FFmpeg: libswscale/rgb2rgb.c Source File

00001 /*
00002  * software RGB to RGB converter
00003  * pluralize by software PAL8 to RGB converter
00004  *              software YUV to YUV converter
00005  *              software YUV to RGB converter
00006  * Written by Nick Kurshev.
00007  * palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)
00008  *
00009  * This file is part of FFmpeg.
00010  *
00011  * FFmpeg is free software; you can redistribute it and/or
00012  * modify it under the terms of the GNU Lesser General Public
00013  * License as published by the Free Software Foundation; either
00014  * version 2.1 of the License, or (at your option) any later version.
00015  *
00016  * FFmpeg is distributed in the hope that it will be useful,
00017  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00018  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00019  * Lesser General Public License for more details.
00020  *
00021  * You should have received a copy of the GNU Lesser General Public
00022  * License along with FFmpeg; if not, write to the Free Software
00023  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00024  */
00025 #include <inttypes.h>
00026 #include "config.h"
00027 #include "libavutil/x86_cpu.h"
00028 #include "libavutil/bswap.h"
00029 #include "rgb2rgb.h"
00030 #include "swscale.h"
00031 #include "swscale_internal.h"
00032 
00033 #define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients
00034 
00035 void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
00036 void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
00037 void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
00038 void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
00039 void (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
00040 void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
00041 void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
00042 void (*rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
00043 void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
00044 void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
00045 void (*rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
00046 void (*rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size);
00047 void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
00048 void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
00049 void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
00050 void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
00051 void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
00052 void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
00053 
00054 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
00055                    long width, long height,
00056                    long lumStride, long chromStride, long dstStride);
00057 void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
00058                    long width, long height,
00059                    long lumStride, long chromStride, long dstStride);
00060 void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
00061                       long width, long height,
00062                       long lumStride, long chromStride, long dstStride);
00063 void (*yuv422ptouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
00064                       long width, long height,
00065                       long lumStride, long chromStride, long dstStride);
00066 void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
00067                    long width, long height,
00068                    long lumStride, long chromStride, long srcStride);
00069 void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
00070                     long width, long height,
00071                     long lumStride, long chromStride, long srcStride);
00072 void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
00073                  long srcStride, long dstStride);
00074 void (*interleaveBytes)(const uint8_t *src1, const uint8_t *src2, uint8_t *dst,
00075                         long width, long height, long src1Stride,
00076                         long src2Stride, long dstStride);
00077 void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
00078                     uint8_t *dst1, uint8_t *dst2,
00079                     long width, long height,
00080                     long srcStride1, long srcStride2,
00081                     long dstStride1, long dstStride2);
00082 void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
00083                      uint8_t *dst,
00084                      long width, long height,
00085                      long srcStride1, long srcStride2,
00086                      long srcStride3, long dstStride);
00087 void (*uyvytoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
00088                      long width, long height,
00089                      long lumStride, long chromStride, long srcStride);
00090 void (*uyvytoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
00091                      long width, long height,
00092                      long lumStride, long chromStride, long srcStride);
00093 void (*yuyvtoyuv420)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
00094                      long width, long height,
00095                      long lumStride, long chromStride, long srcStride);
00096 void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t *src,
00097                      long width, long height,
00098                      long lumStride, long chromStride, long srcStride);
00099 
00100 
00101 #if ARCH_X86
00102 DECLARE_ASM_CONST(8, uint64_t, mmx_null)     = 0x0000000000000000ULL;
00103 DECLARE_ASM_CONST(8, uint64_t, mmx_one)      = 0xFFFFFFFFFFFFFFFFULL;
00104 DECLARE_ASM_CONST(8, uint64_t, mask32b)      = 0x000000FF000000FFULL;
00105 DECLARE_ASM_CONST(8, uint64_t, mask32g)      = 0x0000FF000000FF00ULL;
00106 DECLARE_ASM_CONST(8, uint64_t, mask32r)      = 0x00FF000000FF0000ULL;
00107 DECLARE_ASM_CONST(8, uint64_t, mask32a)      = 0xFF000000FF000000ULL;
00108 DECLARE_ASM_CONST(8, uint64_t, mask32)       = 0x00FFFFFF00FFFFFFULL;
00109 DECLARE_ASM_CONST(8, uint64_t, mask3216br)   = 0x00F800F800F800F8ULL;
00110 DECLARE_ASM_CONST(8, uint64_t, mask3216g)    = 0x0000FC000000FC00ULL;
00111 DECLARE_ASM_CONST(8, uint64_t, mask3215g)    = 0x0000F8000000F800ULL;
00112 DECLARE_ASM_CONST(8, uint64_t, mul3216)      = 0x2000000420000004ULL;
00113 DECLARE_ASM_CONST(8, uint64_t, mul3215)      = 0x2000000820000008ULL;
00114 DECLARE_ASM_CONST(8, uint64_t, mask24b)      = 0x00FF0000FF0000FFULL;
00115 DECLARE_ASM_CONST(8, uint64_t, mask24g)      = 0xFF0000FF0000FF00ULL;
00116 DECLARE_ASM_CONST(8, uint64_t, mask24r)      = 0x0000FF0000FF0000ULL;
00117 DECLARE_ASM_CONST(8, uint64_t, mask24l)      = 0x0000000000FFFFFFULL;
00118 DECLARE_ASM_CONST(8, uint64_t, mask24h)      = 0x0000FFFFFF000000ULL;
00119 DECLARE_ASM_CONST(8, uint64_t, mask24hh)     = 0xffff000000000000ULL;
00120 DECLARE_ASM_CONST(8, uint64_t, mask24hhh)    = 0xffffffff00000000ULL;
00121 DECLARE_ASM_CONST(8, uint64_t, mask24hhhh)   = 0xffffffffffff0000ULL;
00122 DECLARE_ASM_CONST(8, uint64_t, mask15b)      = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
00123 DECLARE_ASM_CONST(8, uint64_t, mask15rg)     = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
00124 DECLARE_ASM_CONST(8, uint64_t, mask15s)      = 0xFFE0FFE0FFE0FFE0ULL;
00125 DECLARE_ASM_CONST(8, uint64_t, mask15g)      = 0x03E003E003E003E0ULL;
00126 DECLARE_ASM_CONST(8, uint64_t, mask15r)      = 0x7C007C007C007C00ULL;
00127 #define mask16b mask15b
00128 DECLARE_ASM_CONST(8, uint64_t, mask16g)      = 0x07E007E007E007E0ULL;
00129 DECLARE_ASM_CONST(8, uint64_t, mask16r)      = 0xF800F800F800F800ULL;
00130 DECLARE_ASM_CONST(8, uint64_t, red_16mask)   = 0x0000f8000000f800ULL;
00131 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
00132 DECLARE_ASM_CONST(8, uint64_t, blue_16mask)  = 0x0000001f0000001fULL;
00133 DECLARE_ASM_CONST(8, uint64_t, red_15mask)   = 0x00007c0000007c00ULL;
00134 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
00135 DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;
00136 #endif /* ARCH_X86 */
00137 
00138 #define RGB2YUV_SHIFT 8
00139 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
00140 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
00141 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
00142 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
00143 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
00144 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
00145 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
00146 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
00147 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
00148 
00149 //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
00150 //plain C versions
00151 #undef HAVE_MMX
00152 #undef HAVE_MMX2
00153 #undef HAVE_AMD3DNOW
00154 #undef HAVE_SSE2
00155 #define HAVE_MMX 0
00156 #define HAVE_MMX2 0
00157 #define HAVE_AMD3DNOW 0
00158 #define HAVE_SSE2 0
00159 #define RENAME(a) a ## _C
00160 #include "rgb2rgb_template.c"
00161 
00162 #if ARCH_X86
00163 
00164 //MMX versions
00165 #undef RENAME
00166 #undef HAVE_MMX
00167 #define HAVE_MMX 1
00168 #define RENAME(a) a ## _MMX
00169 #include "rgb2rgb_template.c"
00170 
00171 //MMX2 versions
00172 #undef RENAME
00173 #undef HAVE_MMX2
00174 #define HAVE_MMX2 1
00175 #define RENAME(a) a ## _MMX2
00176 #include "rgb2rgb_template.c"
00177 
00178 //3DNOW versions
00179 #undef RENAME
00180 #undef HAVE_MMX2
00181 #undef HAVE_AMD3DNOW
00182 #define HAVE_MMX2 0
00183 #define HAVE_AMD3DNOW 1
00184 #define RENAME(a) a ## _3DNOW
00185 #include "rgb2rgb_template.c"
00186 
00187 #endif //ARCH_X86 || ARCH_X86_64
00188 
00189 /*
00190  RGB15->RGB16 original by Strepto/Astral
00191  ported to gcc & bugfixed : A'rpi
00192  MMX2, 3DNOW optimization by Nick Kurshev
00193  32-bit C version, and and&add trick by Michael Niedermayer
00194 */
00195 
00196 void sws_rgb2rgb_init(int flags)
00197 {
00198 #if HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX
00199     if (flags & SWS_CPU_CAPS_MMX2)
00200         rgb2rgb_init_MMX2();
00201     else if (flags & SWS_CPU_CAPS_3DNOW)
00202         rgb2rgb_init_3DNOW();
00203     else if (flags & SWS_CPU_CAPS_MMX)
00204         rgb2rgb_init_MMX();
00205     else
00206 #endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
00207         rgb2rgb_init_C();
00208 }
00209 
00210 #if LIBSWSCALE_VERSION_MAJOR < 1
00211 void palette8topacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
00212 {
00213     sws_convertPalette8ToPacked32(src, dst, num_pixels, palette);
00214 }
00215 
00216 void palette8topacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
00217 {
00218     sws_convertPalette8ToPacked24(src, dst, num_pixels, palette);
00219 }
00220 
00224 void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
00225 {
00226     long i;
00227     for (i=0; i<num_pixels; i++)
00228         ((uint16_t *)dst)[i] = ((const uint16_t *)palette)[src[i]];
00229 }
00230 void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
00231 {
00232     long i;
00233     for (i=0; i<num_pixels; i++)
00234         ((uint16_t *)dst)[i] = bswap_16(((const uint16_t *)palette)[src[i]]);
00235 }
00236 
00237 #endif
00238 
00239 void rgb32to24(const uint8_t *src, uint8_t *dst, long src_size)
00240 {
00241     long i;
00242     long num_pixels = src_size >> 2;
00243     for (i=0; i<num_pixels; i++) {
00244 #if HAVE_BIGENDIAN
00245         /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
00246         dst[3*i + 0] = src[4*i + 1];
00247         dst[3*i + 1] = src[4*i + 2];
00248         dst[3*i + 2] = src[4*i + 3];
00249 #else
00250         dst[3*i + 0] = src[4*i + 2];
00251         dst[3*i + 1] = src[4*i + 1];
00252         dst[3*i + 2] = src[4*i + 0];
00253 #endif
00254     }
00255 }
00256 
00257 void rgb24to32(const uint8_t *src, uint8_t *dst, long src_size)
00258 {
00259     long i;
00260     for (i=0; 3*i<src_size; i++) {
00261 #if HAVE_BIGENDIAN
00262         /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
00263         dst[4*i + 0] = 255;
00264         dst[4*i + 1] = src[3*i + 0];
00265         dst[4*i + 2] = src[3*i + 1];
00266         dst[4*i + 3] = src[3*i + 2];
00267 #else
00268         dst[4*i + 0] = src[3*i + 2];
00269         dst[4*i + 1] = src[3*i + 1];
00270         dst[4*i + 2] = src[3*i + 0];
00271         dst[4*i + 3] = 255;
00272 #endif
00273     }
00274 }
00275 
00276 void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
00277 {
00278     const uint16_t *end;
00279     uint8_t *d = dst;
00280     const uint16_t *s = (const uint16_t *)src;
00281     end = s + src_size/2;
00282     while (s < end) {
00283         register uint16_t bgr;
00284         bgr = *s++;
00285 #if HAVE_BIGENDIAN
00286         *d++ = 255;
00287         *d++ = (bgr&0x1F)<<3;
00288         *d++ = (bgr&0x7E0)>>3;
00289         *d++ = (bgr&0xF800)>>8;
00290 #else
00291         *d++ = (bgr&0xF800)>>8;
00292         *d++ = (bgr&0x7E0)>>3;
00293         *d++ = (bgr&0x1F)<<3;
00294         *d++ = 255;
00295 #endif
00296     }
00297 }
00298 
00299 void rgb16to24(const uint8_t *src, uint8_t *dst, long src_size)
00300 {
00301     const uint16_t *end;
00302     uint8_t *d = dst;
00303     const uint16_t *s = (const uint16_t *)src;
00304     end = s + src_size/2;
00305     while (s < end) {
00306         register uint16_t bgr;
00307         bgr = *s++;
00308         *d++ = (bgr&0xF800)>>8;
00309         *d++ = (bgr&0x7E0)>>3;
00310         *d++ = (bgr&0x1F)<<3;
00311     }
00312 }
00313 
00314 void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
00315 {
00316     long i;
00317     long num_pixels = src_size >> 1;
00318 
00319     for (i=0; i<num_pixels; i++) {
00320         unsigned rgb = ((const uint16_t*)src)[i];
00321         ((uint16_t*)dst)[i] = (rgb>>11) | (rgb&0x7E0) | (rgb<<11);
00322     }
00323 }
00324 
00325 void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
00326 {
00327     long i;
00328     long num_pixels = src_size >> 1;
00329 
00330     for (i=0; i<num_pixels; i++) {
00331         unsigned rgb = ((const uint16_t*)src)[i];
00332         ((uint16_t*)dst)[i] = (rgb>>11) | ((rgb&0x7C0)>>1) | ((rgb&0x1F)<<10);
00333     }
00334 }
00335 
00336 void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
00337 {
00338     const uint16_t *end;
00339     uint8_t *d = dst;
00340     const uint16_t *s = (const uint16_t *)src;
00341     end = s + src_size/2;
00342     while (s < end) {
00343         register uint16_t bgr;
00344         bgr = *s++;
00345 #if HAVE_BIGENDIAN
00346         *d++ = 255;
00347         *d++ = (bgr&0x1F)<<3;
00348         *d++ = (bgr&0x3E0)>>2;
00349         *d++ = (bgr&0x7C00)>>7;
00350 #else
00351         *d++ = (bgr&0x7C00)>>7;
00352         *d++ = (bgr&0x3E0)>>2;
00353         *d++ = (bgr&0x1F)<<3;
00354         *d++ = 255;
00355 #endif
00356     }
00357 }
00358 
00359 void rgb15to24(const uint8_t *src, uint8_t *dst, long src_size)
00360 {
00361     const uint16_t *end;
00362     uint8_t *d = dst;
00363     const uint16_t *s = (const uint16_t *)src;
00364     end = s + src_size/2;
00365     while (s < end) {
00366         register uint16_t bgr;
00367         bgr = *s++;
00368         *d++ = (bgr&0x7C00)>>7;
00369         *d++ = (bgr&0x3E0)>>2;
00370         *d++ = (bgr&0x1F)<<3;
00371     }
00372 }
00373 
00374 void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
00375 {
00376     long i;
00377     long num_pixels = src_size >> 1;
00378 
00379     for (i=0; i<num_pixels; i++) {
00380         unsigned rgb = ((const uint16_t*)src)[i];
00381         ((uint16_t*)dst)[i] = ((rgb&0x7C00)>>10) | ((rgb&0x3E0)<<1) | (rgb<<11);
00382     }
00383 }
00384 
00385 void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
00386 {
00387     long i;
00388     long num_pixels = src_size >> 1;
00389 
00390     for (i=0; i<num_pixels; i++) {
00391         unsigned br;
00392         unsigned rgb = ((const uint16_t*)src)[i];
00393         br = rgb&0x7c1F;
00394         ((uint16_t*)dst)[i] = (br>>10) | (rgb&0x3E0) | (br<<10);
00395     }
00396 }
00397 
00398 void bgr8torgb8(const uint8_t *src, uint8_t *dst, long src_size)
00399 {
00400     long i;
00401     long num_pixels = src_size;
00402     for (i=0; i<num_pixels; i++) {
00403         unsigned b,g,r;
00404         register uint8_t rgb;
00405         rgb = src[i];
00406         r = (rgb&0x07);
00407         g = (rgb&0x38)>>3;
00408         b = (rgb&0xC0)>>6;
00409         dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
00410     }
00411 }
00412 
00413 #define DEFINE_SHUFFLE_BYTES(a, b, c, d)                                \
00414 void shuffle_bytes_##a##b##c##d(const uint8_t *src, uint8_t *dst, long src_size) \
00415 {                                                                       \
00416     long i;                                                             \
00417                                                                         \
00418     for (i = 0; i < src_size; i+=4) {                                   \
00419         dst[i + 0] = src[i + a];                                        \
00420         dst[i + 1] = src[i + b];                                        \
00421         dst[i + 2] = src[i + c];                                        \
00422         dst[i + 3] = src[i + d];                                        \
00423     }                                                                   \
00424 }
00425 
00426 DEFINE_SHUFFLE_BYTES(0, 3, 2, 1);
00427 DEFINE_SHUFFLE_BYTES(1, 2, 3, 0);
00428 DEFINE_SHUFFLE_BYTES(2, 1, 0, 3);
00429 DEFINE_SHUFFLE_BYTES(3, 0, 1, 2);
00430 DEFINE_SHUFFLE_BYTES(3, 2, 1, 0);
00431