dsputil_mmx.c

ffmpeg移植到symbian的全部源代码

/*
 * MMX optimized DSP utils
 * Copyright (c) 2000, 2001 Fabrice Bellard.
 * Copyright (c) 2002-2004 Michael Niedermayer <[e-mail]>
 *
 * 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
 *
 * MMX optimization by Nick Kurshev <[e-mail]>
 */

#include "libavutil/x86_cpu.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/h263.h"
#include "libavcodec/mpegvideo.h"
#include "libavcodec/simple_idct.h"
#include "dsputil_mmx.h"
#include "mmx.h"
#include "vp3dsp_mmx.h"
#include "vp3dsp_sse2.h"
#include "idct_xvid.h"

//#undef NDEBUG
//#include <assert.h>

int mm_flags; /* multimedia extension flags */

/* pixel operations */
DECLARE_ALIGNED_8 (const uint64_t, ff_bone) = 0x0101010101010101ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_wtwo) = 0x0002000200020002ULL;

DECLARE_ALIGNED_16(const uint64_t, ff_pdw_80000000[2]) =
{0x8000000080000000ULL, 0x8000000080000000ULL};

DECLARE_ALIGNED_8 (const uint64_t, ff_pw_3  ) = 0x0003000300030003ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_4  ) = 0x0004000400040004ULL;
DECLARE_ALIGNED_16(const xmm_t,    ff_pw_5  ) = {0x0005000500050005ULL, 0x0005000500050005ULL};
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_8  ) = 0x0008000800080008ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_15 ) = 0x000F000F000F000FULL;
DECLARE_ALIGNED_16(const xmm_t,    ff_pw_16 ) = {0x0010001000100010ULL, 0x0010001000100010ULL};
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_20 ) = 0x0014001400140014ULL;
DECLARE_ALIGNED_16(const xmm_t,    ff_pw_28 ) = {0x001C001C001C001CULL, 0x001C001C001C001CULL};
DECLARE_ALIGNED_16(const xmm_t,    ff_pw_32 ) = {0x0020002000200020ULL, 0x0020002000200020ULL};
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_42 ) = 0x002A002A002A002AULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_64 ) = 0x0040004000400040ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_96 ) = 0x0060006000600060ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_128) = 0x0080008000800080ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_255) = 0x00ff00ff00ff00ffULL;

DECLARE_ALIGNED_8 (const uint64_t, ff_pb_1  ) = 0x0101010101010101ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pb_3  ) = 0x0303030303030303ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pb_7  ) = 0x0707070707070707ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pb_3F ) = 0x3F3F3F3F3F3F3F3FULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pb_A1 ) = 0xA1A1A1A1A1A1A1A1ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pb_FC ) = 0xFCFCFCFCFCFCFCFCULL;

DECLARE_ALIGNED_16(const double, ff_pd_1[2]) = { 1.0, 1.0 };
DECLARE_ALIGNED_16(const double, ff_pd_2[2]) = { 2.0, 2.0 };

#define JUMPALIGN() asm volatile (ASMALIGN(3)::)
#define MOVQ_ZERO(regd)  asm volatile ("pxor %%" #regd ", %%" #regd ::)

#define MOVQ_BFE(regd) \
    asm volatile ( \
    "pcmpeqd %%" #regd ", %%" #regd " \n\t"\
    "paddb %%" #regd ", %%" #regd " \n\t" ::)

#ifndef PIC
#define MOVQ_BONE(regd)  asm volatile ("movq %0, %%" #regd " \n\t" ::"m"(ff_bone))
#define MOVQ_WTWO(regd)  asm volatile ("movq %0, %%" #regd " \n\t" ::"m"(ff_wtwo))
#else
// for shared library it's better to use this way for accessing constants
// pcmpeqd -> -1
#define MOVQ_BONE(regd) \
    asm volatile ( \
    "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
    "psrlw $15, %%" #regd " \n\t" \
    "packuswb %%" #regd ", %%" #regd " \n\t" ::)

#define MOVQ_WTWO(regd) \
    asm volatile ( \
    "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
    "psrlw $15, %%" #regd " \n\t" \
    "psllw $1, %%" #regd " \n\t"::)

#endif

// using regr as temporary and for the output result
// first argument is unmodifed and second is trashed
// regfe is supposed to contain 0xfefefefefefefefe
#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \
    "movq " #rega ", " #regr "  \n\t"\
    "pand " #regb ", " #regr "  \n\t"\
    "pxor " #rega ", " #regb "  \n\t"\
    "pand " #regfe "," #regb "  \n\t"\
    "psrlq $1, " #regb "        \n\t"\
    "paddb " #regb ", " #regr " \n\t"

#define PAVGB_MMX(rega, regb, regr, regfe) \
    "movq " #rega ", " #regr "  \n\t"\
    "por  " #regb ", " #regr "  \n\t"\
    "pxor " #rega ", " #regb "  \n\t"\
    "pand " #regfe "," #regb "  \n\t"\
    "psrlq $1, " #regb "        \n\t"\
    "psubb " #regb ", " #regr " \n\t"

// mm6 is supposed to contain 0xfefefefefefefefe
#define PAVGBP_MMX_NO_RND(rega, regb, regr,  regc, regd, regp) \
    "movq " #rega ", " #regr "  \n\t"\
    "movq " #regc ", " #regp "  \n\t"\
    "pand " #regb ", " #regr "  \n\t"\
    "pand " #regd ", " #regp "  \n\t"\
    "pxor " #rega ", " #regb "  \n\t"\
    "pxor " #regc ", " #regd "  \n\t"\
    "pand %%mm6, " #regb "      \n\t"\
    "pand %%mm6, " #regd "      \n\t"\
    "psrlq $1, " #regb "        \n\t"\
    "psrlq $1, " #regd "        \n\t"\
    "paddb " #regb ", " #regr " \n\t"\
    "paddb " #regd ", " #regp " \n\t"

#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \
    "movq " #rega ", " #regr "  \n\t"\
    "movq " #regc ", " #regp "  \n\t"\
    "por  " #regb ", " #regr "  \n\t"\
    "por  " #regd ", " #regp "  \n\t"\
    "pxor " #rega ", " #regb "  \n\t"\
    "pxor " #regc ", " #regd "  \n\t"\
    "pand %%mm6, " #regb "      \n\t"\
    "pand %%mm6, " #regd "      \n\t"\
    "psrlq $1, " #regd "        \n\t"\
    "psrlq $1, " #regb "        \n\t"\
    "psubb " #regb ", " #regr " \n\t"\
    "psubb " #regd ", " #regp " \n\t"

/***********************************/
/* MMX no rounding */
#define DEF(x, y) x ## _no_rnd_ ## y ##_mmx
#define SET_RND  MOVQ_WONE
#define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX_NO_RND(a, b, c, d, e, f)
#define PAVGB(a, b, c, e)               PAVGB_MMX_NO_RND(a, b, c, e)

#include "dsputil_mmx_rnd.h"

#undef DEF
#undef SET_RND
#undef PAVGBP
#undef PAVGB
/***********************************/
/* MMX rounding */

#define DEF(x, y) x ## _ ## y ##_mmx
#define SET_RND  MOVQ_WTWO
#define PAVGBP(a, b, c, d, e, f)        PAVGBP_MMX(a, b, c, d, e, f)
#define PAVGB(a, b, c, e)               PAVGB_MMX(a, b, c, e)

#include "dsputil_mmx_rnd.h"

#undef DEF
#undef SET_RND
#undef PAVGBP
#undef PAVGB

/***********************************/
/* 3Dnow specific */

#define DEF(x) x ## _3dnow
#define PAVGB "pavgusb"

#include "dsputil_mmx_avg.h"

#undef DEF
#undef PAVGB

/***********************************/
/* MMX2 specific */

#define DEF(x) x ## _mmx2

/* Introduced only in MMX2 set */
#define PAVGB "pavgb"

#include "dsputil_mmx_avg.h"

#undef DEF
#undef PAVGB

#define put_no_rnd_pixels16_mmx put_pixels16_mmx
#define put_no_rnd_pixels8_mmx put_pixels8_mmx
#define put_pixels16_mmx2 put_pixels16_mmx
#define put_pixels8_mmx2 put_pixels8_mmx
#define put_pixels4_mmx2 put_pixels4_mmx
#define put_no_rnd_pixels16_mmx2 put_no_rnd_pixels16_mmx
#define put_no_rnd_pixels8_mmx2 put_no_rnd_pixels8_mmx
#define put_pixels16_3dnow put_pixels16_mmx
#define put_pixels8_3dnow put_pixels8_mmx
#define put_pixels4_3dnow put_pixels4_mmx
#define put_no_rnd_pixels16_3dnow put_no_rnd_pixels16_mmx
#define put_no_rnd_pixels8_3dnow put_no_rnd_pixels8_mmx

/***********************************/
/* standard MMX */

void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
{
    const DCTELEM *p;
    uint8_t *pix;

    /* read the pixels */
    p = block;
    pix = pixels;
    /* unrolled loop */
        asm volatile(
                "movq   %3, %%mm0               \n\t"
                "movq   8%3, %%mm1              \n\t"
                "movq   16%3, %%mm2             \n\t"
                "movq   24%3, %%mm3             \n\t"
                "movq   32%3, %%mm4             \n\t"
                "movq   40%3, %%mm5             \n\t"
                "movq   48%3, %%mm6             \n\t"
                "movq   56%3, %%mm7             \n\t"
                "packuswb %%mm1, %%mm0          \n\t"
                "packuswb %%mm3, %%mm2          \n\t"
                "packuswb %%mm5, %%mm4          \n\t"
                "packuswb %%mm7, %%mm6          \n\t"
                "movq   %%mm0, (%0)             \n\t"
                "movq   %%mm2, (%0, %1)         \n\t"
                "movq   %%mm4, (%0, %1, 2)      \n\t"
                "movq   %%mm6, (%0, %2)         \n\t"
                ::"r" (pix), "r" ((x86_reg)line_size), "r" ((x86_reg)line_size*3), "m"(*p)
                :"memory");
        pix += line_size*4;
        p += 32;

    // if here would be an exact copy of the code above
    // compiler would generate some very strange code
    // thus using "r"
    asm volatile(
            "movq       (%3), %%mm0             \n\t"
            "movq       8(%3), %%mm1            \n\t"
            "movq       16(%3), %%mm2           \n\t"
            "movq       24(%3), %%mm3           \n\t"
            "movq       32(%3), %%mm4           \n\t"
            "movq       40(%3), %%mm5           \n\t"
            "movq       48(%3), %%mm6           \n\t"
            "movq       56(%3), %%mm7           \n\t"
            "packuswb %%mm1, %%mm0              \n\t"
            "packuswb %%mm3, %%mm2              \n\t"
            "packuswb %%mm5, %%mm4              \n\t"
            "packuswb %%mm7, %%mm6              \n\t"
            "movq       %%mm0, (%0)             \n\t"
            "movq       %%mm2, (%0, %1)         \n\t"
            "movq       %%mm4, (%0, %1, 2)      \n\t"
            "movq       %%mm6, (%0, %2)         \n\t"
            ::"r" (pix), "r" ((x86_reg)line_size), "r" ((x86_reg)line_size*3), "r"(p)
            :"memory");
}

static DECLARE_ALIGNED_8(const unsigned char, vector128[8]) =
  { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 };

void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
{
    int i;

    movq_m2r(*vector128, mm1);
    for (i = 0; i < 8; i++) {
        movq_m2r(*(block), mm0);
        packsswb_m2r(*(block + 4), mm0);
        block += 8;
        paddb_r2r(mm1, mm0);
        movq_r2m(mm0, *pixels);
        pixels += line_size;
    }
}

void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size)
{
    const DCTELEM *p;
    uint8_t *pix;
    int i;

    /* read the pixels */
    p = block;
    pix = pixels;