X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=common%2Fdct.c;h=1b525474aea1a81525b5fed33e53fe47b5bea056;hb=69e69197c424bff9e4b90eb5d608f15b59ca77b4;hp=d23a80a031ea971b6ce42fe335819897d16015f6;hpb=271c1947a599ccdc3a509260da2d5cd6699148d7;p=x264 diff --git a/common/dct.c b/common/dct.c index d23a80a0..1b525474 100644 --- a/common/dct.c +++ b/common/dct.c @@ -1,10 +1,10 @@ /***************************************************************************** * dct.c: h264 encoder library ***************************************************************************** - * Copyright (C) 2003 Laurent Aimar - * $Id: dct.c,v 1.1 2004/06/03 19:27:06 fenrir Exp $ + * Copyright (C) 2003-2008 x264 project * - * Authors: Laurent Aimar + * Authors: Loren Merritt + * Laurent Aimar * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -18,45 +18,24 @@ * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA. + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA. *****************************************************************************/ #include "common.h" -#ifdef HAVE_MMXEXT -# include "i386/dct.h" +#ifdef HAVE_MMX +# include "x86/dct.h" #endif #ifdef ARCH_PPC # include "ppc/dct.h" #endif - -static inline int clip_uint8( int a ) -{ - if (a&(~255)) - return (-a)>>31; - else - return a; -} +int x264_dct4_weight2_zigzag[2][16]; +int x264_dct8_weight2_zigzag[2][64]; /* * XXX For all dct dc : input could be equal to output so ... */ -static void dct2x2dc( int16_t d[2][2] ) -{ - int tmp[2][2]; - - tmp[0][0] = d[0][0] + d[0][1]; - tmp[1][0] = d[0][0] - d[0][1]; - tmp[0][1] = d[1][0] + d[1][1]; - tmp[1][1] = d[1][0] - d[1][1]; - - d[0][0] = tmp[0][0] + tmp[0][1]; - d[0][1] = tmp[1][0] + tmp[1][1]; - d[1][0] = tmp[0][0] - tmp[0][1]; - d[1][1] = tmp[1][0] - tmp[1][1]; -} - static void dct4x4dc( int16_t d[4][4] ) { int16_t tmp[4][4]; @@ -84,10 +63,10 @@ static void dct4x4dc( int16_t d[4][4] ) s23 = tmp[i][2] + tmp[i][3]; d23 = tmp[i][2] - tmp[i][3]; - d[0][i] = ( s01 + s23 + 1 ) >> 1; - d[1][i] = ( s01 - s23 + 1 ) >> 1; - d[2][i] = ( d01 - d23 + 1 ) >> 1; - d[3][i] = ( d01 + d23 + 1 ) >> 1; + d[i][0] = ( s01 + s23 + 1 ) >> 1; + d[i][1] = ( s01 - s23 + 1 ) >> 1; + d[i][2] = ( d01 - d23 + 1 ) >> 1; + d[i][3] = ( d01 + d23 + 1 ) >> 1; } } @@ -100,10 +79,10 @@ static void idct4x4dc( int16_t d[4][4] ) for( i = 0; i < 4; i++ ) { - s01 = d[0][i] + d[1][i]; - d01 = d[0][i] - d[1][i]; - s23 = d[2][i] + d[3][i]; - d23 = d[2][i] - d[3][i]; + s01 = d[i][0] + d[i][1]; + d01 = d[i][0] - d[i][1]; + s23 = d[i][2] + d[i][3]; + d23 = d[i][2] - d[i][3]; tmp[0][i] = s01 + s23; tmp[1][i] = s01 - s23; @@ -140,13 +119,13 @@ static inline void pixel_sub_wxh( int16_t *diff, int i_size, } } -static void sub4x4_dct( int16_t dct[4][4], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 ) +static void sub4x4_dct( int16_t dct[4][4], uint8_t *pix1, uint8_t *pix2 ) { int16_t d[4][4]; int16_t tmp[4][4]; int i; - pixel_sub_wxh( (int16_t*)d, 4, pix1, i_pix1, pix2, i_pix2 ); + pixel_sub_wxh( (int16_t*)d, 4, pix1, FENC_STRIDE, pix2, FDEC_STRIDE ); for( i = 0; i < 4; i++ ) { @@ -168,31 +147,31 @@ static void sub4x4_dct( int16_t dct[4][4], uint8_t *pix1, int i_pix1, uint8_t *p const int d03 = tmp[i][0] - tmp[i][3]; const int d12 = tmp[i][1] - tmp[i][2]; - dct[0][i] = s03 + s12; - dct[1][i] = 2*d03 + d12; - dct[2][i] = s03 - s12; - dct[3][i] = d03 - 2*d12; + dct[i][0] = s03 + s12; + dct[i][1] = 2*d03 + d12; + dct[i][2] = s03 - s12; + dct[i][3] = d03 - 2*d12; } } -static void sub8x8_dct( int16_t dct[4][4][4], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 ) +static void sub8x8_dct( int16_t dct[4][4][4], uint8_t *pix1, uint8_t *pix2 ) { - sub4x4_dct( dct[0], &pix1[0], i_pix1, &pix2[0], i_pix2 ); - sub4x4_dct( dct[1], &pix1[4], i_pix1, &pix2[4], i_pix2 ); - sub4x4_dct( dct[2], &pix1[4*i_pix1+0], i_pix1, &pix2[4*i_pix2+0], i_pix2 ); - sub4x4_dct( dct[3], &pix1[4*i_pix1+4], i_pix1, &pix2[4*i_pix2+4], i_pix2 ); + sub4x4_dct( dct[0], &pix1[0], &pix2[0] ); + sub4x4_dct( dct[1], &pix1[4], &pix2[4] ); + sub4x4_dct( dct[2], &pix1[4*FENC_STRIDE+0], &pix2[4*FDEC_STRIDE+0] ); + sub4x4_dct( dct[3], &pix1[4*FENC_STRIDE+4], &pix2[4*FDEC_STRIDE+4] ); } -static void sub16x16_dct( int16_t dct[16][4][4], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 ) +static void sub16x16_dct( int16_t dct[16][4][4], uint8_t *pix1, uint8_t *pix2 ) { - sub8x8_dct( &dct[ 0], pix1, i_pix1, pix2, i_pix2 ); - sub8x8_dct( &dct[ 4], &pix1[8], i_pix1, &pix2[8], i_pix2 ); - sub8x8_dct( &dct[ 8], &pix1[8*i_pix1], i_pix1, &pix2[8*i_pix2], i_pix2 ); - sub8x8_dct( &dct[12], &pix1[8*i_pix1+8], i_pix1, &pix2[8*i_pix2+8], i_pix2 ); + sub8x8_dct( &dct[ 0], &pix1[0], &pix2[0] ); + sub8x8_dct( &dct[ 4], &pix1[8], &pix2[8] ); + sub8x8_dct( &dct[ 8], &pix1[8*FENC_STRIDE+0], &pix2[8*FDEC_STRIDE+0] ); + sub8x8_dct( &dct[12], &pix1[8*FENC_STRIDE+8], &pix2[8*FDEC_STRIDE+8] ); } -static void add4x4_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4] ) +static void add4x4_idct( uint8_t *p_dst, int16_t dct[4][4] ) { int16_t d[4][4]; int16_t tmp[4][4]; @@ -201,10 +180,10 @@ static void add4x4_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4] ) for( i = 0; i < 4; i++ ) { - const int s02 = dct[i][0] + dct[i][2]; - const int d02 = dct[i][0] - dct[i][2]; - const int s13 = dct[i][1] + (dct[i][3]>>1); - const int d13 = (dct[i][1]>>1) - dct[i][3]; + const int s02 = dct[0][i] + dct[2][i]; + const int d02 = dct[0][i] - dct[2][i]; + const int s13 = dct[1][i] + (dct[3][i]>>1); + const int d13 = (dct[1][i]>>1) - dct[3][i]; tmp[i][0] = s02 + s13; tmp[i][1] = d02 + d13; @@ -217,7 +196,7 @@ static void add4x4_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4] ) const int s02 = tmp[0][i] + tmp[2][i]; const int d02 = tmp[0][i] - tmp[2][i]; const int s13 = tmp[1][i] + (tmp[3][i]>>1); - const int d13 = (tmp[1][i]>>1) - tmp[3][i]; + const int d13 = (tmp[1][i]>>1) - tmp[3][i]; d[0][i] = ( s02 + s13 + 32 ) >> 6; d[1][i] = ( d02 + d13 + 32 ) >> 6; @@ -230,26 +209,26 @@ static void add4x4_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4] ) { for( x = 0; x < 4; x++ ) { - p_dst[x] = clip_uint8( p_dst[x] + d[y][x] ); + p_dst[x] = x264_clip_uint8( p_dst[x] + d[y][x] ); } - p_dst += i_dst; + p_dst += FDEC_STRIDE; } } -static void add8x8_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4][4] ) +static void add8x8_idct( uint8_t *p_dst, int16_t dct[4][4][4] ) { - add4x4_idct( p_dst, i_dst, dct[0] ); - add4x4_idct( &p_dst[4], i_dst, dct[1] ); - add4x4_idct( &p_dst[4*i_dst+0], i_dst, dct[2] ); - add4x4_idct( &p_dst[4*i_dst+4], i_dst, dct[3] ); + add4x4_idct( &p_dst[0], dct[0] ); + add4x4_idct( &p_dst[4], dct[1] ); + add4x4_idct( &p_dst[4*FDEC_STRIDE+0], dct[2] ); + add4x4_idct( &p_dst[4*FDEC_STRIDE+4], dct[3] ); } -static void add16x16_idct( uint8_t *p_dst, int i_dst, int16_t dct[16][4][4] ) +static void add16x16_idct( uint8_t *p_dst, int16_t dct[16][4][4] ) { - add8x8_idct( &p_dst[0], i_dst, &dct[0] ); - add8x8_idct( &p_dst[8], i_dst, &dct[4] ); - add8x8_idct( &p_dst[8*i_dst], i_dst, &dct[8] ); - add8x8_idct( &p_dst[8*i_dst+8], i_dst, &dct[12] ); + add8x8_idct( &p_dst[0], &dct[0] ); + add8x8_idct( &p_dst[8], &dct[4] ); + add8x8_idct( &p_dst[8*FDEC_STRIDE+0], &dct[8] ); + add8x8_idct( &p_dst[8*FDEC_STRIDE+8], &dct[12] ); } /**************************************************************************** @@ -273,39 +252,44 @@ static void add16x16_idct( uint8_t *p_dst, int i_dst, int16_t dct[16][4][4] ) const int a5 = d07 - d34 - (d25 + (d25>>1));\ const int a6 = d07 + d34 - (d16 + (d16>>1));\ const int a7 = d16 - d25 + (d34 + (d34>>1));\ - SRC(0) = a0 + a1 ;\ - SRC(1) = a4 + (a7>>2);\ - SRC(2) = a2 + (a3>>1);\ - SRC(3) = a5 + (a6>>2);\ - SRC(4) = a0 - a1 ;\ - SRC(5) = a6 - (a5>>2);\ - SRC(6) = (a2>>1) - a3 ;\ - SRC(7) = (a4>>2) - a7 ;\ + DST(0) = a0 + a1 ;\ + DST(1) = a4 + (a7>>2);\ + DST(2) = a2 + (a3>>1);\ + DST(3) = a5 + (a6>>2);\ + DST(4) = a0 - a1 ;\ + DST(5) = a6 - (a5>>2);\ + DST(6) = (a2>>1) - a3 ;\ + DST(7) = (a4>>2) - a7 ;\ } -static void sub8x8_dct8( int16_t dct[8][8], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 ) +static void sub8x8_dct8( int16_t dct[8][8], uint8_t *pix1, uint8_t *pix2 ) { int i; + int16_t tmp[8][8]; - pixel_sub_wxh( (int16_t*)dct, 8, pix1, i_pix1, pix2, i_pix2 ); + pixel_sub_wxh( (int16_t*)tmp, 8, pix1, FENC_STRIDE, pix2, FDEC_STRIDE ); -#define SRC(x) dct[i][x] +#define SRC(x) tmp[x][i] +#define DST(x) tmp[x][i] for( i = 0; i < 8; i++ ) DCT8_1D #undef SRC +#undef DST -#define SRC(x) dct[x][i] +#define SRC(x) tmp[i][x] +#define DST(x) dct[x][i] for( i = 0; i < 8; i++ ) DCT8_1D #undef SRC +#undef DST } -static void sub16x16_dct8( int16_t dct[4][8][8], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 ) +static void sub16x16_dct8( int16_t dct[4][8][8], uint8_t *pix1, uint8_t *pix2 ) { - sub8x8_dct8( dct[0], pix1, i_pix1, pix2, i_pix2 ); - sub8x8_dct8( dct[1], &pix1[8], i_pix1, &pix2[8], i_pix2 ); - sub8x8_dct8( dct[2], &pix1[8*i_pix1], i_pix1, &pix2[8*i_pix2], i_pix2 ); - sub8x8_dct8( dct[3], &pix1[8*i_pix1+8], i_pix1, &pix2[8*i_pix2+8], i_pix2 ); + sub8x8_dct8( dct[0], &pix1[0], &pix2[0] ); + sub8x8_dct8( dct[1], &pix1[8], &pix2[8] ); + sub8x8_dct8( dct[2], &pix1[8*FENC_STRIDE+0], &pix2[8*FDEC_STRIDE+0] ); + sub8x8_dct8( dct[3], &pix1[8*FENC_STRIDE+8], &pix2[8*FDEC_STRIDE+8] ); } #define IDCT8_1D {\ @@ -335,33 +319,33 @@ static void sub16x16_dct8( int16_t dct[4][8][8], uint8_t *pix1, int i_pix1, uint DST(7, b0 - b7);\ } -static void add8x8_idct8( uint8_t *dst, int i_dst, int16_t dct[8][8] ) +static void add8x8_idct8( uint8_t *dst, int16_t dct[8][8] ) { int i; dct[0][0] += 32; // rounding for the >>6 at the end -#define SRC(x) dct[i][x] -#define DST(x,rhs) dct[i][x] = (rhs) +#define SRC(x) dct[x][i] +#define DST(x,rhs) dct[x][i] = (rhs) for( i = 0; i < 8; i++ ) IDCT8_1D #undef SRC #undef DST -#define SRC(x) dct[x][i] -#define DST(x,rhs) dst[i + x*i_dst] = clip_uint8( dst[i + x*i_dst] + ((rhs) >> 6) ); +#define SRC(x) dct[i][x] +#define DST(x,rhs) dst[i + x*FDEC_STRIDE] = x264_clip_uint8( dst[i + x*FDEC_STRIDE] + ((rhs) >> 6) ); for( i = 0; i < 8; i++ ) IDCT8_1D #undef SRC #undef DST } -static void add16x16_idct8( uint8_t *dst, int i_dst, int16_t dct[4][8][8] ) +static void add16x16_idct8( uint8_t *dst, int16_t dct[4][8][8] ) { - add8x8_idct8( &dst[0], i_dst, dct[0] ); - add8x8_idct8( &dst[8], i_dst, dct[1] ); - add8x8_idct8( &dst[8*i_dst], i_dst, dct[2] ); - add8x8_idct8( &dst[8*i_dst+8], i_dst, dct[3] ); + add8x8_idct8( &dst[0], dct[0] ); + add8x8_idct8( &dst[8], dct[1] ); + add8x8_idct8( &dst[8*FDEC_STRIDE+0], dct[2] ); + add8x8_idct8( &dst[8*FDEC_STRIDE+8], dct[3] ); } @@ -388,43 +372,40 @@ void x264_dct_init( int cpu, x264_dct_function_t *dctf ) dctf->dct4x4dc = dct4x4dc; dctf->idct4x4dc = idct4x4dc; - dctf->dct2x2dc = dct2x2dc; - dctf->idct2x2dc = dct2x2dc; - -#ifdef HAVE_MMXEXT - if( cpu&X264_CPU_MMXEXT ) +#ifdef HAVE_MMX + if( cpu&X264_CPU_MMX ) { - dctf->sub4x4_dct = x264_sub4x4_dct_mmxext; - dctf->sub8x8_dct = x264_sub8x8_dct_mmxext; - dctf->sub16x16_dct = x264_sub16x16_dct_mmxext; - - dctf->add4x4_idct = x264_add4x4_idct_mmxext; - dctf->add8x8_idct = x264_add8x8_idct_mmxext; - dctf->add16x16_idct = x264_add16x16_idct_mmxext; - - dctf->dct4x4dc = x264_dct4x4dc_mmxext; - dctf->idct4x4dc = x264_idct4x4dc_mmxext; + dctf->sub4x4_dct = x264_sub4x4_dct_mmx; + dctf->add4x4_idct = x264_add4x4_idct_mmx; + dctf->dct4x4dc = x264_dct4x4dc_mmx; + dctf->idct4x4dc = x264_idct4x4dc_mmx; #ifndef ARCH_X86_64 - dctf->sub8x8_dct8 = x264_sub8x8_dct8_mmxext; - dctf->sub16x16_dct8 = x264_sub16x16_dct8_mmxext; - - dctf->add8x8_idct8 = x264_add8x8_idct8_mmxext; - dctf->add16x16_idct8= x264_add16x16_idct8_mmxext; + dctf->sub8x8_dct = x264_sub8x8_dct_mmx; + dctf->sub16x16_dct = x264_sub16x16_dct_mmx; + dctf->add8x8_idct = x264_add8x8_idct_mmx; + dctf->add16x16_idct = x264_add16x16_idct_mmx; + + dctf->sub8x8_dct8 = x264_sub8x8_dct8_mmx; + dctf->sub16x16_dct8 = x264_sub16x16_dct8_mmx; + dctf->add8x8_idct8 = x264_add8x8_idct8_mmx; + dctf->add16x16_idct8= x264_add16x16_idct8_mmx; #endif } -#endif -#if defined(HAVE_SSE2) && defined(ARCH_X86_64) if( cpu&X264_CPU_SSE2 ) { dctf->sub8x8_dct8 = x264_sub8x8_dct8_sse2; dctf->sub16x16_dct8 = x264_sub16x16_dct8_sse2; - dctf->add8x8_idct8 = x264_add8x8_idct8_sse2; dctf->add16x16_idct8= x264_add16x16_idct8_sse2; + + dctf->sub8x8_dct = x264_sub8x8_dct_sse2; + dctf->sub16x16_dct = x264_sub16x16_dct_sse2; + dctf->add8x8_idct = x264_add8x8_idct_sse2; + dctf->add16x16_idct = x264_add16x16_idct_sse2; } -#endif +#endif //HAVE_MMX #ifdef ARCH_PPC if( cpu&X264_CPU_ALTIVEC ) @@ -432,7 +413,214 @@ void x264_dct_init( int cpu, x264_dct_function_t *dctf ) dctf->sub4x4_dct = x264_sub4x4_dct_altivec; dctf->sub8x8_dct = x264_sub8x8_dct_altivec; dctf->sub16x16_dct = x264_sub16x16_dct_altivec; + + dctf->add4x4_idct = x264_add4x4_idct_altivec; + dctf->add8x8_idct = x264_add8x8_idct_altivec; + dctf->add16x16_idct = x264_add16x16_idct_altivec; + + dctf->sub8x8_dct8 = x264_sub8x8_dct8_altivec; + dctf->sub16x16_dct8 = x264_sub16x16_dct8_altivec; + + dctf->add8x8_idct8 = x264_add8x8_idct8_altivec; + dctf->add16x16_idct8= x264_add16x16_idct8_altivec; } #endif } +void x264_dct_init_weights( void ) +{ + int i, j; + for( j=0; j<2; j++ ) + { + for( i=0; i<16; i++ ) + x264_dct4_weight2_zigzag[j][i] = x264_dct4_weight2_tab[ x264_zigzag_scan4[j][i] ]; + for( i=0; i<64; i++ ) + x264_dct8_weight2_zigzag[j][i] = x264_dct8_weight2_tab[ x264_zigzag_scan8[j][i] ]; + } +} + + +// gcc pessimizes multi-dimensional arrays here, even with constant indices +#define ZIG(i,y,x) level[i] = dct[0][x*8+y]; +#define ZIGZAG8_FRAME\ + ZIG( 0,0,0) ZIG( 1,0,1) ZIG( 2,1,0) ZIG( 3,2,0)\ + ZIG( 4,1,1) ZIG( 5,0,2) ZIG( 6,0,3) ZIG( 7,1,2)\ + ZIG( 8,2,1) ZIG( 9,3,0) ZIG(10,4,0) ZIG(11,3,1)\ + ZIG(12,2,2) ZIG(13,1,3) ZIG(14,0,4) ZIG(15,0,5)\ + ZIG(16,1,4) ZIG(17,2,3) ZIG(18,3,2) ZIG(19,4,1)\ + ZIG(20,5,0) ZIG(21,6,0) ZIG(22,5,1) ZIG(23,4,2)\ + ZIG(24,3,3) ZIG(25,2,4) ZIG(26,1,5) ZIG(27,0,6)\ + ZIG(28,0,7) ZIG(29,1,6) ZIG(30,2,5) ZIG(31,3,4)\ + ZIG(32,4,3) ZIG(33,5,2) ZIG(34,6,1) ZIG(35,7,0)\ + ZIG(36,7,1) ZIG(37,6,2) ZIG(38,5,3) ZIG(39,4,4)\ + ZIG(40,3,5) ZIG(41,2,6) ZIG(42,1,7) ZIG(43,2,7)\ + ZIG(44,3,6) ZIG(45,4,5) ZIG(46,5,4) ZIG(47,6,3)\ + ZIG(48,7,2) ZIG(49,7,3) ZIG(50,6,4) ZIG(51,5,5)\ + ZIG(52,4,6) ZIG(53,3,7) ZIG(54,4,7) ZIG(55,5,6)\ + ZIG(56,6,5) ZIG(57,7,4) ZIG(58,7,5) ZIG(59,6,6)\ + ZIG(60,5,7) ZIG(61,6,7) ZIG(62,7,6) ZIG(63,7,7)\ + +#define ZIGZAG8_FIELD\ + ZIG( 0,0,0) ZIG( 1,1,0) ZIG( 2,2,0) ZIG( 3,0,1)\ + ZIG( 4,1,1) ZIG( 5,3,0) ZIG( 6,4,0) ZIG( 7,2,1)\ + ZIG( 8,0,2) ZIG( 9,3,1) ZIG(10,5,0) ZIG(11,6,0)\ + ZIG(12,7,0) ZIG(13,4,1) ZIG(14,1,2) ZIG(15,0,3)\ + ZIG(16,2,2) ZIG(17,5,1) ZIG(18,6,1) ZIG(19,7,1)\ + ZIG(20,3,2) ZIG(21,1,3) ZIG(22,0,4) ZIG(23,2,3)\ + ZIG(24,4,2) ZIG(25,5,2) ZIG(26,6,2) ZIG(27,7,2)\ + ZIG(28,3,3) ZIG(29,1,4) ZIG(30,0,5) ZIG(31,2,4)\ + ZIG(32,4,3) ZIG(33,5,3) ZIG(34,6,3) ZIG(35,7,3)\ + ZIG(36,3,4) ZIG(37,1,5) ZIG(38,0,6) ZIG(39,2,5)\ + ZIG(40,4,4) ZIG(41,5,4) ZIG(42,6,4) ZIG(43,7,4)\ + ZIG(44,3,5) ZIG(45,1,6) ZIG(46,2,6) ZIG(47,4,5)\ + ZIG(48,5,5) ZIG(49,6,5) ZIG(50,7,5) ZIG(51,3,6)\ + ZIG(52,0,7) ZIG(53,1,7) ZIG(54,4,6) ZIG(55,5,6)\ + ZIG(56,6,6) ZIG(57,7,6) ZIG(58,2,7) ZIG(59,3,7)\ + ZIG(60,4,7) ZIG(61,5,7) ZIG(62,6,7) ZIG(63,7,7) + +#define ZIGZAG4_FRAME\ + ZIG( 0,0,0) ZIG( 1,0,1) ZIG( 2,1,0) ZIG( 3,2,0)\ + ZIG( 4,1,1) ZIG( 5,0,2) ZIG( 6,0,3) ZIG( 7,1,2)\ + ZIG( 8,2,1) ZIG( 9,3,0) ZIG(10,3,1) ZIG(11,2,2)\ + ZIG(12,1,3) ZIG(13,2,3) ZIG(14,3,2) ZIG(15,3,3) + +#define ZIGZAG4_FIELD\ + ZIG( 0,0,0) ZIG( 1,1,0) ZIG( 2,0,1) ZIG( 3,2,0)\ + ZIG( 4,3,0) ZIG( 5,1,1) ZIG( 6,2,1) ZIG( 7,3,1)\ + ZIG( 8,0,2) ZIG( 9,1,2) ZIG(10,2,2) ZIG(11,3,2)\ + ZIG(12,0,3) ZIG(13,1,3) ZIG(14,2,3) ZIG(15,3,3) + +static void zigzag_scan_8x8_frame( int16_t level[64], int16_t dct[8][8] ) +{ + ZIGZAG8_FRAME +} + +static void zigzag_scan_8x8_field( int16_t level[64], int16_t dct[8][8] ) +{ + ZIGZAG8_FIELD +} + +#undef ZIG +#define ZIG(i,y,x) level[i] = dct[0][x*4+y]; + +static void zigzag_scan_4x4_frame( int16_t level[16], int16_t dct[4][4] ) +{ + ZIGZAG4_FRAME +} + +static void zigzag_scan_4x4_field( int16_t level[16], int16_t dct[4][4] ) +{ + *(uint32_t*)level = *(uint32_t*)dct; + ZIG(2,0,1) ZIG(3,2,0) ZIG(4,3,0) ZIG(5,1,1) + *(uint32_t*)(level+6) = *(uint32_t*)(*dct+6); + *(uint64_t*)(level+8) = *(uint64_t*)(*dct+8); + *(uint64_t*)(level+12) = *(uint64_t*)(*dct+12); +} + +#undef ZIG +#define ZIG(i,y,x) {\ + int oe = x+y*FENC_STRIDE;\ + int od = x+y*FDEC_STRIDE;\ + level[i] = p_src[oe] - p_dst[od];\ +} +#define COPY4x4\ + *(uint32_t*)(p_dst+0*FDEC_STRIDE) = *(uint32_t*)(p_src+0*FENC_STRIDE);\ + *(uint32_t*)(p_dst+1*FDEC_STRIDE) = *(uint32_t*)(p_src+1*FENC_STRIDE);\ + *(uint32_t*)(p_dst+2*FDEC_STRIDE) = *(uint32_t*)(p_src+2*FENC_STRIDE);\ + *(uint32_t*)(p_dst+3*FDEC_STRIDE) = *(uint32_t*)(p_src+3*FENC_STRIDE); +#define COPY8x8\ + *(uint64_t*)(p_dst+0*FDEC_STRIDE) = *(uint64_t*)(p_src+0*FENC_STRIDE);\ + *(uint64_t*)(p_dst+1*FDEC_STRIDE) = *(uint64_t*)(p_src+1*FENC_STRIDE);\ + *(uint64_t*)(p_dst+2*FDEC_STRIDE) = *(uint64_t*)(p_src+2*FENC_STRIDE);\ + *(uint64_t*)(p_dst+3*FDEC_STRIDE) = *(uint64_t*)(p_src+3*FENC_STRIDE);\ + *(uint64_t*)(p_dst+4*FDEC_STRIDE) = *(uint64_t*)(p_src+4*FENC_STRIDE);\ + *(uint64_t*)(p_dst+5*FDEC_STRIDE) = *(uint64_t*)(p_src+5*FENC_STRIDE);\ + *(uint64_t*)(p_dst+6*FDEC_STRIDE) = *(uint64_t*)(p_src+6*FENC_STRIDE);\ + *(uint64_t*)(p_dst+7*FDEC_STRIDE) = *(uint64_t*)(p_src+7*FENC_STRIDE); + +static void zigzag_sub_4x4_frame( int16_t level[16], const uint8_t *p_src, uint8_t *p_dst ) +{ + ZIGZAG4_FRAME + COPY4x4 +} + +static void zigzag_sub_4x4_field( int16_t level[16], const uint8_t *p_src, uint8_t *p_dst ) +{ + ZIGZAG4_FIELD + COPY4x4 +} + +static void zigzag_sub_8x8_frame( int16_t level[64], const uint8_t *p_src, uint8_t *p_dst ) +{ + ZIGZAG8_FRAME + COPY8x8 +} +static void zigzag_sub_8x8_field( int16_t level[64], const uint8_t *p_src, uint8_t *p_dst ) +{ + ZIGZAG8_FIELD + COPY8x8 +} + +#undef ZIG +#undef COPY4x4 + +static void zigzag_interleave_8x8_cavlc( int16_t *dst, int16_t *src ) +{ + int i,j; + for( i=0; i<4; i++ ) + for( j=0; j<16; j++ ) + dst[i*16+j] = src[i+j*4]; +} + +void x264_zigzag_init( int cpu, x264_zigzag_function_t *pf, int b_interlaced ) +{ + if( b_interlaced ) + { + pf->scan_8x8 = zigzag_scan_8x8_field; + pf->scan_4x4 = zigzag_scan_4x4_field; + pf->sub_8x8 = zigzag_sub_8x8_field; + pf->sub_4x4 = zigzag_sub_4x4_field; +#ifdef HAVE_MMX + if( cpu&X264_CPU_MMXEXT ) + pf->scan_4x4 = x264_zigzag_scan_4x4_field_mmxext; +#endif + +#ifdef ARCH_PPC + if( cpu&X264_CPU_ALTIVEC ) + pf->scan_4x4 = x264_zigzag_scan_4x4_field_altivec; +#endif + } + else + { + pf->scan_8x8 = zigzag_scan_8x8_frame; + pf->scan_4x4 = zigzag_scan_4x4_frame; + pf->sub_8x8 = zigzag_sub_8x8_frame; + pf->sub_4x4 = zigzag_sub_4x4_frame; +#ifdef HAVE_MMX + if( cpu&X264_CPU_MMX ) + pf->scan_4x4 = x264_zigzag_scan_4x4_frame_mmx; + if( cpu&X264_CPU_MMXEXT ) + pf->scan_8x8 = x264_zigzag_scan_8x8_frame_mmxext; + if( cpu&X264_CPU_SSE2_IS_FAST ) + pf->scan_8x8 = x264_zigzag_scan_8x8_frame_sse2; + if( cpu&X264_CPU_SSSE3 ) + { + pf->sub_4x4 = x264_zigzag_sub_4x4_frame_ssse3; + pf->scan_8x8 = x264_zigzag_scan_8x8_frame_ssse3; + } + if( cpu&X264_CPU_PHADD_IS_FAST ) + pf->scan_4x4 = x264_zigzag_scan_4x4_frame_ssse3; +#endif + +#ifdef ARCH_PPC + if( cpu&X264_CPU_ALTIVEC ) + pf->scan_4x4 = x264_zigzag_scan_4x4_frame_altivec; +#endif + } + + pf->interleave_8x8_cavlc = zigzag_interleave_8x8_cavlc; +#ifdef HAVE_MMX + if( cpu&X264_CPU_MMX ) + pf->interleave_8x8_cavlc = x264_zigzag_interleave_8x8_cavlc_mmx; +#endif +}