ep(3F)

NAME
     ep, epbase, epprec, epemin, epemax, eptiny, ephuge, epmrsp - FORTRAN
     numeric characteristic constants

SYNOPSIS
     integer*2 shrt
     integer*4 intr
     real*4 sngl
     real*8 dbl
     integer*8 ll

     integer*4 nbase, nprec, nemin, nemax, nhuge
     real*4 ftiny, fhuge, fepsl
     real*8 dtiny, dhuge, depsl
     integer*8 llhuge

     nbase = epbase(shrt)
     nprec = epprec(shrt)
     nhuge = ephuge(shrt)

     nbase = epbase(intr)
     nprec = epprec(intr)
     nhuge = ephuge(intr)

     nbase = epbase(ll)
     nprec = epprec(ll)
     llhuge = ephuge(ll)

     nbase = epbase(sngl)
     nprec = epprec(sngl)
     nemin = epemin(sngl)
     nemax = epemax(sngl)
     ftiny = eptiny(sngl)
     fhuge = ephuge(sngl)
     fepsl = epmrsp(sngl)

     nbase = epbase(dbl)
     nprec = epprec(dbl)
     nemin = epemin(dbl)
     nemax = epemax(dbl)
     dtiny = eptiny(dbl)
     dhuge = ephuge(dbl)
     depsl = epmrsp(dbl)

DESCRIPTION
     This routine is only available with the MIPSpro Fortran 77 compiler.

     ep is the general name for the set of intrinsic functions which return
     the numeric characteristics of each of the standard numeric types.

     epbase returns the base for the type of the argument (for floating
     point types, epbase returns the base of the mantissa).  Currently this
     is 2.

     epprec returns the precision (in terms of power of the base) for the
     type of the argument.  The sign bit (if any) is not included in the
     precision.  The assumed bit (see IEEE floating point specifications),
     if any, is included in the precision.

     epemin and epemax return the minimum (or maximum) exponent for the
     floating point type of the argument.

     eptiny and ephuge return the smallest (or largest) non-zero positive
     number in the type of the argument.

     epmrsp returns the epsilon for the floating point type of the
     argument.

NOTES
     eptiny is not supported for integer types.  It can be assumed to be 1.