h264dsp_mmx.c.svn-base

mediastreamer2是开源的网络传输媒体流的库

/*
 * 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
 */

#include "dsputil_mmx.h"

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

#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), %%mm0  \n\t"
        "movq  48(%0), %%mm3  \n\t"
        "movq  16(%0), %%mm5  \n\t"

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

        "psubw  %%mm3, %%mm5  \n\t"
        "psubw  %%mm3, %%mm0  \n\t"
        "paddw  %%mm7, %%mm5  \n\t"
        "psubw  %%mm7, %%mm0  \n\t"
        "psraw  $1,    %%mm3  \n\t"
        "psraw  $1,    %%mm7  \n\t"
        "psubw  %%mm3, %%mm5  \n\t"
        "psubw  %%mm7, %%mm0  \n\t"

        "movq   %%mm4, %%mm3  \n\t"
        "movq   %%mm1, %%mm7  \n\t"
        "psraw  $2,    %%mm1  \n\t"
        "psraw  $2,    %%mm3  \n\t"
        "paddw  %%mm5, %%mm3  \n\t"
        "psraw  $2,    %%mm5  \n\t"
        "paddw  %%mm0, %%mm1  \n\t"
        "psraw  $2,    %%mm0  \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);
}

#define STORE_DIFF_8P( p, d, t, z )\
        "movq       "#d", "#t" \n"\
        "psraw       $6,  "#p" \n"\
        "punpcklbw  "#z", "#t" \n"\
        "paddsw     "#t", "#p" \n"\
        "packuswb   "#p", "#p" \n"\
        "movq       "#p", "#d" \n"

#define H264_IDCT8_1D_SSE2(a,b,c,d,e,f,g,h)\
        "movdqa     "#c", "#a" \n"\
        "movdqa     "#g", "#e" \n"\
        "psraw       $1,  "#c" \n"\
        "psraw       $1,  "#g" \n"\
        "psubw      "#e", "#c" \n"\
        "paddw      "#a", "#g" \n"\
        "movdqa     "#b", "#e" \n"\
        "psraw       $1,  "#e" \n"\
        "paddw      "#b", "#e" \n"\
        "paddw      "#d", "#e" \n"\
        "paddw      "#f", "#e" \n"\
        "movdqa     "#f", "#a" \n"\
        "psraw       $1,  "#a" \n"\
        "paddw      "#f", "#a" \n"\
        "paddw      "#h", "#a" \n"\
        "psubw      "#b", "#a" \n"\
        "psubw      "#d", "#b" \n"\
        "psubw      "#d", "#f" \n"\
        "paddw      "#h", "#b" \n"\
        "psubw      "#h", "#f" \n"\
        "psraw       $1,  "#d" \n"\
        "psraw       $1,  "#h" \n"\
        "psubw      "#d", "#b" \n"\
        "psubw      "#h", "#f" \n"\
        "movdqa     "#e", "#d" \n"\
        "movdqa     "#a", "#h" \n"\
        "psraw       $2,  "#d" \n"\
        "psraw       $2,  "#h" \n"\
        "paddw      "#f", "#d" \n"\
        "paddw      "#b", "#h" \n"\
        "psraw       $2,  "#f" \n"\
        "psraw       $2,  "#b" \n"\
        "psubw      "#f", "#e" \n"\
        "psubw      "#a", "#b" \n"\
        "movdqa 0x00(%1), "#a" \n"\
        "movdqa 0x40(%1), "#f" \n"\
        SUMSUB_BA(f, a)\
        SUMSUB_BA(g, f)\
        SUMSUB_BA(c, a)\
        SUMSUB_BA(e, g)\
        SUMSUB_BA(b, c)\
        SUMSUB_BA(h, a)\
        SUMSUB_BA(d, f)

static void ff_h264_idct8_add_sse2(uint8_t *dst, int16_t *block, int stride)
{
    asm volatile(
        "movdqa   0x10(%1), %%xmm1 \n"
        "movdqa   0x20(%1), %%xmm2 \n"
        "movdqa   0x30(%1), %%xmm3 \n"
        "movdqa   0x50(%1), %%xmm5 \n"
        "movdqa   0x60(%1), %%xmm6 \n"
        "movdqa   0x70(%1), %%xmm7 \n"
        H264_IDCT8_1D_SSE2(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm6, %%xmm7)
        TRANSPOSE8(%%xmm4, %%xmm1, %%xmm7, %%xmm3, %%xmm5, %%xmm0, %%xmm2, %%xmm6, (%1))
        "paddw          %4, %%xmm4 \n"
        "movdqa     %%xmm4, 0x00(%1) \n"
        "movdqa     %%xmm2, 0x40(%1) \n"
        H264_IDCT8_1D_SSE2(%%xmm4, %%xmm0, %%xmm6, %%xmm3, %%xmm2, %%xmm5, %%xmm7, %%xmm1)
        "movdqa     %%xmm6, 0x60(%1) \n"
        "movdqa     %%xmm7, 0x70(%1) \n"
        "pxor       %%xmm7, %%xmm7 \n"
        STORE_DIFF_8P(%%xmm2, (%0),      %%xmm6, %%xmm7)
        STORE_DIFF_8P(%%xmm0, (%0,%2),   %%xmm6, %%xmm7)
        STORE_DIFF_8P(%%xmm1, (%0,%2,2), %%xmm6, %%xmm7)
        STORE_DIFF_8P(%%xmm3, (%0,%3),   %%xmm6, %%xmm7)
        "lea     (%0,%2,4), %0 \n"
        STORE_DIFF_8P(%%xmm5, (%0),      %%xmm6, %%xmm7)
        STORE_DIFF_8P(%%xmm4, (%0,%2),   %%xmm6, %%xmm7)
        "movdqa   0x60(%1), %%xmm0 \n"
        "movdqa   0x70(%1), %%xmm1 \n"
        STORE_DIFF_8P(%%xmm0, (%0,%2,2), %%xmm6, %%xmm7)
        STORE_DIFF_8P(%%xmm1, (%0,%3),   %%xmm6, %%xmm7)
        :"+r"(dst)
        :"r"(block), "r"((long)stride), "r"(3L*stride), "m"(ff_pw_32)
    );
}

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"