This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH - [tree-ssa] regrouping of expression tree for singlemultiply add.
- From: "Joseph S. Myers" <jsm at polyomino dot org dot uk>
- To: Bradley Lucier <lucier at math dot purdue dot edu>
- Cc: gcc at gcc dot gnu dot org
- Date: Fri, 26 Mar 2004 16:38:04 +0000 (UTC)
- Subject: Re: PATCH - [tree-ssa] regrouping of expression tree for singlemultiply add.
- References: <0C5B1FCC-7ECF-11D8-A83C-003065BA681E@math.purdue.edu>
On Thu, 25 Mar 2004, Bradley Lucier wrote:
> otherwise use -funsafe-math-optimizations. And in the future, will
> there not be pragmas that one can use to designate that certain blocks
> of code will be compiled with certain floating-point properties? If
If someone implements them. Cf. Stephen Moshier in
<http://gcc.gnu.org/ml/gcc/2001-05/msg01624.html> (one of many previous
threads on floating point):
> > My questions is: Why isn't there a switch like
> > -mieee754-compliant-code-no-matter-how-slow-it-is-on-this-buggy-x86-fpu
>
> In a word, because there is no market for it. [...]
>
> The newer C99 standard includes pragmas and support to elaborate the
> IEEE flags. I started a GCC implementation of them, posted some
> patches and a plea for comments a year and a half ago, and there was
> so little interest that I dropped the subject. [...]
I have what I think is the current version of his patches from then, but
while the testcases would probably still be useful a full implementation
of the pragmas would be much larger (involving checking many
transformations of trees and RTL to ensure that floating point flags are
propagated, expressions with different flags but otherwise identical or
fully identical but under FENV_ACCESS aren't considered equivalent, etc.).
--
Joseph S. Myers
jsm@polyomino.org.uk