g
On Tue, Jun 25, 2019 at 05:56:42PM +0200, Richard Henderson wrote: > On 6/25/19 5:37 PM, Mark Cave-Ayland wrote: > > The problem is that in tcg/tcg-op.h we define "DEF(dup2_vec, 1, 2, 0, > > IMPLVEC | > > IMPL(TCG_TARGET_REG_BITS == 32))" and in the last patchset dup2_vec isn't > > introduced > > until towards the end. Unfortunately it's not a simple as bringing the > > patch forward > > within the series to maintain bisectability because the current > > implementation > > depends on VMRG which only appears in the patch just before it... > > Ah, that would explain it. I admit I haven't looked at v5 that closely. > > > Next to try and figure out what exactly is causing the fault. Just a quick > > question > > out of curiosity: is your Power9 system BE or LE? > > The Power9 is LE. It's the kernel determines endianness, not the system. > > I do have access to a Power7 BE system, and that worked last time I checked. > I'll try that again tomorrow to be sure. > > > r~ > -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
signature.asc
Description: PGP signature