diff -r 78a00bc68913 -r 767d3c4153a1 misc/libtremor/tremor/mdct.c --- a/misc/libtremor/tremor/mdct.c Thu Sep 20 23:04:10 2012 +0200 +++ b/misc/libtremor/tremor/mdct.c Fri Sep 21 00:50:04 2012 +0200 @@ -6,14 +6,14 @@ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE * * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. * * * - * THE OggVorbis 'TREMOR' SOURCE CODE IS (C) COPYRIGHT 1994-2002 * + * THE OggVorbis 'TREMOR' SOURCE CODE IS (C) COPYRIGHT 1994-2003 * * BY THE Xiph.Org FOUNDATION http://www.xiph.org/ * * * ******************************************************************** function: normalized modified discrete cosine transform power of two length transform only [64 <= n ] - last mod: $Id: mdct.c,v 1.9 2002/10/16 09:17:39 xiphmont Exp $ + last mod: $Id: mdct.c,v 1.9.6.5 2003/04/29 04:03:27 xiphmont Exp $ Original algorithm adapted long ago from _The use of multirate filter banks for coding of high quality digital audio_, by T. Sporer, @@ -33,23 +33,62 @@ ********************************************************************/ #include "ivorbiscodec.h" -#include "codebook.h" +#include "os.h" #include "misc.h" #include "mdct.h" #include "mdct_lookup.h" +STIN void presymmetry(DATA_TYPE *in,int n2,int step){ + DATA_TYPE *aX; + DATA_TYPE *bX; + LOOKUP_T *T; + int n4=n2>>1; + + aX = in+n2-3; + T = sincos_lookup0; + + do{ + REG_TYPE r0= aX[0]; + REG_TYPE r2= aX[2]; + XPROD31( r0, r2, T[0], T[1], &aX[0], &aX[2] ); T+=step; + aX-=4; + }while(aX>=in+n4); + do{ + REG_TYPE r0= aX[0]; + REG_TYPE r2= aX[2]; + XPROD31( r0, r2, T[1], T[0], &aX[0], &aX[2] ); T-=step; + aX-=4; + }while(aX>=in); + + aX = in+n2-4; + bX = in; + T = sincos_lookup0; + do{ + REG_TYPE ri0= aX[0]; + REG_TYPE ri2= aX[2]; + REG_TYPE ro0= bX[0]; + REG_TYPE ro2= bX[2]; + + XNPROD31( ro2, ro0, T[1], T[0], &aX[0], &aX[2] ); T+=step; + XNPROD31( ri2, ri0, T[0], T[1], &bX[0], &bX[2] ); + + aX-=4; + bX+=4; + }while(aX>=in+n4); + +} /* 8 point butterfly (in place) */ STIN void mdct_butterfly_8(DATA_TYPE *x){ - REG_TYPE r0 = x[4] + x[0]; - REG_TYPE r1 = x[4] - x[0]; - REG_TYPE r2 = x[5] + x[1]; - REG_TYPE r3 = x[5] - x[1]; - REG_TYPE r4 = x[6] + x[2]; - REG_TYPE r5 = x[6] - x[2]; - REG_TYPE r6 = x[7] + x[3]; - REG_TYPE r7 = x[7] - x[3]; + REG_TYPE r0 = x[0] + x[1]; + REG_TYPE r1 = x[0] - x[1]; + REG_TYPE r2 = x[2] + x[3]; + REG_TYPE r3 = x[2] - x[3]; + REG_TYPE r4 = x[4] + x[5]; + REG_TYPE r5 = x[4] - x[5]; + REG_TYPE r6 = x[6] + x[7]; + REG_TYPE r7 = x[6] - x[7]; x[0] = r5 + r3; x[1] = r7 - r1; @@ -64,29 +103,25 @@ /* 16 point butterfly (in place, 4 register) */ STIN void mdct_butterfly_16(DATA_TYPE *x){ - - REG_TYPE r0, r1; - - r0 = x[ 0] - x[ 8]; x[ 8] += x[ 0]; - r1 = x[ 1] - x[ 9]; x[ 9] += x[ 1]; - x[ 0] = MULT31((r0 + r1) , cPI2_8); - x[ 1] = MULT31((r1 - r0) , cPI2_8); + + REG_TYPE r0, r1, r2, r3; + + r0 = x[ 8] - x[ 9]; x[ 8] += x[ 9]; + r1 = x[10] - x[11]; x[10] += x[11]; + r2 = x[ 1] - x[ 0]; x[ 9] = x[ 1] + x[0]; + r3 = x[ 3] - x[ 2]; x[11] = x[ 3] + x[2]; + x[ 0] = MULT31((r0 - r1) , cPI2_8); + x[ 1] = MULT31((r2 + r3) , cPI2_8); + x[ 2] = MULT31((r0 + r1) , cPI2_8); + x[ 3] = MULT31((r3 - r2) , cPI2_8); MB(); - r0 = x[10] - x[ 2]; x[10] += x[ 2]; - r1 = x[ 3] - x[11]; x[11] += x[ 3]; - x[ 2] = r1; x[ 3] = r0; - MB(); - - r0 = x[12] - x[ 4]; x[12] += x[ 4]; - r1 = x[13] - x[ 5]; x[13] += x[ 5]; - x[ 4] = MULT31((r0 - r1) , cPI2_8); - x[ 5] = MULT31((r0 + r1) , cPI2_8); - MB(); - - r0 = x[14] - x[ 6]; x[14] += x[ 6]; - r1 = x[15] - x[ 7]; x[15] += x[ 7]; - x[ 6] = r0; x[ 7] = r1; + r2 = x[12] - x[13]; x[12] += x[13]; + r3 = x[14] - x[15]; x[14] += x[15]; + r0 = x[ 4] - x[ 5]; x[13] = x[ 5] + x[ 4]; + r1 = x[ 7] - x[ 6]; x[15] = x[ 7] + x[ 6]; + x[ 4] = r2; x[ 5] = r1; + x[ 6] = r3; x[ 7] = r0; MB(); mdct_butterfly_8(x); @@ -96,48 +131,40 @@ /* 32 point butterfly (in place, 4 register) */ STIN void mdct_butterfly_32(DATA_TYPE *x){ - REG_TYPE r0, r1; - - r0 = x[30] - x[14]; x[30] += x[14]; - r1 = x[31] - x[15]; x[31] += x[15]; - x[14] = r0; x[15] = r1; - MB(); + REG_TYPE r0, r1, r2, r3; - r0 = x[28] - x[12]; x[28] += x[12]; - r1 = x[29] - x[13]; x[29] += x[13]; - XNPROD31( r0, r1, cPI1_8, cPI3_8, &x[12], &x[13] ); - MB(); - - r0 = x[26] - x[10]; x[26] += x[10]; - r1 = x[27] - x[11]; x[27] += x[11]; - x[10] = MULT31((r0 - r1) , cPI2_8); - x[11] = MULT31((r0 + r1) , cPI2_8); + r0 = x[16] - x[17]; x[16] += x[17]; + r1 = x[18] - x[19]; x[18] += x[19]; + r2 = x[ 1] - x[ 0]; x[17] = x[ 1] + x[ 0]; + r3 = x[ 3] - x[ 2]; x[19] = x[ 3] + x[ 2]; + XNPROD31( r0, r1, cPI3_8, cPI1_8, &x[ 0], &x[ 2] ); + XPROD31 ( r2, r3, cPI1_8, cPI3_8, &x[ 1], &x[ 3] ); MB(); - r0 = x[24] - x[ 8]; x[24] += x[ 8]; - r1 = x[25] - x[ 9]; x[25] += x[ 9]; - XNPROD31( r0, r1, cPI3_8, cPI1_8, &x[ 8], &x[ 9] ); - MB(); - - r0 = x[22] - x[ 6]; x[22] += x[ 6]; - r1 = x[ 7] - x[23]; x[23] += x[ 7]; - x[ 6] = r1; x[ 7] = r0; + r0 = x[20] - x[21]; x[20] += x[21]; + r1 = x[22] - x[23]; x[22] += x[23]; + r2 = x[ 5] - x[ 4]; x[21] = x[ 5] + x[ 4]; + r3 = x[ 7] - x[ 6]; x[23] = x[ 7] + x[ 6]; + x[ 4] = MULT31((r0 - r1) , cPI2_8); + x[ 5] = MULT31((r3 + r2) , cPI2_8); + x[ 6] = MULT31((r0 + r1) , cPI2_8); + x[ 7] = MULT31((r3 - r2) , cPI2_8); MB(); - r0 = x[ 4] - x[20]; x[20] += x[ 4]; - r1 = x[ 5] - x[21]; x[21] += x[ 5]; - XPROD31 ( r0, r1, cPI3_8, cPI1_8, &x[ 4], &x[ 5] ); + r0 = x[24] - x[25]; x[24] += x[25]; + r1 = x[26] - x[27]; x[26] += x[27]; + r2 = x[ 9] - x[ 8]; x[25] = x[ 9] + x[ 8]; + r3 = x[11] - x[10]; x[27] = x[11] + x[10]; + XNPROD31( r0, r1, cPI1_8, cPI3_8, &x[ 8], &x[10] ); + XPROD31 ( r2, r3, cPI3_8, cPI1_8, &x[ 9], &x[11] ); MB(); - r0 = x[ 2] - x[18]; x[18] += x[ 2]; - r1 = x[ 3] - x[19]; x[19] += x[ 3]; - x[ 2] = MULT31((r1 + r0) , cPI2_8); - x[ 3] = MULT31((r1 - r0) , cPI2_8); - MB(); - - r0 = x[ 0] - x[16]; x[16] += x[ 0]; - r1 = x[ 1] - x[17]; x[17] += x[ 1]; - XPROD31 ( r0, r1, cPI1_8, cPI3_8, &x[ 0], &x[ 1] ); + r0 = x[28] - x[29]; x[28] += x[29]; + r1 = x[30] - x[31]; x[30] += x[31]; + r2 = x[12] - x[13]; x[29] = x[13] + x[12]; + r3 = x[15] - x[14]; x[31] = x[15] + x[14]; + x[12] = r0; x[13] = r3; + x[14] = r1; x[15] = r2; MB(); mdct_butterfly_16(x); @@ -147,87 +174,30 @@ /* N/stage point generic N stage butterfly (in place, 2 register) */ STIN void mdct_butterfly_generic(DATA_TYPE *x,int points,int step){ - LOOKUP_T *T = sincos_lookup0; - DATA_TYPE *x1 = x + points - 8; - DATA_TYPE *x2 = x + (points>>1) - 8; - REG_TYPE r0; - REG_TYPE r1; + LOOKUP_T *T = sincos_lookup0; + DATA_TYPE *x1 = x + points - 4; + DATA_TYPE *x2 = x + (points>>1) - 4; + REG_TYPE r0, r1, r2, r3; do{ - r0 = x1[6] - x2[6]; x1[6] += x2[6]; - r1 = x2[7] - x1[7]; x1[7] += x2[7]; - XPROD31( r1, r0, T[0], T[1], &x2[6], &x2[7] ); T+=step; - - r0 = x1[4] - x2[4]; x1[4] += x2[4]; - r1 = x2[5] - x1[5]; x1[5] += x2[5]; - XPROD31( r1, r0, T[0], T[1], &x2[4], &x2[5] ); T+=step; - - r0 = x1[2] - x2[2]; x1[2] += x2[2]; - r1 = x2[3] - x1[3]; x1[3] += x2[3]; - XPROD31( r1, r0, T[0], T[1], &x2[2], &x2[3] ); T+=step; - - r0 = x1[0] - x2[0]; x1[0] += x2[0]; - r1 = x2[1] - x1[1]; x1[1] += x2[1]; - XPROD31( r1, r0, T[0], T[1], &x2[0], &x2[1] ); T+=step; - - x1-=8; x2-=8; + r0 = x1[0] - x1[1]; x1[0] += x1[1]; + r1 = x1[3] - x1[2]; x1[2] += x1[3]; + r2 = x2[1] - x2[0]; x1[1] = x2[1] + x2[0]; + r3 = x2[3] - x2[2]; x1[3] = x2[3] + x2[2]; + XPROD31( r1, r0, T[0], T[1], &x2[0], &x2[2] ); + XPROD31( r2, r3, T[0], T[1], &x2[1], &x2[3] ); T+=step; + x1-=4; + x2-=4; }while(Tsincos_lookup0); - do{ - r0 = x2[6] - x1[6]; x1[6] += x2[6]; - r1 = x2[7] - x1[7]; x1[7] += x2[7]; - XPROD31( r0, r1, T[0], T[1], &x2[6], &x2[7] ); T+=step; - - r0 = x2[4] - x1[4]; x1[4] += x2[4]; - r1 = x2[5] - x1[5]; x1[5] += x2[5]; - XPROD31( r0, r1, T[0], T[1], &x2[4], &x2[5] ); T+=step; - - r0 = x2[2] - x1[2]; x1[2] += x2[2]; - r1 = x2[3] - x1[3]; x1[3] += x2[3]; - XPROD31( r0, r1, T[0], T[1], &x2[2], &x2[3] ); T+=step; - - r0 = x2[0] - x1[0]; x1[0] += x2[0]; - r1 = x2[1] - x1[1]; x1[1] += x2[1]; - XPROD31( r0, r1, T[0], T[1], &x2[0], &x2[1] ); T+=step; - - x1-=8; x2-=8; - }while(Tsincos_lookup0); } @@ -235,15 +205,14 @@ int stages=8-shift; int i,j; - + for(i=0;--stages>0;i++){ for(j=0;j<(1<>i)*j,points>>i,4<<(i+shift)); } - + for(j=0;j>8]|(bitrev[(x&0x0f0)>>4]<<4)|(((int)bitrev[x&0x00f])<<8); } -STIN void mdct_bitreverse(DATA_TYPE *x,int n,int step,int shift){ - +STIN void mdct_bitreverse(DATA_TYPE *x,int n,int shift){ int bit = 0; - DATA_TYPE *w0 = x; - DATA_TYPE *w1 = x = w0+(n>>1); - LOOKUP_T *T = (step>=4)?(sincos_lookup0+(step>>1)):sincos_lookup1; - LOOKUP_T *Ttop = T+1024; - DATA_TYPE r2; + DATA_TYPE *w = x+(n>>1); do{ - DATA_TYPE r3 = bitrev12(bit++); - DATA_TYPE *x0 = x + ((r3 ^ 0xfff)>>shift) -1; - DATA_TYPE *x1 = x + (r3>>shift); + DATA_TYPE b = bitrev12(bit++); + DATA_TYPE *xx = x + (b>>shift); + REG_TYPE r; + + w -= 2; + + if(w>xx){ - REG_TYPE r0 = x0[0] + x1[0]; - REG_TYPE r1 = x1[1] - x0[1]; + r = xx[0]; + xx[0] = w[0]; + w[0] = r; + + r = xx[1]; + xx[1] = w[1]; + w[1] = r; + } + }while(w>x); +} - XPROD32( r0, r1, T[1], T[0], &r2, &r3 ); T+=step; +STIN void mdct_step7(DATA_TYPE *x,int n,int step){ + DATA_TYPE *w0 = x; + DATA_TYPE *w1 = x+(n>>1); + LOOKUP_T *T = (step>=4)?(sincos_lookup0+(step>>1)):sincos_lookup1; + LOOKUP_T *Ttop = T+1024; + REG_TYPE r0, r1, r2, r3; + + do{ + w1 -= 2; - w1 -= 4; + r0 = w0[0] + w1[0]; + r1 = w1[1] - w0[1]; + r2 = MULT32(r0, T[1]) + MULT32(r1, T[0]); + r3 = MULT32(r1, T[1]) - MULT32(r0, T[0]); + T+=step; - r0 = (x0[1] + x1[1])>>1; - r1 = (x0[0] - x1[0])>>1; + r0 = (w0[1] + w1[1])>>1; + r1 = (w0[0] - w1[0])>>1; w0[0] = r0 + r2; w0[1] = r1 + r3; - w1[2] = r0 - r2; - w1[3] = r3 - r1; + w1[0] = r0 - r2; + w1[1] = r3 - r1; - r3 = bitrev12(bit++); - x0 = x + ((r3 ^ 0xfff)>>shift) -1; - x1 = x + (r3>>shift); + w0 += 2; + }while(T>1; - r1 = (x0[0] - x1[0])>>1; - w0[2] = r0 + r2; - w0[3] = r1 + r3; + r0 = (w0[1] + w1[1])>>1; + r1 = (w0[0] - w1[0])>>1; + w0[0] = r0 + r2; + w0[1] = r1 + r3; w1[0] = r0 - r2; w1[1] = r3 - r1; - w0 += 4; - }while(T>shift) -1; - DATA_TYPE *x1 = x + (r3>>shift); - - REG_TYPE r0 = x0[0] + x1[0]; - REG_TYPE r1 = x1[1] - x0[1]; - - T-=step; XPROD32( r0, r1, T[0], T[1], &r2, &r3 ); - - w1 -= 4; - - r0 = (x0[1] + x1[1])>>1; - r1 = (x0[0] - x1[0])>>1; - w0[0] = r0 + r2; - w0[1] = r1 + r3; - w1[2] = r0 - r2; - w1[3] = r3 - r1; - - r3 = bitrev12(bit++); - x0 = x + ((r3 ^ 0xfff)>>shift) -1; - x1 = x + (r3>>shift); - - r0 = x0[0] + x1[0]; - r1 = x1[1] - x0[1]; - - T-=step; XPROD32( r0, r1, T[0], T[1], &r2, &r3 ); - - r0 = (x0[1] + x1[1])>>1; - r1 = (x0[0] - x1[0])>>1; - w0[2] = r0 + r2; - w0[3] = r1 + r3; - w1[0] = r0 - r2; - w1[1] = r3 - r1; - - w0 += 4; + w0 += 2; }while(w0>1; - int n4=n>>2; - DATA_TYPE *iX; - DATA_TYPE *oX; +STIN void mdct_step8(DATA_TYPE *x, int n, int step){ LOOKUP_T *T; LOOKUP_T *V; + DATA_TYPE *iX =x+(n>>1); + step>>=2; + + switch(step) { + default: + T=(step>=4)?(sincos_lookup0+(step>>1)):sincos_lookup1; + do{ + REG_TYPE r0 = x[0]; + REG_TYPE r1 = -x[1]; + XPROD31( r0, r1, T[0], T[1], x, x+1); T+=step; + x +=2; + }while(x>1; + t1 = (*T++)>>1; + do{ + r0 = x[0]; + r1 = -x[1]; + t0 += (v0 = (*V++)>>1); + t1 += (v1 = (*V++)>>1); + XPROD31( r0, r1, t0, t1, x, x+1 ); + + r0 = x[2]; + r1 = -x[3]; + v0 += (t0 = (*T++)>>1); + v1 += (t1 = (*T++)>>1); + XPROD31( r0, r1, v0, v1, x+2, x+3 ); + + x += 4; + }while(x>2); + t1 += (q1 = (v1-t1)>>2); + r0 = x[0]; + r1 = -x[1]; + XPROD31( r0, r1, t0, t1, x, x+1 ); + t0 = v0-q0; + t1 = v1-q1; + r0 = x[2]; + r1 = -x[3]; + XPROD31( r0, r1, t0, t1, x+2, x+3 ); + + t0 = *T++; + t1 = *T++; + v0 += (q0 = (t0-v0)>>2); + v1 += (q1 = (t1-v1)>>2); + r0 = x[4]; + r1 = -x[5]; + XPROD31( r0, r1, v0, v1, x+4, x+5 ); + v0 = t0-q0; + v1 = t1-q1; + r0 = x[6]; + r1 = -x[7]; + XPROD31( r0, r1, v0, v1, x+5, x+6 ); + + x+=8; + }while(x>1,step); + mdct_butterflies(in,n>>1,shift); + mdct_bitreverse(in,n,shift); + mdct_step7(in,n,step); + mdct_step8(in,n,step); +} - do{ - oX-=4; - XPROD31( iX[4], iX[6], T[0], T[1], &oX[2], &oX[3] ); T+=step; - XPROD31( iX[0], iX[2], T[0], T[1], &oX[0], &oX[1] ); T+=step; - iX-=8; - }while(iX>=in+n4); - do{ - oX-=4; - XPROD31( iX[4], iX[6], T[1], T[0], &oX[2], &oX[3] ); T-=step; - XPROD31( iX[0], iX[2], T[1], T[0], &oX[0], &oX[1] ); T-=step; - iX-=8; - }while(iX>=in); +void mdct_shift_right(int n, DATA_TYPE *in, DATA_TYPE *right){ + int i; + n>>=2; + in+=1; - iX = in+n2-8; - oX = out+n2+n4; - T = sincos_lookup0; - - do{ - T+=step; XNPROD31( iX[6], iX[4], T[0], T[1], &oX[0], &oX[1] ); - T+=step; XNPROD31( iX[2], iX[0], T[0], T[1], &oX[2], &oX[3] ); - iX-=8; - oX+=4; - }while(iX>=in+n4); - do{ - T-=step; XNPROD31( iX[6], iX[4], T[1], T[0], &oX[0], &oX[1] ); - T-=step; XNPROD31( iX[2], iX[0], T[1], T[0], &oX[2], &oX[3] ); - iX-=8; - oX+=4; - }while(iX>=in); + for(i=0;i>=2; - { - DATA_TYPE *oX1=out+n2+n4; - DATA_TYPE *oX2=out+n2+n4; - DATA_TYPE *iX =out; +void mdct_unroll_lap(int n0,int n1, + int lW,int W, + DATA_TYPE *in, + DATA_TYPE *right, + LOOKUP_T *w0, + LOOKUP_T *w1, + ogg_int16_t *out, + int step, + int start, /* samples, this frame */ + int end /* samples, this frame */){ - switch(step) { - default: { - T=(step>=4)?(sincos_lookup0+(step>>1)):sincos_lookup1; - do{ - oX1-=4; - XPROD31( iX[0], -iX[1], T[0], T[1], &oX1[3], &oX2[0] ); T+=step; - XPROD31( iX[2], -iX[3], T[0], T[1], &oX1[2], &oX2[1] ); T+=step; - XPROD31( iX[4], -iX[5], T[0], T[1], &oX1[1], &oX2[2] ); T+=step; - XPROD31( iX[6], -iX[7], T[0], T[1], &oX1[0], &oX2[3] ); T+=step; - oX2+=4; - iX+=8; - }while(iX>1 : n0>>1); + DATA_TYPE *r=right+(lW ? n1>>2 : n0>>2); + DATA_TYPE *post; + LOOKUP_T *wR=(W && lW ? w1+(n1>>1) : w0+(n0>>1)); + LOOKUP_T *wL=(W && lW ? w1 : w0 ); - case 1: { - /* linear interpolation between table values: offset=0.5, step=1 */ - REG_TYPE t0,t1,v0,v1; - T = sincos_lookup0; - V = sincos_lookup1; - t0 = (*T++)>>1; - t1 = (*T++)>>1; - do{ - oX1-=4; + int preLap=(lW && !W ? (n1>>2)-(n0>>2) : 0 ); + int halfLap=(lW && W ? (n1>>2) : (n0>>2) ); + int postLap=(!lW && W ? (n1>>2)-(n0>>2) : 0 ); + int n,off; - t0 += (v0 = (*V++)>>1); - t1 += (v1 = (*V++)>>1); - XPROD31( iX[0], -iX[1], t0, t1, &oX1[3], &oX2[0] ); - v0 += (t0 = (*T++)>>1); - v1 += (t1 = (*T++)>>1); - XPROD31( iX[2], -iX[3], v0, v1, &oX1[2], &oX2[1] ); - t0 += (v0 = (*V++)>>1); - t1 += (v1 = (*V++)>>1); - XPROD31( iX[4], -iX[5], t0, t1, &oX1[1], &oX2[2] ); - v0 += (t0 = (*T++)>>1); - v1 += (t1 = (*T++)>>1); - XPROD31( iX[6], -iX[7], v0, v1, &oX1[0], &oX2[3] ); - - oX2+=4; - iX+=8; - }while(iX>2); - t1 += (q1 = (v1-t1)>>2); - XPROD31( iX[0], -iX[1], t0, t1, &oX1[3], &oX2[0] ); - t0 = v0-q0; - t1 = v1-q1; - XPROD31( iX[2], -iX[3], t0, t1, &oX1[2], &oX2[1] ); + /* preceeding direct-copy lapping from previous frame, if any */ + if(preLap){ + n = (endpost){ + *out = CLIP_TO_15((*--r)>>9); + out+=step; + } + } + + /* cross-lap; two halves due to wrap-around */ + n = (endpost){ + l-=2; + *out = CLIP_TO_15((MULT31(*--r,*--wR) + MULT31(*l,*wL++))>>9); + out+=step; + } - t0 = *T++; - t1 = *T++; - v0 += (q0 = (t0-v0)>>2); - v1 += (q1 = (t1-v1)>>2); - XPROD31( iX[4], -iX[5], v0, v1, &oX1[1], &oX2[2] ); - v0 = t0-q0; - v1 = t1-q1; - XPROD31( iX[6], -iX[7], v0, v1, &oX1[0], &oX2[3] ); - - oX2+=4; - iX+=8; - }while(iX>9); + out+=step; + l+=2; + } - do{ - oX1-=4; - iX-=4; - - oX2[0] = -(oX1[3] = iX[3]); - oX2[1] = -(oX1[2] = iX[2]); - oX2[2] = -(oX1[1] = iX[1]); - oX2[3] = -(oX1[0] = iX[0]); - - oX2+=4; - }while(oX2oX2); + /* preceeding direct-copy lapping from previous frame, if any */ + if(postLap){ + n = (end>9); + out+=step; + l+=2; + } } }