[Bug math/19058] [x86_64] Link fail with -fopenmp and -flto

["joseph at codesourcery dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=19058

--- Comment #17 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
On Wed, 10 Feb 2016, hjl.tools at gmail dot com wrote:

> https://sourceware.org/bugzilla/show_bug.cgi?id=19058
> 
> --- Comment #16 from H.J. Lu <hjl.tools at gmail dot com> ---
> (In reply to joseph@codesourcery.com from comment #15)
> > On Wed, 10 Feb 2016, hjl.tools at gmail dot com wrote:
> > 
> > > This only fails with LTO.  Why is libmvec_nonshared.a used in PR 19590 without
> > > LTO?
> > 
> > commit 977a30801f61b7bbc27e8f185c1e7eb49675c60c removed the asm statements 
> > in sysdeps/x86/fpu/bits/math-vector.h that previously meant the 
> > _Z___*_finite references wouldn't appear without LTO, so resulting in 
> > libmvec_nonshared.a being needed not just in the LTO case.
> 
> Are you saying that the alias is always used for ALL compilers, with and
> without limitation?

If the compiler has the hypothetical future feature - that I requested the 
work be done on when originally reviewing the libmvec patches - to allow 
more flexibility in what name the vector function name is based on, and if 
bits/math-vector.h is updated to use that future feature, then of course 
the alias is not used.

Otherwise, if the bits/math-finite.h declarations are active together with 
bits/math-vector.h, then these libmvec_nonshared.a wrappers will be used 
(as a workaround - things weren't meant to take so long to get the 
compiler feature done).

-- 
You are receiving this mail because:
You are on the CC list for the bug.