class HCL_FPArch_d

HCL_FPArch_d is an interface class to the LAPACK DLAMCH, which determines the floating point characteristics of machine on which the code is running

Inheritance:

Public Methods

ostream&

Write ( ostream & str ) const Write displays the values of all parameters on the given stream.

Public

Machine parameters double eps eps is the relative machine precision. double sfmin sfmin is the safe minimum (1/sfmin does not overflow). int base base is the base of the floating point arithmetic. double prec prec = eps*base. int t t is the number of digits in the mantissa. int rnd rnd is 1 if rounding occurs in addition, 0 otherwise. int emin emin is the minimum exponent before (gradual) underflow. double rmin rmin is the underflow threshold: base raised to the power (emin-1) int emax emax is the largest exponent before overflow. double rmax rmax is the overflow threshold: (base to the power emax)*(1-eps).

Additional parameters double small small equals sfmin, unless LOG10(1/sfmin) is large, in which case small is the square root of sfmin double large large equals 1/sfmin, unless LOG10(1/sfmin) is large, in which case large is the square root of 1/sfmin

Machine parameters

double eps

eps is the relative machine precision.

double sfmin

sfmin is the safe minimum (1/sfmin does not overflow).

int base

base is the base of the floating point arithmetic.

double prec

prec = eps*base.

int t

t is the number of digits in the mantissa.

int rnd

rnd is 1 if rounding occurs in addition, 0 otherwise.

int emin

emin is the minimum exponent before (gradual) underflow.

double rmin

rmin is the underflow threshold: base raised to the power (emin-1)

int emax

emax is the largest exponent before overflow.

double rmax

rmax is the overflow threshold: (base to the power emax)*(1-eps).

Additional parameters

The following constants provide reasonable choices for the values of UNDERFLOW and OVERFLOW while also compensating for poor arithmetic in the upper half of the exponent range on certain machines (such as Crays.) (REFERENCE: LAPCK routine DLABAD.)

double small

small equals sfmin, unless LOG10(1/sfmin) is large, in which case small is the square root of sfmin

double large

large equals 1/sfmin, unless LOG10(1/sfmin) is large, in which case large is the square root of 1/sfmin

ostream& Write( ostream & str ) const

Write displays the values of all parameters on the given stream.

alphabetic index hierarchy of classes

this page has been generated automatically by doc++

(c)opyright by Malte Zöckler, Roland Wunderling
contact: doc++@zib.de