This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Make ieee754 fma tolerate architectures without exceptionsupport.


On Mon, 5 Nov 2012, Chris Metcalf wrote:

> On 11/5/2012 1:16 PM, Joseph S. Myers wrote:
> > On Mon, 5 Nov 2012, Chris Metcalf wrote:
> >
> >> +/* For IEEE754 architectures without exception support, tolerate imprecision
> >> +   in __fma but allow using this implementation instead of the generic one.
> > It really shouldn't be hard to adapt Bruno Haible's gnulib fma / fmaf / 
> > fmal to replace the generic versions with something accurate if slow (bug 
> > 13304) - but this does require approval to use the code under LGPLv2.1+ 
> > (the glibc license).
> >
> > (I don't know if the gnulib version passes all the tests I've added to 
> > glibc lately for various fma bugs I've fixed - there's still one left to 
> > fix, bug 14793 - but any problems those tests show ought to be easy to 
> > fix.)
> 
> Meanwhile, does my proposed change seem plausible?  It unbreaks the libm
> build for tile, and the tests continue to report 0 or 1 ULP for all the fma
> tests, much better than the generic fma.

I'd have thought that putting something in your math_private.h to allow 
this file to build, when it's used outside the circumstances for which it 
was designed, would be better.

-- 
Joseph S. Myers
joseph@codesourcery.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]