weird optimization in sin+cos, x86 backend

[Joseph S. Myers]

On Thu, 9 Feb 2012, Andrew Haley wrote:

> > No, the point of the separate project would be to be used by both glibc 
> > and GCC (and possibly other GNU projects such as GSL) - because 
> > cooperation among the various projects wanting such functions is the right 
> > way to do things.
> 
> Well, yes, but I don't quite understand why libm isn't obviously a
> sub-project of libc.

Because glibc's libm is for general purpose use - striking a reasonable 
balance between speed, size and accuracy rather than providing lots of 
different variants including vector versions; it makes sense for glibc's 
libm to be a subset of a wider project rather than for glibc to maintain 
lots of code that isn't actually of use for glibc itself.  And the 
expertise for maintaining it is somewhat different for that for most of 
glibc.

-- 
Joseph S. Myers
joseph@codesourcery.com