[Bug math/20024] [x86_64] vectorized sincos trashes the stack

andrew.n.senkevich at gmail dot com sourceware-bugzilla@sourceware.org
Thu May 26 18:12:00 GMT 2016


https://sourceware.org/bugzilla/show_bug.cgi?id=20024

--- Comment #7 from Andrew Senkevich <andrew.n.senkevich at gmail dot com> ---
(In reply to joseph@codesourcery.com from comment #6)
> Changing exported symbols on release branches is a bad idea.  The 
> GLIBC_2.22 symbols should remain (and I suppose be fixed so they work).  
> The intended, more efficient symbols with a different name and ABI should 
> be added at version GLIBC_2.24 (and then the old ones should become compat 
> symbols so new links can't use them).

Due to GCC bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71300 
AVX version of void sincos(double, double*, double*); 
declared with 
#pragma omp declare simd notinbranch
will break testsuite with segfault of math/test-double-vlen4 test.

Another words sincos AVX vvv implementation will always cause segfault if its
call created with not fixed GCC.

So it seems better to fix GLIBC_2.22 and GLIBC_2.23 ABI.  Joseph, what do you
think?

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


More information about the Glibc-bugs mailing list