[RFC] How to add vector math functions to Glibc

[Joseph S. Myers]

On Wed, 1 Oct 2014, Andrew Senkevich wrote:

> > In general, patch submissions should be minimal (subject to bisectability)
> > - if pieces can sensibly be separated out, they should be, and each piece
> > should be given a meaningful subject (which will be the summary line of
> > the git commit message) describing what that piece does.  It's entirely
> > plausible there are people concerned about a change to build requirements
> > who aren't concerned about vector functions.
> 
> Is it OK to send patch with such update, containing also deletion of
> configure checks about AVX2 support as well as according preprocessor
> directive for hiding AVX2 codes? May be something else need to be
> updated?

I advise keeping architecture-specific removal of configure checks 
separate from architecture-independent increases in minimum versions.

The AVX2 checks appear to be compiler tests, not binutils tests, so they 
could only be removed after an increase of minimum GCC version for 
building glibc to 4.7.  Again, discussion of minimum GCC versions 
(architecture-independent) is best done in a separate thread that is 
explicitly about that question and that question only, but I'm not sure if 
there would be a consensus for 4.7 or only for 4.6 as new minimum version.  
And removal of configure checks that are obsolete with the new minimum 
version might still best be separate from the patch that actually 
increases the minimum.

-- 
Joseph S. Myers
joseph@codesourcery.com