This is the mail archive of the
mailing list for the GCC project.
Re: Floating point trouble with x86's extended precision
- From: dewar at gnat dot com (Robert Dewar)
- To: lucier at math dot purdue dot edu, reichelt at igpm dot rwth-aachen dot de
- Cc: gcc at gcc dot gnu dot org, wilson at tuliptree dot org
- Date: Fri, 22 Aug 2003 21:31:07 -0400 (EDT)
- Subject: Re: Floating point trouble with x86's extended precision
> You don't think it's a bug.
> I, and other people, think that the inability to get predictable
> behavior from GCC (even from an option that you might not like to use
> because of speed or other issues) is a bug.
> And this is just about the end of the story.
It seems silly to me to argue about whether or not this is a bug. Formally
it probably isn't wrt at least most of the language standards, but so
The issue is not what the language standards require, but rather what is
useful to a sufficient number of users to be worth implementing.
> I, and at least some
> other people, think that GCC should include at least one model of
> floating-point arithmetic for the x86 that allows one to deduce the
> types and precisions of all variables, operations, etc., by examining
> the source code. It does not (and it probably will not for the
> foreseeable future).
Sounds like a useful feature to include this model, I think you should
implement such a model, and we can see how useful it is in practice.
> Telling people that not having a predictable
> model of what their program will do is a feature, not a bug, doesn't
> cut it.
Neither does simply complaining :-) All that is needed here is someone
to do the work. Those interested in this feature should either implement
it, or pay someone to implement it!