(j3.2006) exception message on termination

[Keith Bierman]

The combination of underflow and inexact indicates some loss of accuracy;
for whatever that's worth.

In a small section of code that's useful and significant. In a full sized
program it's pretty meaningless because it will almost always happen.

Counts would be useful (especially on systems where the handling is done
partially in software) but expensive.

In terms of cross platform consistency, I think that having the default
clearly be off and providing some informative notes about how to use the
exception reporting meaningfully would be optimal.

Keith Bierman
khbkhb at gmail.com
kbiermank AIM
303 997 2749


On Wed, Jun 12, 2013 at 9:51 AM, Rafik Zurob <rzurob at ca.ibm.com> wrote:

> Just to expand on what Daniel said, the reporting of fp exceptions on STOP
> and ERROR STOP is OFF by default in F77, F90, F95 modes.  It's ON by
> default in F2003 and F2008 modes.
>
> We haven't received complaints from users about the output, but our test
> team complained about inexact, since almost everything turns it on.  I
> think underflow is important enough (for some situations as Robert
> mentioned) that we shouldn't delete it or make it optional (if set) in the
> output.  As for inexact, while I can't think of a useful practical case
> for it, it is something that the IEEE modules (and IEEE-compliant
> hardware) can trap on.  So excluding it from the STOP/ERROR STOP output
> will be a bit of a red herring.
>
> Regards
>
> Rafik
>
> j3-bounces at mailman.j3-fortran.org wrote on 12/06/2013 10:30:42 AM:
>
> > From: Daniel C Chen/Toronto/IBM at IBMCA
> > To: robert.corbett at oracle.com, fortran standards email list for J3
> > <j3 at mailman.j3-fortran.org>,
> > Cc: j3-bounces at mailman.j3-fortran.org
> > Date: 12/06/2013 10:32 AM
> > Subject: Re: (j3.2006) exception message on termination
> > Sent by: j3-bounces at mailman.j3-fortran.org
> >
> > XLF compiler has an option to report fp exceptions including
> > underflow and inexact.  It is default OFF.
> >
> > Daniel
> >
> > XL Fortran Development - IBM Toronto Software Lab
> > Phone: 905-413-3056
> > Tie: 969-3056
> > Email: cdchen at ca.ibm.com
> > http://www.ibm.com/software/awdtools/fortran/xlfortran
> >
> > [image removed] Robert Corbett ---06/12/2013 04:06:43---Thinking
> > back on it, most of the complaints we received were about underflow
> > also.  Underflow is us
> >
> > From: Robert Corbett <robert.corbett at oracle.com>
> > To: fortran standards email list for J3 <j3 at mailman.j3-fortran.org>,
> > Date: 06/12/2013 04:06
> > Subject: Re: (j3.2006) exception message on termination
> > Sent by: j3-bounces at mailman.j3-fortran.org
> >
> >
> >
> > Thinking back on it, most of the complaints we received were about
> underflow
> > also.  Underflow is usually harmless in codes such as root finders and
> > quadrature routines.  On the under hand, underflow can cause silent
> > bad results
> > in some matrix computations.
> >
> > Robert Corbett
> >
> >
> >
> > On 06/12/13 00:38, Malcolm Cohen wrote:
> > > The only complaint we got was for reporting underflow, so we added
> > an option to
> > > suppress that one.
> > >
> > > We never report inexact.  I agree it is completely pointless to
> > report inexact,
> > > in my opinion it was an oversight not to make than an exception to
> > the reporting
> > > rule.
> > >
> > > Cheers,
> > >
> > > -----Original Message-----
> > > From: Robert Corbett
> > > Date: ?? 25?6?12? 12:59
> > > To: j3
> > > Subject: (j3.2006) exception message on termination
> > >
> > > Can we delete the second and third sentences of paragraph 2 in
> > Clause 8.4 in the
> > > upcoming revision?  They state
> > >
> > >        If an exception (14) is signaling on that image,
> > >        the processor shall issue a warning indicating
> > >        which exceptions are signaling; this warning
> > >        shall be on the unit identified by the named
> > >        constant ERROR_UNIT (13.8.2.8).  It is recommended
> > >        that the stop code is made available by formatted
> > >        output to the same unit.
> > >
> > > Sun FORTRAN 77 produced such output.  It was a constant source of
> complaints
> > > from users.  As bad as Sun FORTRAN 77 was in this regard, it did not
> report
> > > inexact exceptions, as is required by the text of the standard.
> > >
> > > Robert Corbett
> > > _______________________________________________
> > > J3 mailing list
> > > J3 at mailman.j3-fortran.org
> > > http://mailman.j3-fortran.org/mailman/listinfo/j3
> > >
> > >
> ________________________________________________________________________
> > > This e-mail has been scanned for all viruses by Star.
> > >
> ________________________________________________________________________
> > >
> >
> > _______________________________________________
> > J3 mailing list
> > J3 at mailman.j3-fortran.org
> > http://mailman.j3-fortran.org/mailman/listinfo/j3
> > _______________________________________________
> > J3 mailing list
> > J3 at mailman.j3-fortran.org
> > http://mailman.j3-fortran.org/mailman/listinfo/j3
>
> _______________________________________________
> J3 mailing list
> J3 at mailman.j3-fortran.org
> http://mailman.j3-fortran.org/mailman/listinfo/j3
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.j3-fortran.org/pipermail/j3/attachments/20130612/15d9e33f/attachment.html