dsputil_mmx.c

这是著名的TCPMP播放器在WINDWOWS,和WINCE下编译通过的源程序.笔者对其中的LIBMAD库做了针对ARM MPU的优化. 并增加了词幕功能.

/*
 * MMX optimized DSP utils
 * Copyright (c) 2000, 2001 Fabrice Bellard.
 * Copyright (c) 2002-2004 Michael Niedermayer <[e-mail]>
 *
 * This library 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 of the License, or (at your option) any later version.
 *
 * This library 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 this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 * MMX optimization by Nick Kurshev <[e-mail]>
 */

#include "../dsputil.h"
#include "../simple_idct.h"
#include "../mpegvideo.h"
#include "mmx.h"

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

extern const uint8_t ff_h263_loop_filter_strength[32];
extern void ff_idct_xvid_mmx(short *block);
extern void ff_idct_xvid_mmx2(short *block);

int mm_flags; /* multimedia extension flags */

/* pixel operations */
static const uint64_t mm_bone attribute_used __attribute__ ((aligned(8))) = 0x0101010101010101ULL;
static const uint64_t mm_wone attribute_used __attribute__ ((aligned(8))) = 0x0001000100010001ULL;
static const uint64_t mm_wtwo attribute_used __attribute__ ((aligned(8))) = 0x0002000200020002ULL;

static const uint64_t ff_pw_20 attribute_used __attribute__ ((aligned(8))) = 0x0014001400140014ULL;
static const uint64_t ff_pw_3  attribute_used __attribute__ ((aligned(8))) = 0x0003000300030003ULL;
static const uint64_t ff_pw_4  attribute_used __attribute__ ((aligned(8))) = 0x0004000400040004ULL;
static const uint64_t ff_pw_5  attribute_used __attribute__ ((aligned(8))) = 0x0005000500050005ULL;
static const uint64_t ff_pw_16 attribute_used __attribute__ ((aligned(8))) = 0x0010001000100010ULL;
static const uint64_t ff_pw_32 attribute_used __attribute__ ((aligned(8))) = 0x0020002000200020ULL;
static const uint64_t ff_pw_64 attribute_used __attribute__ ((aligned(8))) = 0x0040004000400040ULL;
static const uint64_t ff_pw_15 attribute_used __attribute__ ((aligned(8))) = 0x000F000F000F000FULL;

static const uint64_t ff_pb_3F attribute_used __attribute__ ((aligned(8))) = 0x3F3F3F3F3F3F3F3FULL;
static const uint64_t ff_pb_FC attribute_used __attribute__ ((aligned(8))) = 0xFCFCFCFCFCFCFCFCULL;

#define JUMPALIGN() __asm __volatile (".balign 8"::)
#define MOVQ_ZERO(regd)  __asm __volatile ("pxor %%" #regd ", %%" #regd ::)

#define MOVQ_WONE(regd) \
    __asm __volatile ( \
    "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
    "psrlw $15, %%" #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"(mm_bone))
#define MOVQ_WTWO(regd)  __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_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
/* for Athlons PAVGUSB is prefered */
#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

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

#ifdef CONFIG_ENCODERS
static void get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size)
{
    asm volatile(
        "mov $-128, %%"REG_a"	\n\t"
        "pxor %%mm7, %%mm7	\n\t"
        ".balign 16		\n\t"
        "1:			\n\t"
        "movq (%0), %%mm0	\n\t"
        "movq (%0, %2), %%mm2	\n\t"
        "movq %%mm0, %%mm1	\n\t"
        "movq %%mm2, %%mm3	\n\t"
        "punpcklbw %%mm7, %%mm0	\n\t"
        "punpckhbw %%mm7, %%mm1	\n\t"
        "punpcklbw %%mm7, %%mm2	\n\t"
        "punpckhbw %%mm7, %%mm3	\n\t"
        "movq %%mm0, (%1, %%"REG_a")\n\t"
        "movq %%mm1, 8(%1, %%"REG_a")\n\t"
        "movq %%mm2, 16(%1, %%"REG_a")\n\t"
        "movq %%mm3, 24(%1, %%"REG_a")\n\t"
        "add %3, %0		\n\t"
        "add $32, %%"REG_a"	\n\t"
        "js 1b			\n\t"
        : "+r" (pixels)
        : "r" (block+64), "r" ((long)line_size), "r" ((long)line_size*2)
        : "%"REG_a
    );
}

static inline void diff_pixels_mmx(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride)
{
    asm volatile(
        "pxor %%mm7, %%mm7	\n\t"
        "mov $-128, %%"REG_a"	\n\t"
        ".balign 16		\n\t"
        "1:			\n\t"
        "movq (%0), %%mm0	\n\t"
        "movq (%1), %%mm2	\n\t"
        "movq %%mm0, %%mm1	\n\t"
        "movq %%mm2, %%mm3	\n\t"
        "punpcklbw %%mm7, %%mm0	\n\t"
        "punpckhbw %%mm7, %%mm1	\n\t"
        "punpcklbw %%mm7, %%mm2	\n\t"
        "punpckhbw %%mm7, %%mm3	\n\t"
        "psubw %%mm2, %%mm0	\n\t"
        "psubw %%mm3, %%mm1	\n\t"
        "movq %%mm0, (%2, %%"REG_a")\n\t"
        "movq %%mm1, 8(%2, %%"REG_a")\n\t"
        "add %3, %0		\n\t"
        "add %3, %1		\n\t"
        "add $16, %%"REG_a"	\n\t"
        "jnz 1b			\n\t"
        : "+r" (s1), "+r" (s2)
        : "r" (block+64), "r" ((long)stride)
        : "%"REG_a
    );
}
#endif //CONFIG_ENCODERS

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" ((long)line_size), "r" ((long)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" ((long)line_size), "r" ((long)line_size*3), "r"(p)
	    :"memory");
}

static const unsigned char __align8 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;
    MOVQ_ZERO(mm7);
    i = 4;
    do {
	__asm __volatile(
		"movq	(%2), %%mm0\n\t"
		"movq	8(%2), %%mm1\n\t"
		"movq	16(%2), %%mm2\n\t"
		"movq	24(%2), %%mm3\n\t"
		"movq	%0, %%mm4\n\t"
		"movq	%1, %%mm6\n\t"
		"movq	%%mm4, %%mm5\n\t"
		"punpcklbw %%mm7, %%mm4\n\t"
		"punpckhbw %%mm7, %%mm5\n\t"
		"paddsw	%%mm4, %%mm0\n\t"
		"paddsw	%%mm5, %%mm1\n\t"
		"movq	%%mm6, %%mm5\n\t"
		"punpcklbw %%mm7, %%mm6\n\t"
		"punpckhbw %%mm7, %%mm5\n\t"
		"paddsw	%%mm6, %%mm2\n\t"
		"paddsw	%%mm5, %%mm3\n\t"
		"packuswb %%mm1, %%mm0\n\t"
		"packuswb %%mm3, %%mm2\n\t"
		"movq	%%mm0, %0\n\t"
		"movq	%%mm2, %1\n\t"
		:"+m"(*pix), "+m"(*(pix+line_size))
		:"r"(p)
		:"memory");
        pix += line_size*2;
        p += 16;
    } while (--i);
}

static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    __asm __volatile(
	 "lea (%3, %3), %%"REG_a"	\n\t"
	 ".balign 8			\n\t"
	 "1:				\n\t"
	 "movd (%1), %%mm0		\n\t"
	 "movd (%1, %3), %%mm1		\n\t"
	 "movd %%mm0, (%2)		\n\t"
	 "movd %%mm1, (%2, %3)		\n\t"
	 "add %%"REG_a", %1		\n\t"
	 "add %%"REG_a", %2		\n\t"
	 "movd (%1), %%mm0		\n\t"
	 "movd (%1, %3), %%mm1		\n\t"
	 "movd %%mm0, (%2)		\n\t"
	 "movd %%mm1, (%2, %3)		\n\t"
	 "add %%"REG_a", %1		\n\t"
	 "add %%"REG_a", %2		\n\t"
	 "subl $4, %0			\n\t"
	 "jnz 1b			\n\t"
	 : "+g"(h), "+r" (pixels),  "+r" (block)
	 : "r"((long)line_size)
	 : "%"REG_a, "memory"
	);
}

static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    __asm __volatile(
	 "lea (%3, %3), %%"REG_a"	\n\t"
	 ".balign 8			\n\t"
	 "1:				\n\t"