Hi, on 2024/6/14 11:58, Peter Bergner wrote: > On 6/13/24 9:34 PM, Kewen.Lin wrote: >> on 2024/6/14 05:16, Carl Love wrote: > >>> /* { dg-options "-mvsx" } */ >>> /* { dg-additional-options "-mdejagnu-cpu=power8" { target { ! >>> has_arch_pwr8 } } } */ > > With the above, we're going to compile and run this test case with > -mcpu=power8 > or higher, which means we could have P8, P9 or even P10 instructions emitted. > > > >>> /* { dg-require-effective-target powerpc_vsx } */ >> >> Since you changed this for "run", I think you also want >> s/powerpc_vsx/vsx_hw/ . > > ...which means we'd need p8vector_hw, p9vector_hw or ... here.
Ah, good catch! Yes, it would require some harder guard. > > > Should we just always compile with -mcpu=power8 and then check for p8vector_hw > to make our lives easier? Ala... > > > /* { dg-options "-mdejagnu-cpu=power8" } */ > ... > /* { dg-require-effective-target p8vector_hw } */ > > > Note I've removed -mvsx, since that is implied by -mcpu=power8 and no > need for dg-additional-options. Maybe we want to add -O2 as well? > Thoughts? Both sounds reasonable to me, it looks useless to distinguish p8 or p8-up for this test case. BR, Kewen