On Tue, 2019-04-30 at 21:00 +0200, Pavel Labath wrote:
> On 30/04/2019 20:24, Jonas Devlieghere wrote:
> > Hey Pavel,
> >
> > Apologies if I was too hasty, given the number of XFAILs I assumed the
> > test was a WIP and the missing darwin was an oversight.
> >
> > Here's the output on my machine: https://reviews.llvm.org/P8142
> >
> No worries. Darwin is actually the one place which has this implemented.
> However, judging by your output, that support isn't fully functional
> either (you have %zmm16-32, but not their %ymm and %xmm counterparts).
>
> Michal, given this, I think we should remove the ymm and xmm
> expectations from this test (or leave them commented out with a note),
> so that we at least test the part of this feature that is actually
> working. Otherwise we'll end up with a test that is XFAILed everywhere..
>
Hmm, I think the main point of this was that AVX-512 increases
the register count from 16 to 32, so this test implicitly covers
{x,y}mm{16..32}. Of course, I can split this into separate YMM and XMM
tests for AVX512. I can see how this would also break NetBSD if ZMM
support is introduced given the ptrace() implementation.
--
Best regards,
Michał Górny
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
