rgb2rgb.c.svn-base

ffmpeg最新源码

/*
 * software RGB to RGB converter
 * pluralize by software PAL8 to RGB converter
 *              software YUV to YUV converter
 *              software YUV to RGB converter
 * Written by Nick Kurshev.
 * palette & YUV & runtime CPU stuff by Michael ([e-mail])
 *
 * 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
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * The C code (not assembly, MMX, ...) of this file can be used
 * under the LGPL license.
 */
#include <inttypes.h>
#include "config.h"
#include "libavutil/x86_cpu.h"
#include "libavutil/bswap.h"
#include "rgb2rgb.h"
#include "swscale.h"
#include "swscale_internal.h"

#define FAST_BGR2YV12 // use 7-bit instead of 15-bit coefficients

void (*rgb24to32)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32to24)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb15to24)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb16to15)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb16to24)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size);
//void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size);
//void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);

void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
                   long width, long height,
                   long lumStride, long chromStride, long dstStride);
void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
                   long width, long height,
                   long lumStride, long chromStride, long dstStride);
void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
                      long width, long height,
                      long lumStride, long chromStride, long dstStride);
void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
                   long width, long height,
                   long lumStride, long chromStride, long srcStride);
void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
                    long width, long height,
                    long lumStride, long chromStride, long srcStride);
void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
                 long srcStride, long dstStride);
void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
                        long width, long height, long src1Stride,
                        long src2Stride, long dstStride);
void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
                    uint8_t *dst1, uint8_t *dst2,
                    long width, long height,
                    long srcStride1, long srcStride2,
                    long dstStride1, long dstStride2);
void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
                     uint8_t *dst,
                     long width, long height,
                     long srcStride1, long srcStride2,
                     long srcStride3, long dstStride);

#if defined(ARCH_X86) && defined(CONFIG_GPL)
DECLARE_ASM_CONST(8, uint64_t, mmx_null)     = 0x0000000000000000ULL;
DECLARE_ASM_CONST(8, uint64_t, mmx_one)      = 0xFFFFFFFFFFFFFFFFULL;
DECLARE_ASM_CONST(8, uint64_t, mask32b)      = 0x000000FF000000FFULL;
DECLARE_ASM_CONST(8, uint64_t, mask32g)      = 0x0000FF000000FF00ULL;
DECLARE_ASM_CONST(8, uint64_t, mask32r)      = 0x00FF000000FF0000ULL;
DECLARE_ASM_CONST(8, uint64_t, mask32)       = 0x00FFFFFF00FFFFFFULL;
DECLARE_ASM_CONST(8, uint64_t, mask3216br)   = 0x00F800F800F800F8ULL;
DECLARE_ASM_CONST(8, uint64_t, mask3216g)    = 0x0000FC000000FC00ULL;
DECLARE_ASM_CONST(8, uint64_t, mask3215g)    = 0x0000F8000000F800ULL;
DECLARE_ASM_CONST(8, uint64_t, mul3216)      = 0x2000000420000004ULL;
DECLARE_ASM_CONST(8, uint64_t, mul3215)      = 0x2000000820000008ULL;
DECLARE_ASM_CONST(8, uint64_t, mask24b)      = 0x00FF0000FF0000FFULL;
DECLARE_ASM_CONST(8, uint64_t, mask24g)      = 0xFF0000FF0000FF00ULL;
DECLARE_ASM_CONST(8, uint64_t, mask24r)      = 0x0000FF0000FF0000ULL;
DECLARE_ASM_CONST(8, uint64_t, mask24l)      = 0x0000000000FFFFFFULL;
DECLARE_ASM_CONST(8, uint64_t, mask24h)      = 0x0000FFFFFF000000ULL;
DECLARE_ASM_CONST(8, uint64_t, mask24hh)     = 0xffff000000000000ULL;
DECLARE_ASM_CONST(8, uint64_t, mask24hhh)    = 0xffffffff00000000ULL;
DECLARE_ASM_CONST(8, uint64_t, mask24hhhh)   = 0xffffffffffff0000ULL;
DECLARE_ASM_CONST(8, uint64_t, mask15b)      = 0x001F001F001F001FULL; /* 00000000 00011111  xxB */
DECLARE_ASM_CONST(8, uint64_t, mask15rg)     = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000  RGx */
DECLARE_ASM_CONST(8, uint64_t, mask15s)      = 0xFFE0FFE0FFE0FFE0ULL;
DECLARE_ASM_CONST(8, uint64_t, mask15g)      = 0x03E003E003E003E0ULL;
DECLARE_ASM_CONST(8, uint64_t, mask15r)      = 0x7C007C007C007C00ULL;
#define mask16b mask15b
DECLARE_ASM_CONST(8, uint64_t, mask16g)      = 0x07E007E007E007E0ULL;
DECLARE_ASM_CONST(8, uint64_t, mask16r)      = 0xF800F800F800F800ULL;
DECLARE_ASM_CONST(8, uint64_t, red_16mask)   = 0x0000f8000000f800ULL;
DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
DECLARE_ASM_CONST(8, uint64_t, blue_16mask)  = 0x0000001f0000001fULL;
DECLARE_ASM_CONST(8, uint64_t, red_15mask)   = 0x00007c0000007c00ULL;
DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;

#if 0
static volatile uint64_t __attribute__((aligned(8))) b5Dither;
static volatile uint64_t __attribute__((aligned(8))) g5Dither;
static volatile uint64_t __attribute__((aligned(8))) g6Dither;
static volatile uint64_t __attribute__((aligned(8))) r5Dither;

static uint64_t __attribute__((aligned(8))) dither4[2]={
    0x0103010301030103LL,
    0x0200020002000200LL,};

static uint64_t __attribute__((aligned(8))) dither8[2]={
    0x0602060206020602LL,
    0x0004000400040004LL,};
#endif
#endif /* defined(ARCH_X86) */

#define RGB2YUV_SHIFT 8
#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))

//Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
//plain C versions
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef HAVE_SSE2
#define RENAME(a) a ## _C
#include "rgb2rgb_template.c"

#if defined(ARCH_X86) && defined(CONFIG_GPL)

//MMX versions
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef HAVE_SSE2
#define RENAME(a) a ## _MMX
#include "rgb2rgb_template.c"

//MMX2 versions
#undef RENAME
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
#undef HAVE_SSE2
#define RENAME(a) a ## _MMX2
#include "rgb2rgb_template.c"

//3DNOW versions
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
#undef HAVE_SSE2
#define RENAME(a) a ## _3DNOW
#include "rgb2rgb_template.c"

#endif //ARCH_X86 || ARCH_X86_64

/*
 RGB15->RGB16 original by Strepto/Astral
 ported to gcc & bugfixed : A'rpi
 MMX2, 3DNOW optimization by Nick Kurshev
 32-bit C version, and and&add trick by Michael Niedermayer
*/

void sws_rgb2rgb_init(int flags){
#if (defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX))  && defined(CONFIG_GPL)
    if (flags & SWS_CPU_CAPS_MMX2)
        rgb2rgb_init_MMX2();
    else if (flags & SWS_CPU_CAPS_3DNOW)
        rgb2rgb_init_3DNOW();
    else if (flags & SWS_CPU_CAPS_MMX)
        rgb2rgb_init_MMX();
    else
#endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
        rgb2rgb_init_C();
}

/**
 * Palette is assumed to contain BGR32.
 */
void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{
    long i;

/*
    for (i=0; i<num_pixels; i++)
        ((unsigned *)dst)[i] = ((unsigned *)palette)[src[i]];
*/

    for (i=0; i<num_pixels; i++)
    {
        #ifdef WORDS_BIGENDIAN
            dst[3]= palette[src[i]*4+2];
            dst[2]= palette[src[i]*4+1];
            dst[1]= palette[src[i]*4+0];
        #else
        //FIXME slow?
            dst[0]= palette[src[i]*4+2];
            dst[1]= palette[src[i]*4+1];
            dst[2]= palette[src[i]*4+0];
            //dst[3]= 0; /* do we need this cleansing? */
        #endif
        dst+= 4;
    }
}

void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{
    long i;
    for (i=0; i<num_pixels; i++)
    {
        #ifdef WORDS_BIGENDIAN
            dst[3]= palette[src[i]*4+0];
            dst[2]= palette[src[i]*4+1];
            dst[1]= palette[src[i]*4+2];
        #else
            //FIXME slow?
            dst[0]= palette[src[i]*4+0];
            dst[1]= palette[src[i]*4+1];
            dst[2]= palette[src[i]*4+2];
            //dst[3]= 0; /* do we need this cleansing? */
        #endif

        dst+= 4;
    }
}

/**
 * Palette is assumed to contain BGR32.
 */
void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{
    long i;
/*