rgb2rgb_template.c

ffmpeg移植到symbian的全部源代码

        "punpckhwd     %5, %%mm5    \n\t"
        "psllq         $8, %%mm1    \n\t"
        "psllq        $16, %%mm2    \n\t"
        "por        %%mm1, %%mm0    \n\t"
        "por        %%mm2, %%mm0    \n\t"
        "psllq         $8, %%mm4    \n\t"
        "psllq        $16, %%mm5    \n\t"
        "por        %%mm4, %%mm3    \n\t"
        "por        %%mm5, %%mm3    \n\t"
        :"=m"(*d)
        :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null)
        :"memory");
        /* borrowed 32 to 24 */
        asm volatile(
        "movq       %%mm0, %%mm4    \n\t"
        "movq       %%mm3, %%mm5    \n\t"
        "movq       %%mm6, %%mm0    \n\t"
        "movq       %%mm7, %%mm1    \n\t"

        "movq       %%mm4, %%mm6    \n\t"
        "movq       %%mm5, %%mm7    \n\t"
        "movq       %%mm0, %%mm2    \n\t"
        "movq       %%mm1, %%mm3    \n\t"

        "psrlq         $8, %%mm2    \n\t"
        "psrlq         $8, %%mm3    \n\t"
        "psrlq         $8, %%mm6    \n\t"
        "psrlq         $8, %%mm7    \n\t"
        "pand          %2, %%mm0    \n\t"
        "pand          %2, %%mm1    \n\t"
        "pand          %2, %%mm4    \n\t"
        "pand          %2, %%mm5    \n\t"
        "pand          %3, %%mm2    \n\t"
        "pand          %3, %%mm3    \n\t"
        "pand          %3, %%mm6    \n\t"
        "pand          %3, %%mm7    \n\t"
        "por        %%mm2, %%mm0    \n\t"
        "por        %%mm3, %%mm1    \n\t"
        "por        %%mm6, %%mm4    \n\t"
        "por        %%mm7, %%mm5    \n\t"

        "movq       %%mm1, %%mm2    \n\t"
        "movq       %%mm4, %%mm3    \n\t"
        "psllq        $48, %%mm2    \n\t"
        "psllq        $32, %%mm3    \n\t"
        "pand          %4, %%mm2    \n\t"
        "pand          %5, %%mm3    \n\t"
        "por        %%mm2, %%mm0    \n\t"
        "psrlq        $16, %%mm1    \n\t"
        "psrlq        $32, %%mm4    \n\t"
        "psllq        $16, %%mm5    \n\t"
        "por        %%mm3, %%mm1    \n\t"
        "pand          %6, %%mm5    \n\t"
        "por        %%mm5, %%mm4    \n\t"

        MOVNTQ"     %%mm0,   %0     \n\t"
        MOVNTQ"     %%mm1,  8%0     \n\t"
        MOVNTQ"     %%mm4, 16%0"

        :"=m"(*d)
        :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh)
        :"memory");
        d += 24;
        s += 8;
    }
    asm volatile(SFENCE:::"memory");
    asm volatile(EMMS:::"memory");
#endif
    while (s < end)
    {
        register uint16_t bgr;
        bgr = *s++;
        *d++ = (bgr&0x1F)<<3;
        *d++ = (bgr&0x7E0)>>3;
        *d++ = (bgr&0xF800)>>8;
    }
}

static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size)
{
    const uint16_t *end;
#ifdef HAVE_MMX
    const uint16_t *mm_end;
#endif
    uint8_t *d = dst;
    const uint16_t *s = (const uint16_t *)src;
    end = s + src_size/2;
#ifdef HAVE_MMX
    asm volatile(PREFETCH"    %0"::"m"(*s):"memory");
    asm volatile("pxor    %%mm7,%%mm7    \n\t":::"memory");
    mm_end = end - 3;
    while (s < mm_end)
    {
        asm volatile(
        PREFETCH"    32%1           \n\t"
        "movq          %1, %%mm0    \n\t"
        "movq          %1, %%mm1    \n\t"
        "movq          %1, %%mm2    \n\t"
        "pand          %2, %%mm0    \n\t"
        "pand          %3, %%mm1    \n\t"
        "pand          %4, %%mm2    \n\t"
        "psllq         $3, %%mm0    \n\t"
        "psrlq         $2, %%mm1    \n\t"
        "psrlq         $7, %%mm2    \n\t"
        "movq       %%mm0, %%mm3    \n\t"
        "movq       %%mm1, %%mm4    \n\t"
        "movq       %%mm2, %%mm5    \n\t"
        "punpcklwd  %%mm7, %%mm0    \n\t"
        "punpcklwd  %%mm7, %%mm1    \n\t"
        "punpcklwd  %%mm7, %%mm2    \n\t"
        "punpckhwd  %%mm7, %%mm3    \n\t"
        "punpckhwd  %%mm7, %%mm4    \n\t"
        "punpckhwd  %%mm7, %%mm5    \n\t"
        "psllq         $8, %%mm1    \n\t"
        "psllq        $16, %%mm2    \n\t"
        "por        %%mm1, %%mm0    \n\t"
        "por        %%mm2, %%mm0    \n\t"
        "psllq         $8, %%mm4    \n\t"
        "psllq        $16, %%mm5    \n\t"
        "por        %%mm4, %%mm3    \n\t"
        "por        %%mm5, %%mm3    \n\t"
        MOVNTQ"     %%mm0,  %0      \n\t"
        MOVNTQ"     %%mm3, 8%0      \n\t"
        :"=m"(*d)
        :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r)
        :"memory");
        d += 16;
        s += 4;
    }
    asm volatile(SFENCE:::"memory");
    asm volatile(EMMS:::"memory");
#endif
    while (s < end)
    {
#if 0 //slightly slower on Athlon
        int bgr= *s++;
        *((uint32_t*)d)++ = ((bgr&0x1F)<<3) + ((bgr&0x3E0)<<6) + ((bgr&0x7C00)<<9);
#else
        register uint16_t bgr;
        bgr = *s++;
#ifdef WORDS_BIGENDIAN
        *d++ = 0;
        *d++ = (bgr&0x7C00)>>7;
        *d++ = (bgr&0x3E0)>>2;
        *d++ = (bgr&0x1F)<<3;
#else
        *d++ = (bgr&0x1F)<<3;
        *d++ = (bgr&0x3E0)>>2;
        *d++ = (bgr&0x7C00)>>7;
        *d++ = 0;
#endif

#endif
    }
}

static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size)
{
    const uint16_t *end;
#ifdef HAVE_MMX
    const uint16_t *mm_end;
#endif
    uint8_t *d = dst;
    const uint16_t *s = (const uint16_t*)src;
    end = s + src_size/2;
#ifdef HAVE_MMX
    asm volatile(PREFETCH"    %0"::"m"(*s):"memory");
    asm volatile("pxor    %%mm7,%%mm7    \n\t":::"memory");
    mm_end = end - 3;
    while (s < mm_end)
    {
        asm volatile(
        PREFETCH"    32%1           \n\t"
        "movq          %1, %%mm0    \n\t"
        "movq          %1, %%mm1    \n\t"
        "movq          %1, %%mm2    \n\t"
        "pand          %2, %%mm0    \n\t"
        "pand          %3, %%mm1    \n\t"
        "pand          %4, %%mm2    \n\t"
        "psllq         $3, %%mm0    \n\t"
        "psrlq         $3, %%mm1    \n\t"
        "psrlq         $8, %%mm2    \n\t"
        "movq       %%mm0, %%mm3    \n\t"
        "movq       %%mm1, %%mm4    \n\t"
        "movq       %%mm2, %%mm5    \n\t"
        "punpcklwd  %%mm7, %%mm0    \n\t"
        "punpcklwd  %%mm7, %%mm1    \n\t"
        "punpcklwd  %%mm7, %%mm2    \n\t"
        "punpckhwd  %%mm7, %%mm3    \n\t"
        "punpckhwd  %%mm7, %%mm4    \n\t"
        "punpckhwd  %%mm7, %%mm5    \n\t"
        "psllq         $8, %%mm1    \n\t"
        "psllq        $16, %%mm2    \n\t"
        "por        %%mm1, %%mm0    \n\t"
        "por        %%mm2, %%mm0    \n\t"
        "psllq         $8, %%mm4    \n\t"
        "psllq        $16, %%mm5    \n\t"
        "por        %%mm4, %%mm3    \n\t"
        "por        %%mm5, %%mm3    \n\t"
        MOVNTQ"     %%mm0, %0       \n\t"
        MOVNTQ"     %%mm3, 8%0      \n\t"
        :"=m"(*d)
        :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r)
        :"memory");
        d += 16;
        s += 4;
    }
    asm volatile(SFENCE:::"memory");
    asm volatile(EMMS:::"memory");
#endif
    while (s < end)
    {
        register uint16_t bgr;
        bgr = *s++;
#ifdef WORDS_BIGENDIAN
        *d++ = 0;
        *d++ = (bgr&0xF800)>>8;
        *d++ = (bgr&0x7E0)>>3;
        *d++ = (bgr&0x1F)<<3;
#else
        *d++ = (bgr&0x1F)<<3;
        *d++ = (bgr&0x7E0)>>3;
        *d++ = (bgr&0xF800)>>8;
        *d++ = 0;
#endif
    }
}

static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size)
{
    long idx = 15 - src_size;
    const uint8_t *s = src-idx;
    uint8_t *d = dst-idx;
#ifdef HAVE_MMX
    asm volatile(
    "test          %0, %0           \n\t"
    "jns           2f               \n\t"
    PREFETCH"       (%1, %0)        \n\t"
    "movq          %3, %%mm7        \n\t"
    "pxor          %4, %%mm7        \n\t"
    "movq       %%mm7, %%mm6        \n\t"
    "pxor          %5, %%mm7        \n\t"
    ASMALIGN(4)
    "1:                             \n\t"
    PREFETCH"     32(%1, %0)        \n\t"
    "movq           (%1, %0), %%mm0 \n\t"
    "movq          8(%1, %0), %%mm1 \n\t"
# ifdef HAVE_MMX2
    "pshufw      $177, %%mm0, %%mm3 \n\t"
    "pshufw      $177, %%mm1, %%mm5 \n\t"
    "pand       %%mm7, %%mm0        \n\t"
    "pand       %%mm6, %%mm3        \n\t"
    "pand       %%mm7, %%mm1        \n\t"
    "pand       %%mm6, %%mm5        \n\t"
    "por        %%mm3, %%mm0        \n\t"
    "por        %%mm5, %%mm1        \n\t"
# else
    "movq       %%mm0, %%mm2        \n\t"
    "movq       %%mm1, %%mm4        \n\t"
    "pand       %%mm7, %%mm0        \n\t"
    "pand       %%mm6, %%mm2        \n\t"
    "pand       %%mm7, %%mm1        \n\t"
    "pand       %%mm6, %%mm4        \n\t"
    "movq       %%mm2, %%mm3        \n\t"
    "movq       %%mm4, %%mm5        \n\t"
    "pslld        $16, %%mm2        \n\t"
    "psrld        $16, %%mm3        \n\t"
    "pslld        $16, %%mm4        \n\t"
    "psrld        $16, %%mm5        \n\t"
    "por        %%mm2, %%mm0        \n\t"
    "por        %%mm4, %%mm1        \n\t"
    "por        %%mm3, %%mm0        \n\t"
    "por        %%mm5, %%mm1        \n\t"
# endif
    MOVNTQ"     %%mm0,  (%2, %0)    \n\t"
    MOVNTQ"     %%mm1, 8(%2, %0)    \n\t"
    "add          $16, %0           \n\t"
    "js            1b               \n\t"
    SFENCE"                         \n\t"
    EMMS"                           \n\t"
    "2:                             \n\t"
    : "+&r"(idx)
    : "r" (s), "r" (d), "m" (mask32b), "m" (mask32r), "m" (mmx_one)
    : "memory");
#endif
    for (; idx<15; idx+=4) {
        register int v = *(const uint32_t *)&s[idx], g = v & 0xff00ff00;
        v &= 0xff00ff;
        *(uint32_t *)&d[idx] = (v>>16) + g + (v<<16);
    }
}

static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)
{
    unsigned i;
#ifdef HAVE_MMX
    long mmx_size= 23 - src_size;
    asm volatile (
    "test             %%"REG_a", %%"REG_a"          \n\t"
    "jns                     2f                     \n\t"
    "movq     "MANGLE(mask24r)", %%mm5              \n\t"
    "movq     "MANGLE(mask24g)", %%mm6              \n\t"
    "movq     "MANGLE(mask24b)", %%mm7              \n\t"
    ASMALIGN(4)
    "1:                                             \n\t"
    PREFETCH" 32(%1, %%"REG_a")                     \n\t"
    "movq       (%1, %%"REG_a"), %%mm0              \n\t" // BGR BGR BG
    "movq       (%1, %%"REG_a"), %%mm1              \n\t" // BGR BGR BG
    "movq      2(%1, %%"REG_a"), %%mm2              \n\t" // R BGR BGR B
    "psllq                  $16, %%mm0              \n\t" // 00 BGR BGR
    "pand                 %%mm5, %%mm0              \n\t"
    "pand                 %%mm6, %%mm1              \n\t"
    "pand                 %%mm7, %%mm2              \n\t"
    "por                  %%mm0, %%mm1              \n\t"
    "por                  %%mm2, %%mm1              \n\t"
    "movq      6(%1, %%"REG_a"), %%mm0              \n\t" // BGR BGR BG
    MOVNTQ"               %%mm1,   (%2, %%"REG_a")  \n\t" // RGB RGB RG
    "movq      8(%1, %%"REG_a"), %%mm1              \n\t" // R BGR BGR B
    "movq     10(%1, %%"REG_a"), %%mm2              \n\t" // GR BGR BGR
    "pand                 %%mm7, %%mm0              \n\t"
    "pand                 %%mm5, %%mm1              \n\t"
    "pand                 %%mm6, %%mm2              \n\t"
    "por                  %%mm0, %%mm1              \n\t"
    "por                  %%mm2, %%mm1              \n\t"
    "movq     14(%1, %%"REG_a"), %%mm0              \n\t" // R BGR BGR B
    MOVNTQ"               %%mm1,  8(%2, %%"REG_a")  \n\t" // B RGB RGB R
    "movq     16(%1, %%"REG_a"), %%mm1              \n\t" // GR BGR BGR
    "movq     18(%1, %%"REG_a"), %%mm2              \n\t" // BGR BGR BG
    "pand                 %%mm6, %%mm0              \n\t"
    "pand                 %%mm7, %%mm1              \n\t"
    "pand                 %%mm5, %%mm2              \n\t"
    "por                  %%mm0, %%mm1              \n\t"
    "por                  %%mm2, %%mm1              \n\t"
    MOVNTQ"               %%mm1, 16(%2, %%"REG_a")  \n\t"
    "add                    $24, %%"REG_a"          \n\t"
    " js                     1b                     \n\t"
    "2:                                             \n\t"
    : "+a" (mmx_size)
    : "r" (src-mmx_size), "r"(dst-mmx_size)
    );

    asm volatile(SFENCE:::"memory");
    asm volatile(EMMS:::"memory");

    if (mmx_size==23) return; //finished, was multiple of 8

    src+= src_size;
    dst+= src_size;
    src_size= 23-mmx_size;
    src-= src_size;
    dst-= src_size;
#endif
    for (i=0; i<src_size; i+=3)
    {
        register uint8_t x;
        x          = src[i + 2];
        dst[i + 1] = src[i + 1];
        dst[i + 2] = src[i + 0];
        dst[i + 0] = x;
    }
}

static inline void RENAME(yuvPlanartoyuy2)(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, long vertLumPerChroma)
{
    long y;
    const long chromWidth= width>>1;
    for (y=0; y<height; y++)
    {
#ifdef HAVE_MMX
//FIXME handle 2 lines at once (fewer prefetches, reuse some chroma, but very likely memory-limited anyway)
        asm volatile(
        "xor                 %%"REG_a", %%"REG_a"   \n\t"
        ASMALIGN(4)
        "1:                                         \n\t"
        PREFETCH"    32(%1, %%"REG_a", 2)           \n\t"
        PREFETCH"    32(%2, %%"REG_a")              \n\t"
        PREFETCH"    32(%3, %%"REG_a")              \n\t"