ffmpeg/libavcodec/x86/dsputilenc_mmx.c

/*
 * MMX optimized DSP utils
 * Copyright (c) 2000, 2001 Fabrice Bellard
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
 *
 * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
 *
 * 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 "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavutil/x86/asm.h"
#include "libavutil/x86/cpu.h"
#include "libavcodec/dct.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/mpegvideo.h"
#include "libavcodec/mathops.h"
#include "dsputil_x86.h"

void ff_get_pixels_mmx(int16_t *block, const uint8_t *pixels, int line_size);
void ff_get_pixels_sse2(int16_t *block, const uint8_t *pixels, int line_size);
void ff_diff_pixels_mmx(int16_t *block, const uint8_t *s1, const uint8_t *s2,
                        int stride);
int ff_pix_sum16_mmx(uint8_t *pix, int line_size);
int ff_pix_norm1_mmx(uint8_t *pix, int line_size);

#if HAVE_INLINE_ASM

static int sse8_mmx(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int tmp;

    __asm__ volatile (
        "movl         %4, %%ecx          \n"
        "shr          $1, %%ecx          \n"
        "pxor      %%mm0, %%mm0          \n" /* mm0 = 0 */
        "pxor      %%mm7, %%mm7          \n" /* mm7 holds the sum */
        "1:                              \n"
        "movq       (%0), %%mm1          \n" /* mm1 = pix1[0][0 - 7] */
        "movq       (%1), %%mm2          \n" /* mm2 = pix2[0][0 - 7] */
        "movq   (%0, %3), %%mm3          \n" /* mm3 = pix1[1][0 - 7] */
        "movq   (%1, %3), %%mm4          \n" /* mm4 = pix2[1][0 - 7] */

        /* todo: mm1-mm2, mm3-mm4 */
        /* algo: subtract mm1 from mm2 with saturation and vice versa */
        /*       OR the results to get absolute difference */
        "movq      %%mm1, %%mm5          \n"
        "movq      %%mm3, %%mm6          \n"
        "psubusb   %%mm2, %%mm1          \n"
        "psubusb   %%mm4, %%mm3          \n"
        "psubusb   %%mm5, %%mm2          \n"
        "psubusb   %%mm6, %%mm4          \n"

        "por       %%mm1, %%mm2          \n"
        "por       %%mm3, %%mm4          \n"

        /* now convert to 16-bit vectors so we can square them */
        "movq      %%mm2, %%mm1          \n"
        "movq      %%mm4, %%mm3          \n"

        "punpckhbw %%mm0, %%mm2          \n"
        "punpckhbw %%mm0, %%mm4          \n"
        "punpcklbw %%mm0, %%mm1          \n" /* mm1 now spread over (mm1, mm2) */
        "punpcklbw %%mm0, %%mm3          \n" /* mm4 now spread over (mm3, mm4) */

        "pmaddwd   %%mm2, %%mm2          \n"
        "pmaddwd   %%mm4, %%mm4          \n"
        "pmaddwd   %%mm1, %%mm1          \n"
        "pmaddwd   %%mm3, %%mm3          \n"

        "lea (%0, %3, 2), %0             \n" /* pix1 += 2 * line_size */
        "lea (%1, %3, 2), %1             \n" /* pix2 += 2 * line_size */

        "paddd     %%mm2, %%mm1          \n"
        "paddd     %%mm4, %%mm3          \n"
        "paddd     %%mm1, %%mm7          \n"
        "paddd     %%mm3, %%mm7          \n"

        "decl      %%ecx                 \n"
        "jnz       1b                    \n"

        "movq      %%mm7, %%mm1          \n"
        "psrlq       $32, %%mm7          \n" /* shift hi dword to lo */
        "paddd     %%mm7, %%mm1          \n"
        "movd      %%mm1, %2             \n"
        : "+r" (pix1), "+r" (pix2), "=r" (tmp)
        : "r" ((x86_reg) line_size), "m" (h)
        : "%ecx");

    return tmp;
}

static int sse16_mmx(void *v, uint8_t *pix1, uint8_t *pix2,
                     int line_size, int h)
{
    int tmp;

    __asm__ volatile (
        "movl %4, %%ecx\n"
        "pxor %%mm0, %%mm0\n"    /* mm0 = 0 */
        "pxor %%mm7, %%mm7\n"    /* mm7 holds the sum */
        "1:\n"
        "movq (%0), %%mm1\n"     /* mm1 = pix1[0 -  7] */
        "movq (%1), %%mm2\n"     /* mm2 = pix2[0 -  7] */
        "movq 8(%0), %%mm3\n"    /* mm3 = pix1[8 - 15] */
        "movq 8(%1), %%mm4\n"    /* mm4 = pix2[8 - 15] */

        /* todo: mm1-mm2, mm3-mm4 */
        /* algo: subtract mm1 from mm2 with saturation and vice versa */
        /*       OR the results to get absolute difference */
        "movq %%mm1, %%mm5\n"
        "movq %%mm3, %%mm6\n"
        "psubusb %%mm2, %%mm1\n"
        "psubusb %%mm4, %%mm3\n"
        "psubusb %%mm5, %%mm2\n"
        "psubusb %%mm6, %%mm4\n"

        "por %%mm1, %%mm2\n"
        "por %%mm3, %%mm4\n"

        /* now convert to 16-bit vectors so we can square them */
        "movq %%mm2, %%mm1\n"
        "movq %%mm4, %%mm3\n"

        "punpckhbw %%mm0, %%mm2\n"
        "punpckhbw %%mm0, %%mm4\n"
        "punpcklbw %%mm0, %%mm1\n" /* mm1 now spread over (mm1, mm2) */
        "punpcklbw %%mm0, %%mm3\n" /* mm4 now spread over (mm3, mm4) */

        "pmaddwd %%mm2, %%mm2\n"
        "pmaddwd %%mm4, %%mm4\n"
        "pmaddwd %%mm1, %%mm1\n"
        "pmaddwd %%mm3, %%mm3\n"

        "add %3, %0\n"
        "add %3, %1\n"

        "paddd %%mm2, %%mm1\n"
        "paddd %%mm4, %%mm3\n"
        "paddd %%mm1, %%mm7\n"
        "paddd %%mm3, %%mm7\n"

        "decl %%ecx\n"
        "jnz 1b\n"

        "movq %%mm7, %%mm1\n"
        "psrlq $32, %%mm7\n"    /* shift hi dword to lo */
        "paddd %%mm7, %%mm1\n"
        "movd %%mm1, %2\n"
        : "+r" (pix1), "+r" (pix2), "=r" (tmp)
        : "r" ((x86_reg) line_size), "m" (h)
        : "%ecx");

    return tmp;
}

static int hf_noise8_mmx(uint8_t *pix1, int line_size, int h)
{
    int tmp;

    __asm__ volatile (
        "movl %3, %%ecx\n"
        "pxor %%mm7, %%mm7\n"
        "pxor %%mm6, %%mm6\n"

        "movq (%0), %%mm0\n"
        "movq %%mm0, %%mm1\n"
        "psllq $8, %%mm0\n"
        "psrlq $8, %%mm1\n"
        "psrlq $8, %%mm0\n"
        "movq %%mm0, %%mm2\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm0\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm2\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm0\n"
        "psubw %%mm3, %%mm2\n"

        "add %2, %0\n"

        "movq (%0), %%mm4\n"
        "movq %%mm4, %%mm1\n"
        "psllq $8, %%mm4\n"
        "psrlq $8, %%mm1\n"
        "psrlq $8, %%mm4\n"
        "movq %%mm4, %%mm5\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm4\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm5\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm4\n"
        "psubw %%mm3, %%mm5\n"
        "psubw %%mm4, %%mm0\n"
        "psubw %%mm5, %%mm2\n"
        "pxor %%mm3, %%mm3\n"
        "pxor %%mm1, %%mm1\n"
        "pcmpgtw %%mm0, %%mm3\n\t"
        "pcmpgtw %%mm2, %%mm1\n\t"
        "pxor %%mm3, %%mm0\n"
        "pxor %%mm1, %%mm2\n"
        "psubw %%mm3, %%mm0\n"
        "psubw %%mm1, %%mm2\n"
        "paddw %%mm0, %%mm2\n"
        "paddw %%mm2, %%mm6\n"

        "add %2, %0\n"
        "1:\n"

        "movq (%0), %%mm0\n"
        "movq %%mm0, %%mm1\n"
        "psllq $8, %%mm0\n"
        "psrlq $8, %%mm1\n"
        "psrlq $8, %%mm0\n"
        "movq %%mm0, %%mm2\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm0\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm2\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm0\n"
        "psubw %%mm3, %%mm2\n"
        "psubw %%mm0, %%mm4\n"
        "psubw %%mm2, %%mm5\n"
        "pxor  %%mm3, %%mm3\n"
        "pxor  %%mm1, %%mm1\n"
        "pcmpgtw %%mm4, %%mm3\n\t"
        "pcmpgtw %%mm5, %%mm1\n\t"
        "pxor  %%mm3, %%mm4\n"
        "pxor  %%mm1, %%mm5\n"
        "psubw %%mm3, %%mm4\n"
        "psubw %%mm1, %%mm5\n"
        "paddw %%mm4, %%mm5\n"
        "paddw %%mm5, %%mm6\n"

        "add %2, %0\n"

        "movq (%0), %%mm4\n"
        "movq      %%mm4, %%mm1\n"
        "psllq $8, %%mm4\n"
        "psrlq $8, %%mm1\n"
        "psrlq $8, %%mm4\n"
        "movq      %%mm4, %%mm5\n"
        "movq      %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm4\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm5\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw     %%mm1, %%mm4\n"
        "psubw     %%mm3, %%mm5\n"
        "psubw     %%mm4, %%mm0\n"
        "psubw     %%mm5, %%mm2\n"
        "pxor      %%mm3, %%mm3\n"
        "pxor      %%mm1, %%mm1\n"
        "pcmpgtw   %%mm0, %%mm3\n\t"
        "pcmpgtw   %%mm2, %%mm1\n\t"
        "pxor      %%mm3, %%mm0\n"
        "pxor      %%mm1, %%mm2\n"
        "psubw     %%mm3, %%mm0\n"
        "psubw     %%mm1, %%mm2\n"
        "paddw     %%mm0, %%mm2\n"
        "paddw     %%mm2, %%mm6\n"

        "add  %2, %0\n"
        "subl $2, %%ecx\n"
        " jnz 1b\n"

        "movq      %%mm6, %%mm0\n"
        "punpcklwd %%mm7, %%mm0\n"
        "punpckhwd %%mm7, %%mm6\n"
        "paddd     %%mm0, %%mm6\n"

        "movq  %%mm6, %%mm0\n"
        "psrlq $32,   %%mm6\n"
        "paddd %%mm6, %%mm0\n"
        "movd  %%mm0, %1\n"
        : "+r" (pix1), "=r" (tmp)
        : "r" ((x86_reg) line_size), "g" (h - 2)
        : "%ecx");

    return tmp;
}

static int hf_noise16_mmx(uint8_t *pix1, int line_size, int h)
{
    int tmp;
    uint8_t *pix = pix1;

    __asm__ volatile (
        "movl %3, %%ecx\n"
        "pxor %%mm7, %%mm7\n"
        "pxor %%mm6, %%mm6\n"

        "movq (%0), %%mm0\n"
        "movq 1(%0), %%mm1\n"
        "movq %%mm0, %%mm2\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm0\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm2\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm0\n"
        "psubw %%mm3, %%mm2\n"

        "add %2, %0\n"

        "movq (%0), %%mm4\n"
        "movq 1(%0), %%mm1\n"
        "movq %%mm4, %%mm5\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm4\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm5\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm4\n"
        "psubw %%mm3, %%mm5\n"
        "psubw %%mm4, %%mm0\n"
        "psubw %%mm5, %%mm2\n"
        "pxor %%mm3, %%mm3\n"
        "pxor %%mm1, %%mm1\n"
        "pcmpgtw %%mm0, %%mm3\n\t"
        "pcmpgtw %%mm2, %%mm1\n\t"
        "pxor %%mm3, %%mm0\n"
        "pxor %%mm1, %%mm2\n"
        "psubw %%mm3, %%mm0\n"
        "psubw %%mm1, %%mm2\n"
        "paddw %%mm0, %%mm2\n"
        "paddw %%mm2, %%mm6\n"

        "add %2, %0\n"
        "1:\n"

        "movq (%0), %%mm0\n"
        "movq 1(%0), %%mm1\n"
        "movq %%mm0, %%mm2\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm0\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm2\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm0\n"
        "psubw %%mm3, %%mm2\n"
        "psubw %%mm0, %%mm4\n"
        "psubw %%mm2, %%mm5\n"
        "pxor %%mm3, %%mm3\n"
        "pxor %%mm1, %%mm1\n"
        "pcmpgtw %%mm4, %%mm3\n\t"
        "pcmpgtw %%mm5, %%mm1\n\t"
        "pxor %%mm3, %%mm4\n"
        "pxor %%mm1, %%mm5\n"
        "psubw %%mm3, %%mm4\n"
        "psubw %%mm1, %%mm5\n"
        "paddw %%mm4, %%mm5\n"
        "paddw %%mm5, %%mm6\n"

        "add %2, %0\n"

        "movq (%0), %%mm4\n"
        "movq 1(%0), %%mm1\n"
        "movq %%mm4, %%mm5\n"
        "movq %%mm1, %%mm3\n"
        "punpcklbw %%mm7, %%mm4\n"
        "punpcklbw %%mm7, %%mm1\n"
        "punpckhbw %%mm7, %%mm5\n"
        "punpckhbw %%mm7, %%mm3\n"
        "psubw %%mm1, %%mm4\n"
        "psubw %%mm3, %%mm5\n"
        "psubw %%mm4, %%mm0\n"
        "psubw %%mm5, %%mm2\n"
        "pxor %%mm3, %%mm3\n"
        "pxor %%mm1, %%mm1\n"
        "pcmpgtw %%mm0, %%mm3\n\t"
        "pcmpgtw %%mm2, %%mm1\n\t"
        "pxor %%mm3, %%mm0\n"
        "pxor %%mm1, %%mm2\n"
        "psubw %%mm3, %%mm0\n"
        "psubw %%mm1, %%mm2\n"
        "paddw %%mm0, %%mm2\n"
        "paddw %%mm2, %%mm6\n"

        "add %2, %0\n"
        "subl $2, %%ecx\n"
        " jnz 1b\n"

        "movq %%mm6, %%mm0\n"
        "punpcklwd %%mm7, %%mm0\n"
        "punpckhwd %%mm7, %%mm6\n"
        "paddd %%mm0, %%mm6\n"

        "movq %%mm6, %%mm0\n"
        "psrlq $32, %%mm6\n"
        "paddd %%mm6, %%mm0\n"
        "movd %%mm0, %1\n"
        : "+r" (pix1), "=r" (tmp)
        : "r" ((x86_reg) line_size), "g" (h - 2)
        : "%ecx");

    return tmp + hf_noise8_mmx(pix + 8, line_size, h);
}

static int nsse16_mmx(void *p, uint8_t *pix1, uint8_t *pix2,
                      int line_size, int h)
{
    MpegEncContext *c = p;
    int score1, score2;

    if (c)
        score1 = c->dsp.sse[0](c, pix1, pix2, line_size, h);
    else
        score1 = sse16_mmx(c, pix1, pix2, line_size, h);
    score2 = hf_noise16_mmx(pix1, line_size, h) -
             hf_noise16_mmx(pix2, line_size, h);

    if (c)
        return score1 + FFABS(score2) * c->avctx->nsse_weight;
    else
        return score1 + FFABS(score2) * 8;
}

static int nsse8_mmx(void *p, uint8_t *pix1, uint8_t *pix2,
                     int line_size, int h)
{
    MpegEncContext *c = p;
    int score1 = sse8_mmx(c, pix1, pix2, line_size, h);
    int score2 = hf_noise8_mmx(pix1, line_size, h) -
                 hf_noise8_mmx(pix2, line_size, h);

    if (c)
        return score1 + FFABS(score2) * c->avctx->nsse_weight;
    else
        return score1 + FFABS(score2) * 8;
}

static int vsad_intra16_mmx(void *v, uint8_t *pix, uint8_t *dummy,
                            int line_size, int h)
{
    int tmp;

    av_assert2((((int) pix) & 7) == 0);
    av_assert2((line_size & 7) == 0);

#define SUM(in0, in1, out0, out1)               \
    "movq (%0), %%mm2\n"                        \
    "movq 8(%0), %%mm3\n"                       \
    "add %2,%0\n"                               \
    "movq %%mm2, " #out0 "\n"                   \
    "movq %%mm3, " #out1 "\n"                   \
    "psubusb " #in0 ", %%mm2\n"                 \
    "psubusb " #in1 ", %%mm3\n"                 \
    "psubusb " #out0 ", " #in0 "\n"             \
    "psubusb " #out1 ", " #in1 "\n"             \
    "por %%mm2, " #in0 "\n"                     \
    "por %%mm3, " #in1 "\n"                     \
    "movq " #in0 ", %%mm2\n"                    \
    "movq " #in1 ", %%mm3\n"                    \
    "punpcklbw %%mm7, " #in0 "\n"               \
    "punpcklbw %%mm7, " #in1 "\n"               \
    "punpckhbw %%mm7, %%mm2\n"                  \
    "punpckhbw %%mm7, %%mm3\n"                  \
    "paddw " #in1 ", " #in0 "\n"                \
    "paddw %%mm3, %%mm2\n"                      \
    "paddw %%mm2, " #in0 "\n"                   \
    "paddw " #in0 ", %%mm6\n"


    __asm__ volatile (
        "movl    %3, %%ecx\n"
        "pxor %%mm6, %%mm6\n"
        "pxor %%mm7, %%mm7\n"
        "movq  (%0), %%mm0\n"
        "movq 8(%0), %%mm1\n"
        "add %2, %0\n"
        "jmp 2f\n"
        "1:\n"

        SUM(%%mm4, %%mm5, %%mm0, %%mm1)
        "2:\n"
        SUM(%%mm0, %%mm1, %%mm4, %%mm5)

        "subl $2, %%ecx\n"
        "jnz 1b\n"

        "movq  %%mm6, %%mm0\n"
        "psrlq $32,   %%mm6\n"
        "paddw %%mm6, %%mm0\n"
        "movq  %%mm0, %%mm6\n"
        "psrlq $16,   %%mm0\n"
        "paddw %%mm6, %%mm0\n"
        "movd  %%mm0, %1\n"
        : "+r" (pix), "=r" (tmp)
        : "r" ((x86_reg) line_size), "m" (h)
        : "%ecx");

    return tmp & 0xFFFF;
}
#undef SUM

static int vsad_intra16_mmxext(void *v, uint8_t *pix, uint8_t *dummy,
                               int line_size, int h)
{
    int tmp;

    av_assert2((((int) pix) & 7) == 0);
    av_assert2((line_size & 7) == 0);

#define SUM(in0, in1, out0, out1)               \
    "movq (%0), " #out0 "\n"                    \
    "movq 8(%0), " #out1 "\n"                   \
    "add %2, %0\n"                              \
    "psadbw " #out0 ", " #in0 "\n"              \
    "psadbw " #out1 ", " #in1 "\n"              \
    "paddw " #in1 ", " #in0 "\n"                \
    "paddw " #in0 ", %%mm6\n"

    __asm__ volatile (
        "movl %3, %%ecx\n"
        "pxor %%mm6, %%mm6\n"
        "pxor %%mm7, %%mm7\n"
        "movq (%0), %%mm0\n"
        "movq 8(%0), %%mm1\n"
        "add %2, %0\n"
        "jmp 2f\n"
        "1:\n"

        SUM(%%mm4, %%mm5, %%mm0, %%mm1)
        "2:\n"
        SUM(%%mm0, %%mm1, %%mm4, %%mm5)

        "subl $2, %%ecx\n"
        "jnz 1b\n"

        "movd %%mm6, %1\n"
        : "+r" (pix), "=r" (tmp)
        : "r" ((x86_reg) line_size), "m" (h)
        : "%ecx");

    return tmp;
}
#undef SUM

static int vsad16_mmx(void *v, uint8_t *pix1, uint8_t *pix2,
                      int line_size, int h)
{
    int tmp;

    av_assert2((((int) pix1) & 7) == 0);
    av_assert2((((int) pix2) & 7) == 0);
    av_assert2((line_size & 7) == 0);

#define SUM(in0, in1, out0, out1)       \
    "movq (%0), %%mm2\n"                \
    "movq (%1), " #out0 "\n"            \
    "movq 8(%0), %%mm3\n"               \
    "movq 8(%1), " #out1 "\n"           \
    "add %3, %0\n"                      \
    "add %3, %1\n"                      \
    "psubb " #out0 ", %%mm2\n"          \
    "psubb " #out1 ", %%mm3\n"          \
    "pxor %%mm7, %%mm2\n"               \
    "pxor %%mm7, %%mm3\n"               \
    "movq %%mm2, " #out0 "\n"           \
    "movq %%mm3, " #out1 "\n"           \
    "psubusb " #in0 ", %%mm2\n"         \
    "psubusb " #in1 ", %%mm3\n"         \
    "psubusb " #out0 ", " #in0 "\n"     \
    "psubusb " #out1 ", " #in1 "\n"     \
    "por %%mm2, " #in0 "\n"             \
    "por %%mm3, " #in1 "\n"             \
    "movq " #in0 ", %%mm2\n"            \
    "movq " #in1 ", %%mm3\n"            \
    "punpcklbw %%mm7, " #in0 "\n"       \
    "punpcklbw %%mm7, " #in1 "\n"       \
    "punpckhbw %%mm7, %%mm2\n"          \
    "punpckhbw %%mm7, %%mm3\n"          \
    "paddw " #in1 ", " #in0 "\n"        \
    "paddw %%mm3, %%mm2\n"              \
    "paddw %%mm2, " #in0 "\n"           \
    "paddw " #in0 ", %%mm6\n"


    __asm__ volatile (
        "movl %4, %%ecx\n"
        "pxor %%mm6, %%mm6\n"
        "pcmpeqw %%mm7, %%mm7\n"
        "psllw $15, %%mm7\n"
        "packsswb %%mm7, %%mm7\n"
        "movq (%0), %%mm0\n"
        "movq (%1), %%mm2\n"
        "movq 8(%0), %%mm1\n"
        "movq 8(%1), %%mm3\n"
        "add %3, %0\n"
        "add %3, %1\n"
        "psubb %%mm2, %%mm0\n"
        "psubb %%mm3, %%mm1\n"
        "pxor %%mm7, %%mm0\n"
        "pxor %%mm7, %%mm1\n"
        "jmp 2f\n"
        "1:\n"

        SUM(%%mm4, %%mm5, %%mm0, %%mm1)
        "2:\n"
        SUM(%%mm0, %%mm1, %%mm4, %%mm5)

        "subl $2, %%ecx\n"
        "jnz 1b\n"

        "movq %%mm6, %%mm0\n"
        "psrlq $32, %%mm6\n"
        "paddw %%mm6, %%mm0\n"
        "movq %%mm0, %%mm6\n"
        "psrlq $16, %%mm0\n"
        "paddw %%mm6, %%mm0\n"
        "movd %%mm0, %2\n"
        : "+r" (pix1), "+r" (pix2), "=r" (tmp)
        : "r" ((x86_reg) line_size), "m" (h)
        : "%ecx");

    return tmp & 0x7FFF;
}
#undef SUM

static int vsad16_mmxext(void *v, uint8_t *pix1, uint8_t *pix2,
                         int line_size, int h)
{
    int tmp;

    av_assert2((((int) pix1) & 7) == 0);
    av_assert2((((int) pix2) & 7) == 0);
    av_assert2((line_size & 7) == 0);

#define SUM(in0, in1, out0, out1)               \
    "movq (%0), " #out0 "\n"                    \
    "movq (%1), %%mm2\n"                        \
    "movq 8(%0), " #out1 "\n"                   \
    "movq 8(%1), %%mm3\n"                       \
    "add %3, %0\n"                              \
    "add %3, %1\n"                              \
    "psubb %%mm2, " #out0 "\n"                  \
    "psubb %%mm3, " #out1 "\n"                  \
    "pxor %%mm7, " #out0 "\n"                   \
    "pxor %%mm7, " #out1 "\n"                   \
    "psadbw " #out0 ", " #in0 "\n"              \
    "psadbw " #out1 ", " #in1 "\n"              \
    "paddw " #in1 ", " #in0 "\n"                \
    "paddw " #in0 ", %%mm6\n    "

    __asm__ volatile (
        "movl %4, %%ecx\n"
        "pxor %%mm6, %%mm6\n"
        "pcmpeqw %%mm7, %%mm7\n"
        "psllw $15, %%mm7\n"
        "packsswb %%mm7, %%mm7\n"
        "movq (%0), %%mm0\n"
        "movq (%1), %%mm2\n"
        "movq 8(%0), %%mm1\n"
        "movq 8(%1), %%mm3\n"
        "add %3, %0\n"
        "add %3, %1\n"
        "psubb %%mm2, %%mm0\n"
        "psubb %%mm3, %%mm1\n"
        "pxor %%mm7, %%mm0\n"
        "pxor %%mm7, %%mm1\n"
        "jmp 2f\n"
        "1:\n"

        SUM(%%mm4, %%mm5, %%mm0, %%mm1)
        "2:\n"
        SUM(%%mm0, %%mm1, %%mm4, %%mm5)

        "subl $2, %%ecx\n"
        "jnz 1b\n"

        "movd %%mm6, %2\n"
        : "+r" (pix1), "+r" (pix2), "=r" (tmp)
        : "r" ((x86_reg) line_size), "m" (h)
        : "%ecx");

    return tmp;
}
#undef SUM

static void diff_bytes_mmx(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w)
{
    x86_reg i = 0;

    if (w >= 16)
    __asm__ volatile (
        "1:                             \n\t"
        "movq  (%2, %0), %%mm0          \n\t"
        "movq  (%1, %0), %%mm1          \n\t"
        "psubb %%mm0, %%mm1             \n\t"
        "movq %%mm1, (%3, %0)           \n\t"
        "movq 8(%2, %0), %%mm0          \n\t"
        "movq 8(%1, %0), %%mm1          \n\t"
        "psubb %%mm0, %%mm1             \n\t"
        "movq %%mm1, 8(%3, %0)          \n\t"
        "add $16, %0                    \n\t"
        "cmp %4, %0                     \n\t"
        " jb 1b                         \n\t"
        : "+r" (i)
        : "r" (src1), "r" (src2), "r" (dst), "r" ((x86_reg) w - 15));

    for (; i < w; i++)
        dst[i + 0] = src1[i + 0] - src2[i + 0];
}

static void sub_hfyu_median_prediction_mmxext(uint8_t *dst, const uint8_t *src1,
                                              const uint8_t *src2, int w,
                                              int *left, int *left_top)
{
    x86_reg i = 0;
    uint8_t l, lt;

    __asm__ volatile (
        "movq  (%1, %0), %%mm0          \n\t" // LT
        "psllq $8, %%mm0                \n\t"
        "1:                             \n\t"
        "movq  (%1, %0), %%mm1          \n\t" // T
        "movq  -1(%2, %0), %%mm2        \n\t" // L
        "movq  (%2, %0), %%mm3          \n\t" // X
        "movq %%mm2, %%mm4              \n\t" // L
        "psubb %%mm0, %%mm2             \n\t"
        "paddb %%mm1, %%mm2             \n\t" // L + T - LT
        "movq %%mm4, %%mm5              \n\t" // L
        "pmaxub %%mm1, %%mm4            \n\t" // max(T, L)
        "pminub %%mm5, %%mm1            \n\t" // min(T, L)
        "pminub %%mm2, %%mm4            \n\t"
        "pmaxub %%mm1, %%mm4            \n\t"
        "psubb %%mm4, %%mm3             \n\t" // dst - pred
        "movq %%mm3, (%3, %0)           \n\t"
        "add $8, %0                     \n\t"
        "movq -1(%1, %0), %%mm0         \n\t" // LT
        "cmp %4, %0                     \n\t"
        " jb 1b                         \n\t"
        : "+r" (i)
        : "r" (src1), "r" (src2), "r" (dst), "r" ((x86_reg) w));

    l  = *left;
    lt = *left_top;

    dst[0] = src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt) & 0xFF);

    *left_top = src1[w - 1];
    *left     = src2[w - 1];
}

#define MMABS_MMX(a,z)                          \
    "pxor "    #z ", " #z "             \n\t"   \
    "pcmpgtw " #a ", " #z "             \n\t"   \
    "pxor "    #z ", " #a "             \n\t"   \
    "psubw "   #z ", " #a "             \n\t"

#define MMABS_MMXEXT(a, z)                      \
    "pxor "    #z ", " #z "             \n\t"   \
    "psubw "   #a ", " #z "             \n\t"   \
    "pmaxsw "  #z ", " #a "             \n\t"

#define MMABS_SSSE3(a,z)                        \
    "pabsw "   #a ", " #a "             \n\t"

#define MMABS_SUM(a,z, sum)                     \
    MMABS(a,z)                                  \
    "paddusw " #a ", " #sum "           \n\t"

/* FIXME: HSUM_* saturates at 64k, while an 8x8 hadamard or dct block can get
 * up to about 100k on extreme inputs. But that's very unlikely to occur in
 * natural video, and it's even more unlikely to not have any alternative
 * mvs/modes with lower cost. */
#define HSUM_MMX(a, t, dst)                     \
    "movq    " #a ", " #t "             \n\t"   \
    "psrlq      $32, " #a "             \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "movq    " #a ", " #t "             \n\t"   \
    "psrlq      $16, " #a "             \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "movd    " #a ", " #dst "           \n\t"   \

#define HSUM_MMXEXT(a, t, dst)                  \
    "pshufw   $0x0E, " #a ", " #t "     \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "pshufw   $0x01, " #a ", " #t "     \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "movd    " #a ", " #dst "           \n\t"   \

#define HSUM_SSE2(a, t, dst)                    \
    "movhlps " #a ", " #t "             \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "pshuflw  $0x0E, " #a ", " #t "     \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "pshuflw  $0x01, " #a ", " #t "     \n\t"   \
    "paddusw " #t ", " #a "             \n\t"   \
    "movd    " #a ", " #dst "           \n\t"   \

#define DCT_SAD4(m, mm, o)                      \
    "mov"#m" "#o" +  0(%1), " #mm "2    \n\t"   \
    "mov"#m" "#o" + 16(%1), " #mm "3    \n\t"   \
    "mov"#m" "#o" + 32(%1), " #mm "4    \n\t"   \
    "mov"#m" "#o" + 48(%1), " #mm "5    \n\t"   \
    MMABS_SUM(mm ## 2, mm ## 6, mm ## 0)        \
    MMABS_SUM(mm ## 3, mm ## 7, mm ## 1)        \
    MMABS_SUM(mm ## 4, mm ## 6, mm ## 0)        \
    MMABS_SUM(mm ## 5, mm ## 7, mm ## 1)        \

#define DCT_SAD_MMX                             \
    "pxor    %%mm0, %%mm0               \n\t"   \
    "pxor    %%mm1, %%mm1               \n\t"   \
    DCT_SAD4(q, %%mm, 0)                        \
    DCT_SAD4(q, %%mm, 8)                        \
    DCT_SAD4(q, %%mm, 64)                       \
    DCT_SAD4(q, %%mm, 72)                       \
    "paddusw %%mm1, %%mm0               \n\t"   \
    HSUM(%%mm0, %%mm1, %0)

#define DCT_SAD_SSE2                            \
    "pxor    %%xmm0, %%xmm0             \n\t"   \
    "pxor    %%xmm1, %%xmm1             \n\t"   \
    DCT_SAD4(dqa, %%xmm, 0)                     \
    DCT_SAD4(dqa, %%xmm, 64)                    \
    "paddusw %%xmm1, %%xmm0             \n\t"   \
    HSUM(%%xmm0, %%xmm1, %0)

#define DCT_SAD_FUNC(cpu)                           \
static int sum_abs_dctelem_ ## cpu(int16_t *block)  \
{                                                   \
    int sum;                                        \
    __asm__ volatile (                              \
        DCT_SAD                                     \
        :"=r"(sum)                                  \
        :"r"(block));                               \
    return sum & 0xFFFF;                            \
}

#define DCT_SAD         DCT_SAD_MMX
#define HSUM(a, t, dst) HSUM_MMX(a, t, dst)
#define MMABS(a, z)     MMABS_MMX(a, z)
DCT_SAD_FUNC(mmx)
#undef MMABS
#undef HSUM

#define HSUM(a, t, dst) HSUM_MMXEXT(a, t, dst)
#define MMABS(a, z)     MMABS_MMXEXT(a, z)
DCT_SAD_FUNC(mmxext)
#undef HSUM
#undef DCT_SAD

#define DCT_SAD         DCT_SAD_SSE2
#define HSUM(a, t, dst) HSUM_SSE2(a, t, dst)
DCT_SAD_FUNC(sse2)
#undef MMABS

#if HAVE_SSSE3_INLINE
#define MMABS(a, z)     MMABS_SSSE3(a, z)
DCT_SAD_FUNC(ssse3)
#undef MMABS
#endif
#undef HSUM
#undef DCT_SAD

static int ssd_int8_vs_int16_mmx(const int8_t *pix1, const int16_t *pix2,
                                 int size)
{
    int sum;
    x86_reg i = size;

    __asm__ volatile (
        "pxor %%mm4, %%mm4 \n"
        "1: \n"
        "sub $8, %0 \n"
        "movq (%2, %0), %%mm2 \n"
        "movq (%3, %0, 2), %%mm0 \n"
        "movq 8(%3, %0, 2), %%mm1 \n"
        "punpckhbw %%mm2, %%mm3 \n"
        "punpcklbw %%mm2, %%mm2 \n"
        "psraw $8, %%mm3 \n"
        "psraw $8, %%mm2 \n"
        "psubw %%mm3, %%mm1 \n"
        "psubw %%mm2, %%mm0 \n"
        "pmaddwd %%mm1, %%mm1 \n"
        "pmaddwd %%mm0, %%mm0 \n"
        "paddd %%mm1, %%mm4 \n"
        "paddd %%mm0, %%mm4 \n"
        "jg 1b \n"
        "movq %%mm4, %%mm3 \n"
        "psrlq $32, %%mm3 \n"
        "paddd %%mm3, %%mm4 \n"
        "movd %%mm4, %1 \n"
        : "+r" (i), "=r" (sum)
        : "r" (pix1), "r" (pix2));

    return sum;
}

#define PHADDD(a, t)                            \
    "movq  " #a ", " #t "               \n\t"   \
    "psrlq    $32, " #a "               \n\t"   \
    "paddd " #t ", " #a "               \n\t"

/*
 * pmulhw:   dst[0 - 15] = (src[0 - 15] * dst[0 - 15])[16 - 31]
 * pmulhrw:  dst[0 - 15] = (src[0 - 15] * dst[0 - 15] + 0x8000)[16 - 31]
 * pmulhrsw: dst[0 - 15] = (src[0 - 15] * dst[0 - 15] + 0x4000)[15 - 30]
 */
#define PMULHRW(x, y, s, o)                     \
    "pmulhw " #s ", " #x "              \n\t"   \
    "pmulhw " #s ", " #y "              \n\t"   \
    "paddw  " #o ", " #x "              \n\t"   \
    "paddw  " #o ", " #y "              \n\t"   \
    "psraw      $1, " #x "              \n\t"   \
    "psraw      $1, " #y "              \n\t"
#define DEF(x) x ## _mmx
#define SET_RND MOVQ_WONE
#define SCALE_OFFSET 1

#include "dsputil_qns_template.c"

#undef DEF
#undef SET_RND
#undef SCALE_OFFSET
#undef PMULHRW

#define DEF(x) x ## _3dnow
#define SET_RND(x)
#define SCALE_OFFSET 0
#define PMULHRW(x, y, s, o)                     \
    "pmulhrw " #s ", " #x "             \n\t"   \
    "pmulhrw " #s ", " #y "             \n\t"

#include "dsputil_qns_template.c"

#undef DEF
#undef SET_RND
#undef SCALE_OFFSET
#undef PMULHRW

#if HAVE_SSSE3_INLINE
#undef PHADDD
#define DEF(x) x ## _ssse3
#define SET_RND(x)
#define SCALE_OFFSET -1

#define PHADDD(a, t)                            \
    "pshufw $0x0E, " #a ", " #t "       \n\t"   \
    /* faster than phaddd on core2 */           \
    "paddd " #t ", " #a "               \n\t"

#define PMULHRW(x, y, s, o)                     \
    "pmulhrsw " #s ", " #x "            \n\t"   \
    "pmulhrsw " #s ", " #y "            \n\t"

#include "dsputil_qns_template.c"

#undef DEF
#undef SET_RND
#undef SCALE_OFFSET
#undef PMULHRW
#undef PHADDD
#endif /* HAVE_SSSE3_INLINE */

#endif /* HAVE_INLINE_ASM */

int ff_sse16_sse2(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h);

#define hadamard_func(cpu)                                              \
int ff_hadamard8_diff_ ## cpu(void *s, uint8_t *src1, uint8_t *src2,    \
                              int stride, int h);                       \
int ff_hadamard8_diff16_ ## cpu(void *s, uint8_t *src1, uint8_t *src2,  \
                                int stride, int h);

hadamard_func(mmx)
hadamard_func(mmxext)
hadamard_func(sse2)
hadamard_func(ssse3)

av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
{
    int cpu_flags = av_get_cpu_flags();
    const int dct_algo = avctx->dct_algo;

#if HAVE_YASM
    int bit_depth = avctx->bits_per_raw_sample;

    if (EXTERNAL_MMX(cpu_flags)) {
        if (bit_depth <= 8)
            c->get_pixels = ff_get_pixels_mmx;
        c->diff_pixels = ff_diff_pixels_mmx;
        c->pix_sum     = ff_pix_sum16_mmx;
        c->pix_norm1   = ff_pix_norm1_mmx;
    }

    if (EXTERNAL_SSE2(cpu_flags))
        if (bit_depth <= 8)
            c->get_pixels = ff_get_pixels_sse2;
#endif /* HAVE_YASM */

#if HAVE_INLINE_ASM
    if (INLINE_MMX(cpu_flags)) {
        if (avctx->bits_per_raw_sample <= 8 &&
            (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX))
            c->fdct = ff_fdct_mmx;

        c->diff_bytes      = diff_bytes_mmx;
        c->sum_abs_dctelem = sum_abs_dctelem_mmx;

        c->sse[0]  = sse16_mmx;
        c->sse[1]  = sse8_mmx;
        c->vsad[4] = vsad_intra16_mmx;

        c->nsse[0] = nsse16_mmx;
        c->nsse[1] = nsse8_mmx;
        if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
            c->vsad[0]      = vsad16_mmx;
            c->try_8x8basis = try_8x8basis_mmx;
        }
        c->add_8x8basis = add_8x8basis_mmx;

        c->ssd_int8_vs_int16 = ssd_int8_vs_int16_mmx;
    }

    if (INLINE_AMD3DNOW(cpu_flags)) {
        if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
            c->try_8x8basis = try_8x8basis_3dnow;
        }
        c->add_8x8basis = add_8x8basis_3dnow;
    }

    if (INLINE_MMXEXT(cpu_flags)) {
        if (avctx->bits_per_raw_sample <= 8 &&
            (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX))
            c->fdct = ff_fdct_mmxext;

        c->sum_abs_dctelem = sum_abs_dctelem_mmxext;
        c->vsad[4]         = vsad_intra16_mmxext;

        if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
            c->vsad[0] = vsad16_mmxext;
        }

        c->sub_hfyu_median_prediction = sub_hfyu_median_prediction_mmxext;
    }

    if (INLINE_SSE2(cpu_flags)) {
        if (avctx->bits_per_raw_sample <= 8 &&
            (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX))
            c->fdct = ff_fdct_sse2;

        c->sum_abs_dctelem = sum_abs_dctelem_sse2;
    }

#if HAVE_SSSE3_INLINE
    if (INLINE_SSSE3(cpu_flags)) {
        if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
            c->try_8x8basis = try_8x8basis_ssse3;
        }
        c->add_8x8basis    = add_8x8basis_ssse3;
        c->sum_abs_dctelem = sum_abs_dctelem_ssse3;
    }
#endif
#endif /* HAVE_INLINE_ASM */

    if (EXTERNAL_MMX(cpu_flags)) {
        c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx;
        c->hadamard8_diff[1] = ff_hadamard8_diff_mmx;
    }

    if (EXTERNAL_MMXEXT(cpu_flags)) {
        c->hadamard8_diff[0] = ff_hadamard8_diff16_mmxext;
        c->hadamard8_diff[1] = ff_hadamard8_diff_mmxext;
    }

    if (EXTERNAL_SSE2(cpu_flags)) {
        c->sse[0] = ff_sse16_sse2;

#if HAVE_ALIGNED_STACK
        c->hadamard8_diff[0] = ff_hadamard8_diff16_sse2;
        c->hadamard8_diff[1] = ff_hadamard8_diff_sse2;
#endif
    }

    if (EXTERNAL_SSSE3(cpu_flags) && HAVE_ALIGNED_STACK) {
        c->hadamard8_diff[0] = ff_hadamard8_diff16_ssse3;
        c->hadamard8_diff[1] = ff_hadamard8_diff_ssse3;
    }

    ff_dsputil_init_pix_mmx(c, avctx);
}