[Bug math/14412] Removal of sysdeps/x86_64/fpu/s_sincos.S causes regressions

["siddhesh at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=14412

--- Comment #39 from Siddhesh Poyarekar <siddhesh at redhat dot com> 2013-04-29 14:19:35 UTC ---
(In reply to comment #36)
> (In reply to comment #30)
> > Your patch doesn't work for me. I've re-build glibc and then
> > re-compiled poppler with "-ffinite-math-only". But I still
> > only see calls to __sin and __cos with "perf top".
> 
> Looks like your code does not call sincos directly.  sin and cos calls on the
> same input seem to get optimized into sincos, so I'm not sure if
> ffinite-math-only can detect that.  Also, the patch also changes an installed
> header, so you need to somehow include that during your compilation.

I've just verified locally that if gcc's transformation of sin + cos into
sincos works fine with the further transformation into __sincos_finite with
-ffinite-math-only, so it's most likely that your code is not using the header
file changes.  Easiest way to test this is to add the patch on top of your
distribution glibc, build and install it.

BTW, the question of whether -ffinite-math-only is the right option to do this
kind of thing is still open, so it's likely that the resulting fix will be
different.  The crux would be similar though, i.e. a fast function assembly
that is not always accurate.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.