usr/src/lib/libm/common_source/infnan.3

.TH INFNAN 3M  "20 August 1985"
.UC 4
.ds nn \fINaN\fR
.SH NAME
.nf
.ta \w'infnan \- 'u+1n
infnan \-       signals invalid floating\-point operations on a VAX
        TEMPORARY! ... Subject to Change.
.ta
.fi
.SH SYNOPSIS
.nf
.B #include <math.h>
.PP
.B double infnan(iarg)
.B int iarg;
.SH DESCRIPTION
At some time in the future, some of the useful properties of
the Infinities and \*(nns in the IEEE standard 754 for Binary
Floating\-Point Arithmetic will be simulated in UNIX on the
DEC VAX by using its Reserved Operands.  Meanwhile, the
Invalid, Overflow and Divide\-by\-Zero exceptions of the
IEEE standard are being approximated on a VAX by calls to a
procedure \fIinfnan\fR in appropriate places in \fIlibm\fR.  When
better exception\-handling is implemented in UNIX, only
\fIinfnan\fR among the codes in \fIlibm\fR will have to be changed.
And users of \fIlibm\fR can design their own \fIinfnan\fR now to
insulate themselves from future changes.
.PP
Whenever an elementary function code in \fIlibm\fR has to
simulate one of the aforementioned IEEE exceptions, it calls
infnan(iarg) with an appropriate value of \fIiarg\fR.  Then a
reserved operand fault stops computation.  But \fIinfnan\fR could
be replaced by a function with the same name that returns
some plausible value, assigns an apt value to the global
variable \fIerrno\fR, 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 IEEE standard, and the default
result it prescribes.  The third column shows what value is
given to \fIiarg\fR by functions in \fIlibm\fR when they
invoke infnan(iarg) under analogous circumstances on a VAX.
Currently \fIinfnan\fR stops computation under all those
circumstances.  The last two columns offer an alternative;
they suggest a setting for \fIerrno\fR and a value for a
revised \fIinfnan\fR to return.  And a C program to
implement that suggestion follows.
.sp 0.5
.RS
.nf
.ta \w'Div\-by\-0'u+2n +\w'+Infinity'u+1n +\w'+ERANGE'u+1n +\w'ERANGE or EDOM'u+4n +\w'+HUGE'u+1n
IEEE    IEEE
Signal  Default \fIiarg\fR      \fIerrno\fR     \fIinfnan\fR
.if t \
\l'4i'
.if n \
\l'5i'
Invalid \*(nn   EDOM    EDOM    0
.if n \{\
Overflow        \(+-Infinity    ERANGE  ERANGE  HUGE
Div\-by\-0      \(+-Infinity    \(+-ERANGE      ERANGE or EDOM  \(+-HUGE \}
.if t \{\
Overflow        \(+-\(if        ERANGE  ERANGE  HUGE
Div\-by\-0      \(+-\(if        \(+-ERANGE      ERANGE or EDOM  \(+-HUGE \}
.sp 0.5
        (HUGE = 1.7e38 ... nearly  2.0**127)
.ta
.fi
.RE
.sp 0.5
.PP
.nf
.ta \w'\fBextern int\fR'u+1n +\w'\fBdefault:\fR'u+1n +\w'\fB\-ERANGE:\fR'u+1n +\w'\fBerrno = ERANGE;\fR'u+1n
ALTERNATIVE  \fIinfnan\fR:\fB
.sp 0.5
#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);
        }
}\fR
.ta
.fi
.SH SEE ALSO
intro(3M), intro(2), signal(3).
.PP
ERANGE and EDOM are defined in <errno.h>.  See intro(2)
for interpretations of EDOM and ERANGE from the operating
system's point of view.