view x86/dsputil_h264_template_ssse3.c @ 11190:afc60af2d008 libavcodec

2x faster ff_h264_init_cabac_states(), 4k cpu cycles less. Sadly this is just per slice so the speedup with normal files should be negligible.

/*
 * Copyright (c) 2008 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
 */

/**
 * SSSE3 optimized version of (put|avg)_h264_chroma_mc8.
 * H264_CHROMA_MC8_TMPL must be defined to the desired function name
 * H264_CHROMA_MC8_MV0 must be defined to a (put|avg)_pixels8 function
 * AVG_OP must be defined to empty for put and the identify for avg
 */
static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y, int rnd)
{
    if(y==0 && x==0) {
        /* no filter needed */
        H264_CHROMA_MC8_MV0(dst, src, stride, h);
        return;
    }

    assert(x<8 && y<8 && x>=0 && y>=0);

    if(y==0 || x==0)
    {
        /* 1 dimensional filter only */
        __asm__ volatile(
            "movd %0, %%xmm7 \n\t"
            "movq %1, %%xmm6 \n\t"
            "pshuflw $0, %%xmm7, %%xmm7 \n\t"
            "movlhps %%xmm6, %%xmm6 \n\t"
            "movlhps %%xmm7, %%xmm7 \n\t"
            :: "r"(255*(x+y)+8), "m"(*(rnd?&ff_pw_4:&ff_pw_3))
        );

        if(x) {
            __asm__ volatile(
                "1: \n\t"
                "movq (%1), %%xmm0 \n\t"
                "movq 1(%1), %%xmm1 \n\t"
                "movq (%1,%3), %%xmm2 \n\t"
                "movq 1(%1,%3), %%xmm3 \n\t"
                "punpcklbw %%xmm1, %%xmm0 \n\t"
                "punpcklbw %%xmm3, %%xmm2 \n\t"
                "pmaddubsw %%xmm7, %%xmm0 \n\t"
                "pmaddubsw %%xmm7, %%xmm2 \n\t"
         AVG_OP("movq (%0), %%xmm4 \n\t")
         AVG_OP("movhps (%0,%3), %%xmm4 \n\t")
                "paddw %%xmm6, %%xmm0 \n\t"
                "paddw %%xmm6, %%xmm2 \n\t"
                "psrlw $3, %%xmm0 \n\t"
                "psrlw $3, %%xmm2 \n\t"
                "packuswb %%xmm2, %%xmm0 \n\t"
         AVG_OP("pavgb %%xmm4, %%xmm0 \n\t")
                "movq %%xmm0, (%0) \n\t"
                "movhps %%xmm0, (%0,%3) \n\t"
                "sub $2, %2 \n\t"
                "lea (%1,%3,2), %1 \n\t"
                "lea (%0,%3,2), %0 \n\t"
                "jg 1b \n\t"
                :"+r"(dst), "+r"(src), "+r"(h)
                :"r"((x86_reg)stride)
            );
        } else {
            __asm__ volatile(
                "1: \n\t"
                "movq (%1), %%xmm0 \n\t"
                "movq (%1,%3), %%xmm1 \n\t"
                "movdqa %%xmm1, %%xmm2 \n\t"
                "movq (%1,%3,2), %%xmm3 \n\t"
                "punpcklbw %%xmm1, %%xmm0 \n\t"
                "punpcklbw %%xmm3, %%xmm2 \n\t"
                "pmaddubsw %%xmm7, %%xmm0 \n\t"
                "pmaddubsw %%xmm7, %%xmm2 \n\t"
         AVG_OP("movq (%0), %%xmm4 \n\t")
         AVG_OP("movhps (%0,%3), %%xmm4 \n\t")
                "paddw %%xmm6, %%xmm0 \n\t"
                "paddw %%xmm6, %%xmm2 \n\t"
                "psrlw $3, %%xmm0 \n\t"
                "psrlw $3, %%xmm2 \n\t"
                "packuswb %%xmm2, %%xmm0 \n\t"
         AVG_OP("pavgb %%xmm4, %%xmm0 \n\t")
                "movq %%xmm0, (%0) \n\t"
                "movhps %%xmm0, (%0,%3) \n\t"
                "sub $2, %2 \n\t"
                "lea (%1,%3,2), %1 \n\t"
                "lea (%0,%3,2), %0 \n\t"
                "jg 1b \n\t"
                :"+r"(dst), "+r"(src), "+r"(h)
                :"r"((x86_reg)stride)
            );
        }
        return;
    }

    /* general case, bilinear */
    __asm__ volatile(
        "movd %0, %%xmm7 \n\t"
        "movd %1, %%xmm6 \n\t"
        "movdqa %2, %%xmm5 \n\t"
        "pshuflw $0, %%xmm7, %%xmm7 \n\t"
        "pshuflw $0, %%xmm6, %%xmm6 \n\t"
        "movlhps %%xmm7, %%xmm7 \n\t"
        "movlhps %%xmm6, %%xmm6 \n\t"
        :: "r"((x*255+8)*(8-y)), "r"((x*255+8)*y), "m"(*(rnd?&ff_pw_32:&ff_pw_28))
    );

    __asm__ volatile(
        "movq (%1), %%xmm0 \n\t"
        "movq 1(%1), %%xmm1 \n\t"
        "punpcklbw %%xmm1, %%xmm0 \n\t"
        "add %3, %1 \n\t"
        "1: \n\t"
        "movq (%1), %%xmm1 \n\t"
        "movq 1(%1), %%xmm2 \n\t"
        "movq (%1,%3), %%xmm3 \n\t"
        "movq 1(%1,%3), %%xmm4 \n\t"
        "lea (%1,%3,2), %1 \n\t"
        "punpcklbw %%xmm2, %%xmm1 \n\t"
        "punpcklbw %%xmm4, %%xmm3 \n\t"
        "movdqa %%xmm1, %%xmm2 \n\t"
        "movdqa %%xmm3, %%xmm4 \n\t"
        "pmaddubsw %%xmm7, %%xmm0 \n\t"
        "pmaddubsw %%xmm6, %%xmm1 \n\t"
        "pmaddubsw %%xmm7, %%xmm2 \n\t"
        "pmaddubsw %%xmm6, %%xmm3 \n\t"
        "paddw %%xmm5, %%xmm0 \n\t"
        "paddw %%xmm5, %%xmm2 \n\t"
        "paddw %%xmm0, %%xmm1 \n\t"
        "paddw %%xmm2, %%xmm3 \n\t"
        "movdqa %%xmm4, %%xmm0 \n\t"
        "psrlw $6, %%xmm1 \n\t"
        "psrlw $6, %%xmm3 \n\t"
 AVG_OP("movq (%0), %%xmm2 \n\t")
 AVG_OP("movhps (%0,%3), %%xmm2 \n\t")
        "packuswb %%xmm3, %%xmm1 \n\t"
 AVG_OP("pavgb %%xmm2, %%xmm1 \n\t")
        "movq %%xmm1, (%0)\n\t"
        "movhps %%xmm1, (%0,%3)\n\t"
        "sub $2, %2 \n\t"
        "lea (%0,%3,2), %0 \n\t"
        "jg 1b \n\t"
        :"+r"(dst), "+r"(src), "+r"(h)
        :"r"((x86_reg)stride)
    );
}

static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 4*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y)
{
    __asm__ volatile(
        "movd %0, %%mm7 \n\t"
        "movd %1, %%mm6 \n\t"
        "movq %2, %%mm5 \n\t"
        "pshufw $0, %%mm7, %%mm7 \n\t"
        "pshufw $0, %%mm6, %%mm6 \n\t"
        :: "r"((x*255+8)*(8-y)), "r"((x*255+8)*y), "m"(ff_pw_32)
    );

    __asm__ volatile(
        "movd (%1), %%mm0 \n\t"
        "punpcklbw 1(%1), %%mm0 \n\t"
        "add %3, %1 \n\t"
        "1: \n\t"
        "movd (%1), %%mm1 \n\t"
        "movd (%1,%3), %%mm3 \n\t"
        "punpcklbw 1(%1), %%mm1 \n\t"
        "punpcklbw 1(%1,%3), %%mm3 \n\t"
        "lea (%1,%3,2), %1 \n\t"
        "movq %%mm1, %%mm2 \n\t"
        "movq %%mm3, %%mm4 \n\t"
        "pmaddubsw %%mm7, %%mm0 \n\t"
        "pmaddubsw %%mm6, %%mm1 \n\t"
        "pmaddubsw %%mm7, %%mm2 \n\t"
        "pmaddubsw %%mm6, %%mm3 \n\t"
        "paddw %%mm5, %%mm0 \n\t"
        "paddw %%mm5, %%mm2 \n\t"
        "paddw %%mm0, %%mm1 \n\t"
        "paddw %%mm2, %%mm3 \n\t"
        "movq %%mm4, %%mm0 \n\t"
        "psrlw $6, %%mm1 \n\t"
        "psrlw $6, %%mm3 \n\t"
        "packuswb %%mm1, %%mm1 \n\t"
        "packuswb %%mm3, %%mm3 \n\t"
 AVG_OP("pavgb (%0), %%mm1 \n\t")
 AVG_OP("pavgb (%0,%3), %%mm3 \n\t")
        "movd %%mm1, (%0)\n\t"
        "movd %%mm3, (%0,%3)\n\t"
        "sub $2, %2 \n\t"
        "lea (%0,%3,2), %0 \n\t"
        "jg 1b \n\t"
        :"+r"(dst), "+r"(src), "+r"(h)
        :"r"((x86_reg)stride)
    );
}