This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Question on g++ compile options / numerical differences




-----Original Message-----
From: Dockeen [mailto:dockeen@mchsi.com]
Sent: Monday, May 20, 2002 6:06 AM
To: Alexandre Oliva
Subject: RE: Question on g++ compile options / numerical differences


Thanks for your reply.  I am probably stuck with running it down myself
as the code I am using is not easily distributable.  I mainly asked the 
list in case it was one of those "Wayne, everyone knows its a bad idea
to use that option, what rock have you been living under" cases.

So, again, many sincere thanks for your time and patience. Have a great
day!

Wayne (notso) Keen

-----Original Message-----
From: Alexandre Oliva [mailto:aoliva@redhat.com]
Sent: Monday, May 20, 2002 4:48 PM
To: Dockeen
Cc: gcc@gcc.gnu.org; gcc-help@gcc.gnu.org
Subject: Re: Question on g++ compile options / numerical differences


On May 19, 2002, "Dockeen" <dockeen@mchsi.com> wrote:

> Is there something transparently stupid in the additional
> options above that might lead to numerical calculations going
> bad?

Well, depending on the numerical stability of your code, -ffast-math's
effects may range from none to dramatic.  Also, the possibility of one
of the other flags you're using being exposing a bug in GCC cannot be
ruled out, but it would require some work on your end to figure that
out and provide us with a testcase that demonstrates the problem.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]