static force_inline __vector4
pix_multiply (__vector4 p, __vector4 a)
{
__vector4 hi, lo, mod;
__vector4 hiLow, hiHigh, modLow, modHigh, loLow, loHigh;
__vector4 hiLowFP, hiHighFP, modLowFP, modHighFP, loLowFP, loHighFP;
__vector4 himodLow, himodHigh, lomodLow, lomodHigh;
__vector4 zeroVector = *(__vector4*)(&zeroVectori);
__vector4 zeroEightVector = *(__vector4*)(&zeroEightVectori);
__vector4 permVecLo = *(__vector4*)(&permVecLoi);
__vector4 permVecHi = *(__vector4*)(&permVecHii);
__vector4 permVec1Vec2 = *(__vector4*)(&permVec1Vec2i);
/* unpack to short */
hi = __vmrghb(zeroVector,p);
mod = __vmrghb(zeroVector,a);
//+ What we want to do here is to multiply 8 unsigned shorts of the hi with 8 unsigned shorts of mod.
/* Extract the hi vector into 4 Unsigned int by using 4 Lower unsigned shorts*/
hiLow = __vperm(hi,zeroVector, permVecLo);
/* Extract the hi vector into 4 Unsigned int by using 4 Upper unsigned shorts*/
hiHigh = __vperm(hi,zeroVector, permVecHi);
/* Extract the mod vector into 4 Unsigned int by using 4 Lower unsigned shorts*/
modLow = __vperm(mod,zeroVector, permVecLo);
/* Extract the mod vector into 4 Unsigned int by using 4 Upper unsigned shorts*/
modHigh = __vperm(mod,zeroVector, permVecHi);
/* Convert the 4 unsigned ints to floating point by treating them as Fixed point*/
hiLowFP = __vcfux(hiLow,0);
hiHighFP = __vcfux(hiHigh,0);
modLowFP = __vcfux(modLow,0);
modHighFP = __vcfux(modHigh,0);
/* Multiply the floating points */
himodLow = __vmaddfp(hiLowFP, modLowFP,zeroVector);
himodHigh = __vmaddfp(hiHighFP, modHighFP,zeroVector);
/* Convert the floating points to Fixed Point with zero digits after radix point - Effectively an unsigned int*/
himodLow = __vctuxs(himodLow,0);
himodHigh = __vctuxs(himodHigh,0);
/* Fuse the multiplication together to get the final product*/
hi = __vperm(himodLow,himodHigh,permVec1Vec2);
//-
hi = __vadduhm(hi,zeroEightVector);
hi = __vadduhs(hi, __vsrh(hi, __vspltish (8)));
hi = __vsrh (hi, __vspltish (8));
/* unpack to short */
lo = __vmrglb(zeroVector,p);
mod = __vmrglb(zeroVector,a);
//+ Comments from few lines above applicable here.
loLow = __vperm(lo,zeroVector,permVecLo);
loHigh = __vperm(lo,zeroVector, permVecHi);
modLow = __vperm(mod,zeroVector, permVecLo);
modHigh = __vperm(mod,zeroVector, permVecHi);
loLowFP = __vcfux(loLow,0);
loHighFP = __vcfux(loHigh,0);
modLowFP = __vcfux(modLow,0);
modHighFP = __vcfux(modHigh,0);
lomodLow = __vmaddfp(loLowFP, modLowFP,zeroVector);
lomodHigh = __vmaddfp(loHighFP, modHighFP,zeroVector);
lomodLow = __vctuxs(lomodLow,0);
lomodHigh = __vctuxs(lomodHigh,0);
lo = __vperm(lomodLow,lomodHigh,permVec1Vec2);
//-
lo = __vadduhm(lo,zeroEightVector);
lo = __vadduhs (lo, __vsrh (lo, __vspltish (8)));
lo = __vsrh (lo, __vspltish (8));
return __vpkuhus (hi, lo);
}
{((8 - x) * (8 - y)),
(( x) * (8 - y)),
((8 - x) * ( y)),
(( x) * ( y))};
register int i;
__vector4 fperm;
const __vector4 vABCD = __lvx(ABCD,0);
const __vector4 vA = __vsplth(vABCD, 1);
const __vector4 vB = __vsplth(vABCD, 3);
const __vector4 vC = __vsplth(vABCD, 5);
const __vector4 vD = __vsplth(vABCD, 7);
//LOAD_ZERO;
__vector4 zerov = __vzero();
const __vector4 v32ss = __vslh(__vspltish(1),__vspltish(5));
const __vector4 v6us = __vspltish(6);
register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1;
register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0;
__vector4 vsrcAuc, av_uninit(vsrcBuc), vsrcperm0, vsrcperm1;
__vector4 vsrc0uc, vsrc1uc;
__vector4 vsrc0ssH, vsrc1ssH;
__vector4 vsrcCuc, vsrc2uc, vsrc3uc;
__vector4 vsrc2ssH, vsrc3ssH, psum;
__vector4 vdst, ppsum, vfdst, fsum;
if (((unsigned long)dst) % 16 == 0) {
fperm = {
0x10, 0x11, 0x12, 0x13,
0x14, 0x15, 0x16, 0x17,
