_CONST(8, uint64_t, mask3215g)    = 0x0000F8000000F800ULL;
 113 DECLARE_ASM_CONST(8, uint64_t, mul3216)      = 0x2000000420000004ULL;
 114 DECLARE_ASM_CONST(8, uint64_t, mul3215)      = 0x2000000820000008ULL;
 115 DECLARE_ASM_CONST(8, uint64_t, mask24b)      = 0x00FF0000FF0000FFULL;
 116 DECLARE_ASM_CONST(8, uint64_t, mask24g)      = 0xFF0000FF0000FF00ULL;
 117 DECLARE_ASM_CONST(8, uint64_t, mask24r)      = 0x0000FF0000FF0000ULL;
 118 DECLARE_ASM_CONST(8, uint64_t, mask24l)      = 0x0000000000FFFFFFULL;
 119 DECLARE_ASM_CONST(8, uint64_t, mask24h)      = 0x0000FFFFFF000000ULL;
 120 DECLARE_ASM_CONST(8, uint64_t, mask24hh)     = 0xffff000000000000ULL;
 121 DECLARE_ASM_CONST(8, uint64_t, mask24hhh)    = 0xffffffff00000000ULL;
 122 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh)   = 0xffffffffffff0000ULL;
 123 DECLARE_ASM_CONST(8, uint64_t, mask15b)      = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
 124 DECLARE_ASM_CONST(8, uint64_t, mask15rg)     = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
 125 DECLARE_ASM_CONST(8, uint64_t, mask15s)      = 0xFFE0FFE0FFE0FFE0ULL;
 126 DECLARE_ASM_CONST(8, uint64_t, mask15g)      = 0x03E003E003E003E0ULL;
 127 DECLARE_ASM_CONST(8, uint64_t, mask15r)      = 0x7C007C007C007C00ULL;
 128 #define mask16b mask15b
 129 DECLARE_ASM_CONST(8, uint64_t, mask16g)      = 0x07E007E007E007E0ULL;
 130 DECLARE_ASM_CONST(8, uint64_t, mask16r)      = 0xF800F800F800F800ULL;
 131 DECLARE_ASM_CONST(8, uint64_t, red_16mask)   = 0x0000f8000000f800ULL;
 132 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
 133 DECLARE_ASM_CONST(8, uint64_t, blue_16mask)  = 0x0000001f0000001fULL;
 134 DECLARE_ASM_CONST(8, uint64_t, red_15mask)   = 0x00007c0000007c00ULL;
 135 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
 136 DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;
 137 #endif /* ARCH_X86 */
 138
 139 #define RGB2YUV_SHIFT 8
 140 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
 141 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
 142 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 143 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
 144 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
 145 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
 146 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
 147 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
 148 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
 149
 150 //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
 151 //plain C versions
 152 #define COMPILE_TEMPLATE_MMX 0
 153 #define COMPILE_TEMPLATE_MMX2 0
 154 #define COMPILE_TEMPLATE_AMD3DNOW 0
 155 #define COMPILE_TEMPLATE_SSE2 0
 156 #define RENAME(a) a ## _C
 157 #include "rgb2rgb_template.c"
 158
 159 #if ARCH_X86
 160
 161 //MMX versions
 162 #undef RENAME
 163 #undef COMPILE_TEMPLATE_MMX
 164 #define COMPILE_TEMPLATE_MMX 1
 165 #define RENAME(a) a ## _MMX
 166 #include "rgb2rgb_template.c"
 167
 168 //MMX2 versions
 169 #undef RENAME
 170 #undef COMPILE_TEMPLATE_MMX2
 171 #define COMPILE_TEMPLATE_MMX2 1
 172 #define RENAME(a) a ## _MMX2
 173 #include "rgb2rgb_template.c"
 174
 175 //SSE2 versions
 176 #undef RENAME
 177 #undef COMPILE_TEMPLATE_SSE2
 178 #define COMPILE_TEMPLATE_SSE2 1
 179 #define RENAME(a) a ## _SSE2
 180 #include "rgb2rgb_template.c"
 181
 182 //3DNOW versions
 183 #undef RENAME
 184 #undef COMPILE_TEMPLATE_MMX2
 185 #undef COMPILE_TEMPLATE_SSE2
 186 #undef COMPILE_TEMPLATE_AMD3DNOW
 187 #define COMPILE_TEMPLATE_MMX2 0
 188 #define COMPILE_TEMPLATE_SSE2 1
 189 #define COMPILE_TEMPLATE_AMD3DNOW 1
 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 {
 204 #if HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX
 205     if (flags & SWS_CPU_CAPS_SSE2)
 206         rgb2rgb_init_SSE2();
 207     else if (flags & SWS_CPU_CAPS_MMX2)
 208         rgb2rgb_init_MMX2();
 209     else if (flags & SWS_CPU_CAPS_3DNOW)
 210         rgb2rgb_init_3DNOW();
 211     else if (flags & SWS_CPU_CAPS_MMX)
 212         rgb2rgb_init_MMX();
 213     else
 214 #endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
 215         rgb2rgb_init_C();
 216 }
 217
 218 #if LIBSWSCALE_VERSION_MAJOR < 1
 219 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 220 {
 221     sws_convertPalette8ToPacked32(src, dst, num_pixels, palette);
 222 }
 223
 224 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 225 {
 226     sws_convertPalette8ToPacked24(src, dst, num_pixels, palette);
 227 }
 228
 229 /**
 230  * Palette is assumed to contain BGR16, see rgb32to16 to convert the palette.
 231  */
 232 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 233 {
 234     long i;
 235     for (i=0; i<num_pixels; i++)
 236         ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
 237 }
 238 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
 239 {
 240     long i;
 241     for (i=0; i<num_pixels; i++)
 242         ((uint16_t *)dst)[i] = av_bswap16(((const uint16_t *)palette)[src[i]]);
 243 }
 244 #endif
 245
 246 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
 247 {
 248     long i;
 249     long num_pixels = src_size >> 2;
 250     for (i=0; i<num_pixels; i++) {
 251 #if HAVE_BIGENDIAN
 252         /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
 253         dst[3*i + 0] = src[4*i + 1];
 254         dst[3*i + 1] = src[4*i + 2];
 255         dst[3*i + 2] = src[4*i + 3];
 256 #else
 257         dst[3*i + 0] = src[4*i + 2];
 258         dst[3*i + 1] = src[4*i + 1];
 259         dst[3*i + 2] = src[4*i + 0];
 260 #endif
 261     }
 262 }
 263
 264 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
 265 {
 266     long i;
 267     for (i=0; 3*i<src_size; i++) {
 268 #if HAVE_BIGENDIAN
 269         /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
 270         dst[4*i + 0] = 255;
 271         dst[4*i + 1] = src[3*i + 0];
 272         dst[4*i + 2] = src[3*i + 1];
 273         dst[4*i + 3] = src[3*i + 2];
 274 #else
 275         dst[4*i + 0] = src[3*i + 2];
 276         dst[4*i + 1] = src[3*i + 1];
 277         dst[4*i + 2] = src[3*i + 0];
 278         dst[4*i + 3] = 255;
 279 #endif
 280     }
 281 }
 282
 283 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 284 {
 285     const uint16_t *end;
 286     uint8_t *d = dst;
 287     const uint16_t *s = (const uint16_t *)src;
 288     end = s + src_size/2;
 289     while (s < end) {
 290         register uint16_t bgr;
 291         bgr = *s++;
 292 #if HAVE_BIGENDIAN
 293         *d++ = 255;
 294         *d++ = (bgr&0x1F)<<3;
 295         *d++ = (bgr&0x7E0)>>3;
 296         *d++ = (bgr&0xF800)>>8;
 297 #else
 298         *d++ = (bgr&0xF800)>>8;
 299         *d++ = (bgr&0x7E0)>>3;
 300         *d++ = (bgr&0x1F)<<3;
 301         *d++ = 255;
 302 #endif
 303     }
 304 }
 305
 306 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
 307 {
 308     const uint16_t *end;
 309     uint8_t *d = dst;
 310     const uint16_t *s = (const uint16_t *)src;
 311     end = s + src_size/2;
 312     while (s < end) {
 313         register uint16_t bgr;
 314         bgr = *s++;
 315         *d++ = (bgr&0xF800)>>8;
 316         *d++ = (bgr&0x7E0)>>3;
 317         *d++ = (bgr&0x1F)<<3;
 318     }
 319 }
 320
 321 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 322 {
 323     long i;
 324     long num_pixels = src_size >> 1;
 325
 326     for (i=0; i<num_pixels; i++) {
 327         unsigned rgb = ((const uint16_t*)src)[i];
 328         ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
 329     }
 330 }
 331
 332 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 333 {
 334     long i;
 335     long num_pixels = src_size >> 1;
 336
 337     for (i=0; i<num_pixels; i++) {
 338         unsigned rgb = ((const uint16_t*)src)[i];
 339         ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
 340     }
 341 }
 342
 343 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
 344 {
 345     const uint16_t *end;
 346     uint8_t *d = dst;
 347     const uint16_t *s = (const uint16_t *)src;
 348     end = s + src_size/2;
 349     while (s < end) {
 350         register uint16_t bgr;
 351         bgr = *s++;
 352 #if HAVE_BIGENDIAN
 353         *d++ = 255;
 354         *d++ = (bgr&0x1F)<<3;
 355         *d++ = (bgr&0x3E0)>>2;
 356         *d++ = (bgr&0x7C00)>>7;
 357 #else
 358         *d++ = (bgr&0x7C00)>>7;
 359         *d++ = (bgr&0x3E0)>>2;
 360         *d++ = (bgr&0x1F)<<3;
 361         *d++ = 255;
 362 #endif
 363     }
 364 }
 365
 366 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
 367 {
 368     const uint16_t *end;
 369     uint8_t *d = dst;
 370     const uint16_t *s = (const uint16_t *)src;
 371     end = s + src_size/2;
 372     while (s < end) {
 373         register uint16_t bgr;
 374         bgr = *s++;
 375         *d++ = (bgr&0x7C00)>>7;
 376         *d++ = (bgr&0x3E0)>>2;
 377         *d++ = (bgr&0x1F)<<3;
 378     }
 379 }
 380
 381 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
 382 {
 383     long i;
 384     long num_pixels = src_size >> 1;
 385
 386     for (i=0; i<num_pixels; i++) {
 387         unsigned rgb = ((const uint16_t*)src)[i];
 388         ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
 389     }
 390 }
 391
 392 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
 393 {
 394     long i;
 395     long num_pixels = src_size >> 1;
 396
 397     for (i=0; i<num_pixels; i++) {
 398         unsigned br;
 399         unsigned rgb = ((const uint16_t*)src)[i];
 400         br = rgb&0x7c1F;
 401         ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
 402     }
 403 }
 404
 405 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
 406 {
 407     long i;
 408     long num_pixels = src_size;
 409     for (i=0; i<num_pixels; i++) {
 410         unsigned b,g,r;
 411         register uint8_t rgb;
 412         rgb = src[i];
 413         r = (rgb&0x07);
 414         g = (rgb&0x38)>>3;
 415         b = (rgb&0xC0)>>6;
 416         dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
 417     }
 418 }
 419
 420 #define DEFINE_SHUFFLE_BYTES(a, b, c, d)                                \
 421 void shuffle_bytes_##a##b##c##d(const uint8_t *src, uint8_t *dst, long src_size) \
 422 {                                                                       \
 423     long i;                                                             \
 424                                                                         \
 425     for (i = 0; i < src_size; i+=4) {                                   \
 426         dst[i + 0] = src[i + a];                                        \
 427         dst[i + 1] = src[i + b];                                        \
 428         dst[i + 2] = src[i + c];                                        \
 429         dst[i + 3] = src[i + d];                                        \
 430     }                                                                   \
 431 }
 432
 433 DEFINE_SHUFFLE_BYTES(0, 3, 2, 1);
 434 DEFINE_SHUFFLE_BYTES(1, 2, 3, 0);
 435 DEFINE_SHUFFLE_BYTES(3, 0, 1, 2);
 436 DEFINE_SHUFFLE_BYTES(3, 2, 1, 0);
 437