This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 2/14] [x86_64] Vector math functions (added vector cos tests)
- From: Joseph Myers <joseph at codesourcery dot com>
- To: Andrew Senkevich <andrew dot n dot senkevich at gmail dot com>
- Cc: libc-alpha <libc-alpha at sourceware dot org>
- Date: Fri, 22 May 2015 17:23:26 +0000
- Subject: Re: [PATCH 2/14] [x86_64] Vector math functions (added vector cos tests)
- Authentication-results: sourceware.org; auth=none
- References: <CAMXFM3uL-z3r0Q2aSpnPQnGLkj1_SDmj+h=WUr4MaaMaRGGftg at mail dot gmail dot com> <alpine dot DEB dot 2 dot 10 dot 1505221547070 dot 16611 at digraph dot polyomino dot org dot uk> <CAMXFM3uxTnzX14LAQu11Oc78jGVsWwLXim3e2uEL9ECb6ah_=w at mail dot gmail dot com>
On Fri, 22 May 2015, Andrew Senkevich wrote:
> 2015-05-22 18:54 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> > Why does vlen8 need a separate -wrappers file when others don't?
>
> Because only for test built with -mavx512f some avx512 instruction was
> emitted before architecture check.
> Is it needed split AVX2 test to prevent changes that may appear in the future?
That suggests to me that *all* these tests should be split that way - the
main program built without special options, the calls to the functions
built with the special options. That would be the safest
architecture-independent approach - have the separate wrappers file in
every case, even if some cases don't actually need it in your x86 tests
right now.
--
Joseph S. Myers
joseph@codesourcery.com