https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118407

--- Comment #12 from Tamar Christina <tnfchris at gcc dot gnu.org> ---
(In reply to Torbjorn SVENSSON from comment #11)
> (In reply to Tamar Christina from comment #10)
> > Any fix would be too big to backport. So I think xfail is the best approach.
> 
> Then the hard thing is to identify when it's suppose to be xfail and when
> not to.
> In r17-2366-gcfc5aa4ed975c4 I see fail for:
> 
> arm/arch=armv7ve/cpu=cortex-a7/float-abi=hard/fpu=neon
> arm/arch=armv7ve+neon/tune=cortex-a7/float-abi=hard/fpu=auto
> thumb/arch=armv7ve/cpu=cortex-a7/float-abi=hard/fpu=neon
> thumb/arch=armv7ve+neon/tune=cortex-a7/float-abi=hard/fpu=auto
> 
> I tried using vect_early_break, but there is nothing for armv7 in that, so
> the test still xpass/fail.
> Any idea?

could you attach the output of -fdump-tree-vect-all for one of the failing
variants?

Reply via email to