On Tue, Jul 31, 2018 at 09:02:56AM +0100, Kyrill Tkachov wrote: > Hi Segher, > > On 30/07/18 18:37, Segher Boessenkool wrote: > >On Mon, Jul 30, 2018 at 03:55:30PM +0100, Kyrill Tkachov wrote: > >>Hi Segher, > >> > >>On 30/07/18 14:14, Segher Boessenkool wrote: > >>>In arm_block_set_aligned_vect 8-bit constants are generated as zero- > >>>extended const_ints, not sign-extended as required. Fix that. > >>> > >>>Tamar tested the patch (see PR); no problems were found. Is this okay > >>>for trunk? > >>> > >>The patch is okay but please add the testcase from the PR to gcc.dg/ > >>or somewhere else generic (it's not arm-specific). > >It only failed with very specific options, which aren't valid on every > >ARM config either I think? > > > >-O3 -mfpu=neon -mfloat-abi=hard -march=armv7-a > > This is a fairly common Arm configuration that is tested by many people by > default, > so you wouldn't need to add any arm-specific directives. Just putting it in > one of > the generic C tests folders would be enough.
Ah ok, so just dg-options -O3 will do? (As before, I have no reasonable way to test this). Segher
