This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: ieee.exp tests containing generic FP tests
- From: Geoffrey Keating <geoffk at geoffk dot org>
- To: Aldy Hernandez <aldyh at redhat dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: 15 Oct 2004 15:18:56 -0700
- Subject: Re: ieee.exp tests containing generic FP tests
- References: <20041015191645.GA11010@redhat.com>
Aldy Hernandez <aldyh@redhat.com> writes:
> Hi Richard. Hi folks.
>
> Most of the tests in execute/ieee/ fail on the e500 (and the Vax for
> that matter), but there are a handful that pass, and which I don't
> consider IEEE compliance tests.
>
> Could someone more IEEE savvy point out which tests could be moved one
> level higher (gcc.c-torture/execute). I volunteer to move them,
> I'm just not sure whether all the ones I think should be moved, should...
>
> In particular, I believe we should keep all the INF, NAN, PZERO,
> NZERO, *ORDERED, tests in ieee/ and move everything else out.
>
> For example, these are a few I believe don't belong in IEEE:
>
> 20001122-1.c
> 20010114-2.c
> 20011123-1.c
> 20030331-1.c
> acc1.c
> etc.
>
>
> OTOH, things like fp-cmp*, mzero*, compare-fp*, etc do obviously belong in
> IEEE.
>
> Agreed? Am I missing something?
Some non-IEEE machines don't just not have NaNs, they also have
arithmetic that differs in some important way from what the IEEE
standard specifies. Tests outside the IEEE directory can't really
rely on any particular floating-point implementation.
As far as I can tell, all these tests really won't work on some
non-IEEE machine.