Bug#372544: fixed in eglibc 2.13-1
On 2011-10-18 07:15:31 +0200, Aurelien Jarno wrote:
> I have reopened the bug, but tagged it moreinfo + unreproducible given
> fma() has been implemented in eglibc 2.13, and that the testcases you
> provided now pass correctly on at least i386 and amd64.
>
> Please provide some more details or testcases.
Ah, I didn't see that the bug that was opened upstream yeaterday
was against an old glibc version! Still, Bruno Haible said:
"I see 6 different implementations of fma(), 4 implementations of fmaf(),
and 4 implementations of fmal() in the glibc source code.
How can you guarantee that all of them are thoroughly tested?
The ones in math/s_fma.c, math/s_fmaf.c, math/s_fmal.c are definitely
buggy."
But I wonder whether Debian supports a platform with such an
implementation.
The old upstream bug http://sourceware.org/bugzilla/show_bug.cgi?id=3268
was also still open.
I'll try to write my own test suite (based on difficult cases and
with the 4 rounding modes).
--
Vincent Lefèvre <vincent@vinc17.net> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / Arénaire project (LIP, ENS-Lyon)
Reply to: