xref: /csrg-svn/lib/libm/common_source/infnan.3 (revision 61306)

.\" Copyright (c) 1985, 1991, 1993
.\"	The Regents of the University of California.  All rights reserved.
.\"
.\" %sccs.include.redist.man%
.\"
.\"     @(#)infnan.3	8.1 (Berkeley) 06/04/93
.\"
.Dd
.Dt INFNAN 3
.Os BSD 4.3
.Sh NAME
.Nm infnan
.Nd signals invalid floating\-point operations on a
.Tn VAX
(temporary)
.Sh SYNOPSIS
.Fd #include <math.h>
.Ft double
.Fn infnan "int iarg"
.Sh DESCRIPTION
At some time in the future, some of the useful properties of
the Infinities and \*(Nas in the
.Tn IEEE
standard 754 for Binary
Floating\-Point Arithmetic will be simulated in
.Tn UNIX
on the
.Tn DEC VAX
by using its Reserved Operands.  Meanwhile, the
Invalid, Overflow and Divide\-by\-Zero exceptions of the
.Tn IEEE
standard are being approximated on a
.Tn VAX
by calls to a
procedure
.Fn infnan
in appropriate places in
.Xr libm 3 .
When
better exception\-handling is implemented in
.Tn UNIX ,
only
.Fn infnan
among the codes in
.Xr libm
will have to be changed.
And users of
.Xr libm
can design their own
.Fn infnan
now to
insulate themselves from future changes.
.Pp
Whenever an elementary function code in
.Xr libm
has to
simulate one of the aforementioned
.Tn IEEE
exceptions, it calls
.Fn infnan iarg
with an appropriate value of
.Fa iarg .
Then a
reserved operand fault stops computation.  But
.Fn infnan
could
be replaced by a function with the same name that returns
some plausible value, assigns an apt value to the global
variable
.Va errno ,
and allows computation to resume.
Alternatively, the Reserved Operand Fault Handler could be
changed to respond by returning that plausible value, etc.
instead of aborting.
.Pp
In the table below, the first two columns show various
exceptions signaled by the
.Tn IEEE
standard, and the default
result it prescribes.  The third column shows what value is
given to
.Fa iarg
by functions in
.Xr libm
when they
invoke
.Fn infnan iarg
under analogous circumstances on a
.Tn VAX .
Currently
.Fn infnan
stops computation under all those
circumstances.  The last two columns offer an alternative;
they suggest a setting for
.Va errno
and a value for a
revised
.Fn infnan
to return.  And a C program to
implement that suggestion follows.
.sp 0.5
.Bd -filled -offset indent
.Bl -column "IEEE Signal" "IEEE Default" XXERANGE ERANGEXXorXXEDOM
.It IEEE Signal	IEEE Default Ta
.Fa iarg Ta
.Va errno Ta
.Fn infnan
.It Invalid	\*(Na Ta
.Dv EDOM	EDOM	0
.It Overflow	\(+-\*(If Ta
.Dv ERANGE	ERANGE	HUGE
.It Div\-by\-0	\(+-Infinity Ta
.Dv \(+-ERANGE	ERANGE or EDOM	\(+-HUGE
.It 	( Ns Dv HUGE No "= 1.7e38 ... nearly  2.0**127)"
.El
.Ed
.Pp
ALTERNATIVE
.Fn infnan :
.Bd -literal -offset indent
#include	<math.h>
#include	<errno.h>
extern int	errno ;
double	infnan(iarg)
int	iarg ;
{
	switch(iarg) {
	case	\0ERANGE:	errno = ERANGE; return(HUGE);
	case	\-ERANGE:	errno = EDOM;	return(\-HUGE);
	default:		errno = EDOM;	return(0);
	}
}
.Ed
.Sh SEE ALSO
.Xr math 3 ,
.Xr intro 2 ,
.Xr signal 3 .
.Pp
.Dv ERANGE
and
.Dv EDOM
are defined in
.Aq Pa errno.h .
(See
.Xr intro 2
for explanation of
.Dv EDOM
and
.Dv ERANGE . )
.Sh HISTORY
The
.Fn infnan
function appeared in
.Bx 4.3 .