h264dsp_mmx.c

ffmpeg的完整源代码和作者自己写的文档。不但有在Linux的工程哦

/*
 * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 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
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser 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
 */


/***********************************/
/* IDCT */

/* in/out: mma=mma+mmb, mmb=mmb-mma */
#define SUMSUB_BA( a, b ) \
    "paddw "#b", "#a" \n\t"\
    "paddw "#b", "#b" \n\t"\
    "psubw "#a", "#b" \n\t"

#define SUMSUB_BADC( a, b, c, d ) \
    "paddw "#b", "#a" \n\t"\
    "paddw "#d", "#c" \n\t"\
    "paddw "#b", "#b" \n\t"\
    "paddw "#d", "#d" \n\t"\
    "psubw "#a", "#b" \n\t"\
    "psubw "#c", "#d" \n\t"

#define SUMSUBD2_AB( a, b, t ) \
    "movq  "#b", "#t" \n\t"\
    "psraw  $1 , "#b" \n\t"\
    "paddw "#a", "#b" \n\t"\
    "psraw  $1 , "#a" \n\t"\
    "psubw "#t", "#a" \n\t"

#define IDCT4_1D( s02, s13, d02, d13, t ) \
    SUMSUB_BA  ( s02, d02 )\
    SUMSUBD2_AB( s13, d13, t )\
    SUMSUB_BADC( d13, s02, s13, d02 )

#define STORE_DIFF_4P( p, t, z ) \
    "psraw      $6,     "#p" \n\t"\
    "movd       (%0),   "#t" \n\t"\
    "punpcklbw "#z",    "#t" \n\t"\
    "paddsw    "#t",    "#p" \n\t"\
    "packuswb  "#z",    "#p" \n\t"\
    "movd      "#p",    (%0) \n\t"

static void ff_h264_idct_add_mmx(uint8_t *dst, int16_t *block, int stride)
{
    /* Load dct coeffs */
    asm volatile(
        "movq   (%0), %%mm0 \n\t"
        "movq  8(%0), %%mm1 \n\t"
        "movq 16(%0), %%mm2 \n\t"
        "movq 24(%0), %%mm3 \n\t"
    :: "r"(block) );

    asm volatile(
        /* mm1=s02+s13  mm2=s02-s13  mm4=d02+d13  mm0=d02-d13 */
        IDCT4_1D( %%mm2, %%mm1, %%mm0, %%mm3, %%mm4 )

        "movq      %0,    %%mm6 \n\t"
        /* in: 1,4,0,2  out: 1,2,3,0 */
        TRANSPOSE4( %%mm3, %%mm1, %%mm0, %%mm2, %%mm4 )

        "paddw     %%mm6, %%mm3 \n\t"

        /* mm2=s02+s13  mm3=s02-s13  mm4=d02+d13  mm1=d02-d13 */
        IDCT4_1D( %%mm4, %%mm2, %%mm3, %%mm0, %%mm1 )

        "pxor %%mm7, %%mm7    \n\t"
    :: "m"(ff_pw_32));

    asm volatile(
    STORE_DIFF_4P( %%mm0, %%mm1, %%mm7)
        "add %1, %0             \n\t"
    STORE_DIFF_4P( %%mm2, %%mm1, %%mm7)
        "add %1, %0             \n\t"
    STORE_DIFF_4P( %%mm3, %%mm1, %%mm7)
        "add %1, %0             \n\t"
    STORE_DIFF_4P( %%mm4, %%mm1, %%mm7)
        : "+r"(dst)
        : "r" ((long)stride)
    );
}

static inline void h264_idct8_1d(int16_t *block)
{
    asm volatile(
        "movq 112(%0), %%mm7  \n\t"
        "movq  80(%0), %%mm5  \n\t"
        "movq  48(%0), %%mm3  \n\t"
        "movq  16(%0), %%mm1  \n\t"

        "movq   %%mm7, %%mm4  \n\t"
        "movq   %%mm3, %%mm6  \n\t"
        "movq   %%mm5, %%mm0  \n\t"
        "movq   %%mm7, %%mm2  \n\t"
        "psraw  $1,    %%mm4  \n\t"
        "psraw  $1,    %%mm6  \n\t"
        "psubw  %%mm7, %%mm0  \n\t"
        "psubw  %%mm6, %%mm2  \n\t"
        "psubw  %%mm4, %%mm0  \n\t"
        "psubw  %%mm3, %%mm2  \n\t"
        "psubw  %%mm3, %%mm0  \n\t"
        "paddw  %%mm1, %%mm2  \n\t"

        "movq   %%mm5, %%mm4  \n\t"
        "movq   %%mm1, %%mm6  \n\t"
        "psraw  $1,    %%mm4  \n\t"
        "psraw  $1,    %%mm6  \n\t"
        "paddw  %%mm5, %%mm4  \n\t"
        "paddw  %%mm1, %%mm6  \n\t"
        "paddw  %%mm7, %%mm4  \n\t"
        "paddw  %%mm5, %%mm6  \n\t"
        "psubw  %%mm1, %%mm4  \n\t"
        "paddw  %%mm3, %%mm6  \n\t"

        "movq   %%mm0, %%mm1  \n\t"
        "movq   %%mm4, %%mm3  \n\t"
        "movq   %%mm2, %%mm5  \n\t"
        "movq   %%mm6, %%mm7  \n\t"
        "psraw  $2,    %%mm6  \n\t"
        "psraw  $2,    %%mm3  \n\t"
        "psraw  $2,    %%mm5  \n\t"
        "psraw  $2,    %%mm0  \n\t"
        "paddw  %%mm6, %%mm1  \n\t"
        "paddw  %%mm2, %%mm3  \n\t"
        "psubw  %%mm4, %%mm5  \n\t"
        "psubw  %%mm0, %%mm7  \n\t"

        "movq  32(%0), %%mm2  \n\t"
        "movq  96(%0), %%mm6  \n\t"
        "movq   %%mm2, %%mm4  \n\t"
        "movq   %%mm6, %%mm0  \n\t"
        "psraw  $1,    %%mm4  \n\t"
        "psraw  $1,    %%mm6  \n\t"
        "psubw  %%mm0, %%mm4  \n\t"
        "paddw  %%mm2, %%mm6  \n\t"

        "movq    (%0), %%mm2  \n\t"
        "movq  64(%0), %%mm0  \n\t"
        SUMSUB_BA( %%mm0, %%mm2 )
        SUMSUB_BA( %%mm6, %%mm0 )
        SUMSUB_BA( %%mm4, %%mm2 )
        SUMSUB_BA( %%mm7, %%mm6 )
        SUMSUB_BA( %%mm5, %%mm4 )
        SUMSUB_BA( %%mm3, %%mm2 )
        SUMSUB_BA( %%mm1, %%mm0 )
        :: "r"(block)
    );
}

static void ff_h264_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
{
    int i;
    int16_t __attribute__ ((aligned(8))) b2[64];

    block[0] += 32;

    for(i=0; i<2; i++){
        DECLARE_ALIGNED_8(uint64_t, tmp);

        h264_idct8_1d(block+4*i);

        asm volatile(
            "movq   %%mm7,    %0   \n\t"
            TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 )
            "movq   %%mm0,  8(%1)  \n\t"
            "movq   %%mm6, 24(%1)  \n\t"
            "movq   %%mm7, 40(%1)  \n\t"
            "movq   %%mm4, 56(%1)  \n\t"
            "movq    %0,    %%mm7  \n\t"
            TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 )
            "movq   %%mm7,   (%1)  \n\t"
            "movq   %%mm1, 16(%1)  \n\t"
            "movq   %%mm0, 32(%1)  \n\t"
            "movq   %%mm3, 48(%1)  \n\t"
            : "=m"(tmp)
            : "r"(b2+32*i)
            : "memory"
        );
    }

    for(i=0; i<2; i++){
        h264_idct8_1d(b2+4*i);

        asm volatile(
            "psraw     $6, %%mm7  \n\t"
            "psraw     $6, %%mm6  \n\t"
            "psraw     $6, %%mm5  \n\t"
            "psraw     $6, %%mm4  \n\t"
            "psraw     $6, %%mm3  \n\t"
            "psraw     $6, %%mm2  \n\t"
            "psraw     $6, %%mm1  \n\t"
            "psraw     $6, %%mm0  \n\t"

            "movq   %%mm7,    (%0)  \n\t"
            "movq   %%mm5,  16(%0)  \n\t"
            "movq   %%mm3,  32(%0)  \n\t"
            "movq   %%mm1,  48(%0)  \n\t"
            "movq   %%mm0,  64(%0)  \n\t"
            "movq   %%mm2,  80(%0)  \n\t"
            "movq   %%mm4,  96(%0)  \n\t"
            "movq   %%mm6, 112(%0)  \n\t"
            :: "r"(b2+4*i)
            : "memory"
        );
    }

    add_pixels_clamped_mmx(b2, dst, stride);
}

static void ff_h264_idct_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
{
    int dc = (block[0] + 32) >> 6;
    asm volatile(
        "movd          %0, %%mm0 \n\t"
        "pshufw $0, %%mm0, %%mm0 \n\t"
        "pxor       %%mm1, %%mm1 \n\t"
        "psubw      %%mm0, %%mm1 \n\t"
        "packuswb   %%mm0, %%mm0 \n\t"
        "packuswb   %%mm1, %%mm1 \n\t"
        ::"r"(dc)
    );
    asm volatile(
        "movd          %0, %%mm2 \n\t"
        "movd          %1, %%mm3 \n\t"
        "movd          %2, %%mm4 \n\t"
        "movd          %3, %%mm5 \n\t"
        "paddusb    %%mm0, %%mm2 \n\t"
        "paddusb    %%mm0, %%mm3 \n\t"
        "paddusb    %%mm0, %%mm4 \n\t"
        "paddusb    %%mm0, %%mm5 \n\t"
        "psubusb    %%mm1, %%mm2 \n\t"
        "psubusb    %%mm1, %%mm3 \n\t"
        "psubusb    %%mm1, %%mm4 \n\t"
        "psubusb    %%mm1, %%mm5 \n\t"
        "movd       %%mm2, %0    \n\t"
        "movd       %%mm3, %1    \n\t"
        "movd       %%mm4, %2    \n\t"
        "movd       %%mm5, %3    \n\t"
        :"+m"(*(uint32_t*)(dst+0*stride)),
         "+m"(*(uint32_t*)(dst+1*stride)),
         "+m"(*(uint32_t*)(dst+2*stride)),
         "+m"(*(uint32_t*)(dst+3*stride))
    );
}

static void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
{
    int dc = (block[0] + 32) >> 6;
    int y;
    asm volatile(
        "movd          %0, %%mm0 \n\t"
        "pshufw $0, %%mm0, %%mm0 \n\t"
        "pxor       %%mm1, %%mm1 \n\t"
        "psubw      %%mm0, %%mm1 \n\t"
        "packuswb   %%mm0, %%mm0 \n\t"
        "packuswb   %%mm1, %%mm1 \n\t"
        ::"r"(dc)
    );
    for(y=2; y--; dst += 4*stride){
    asm volatile(
        "movq          %0, %%mm2 \n\t"
        "movq          %1, %%mm3 \n\t"
        "movq          %2, %%mm4 \n\t"
        "movq          %3, %%mm5 \n\t"
        "paddusb    %%mm0, %%mm2 \n\t"
        "paddusb    %%mm0, %%mm3 \n\t"
        "paddusb    %%mm0, %%mm4 \n\t"
        "paddusb    %%mm0, %%mm5 \n\t"
        "psubusb    %%mm1, %%mm2 \n\t"
        "psubusb    %%mm1, %%mm3 \n\t"
        "psubusb    %%mm1, %%mm4 \n\t"
        "psubusb    %%mm1, %%mm5 \n\t"
        "movq       %%mm2, %0    \n\t"
        "movq       %%mm3, %1    \n\t"
        "movq       %%mm4, %2    \n\t"
        "movq       %%mm5, %3    \n\t"
        :"+m"(*(uint64_t*)(dst+0*stride)),
         "+m"(*(uint64_t*)(dst+1*stride)),
         "+m"(*(uint64_t*)(dst+2*stride)),
         "+m"(*(uint64_t*)(dst+3*stride))
    );
    }
}


/***********************************/
/* deblocking */

// out: o = |x-y|>a
// clobbers: t
#define DIFF_GT_MMX(x,y,a,o,t)\
    "movq     "#y", "#t"  \n\t"\
    "movq     "#x", "#o"  \n\t"\
    "psubusb  "#x", "#t"  \n\t"\
    "psubusb  "#y", "#o"  \n\t"\
    "por      "#t", "#o"  \n\t"\
    "psubusb  "#a", "#o"  \n\t"

// out: o = |x-y|>a
// clobbers: t
#define DIFF_GT2_MMX(x,y,a,o,t)\
    "movq     "#y", "#t"  \n\t"\
    "movq     "#x", "#o"  \n\t"\
    "psubusb  "#x", "#t"  \n\t"\
    "psubusb  "#y", "#o"  \n\t"\
    "psubusb  "#a", "#t"  \n\t"\
    "psubusb  "#a", "#o"  \n\t"\
    "pcmpeqb  "#t", "#o"  \n\t"\

// in: mm0=p1 mm1=p0 mm2=q0 mm3=q1
// out: mm5=beta-1, mm7=mask
// clobbers: mm4,mm6
#define H264_DEBLOCK_MASK(alpha1, beta1) \
    "pshufw $0, "#alpha1", %%mm4 \n\t"\
    "pshufw $0, "#beta1 ", %%mm5 \n\t"\
    "packuswb  %%mm4, %%mm4      \n\t"\
    "packuswb  %%mm5, %%mm5      \n\t"\
    DIFF_GT_MMX(%%mm1, %%mm2, %%mm4, %%mm7, %%mm6) /* |p0-q0| > alpha-1 */\
    DIFF_GT_MMX(%%mm0, %%mm1, %%mm5, %%mm4, %%mm6) /* |p1-p0| > beta-1 */\
    "por       %%mm4, %%mm7      \n\t"\
    DIFF_GT_MMX(%%mm3, %%mm2, %%mm5, %%mm4, %%mm6) /* |q1-q0| > beta-1 */\
    "por       %%mm4, %%mm7      \n\t"\
    "pxor      %%mm6, %%mm6      \n\t"\
    "pcmpeqb   %%mm6, %%mm7      \n\t"

// in: mm0=p1 mm1=p0 mm2=q0 mm3=q1 mm7=(tc&mask)
// out: mm1=p0' mm2=q0'
// clobbers: mm0,3-6
#define H264_DEBLOCK_P0_Q0(pb_01, pb_3f)\
        "movq    %%mm1              , %%mm5 \n\t"\
        "pxor    %%mm2              , %%mm5 \n\t" /* p0^q0*/\
        "pand    "#pb_01"           , %%mm5 \n\t" /* (p0^q0)&1*/\
        "pcmpeqb %%mm4              , %%mm4 \n\t"\
        "pxor    %%mm4              , %%mm3 \n\t"\
        "pavgb   %%mm0              , %%mm3 \n\t" /* (p1 - q1 + 256)>>1*/\
        "pavgb   "MANGLE(ff_pb_3)"  , %%mm3 \n\t" /*(((p1 - q1 + 256)>>1)+4)>>1 = 64+2+(p1-q1)>>2*/\
        "pxor    %%mm1              , %%mm4 \n\t"\
        "pavgb   %%mm2              , %%mm4 \n\t" /* (q0 - p0 + 256)>>1*/\
        "pavgb   %%mm5              , %%mm3 \n\t"\
        "paddusb %%mm4              , %%mm3 \n\t" /* d+128+33*/\
        "movq    "MANGLE(ff_pb_A1)" , %%mm6 \n\t"\
        "psubusb %%mm3              , %%mm6 \n\t"\
        "psubusb "MANGLE(ff_pb_A1)" , %%mm3 \n\t"\
        "pminub  %%mm7              , %%mm6 \n\t"\
        "pminub  %%mm7              , %%mm3 \n\t"\
        "psubusb %%mm6              , %%mm1 \n\t"\
        "psubusb %%mm3              , %%mm2 \n\t"\
        "paddusb %%mm3              , %%mm1 \n\t"\
        "paddusb %%mm6              , %%mm2 \n\t"

// in: mm0=p1 mm1=p0 mm2=q0 mm3=q1 mm7=(tc&mask) %8=mm_bone
// out: (q1addr) = av_clip( (q2+((p0+q0+1)>>1))>>1, q1-tc0, q1+tc0 )
// clobbers: q2, tmp, tc0
#define H264_DEBLOCK_Q1(p1, q2, q2addr, q1addr, tc0, tmp)\
        "movq     %%mm1,  "#tmp"   \n\t"\
        "pavgb    %%mm2,  "#tmp"   \n\t"\
        "pavgb    "#tmp", "#q2"    \n\t" /* avg(p2,avg(p0,q0)) */\
        "pxor   "q2addr", "#tmp"   \n\t"\
        "pand     %8,     "#tmp"   \n\t" /* (p2^avg(p0,q0))&1 */\
        "psubusb  "#tmp", "#q2"    \n\t" /* (p2+((p0+q0+1)>>1))>>1 */\
        "movq     "#p1",  "#tmp"   \n\t"\
        "psubusb  "#tc0", "#tmp"   \n\t"\
        "paddusb  "#p1",  "#tc0"   \n\t"\