This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH]: PR29335 use MPFR for builtins fma, fmin and fmax
- From: Andrew Pinski <pinskia at gmail dot com>
- To: "Kaveh R. GHAZI" <ghazi at caip dot rutgers dot edu>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 07 Nov 2006 19:21:18 -0800
- Subject: Re: [PATCH]: PR29335 use MPFR for builtins fma, fmin and fmax
- References: <Pine.GSO.4.58.0611072152060.25043@caipclassic.rutgers.edu>
On Tue, 2006-11-07 at 21:57 -0500, Kaveh R. GHAZI wrote:
> This patch uses MPFR for builtins fma, fmin and fmax.
>
> In addition, I updated the test macros to look at the sign of the results
> to distinguish 0.0 vs -0.0. There were lots of pre-existing tests
> expecting 0.0 or -0.0 results, but since 0.0 == -0.0 it wasn't actually
> testing that the correct sign of zero was returned AFAICT. Thankfully
> after adding the check, everything still works. :-)
>
> Bootstrapped on sparc-sun-solaris2.10, no regressions and the new tests
> all pass.
Do you know how much precision fma gets for mpfr? The main question I
am asking is that fma on say powerpc is most likely implemented using
the fused multiple and add instruction which is better than doing a
multiple and then doing an add which is allowed by the C standard IIRC.
Thanks,
Andrew Pinski