_CODE_ACCESS void *memset(void *dst, int fill, size_t len)
{
char *restrict dst1, *restrict dst2;
int pre_bytes, post_bytes, wfill, i;
double dfill1, dfill2;
dst1 = (char *)dst;
/*--------------------------------------------------------------------*/
/* Replicate the 8-bit value in fill into all 4 bytes of wfill */
/*--------------------------------------------------------------------*/
wfill = _pack2 (fill, fill);
wfill = _packl4(wfill, wfill);
dfill1 = _itod (wfill, wfill);
dfill2 = _itod (wfill, wfill);
/*--------------------------------------------------------------------*/
/* Calculate number of bytes to pre-copy to get to an alignment of 8 */
/*--------------------------------------------------------------------*/
pre_bytes = (8 - (int) dst) & 7;
if (len > pre_bytes)
{
len -= pre_bytes;
if (pre_bytes & 1) { *dst1 = fill; dst1 += 1; }
if (pre_bytes & 2) { _amem2(dst1) = wfill; dst1 += 2; }
if (pre_bytes & 4) { _amem4(dst1) = wfill; dst1 += 4; }
}
/*--------------------------------------------------------------------*/
/* Double word fills */
/*--------------------------------------------------------------------*/
post_bytes = len > 0 ? len : 0;
dst2 = dst1 + 8;
if (len > 15)
for (i = 0; i < len >> 4; i++)
{
_amemd8(dst1) = dfill1; dst1 += 16;
_amemd8(dst2) = dfill2; dst2 += 16;
post_bytes -= 16;
}
/*--------------------------------------------------------------------*/
/* Finish transfer with 8, 4, 2 and/or 1-byte writes */
/*--------------------------------------------------------------------*/
if (post_bytes & 8) { _memd8(dst1) = dfill1; dst1 += 8; }
if (post_bytes & 4) { _mem4 (dst1) = wfill; dst1 += 4; }
if (post_bytes & 2) { dst1[0] = wfill;
dst1[1] = wfill; dst1 += 2; }
if (post_bytes & 1) { *dst1 = fill; dst1 += 1; }
return dst;
}
double expdp_c(double a)
{
double Halfe = 0.5; //Zeroe = 0.0, Onee = 1.0, ;
double Maxe = 1.7976931348623157e+308 ;
double LnMaxe= 709.7827138470578 ;
double a0e= 0.249999999999999993;
double a1e= 0.694360001511792852e-2 ;
double a2e= 0.165203300268279130e-4;
double b0e= 0.5 ;
double b1e= 0.555538666969001188e-1 ;
double b2e= 0.495862884905441294e-3 ;
double c1e= 0.693359375 ; /* 355/512 */
double C2e=-2.1219444005469058277e-4 ;
double L2e = 1.4426950408889634074 ; /* log (base 2) of e */
double Ye,Xe,We,Re,Se,Be,Ce,De;
int Ne, upper;
Ye = a;
Ce = Ye*L2e ; /* base e --> base 2 argument */
Ne = _dpint(Ce) ; /* get unbiased exponent as an int */
Se = Ne ; /* double(int N) */
Xe = (Ye - Se*c1e) - Se*C2e ; /* range reduction */
We = Xe*Xe ;
Be = (b2e*We + b1e)*We + b0e ; /* denominator */
De = ((a2e*We + a1e)*We + a0e)*Xe ; /* numerator */
Re = Halfe + divdpMod_ci5( De, (Be-De) ) ;
upper = 1024 + Ne;
upper = _extu(upper,20,0);
Se = _itod(upper,0x00000000);
Ce = Re*Se ; /* scale by power of 2 */
if (a > LnMaxe) Ce = (Maxe) ; /* > LnMax returns MAX */
return (Ce) ;
}
/* -------------------------------------------------------------------- */
/* Inform the compiler by _nasserts the following: */
/* a) The output array is word aligned */
/* b) The width is greater than or equal to 2 */
/* c) The width is a multiple of 2. */
/* -------------------------------------------------------------------- */
_nassert((int) imgout_ptr % 4 == 0);
_nassert((int) width % 2 == 0);
_nassert((int) width >= 2);
/* -------------------------------------------------------------------- */
/* The 1st two mask values of each row are loaded into one int */
/* Reverse order of mask coefficients for rotation. */
/* -------------------------------------------------------------------- */
mask_tmp = _memd8_const((void *) &mask_ptr[5]);
mask3_0 = _itod(_packlh2(_lo(mask_tmp),_lo(mask_tmp)),
_packlh2(_hi(mask_tmp),_hi(mask_tmp)));
mask_tmp = _memd8_const((void *) &mask_ptr[1]);
mask7_4 = _itod(_packlh2(_lo(mask_tmp),_lo(mask_tmp)),
_packlh2(_hi(mask_tmp),_hi(mask_tmp)));
/* -------------------------------------------------------------------- */
/* The last mask values of each row are loaded into an int */
/* -------------------------------------------------------------------- */
mask8 = mask_ptr[0];
mask43 = _packlh2(_lo(mask7_4),_hi(mask3_0));
/* -------------------------------------------------------------------- */
/* mask2_88 contains the last mask of row 3 in its half words */
/* -------------------------------------------------------------------- */
mask88 = _pack2(mask8,mask8);
double logdp_c(double a)
{
// double One=1.0 ;
double Zero = 0.0, Half = 0.5 ;
double srHalf= 0.70710678118654752440 ; /* sqrt(0.5) */
double MIN = 2.2250738585072014e-308 ;
double LnMin = -708.3964185322641 ;
double MAX = 1.7976931348623157e+308 ;
double LnMax = 709.7827138470578 ;
double a0 =-0.64124943423745581147e+2 ;
double a1 = 0.16383943563021534222e+2 ;
double a2 =-0.78956112887491257267e+0 ;
double b0 =-0.76949932108494879777e+3 ;
double b1 = 0.31203222091924532844e+3 ;
double b2 =-0.35667977739034646171e+2 ; /* Note b3 = 1.0 */
double c1 = 0.693359375 ; /* 355/512 */
double c2 =-2.121944400546905827679e-4 ;
double Y,Z,zn,zd,X,W,Rz,Sa,Bd,Cn,Da ;
int N, exp, upper;
Y = a ;
exp = _extu(_hi(Y),1,21);
N = exp - 1022;
upper = _clr(_hi(Y),20,31);
upper = 0x3fe00000 | upper;
Z = _itod(upper, _lo(Y));
if (exp == 0) Z = 0;
if (Z > srHalf)
{ zn=(Z-Half)-Half ; zd=Z*Half+Half ; }
else
{ zn=Z-Half; zd=zn*Half+Half ; N=N-1 ; }
//X = zn/zd;
X = divdpMod_clog(zn,zd);
W = X*X ;
Bd= ( (W+b2) * W + b1) * W + b0 ;
Cn= (W*a2 + a1) * W + a0 ;
Rz= W * divdpMod_clog(Cn,Bd);//Cn / Bd ;
Sa = X + X*Rz ;
Cn= N ;
Da = (Cn*c2 + Sa) + Cn*c1 ;
if (Y <= Zero) Da = 0;//{Z = setnmx(Y) ; Da = (Z) ; }
if (Y < MIN) Da = (LnMin) ;
if (Y > MAX) Da = (LnMax) ;
return (Da) ;
}
