fdct_altivec.c

Trolltech公司发布的图形界面操作系统。可在qt-embedded-2.3.10平台上编译为嵌入式图形界面操作系统。

/* ffmpeg/libavcodec/ppc/fdct_altivec.c, this file is part of the
 * AltiVec optimized library for the FFMPEG Multimedia System
 * Copyright (C) 2003  James Klicman <[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 2.1 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
 */


#include "../common.h"
#include "../dsputil.h"
#include "dsputil_altivec.h"
#include "gcc_fixes.h"


#define vs16(v) ((vector signed short)(v))
#define vs32(v) ((vector signed int)(v))
#define vu8(v)  ((vector unsigned char)(v))
#define vu16(v) ((vector unsigned short)(v))
#define vu32(v) ((vector unsigned int)(v))


#define C1     0.98078525066375732421875000 /* cos(1*PI/16) */
#define C2     0.92387950420379638671875000 /* cos(2*PI/16) */
#define C3     0.83146959543228149414062500 /* cos(3*PI/16) */
#define C4     0.70710676908493041992187500 /* cos(4*PI/16) */
#define C5     0.55557024478912353515625000 /* cos(5*PI/16) */
#define C6     0.38268342614173889160156250 /* cos(6*PI/16) */
#define C7     0.19509032368659973144531250 /* cos(7*PI/16) */
#define SQRT_2 1.41421353816986083984375000 /* sqrt(2)      */


#define W0 -(2 * C2)
#define W1 (2 * C6)
#define W2 (SQRT_2 * C6)
#define W3 (SQRT_2 * C3)
#define W4 (SQRT_2 * (-C1 + C3 + C5 - C7))
#define W5 (SQRT_2 * ( C1 + C3 - C5 + C7))
#define W6 (SQRT_2 * ( C1 + C3 + C5 - C7))
#define W7 (SQRT_2 * ( C1 + C3 - C5 - C7))
#define W8 (SQRT_2 * ( C7 - C3))
#define W9 (SQRT_2 * (-C1 - C3))
#define WA (SQRT_2 * (-C3 - C5))
#define WB (SQRT_2 * ( C5 - C3))


static vector float fdctconsts[3] = {
    (vector float)AVV( W0, W1, W2, W3 ),
    (vector float)AVV( W4, W5, W6, W7 ),
    (vector float)AVV( W8, W9, WA, WB )
};

#define LD_W0 vec_splat(cnsts0, 0)
#define LD_W1 vec_splat(cnsts0, 1)
#define LD_W2 vec_splat(cnsts0, 2)
#define LD_W3 vec_splat(cnsts0, 3)
#define LD_W4 vec_splat(cnsts1, 0)
#define LD_W5 vec_splat(cnsts1, 1)
#define LD_W6 vec_splat(cnsts1, 2)
#define LD_W7 vec_splat(cnsts1, 3)
#define LD_W8 vec_splat(cnsts2, 0)
#define LD_W9 vec_splat(cnsts2, 1)
#define LD_WA vec_splat(cnsts2, 2)
#define LD_WB vec_splat(cnsts2, 3)


#define FDCTROW(b0,b1,b2,b3,b4,b5,b6,b7) /* {{{ */                  \
    x0 = vec_add(b0, b7);               /* x0 = b0 + b7; */         \
    x7 = vec_sub(b0, b7);               /* x7 = b0 - b7; */         \
    x1 = vec_add(b1, b6);               /* x1 = b1 + b6; */         \
    x6 = vec_sub(b1, b6);               /* x6 = b1 - b6; */         \
    x2 = vec_add(b2, b5);               /* x2 = b2 + b5; */         \
    x5 = vec_sub(b2, b5);               /* x5 = b2 - b5; */         \
    x3 = vec_add(b3, b4);               /* x3 = b3 + b4; */         \
    x4 = vec_sub(b3, b4);               /* x4 = b3 - b4; */         \
                                                                    \
    b7 = vec_add(x0, x3);               /* b7 = x0 + x3; */         \
    b1 = vec_add(x1, x2);               /* b1 = x1 + x2; */         \
    b0 = vec_add(b7, b1);               /* b0 = b7 + b1; */         \
    b4 = vec_sub(b7, b1);               /* b4 = b7 - b1; */         \
                                                                    \
    b2 = vec_sub(x0, x3);               /* b2 = x0 - x3; */         \
    b6 = vec_sub(x1, x2);               /* b6 = x1 - x2; */         \
    b5 = vec_add(b6, b2);               /* b5 = b6 + b2; */         \
    cnst = LD_W2;                                                   \
    b5 = vec_madd(cnst, b5, mzero);     /* b5 = b5 * W2; */         \
    cnst = LD_W1;                                                   \
    b2 = vec_madd(cnst, b2, b5);        /* b2 = b5 + b2 * W1; */    \
    cnst = LD_W0;                                                   \
    b6 = vec_madd(cnst, b6, b5);        /* b6 = b5 + b6 * W0; */    \
                                                                    \
    x0 = vec_add(x4, x7);               /* x0 = x4 + x7; */         \
    x1 = vec_add(x5, x6);               /* x1 = x5 + x6; */         \
    x2 = vec_add(x4, x6);               /* x2 = x4 + x6; */         \
    x3 = vec_add(x5, x7);               /* x3 = x5 + x7; */         \
    x8 = vec_add(x2, x3);               /* x8 = x2 + x3; */         \
    cnst = LD_W3;                                                   \
    x8 = vec_madd(cnst, x8, mzero);     /* x8 = x8 * W3; */         \
                                                                    \
    cnst = LD_W8;                                                   \
    x0 = vec_madd(cnst, x0, mzero);     /* x0 *= W8; */             \
    cnst = LD_W9;                                                   \
    x1 = vec_madd(cnst, x1, mzero);     /* x1 *= W9; */             \
    cnst = LD_WA;                                                   \
    x2 = vec_madd(cnst, x2, x8);        /* x2 = x2 * WA + x8; */    \
    cnst = LD_WB;                                                   \
    x3 = vec_madd(cnst, x3, x8);        /* x3 = x3 * WB + x8; */    \
                                                                    \
    cnst = LD_W4;                                                   \
    b7 = vec_madd(cnst, x4, x0);        /* b7 = x4 * W4 + x0; */    \
    cnst = LD_W5;                                                   \
    b5 = vec_madd(cnst, x5, x1);        /* b5 = x5 * W5 + x1; */    \
    cnst = LD_W6;                                                   \
    b3 = vec_madd(cnst, x6, x1);        /* b3 = x6 * W6 + x1; */    \
    cnst = LD_W7;                                                   \
    b1 = vec_madd(cnst, x7, x0);        /* b1 = x7 * W7 + x0; */    \
                                                                    \
    b7 = vec_add(b7, x2);               /* b7 = b7 + x2; */         \
    b5 = vec_add(b5, x3);               /* b5 = b5 + x3; */         \
    b3 = vec_add(b3, x2);               /* b3 = b3 + x2; */         \
    b1 = vec_add(b1, x3);               /* b1 = b1 + x3; */         \
    /* }}} */

#define FDCTCOL(b0,b1,b2,b3,b4,b5,b6,b7) /* {{{ */                  \
    x0 = vec_add(b0, b7);               /* x0 = b0 + b7; */         \
    x7 = vec_sub(b0, b7);               /* x7 = b0 - b7; */         \
    x1 = vec_add(b1, b6);               /* x1 = b1 + b6; */         \
    x6 = vec_sub(b1, b6);               /* x6 = b1 - b6; */         \
    x2 = vec_add(b2, b5);               /* x2 = b2 + b5; */         \
    x5 = vec_sub(b2, b5);               /* x5 = b2 - b5; */         \
    x3 = vec_add(b3, b4);               /* x3 = b3 + b4; */         \
    x4 = vec_sub(b3, b4);               /* x4 = b3 - b4; */         \
                                                                    \
    b7 = vec_add(x0, x3);               /* b7 = x0 + x3; */         \
    b1 = vec_add(x1, x2);               /* b1 = x1 + x2; */         \
    b0 = vec_add(b7, b1);               /* b0 = b7 + b1; */         \
    b4 = vec_sub(b7, b1);               /* b4 = b7 - b1; */         \
                                                                    \
    b2 = vec_sub(x0, x3);               /* b2 = x0 - x3; */         \
    b6 = vec_sub(x1, x2);               /* b6 = x1 - x2; */         \
    b5 = vec_add(b6, b2);               /* b5 = b6 + b2; */         \
    cnst = LD_W2;                                                   \
    b5 = vec_madd(cnst, b5, mzero);     /* b5 = b5 * W2; */         \
    cnst = LD_W1;                                                   \
    b2 = vec_madd(cnst, b2, b5);        /* b2 = b5 + b2 * W1; */    \
    cnst = LD_W0;                                                   \
    b6 = vec_madd(cnst, b6, b5);        /* b6 = b5 + b6 * W0; */    \
                                                                    \
    x0 = vec_add(x4, x7);               /* x0 = x4 + x7; */         \
    x1 = vec_add(x5, x6);               /* x1 = x5 + x6; */         \
    x2 = vec_add(x4, x6);               /* x2 = x4 + x6; */         \
    x3 = vec_add(x5, x7);               /* x3 = x5 + x7; */         \
    x8 = vec_add(x2, x3);               /* x8 = x2 + x3; */         \
    cnst = LD_W3;                                                   \
    x8 = vec_madd(cnst, x8, mzero);     /* x8 = x8 * W3; */         \
                                                                    \
    cnst = LD_W8;                                                   \
    x0 = vec_madd(cnst, x0, mzero);     /* x0 *= W8; */             \
    cnst = LD_W9;                                                   \
    x1 = vec_madd(cnst, x1, mzero);     /* x1 *= W9; */             \
    cnst = LD_WA;                                                   \
    x2 = vec_madd(cnst, x2, x8);        /* x2 = x2 * WA + x8; */    \
    cnst = LD_WB;                                                   \
    x3 = vec_madd(cnst, x3, x8);        /* x3 = x3 * WB + x8; */    \
                                                                    \
    cnst = LD_W4;                                                   \
    b7 = vec_madd(cnst, x4, x0);        /* b7 = x4 * W4 + x0; */    \
    cnst = LD_W5;                                                   \
    b5 = vec_madd(cnst, x5, x1);        /* b5 = x5 * W5 + x1; */    \
    cnst = LD_W6;                                                   \
    b3 = vec_madd(cnst, x6, x1);        /* b3 = x6 * W6 + x1; */    \
    cnst = LD_W7;                                                   \
    b1 = vec_madd(cnst, x7, x0);        /* b1 = x7 * W7 + x0; */    \
                                                                    \
    b7 = vec_add(b7, x2);               /* b7 += x2; */             \
    b5 = vec_add(b5, x3);               /* b5 += x3; */             \
    b3 = vec_add(b3, x2);               /* b3 += x2; */             \
    b1 = vec_add(b1, x3);               /* b1 += x3; */             \
    /* }}} */



/* two dimensional discrete cosine transform */

void fdct_altivec(int16_t *block)
{
POWERPC_PERF_DECLARE(altivec_fdct, 1);
#ifdef ALTIVEC_USE_REFERENCE_C_CODE
POWERPC_PERF_START_COUNT(altivec_fdct, 1);
    void ff_jpeg_fdct_islow(int16_t *block);
    ff_jpeg_fdct_islow(block);
POWERPC_PERF_STOP_COUNT(altivec_fdct, 1);
#else /* ALTIVEC_USE_REFERENCE_C_CODE */
    vector signed short *bp;
    vector float *cp;
    vector float b00, b10, b20, b30, b40, b50, b60, b70;
    vector float b01, b11, b21, b31, b41, b51, b61, b71;
    vector float mzero, cnst, cnsts0, cnsts1, cnsts2;
    vector float x0, x1, x2, x3, x4, x5, x6, x7, x8;

    POWERPC_PERF_START_COUNT(altivec_fdct, 1);


    /* setup constants {{{ */
    /* mzero = -0.0 */
    vu32(mzero) = vec_splat_u32(-1);
    vu32(mzero) = vec_sl(vu32(mzero), vu32(mzero));
    cp = fdctconsts;
    cnsts0 = vec_ld(0, cp); cp++;
    cnsts1 = vec_ld(0, cp); cp++;
    cnsts2 = vec_ld(0, cp);
    /* }}} */


    /* 8x8 matrix transpose (vector short[8]) {{{ */
#define MERGE_S16(hl,a,b) vec_merge##hl(vs16(a), vs16(b))

    bp = (vector signed short*)block;
    vs16(b00) = vec_ld(0,    bp);
    vs16(b40) = vec_ld(16*4, bp);
    vs16(b01) = MERGE_S16(h, b00, b40);
    vs16(b11) = MERGE_S16(l, b00, b40);
    bp++;
    vs16(b10) = vec_ld(0,    bp);
    vs16(b50) = vec_ld(16*4, bp);
    vs16(b21) = MERGE_S16(h, b10, b50);
    vs16(b31) = MERGE_S16(l, b10, b50);
    bp++;
    vs16(b20) = vec_ld(0,    bp);
    vs16(b60) = vec_ld(16*4, bp);
    vs16(b41) = MERGE_S16(h, b20, b60);
    vs16(b51) = MERGE_S16(l, b20, b60);
    bp++;
    vs16(b30) = vec_ld(0,    bp);
    vs16(b70) = vec_ld(16*4, bp);
    vs16(b61) = MERGE_S16(h, b30, b70);
    vs16(b71) = MERGE_S16(l, b30, b70);

    vs16(x0) = MERGE_S16(h, b01, b41);