Re: Floating Exceptions

[Greg Lindahl <lindahl cs virginia edu>]

> I'm using Redhat Linux 5.2 on a AXP-Alpha and have problems with
> floating exceptions. I'm using CPML which prevents some 
> floating exceptions. Still, by divison by a small number e.g. I
> get 
> 
> floating exception (core dumped) 

Which should be no surprise; you can simply exceed the allowed
exponent range.

There are a few ways to solve this:

1) use gcc and -mieee, if it worked right. You seem to have found a
bug; if you can isolate it to a small example, maybe it will get
fixed.

2) Use compaq's C compiler, which isn't out yet.

3) use double precision. Perhaps this will give you enough exponent
range to dodge the problem.

> Also signal.h with a handler routine do not work.

Right. It's hard to get that right...

-- g