> I'm surprised that this worked! > > It looks like you unconditionally add the -march=armv8.2-a+dotprod options, > which should cause you to generate instructions which will not execute on > targets which don't support this instruction. As far as I can see, this is an > execute test, so that should cause undefined instruction exceptions on an > Armv8-A target at the very least.
It's not, there is no dg-do specified, which means it defaults to "compile" This is a straight compilation tests that checks to see if the target can do the reduction. There may be a main, but it's never executed, which is why I don't have a hardware check against it. The unconditional armv8.2+dotprod is for this reason. It doesn't matter what hardware. > > So, not OK in its current form. > > Thanks, > James > > > > > Ok for trunk? > > > > gcc/testsuite > > 2017-09-01 Tamar Christina <tamar.christ...@arm.com> > > > > * gcc.dg/vect/vect-reduc-dot-s8a.c > > (dg-additional-options, dg-require-effective-target): Add +dotprod. > > * gcc.dg/vect/vect-reduc-dot-u8a.c > > (dg-additional-options, dg-require-effective-target): Add +dotprod. > > > > --
