On Mon, Aug 15, 2022 at 3:04 PM Niels Möller <ni...@lysator.liu.se> wrote:
> ni...@lysator.liu.se (Niels Möller) writes:
>
> > Do I understand that code correctly, that qemu never reads the third
> > "vrc" input to the instruction?
>
> Actually, it appears latest qemu is doing this correctly (it's just that
> for some reason, operands are named a->vra, a->vrb, a->rc (sic!), a->vrt
> in the qemu sources, which had me a bit confused). I've tried building
> qemu from source, which was a very easy, and then the tests pass.
>
> I've also checked the sources of the version I have in debian
> ("qemu-ppc64le version 5.2.0 (Debian 1:5.2+dfsg-11+deb11u2)"), and it
> appears it has no support for the vmsumudm instruction at all. The
> confusing part is that (i) the tests fail with incorrect results, rather
> than some kind of unimplemented instruction exception, and (ii) support
> is still advertised so that it is picked up by the fat build logic.
>
It's just a weird behavior, turning unimplemented instructions to no-op or
having an output of zero were the least of it to make it less confusing.
A proper implementation is in qemu-7, currently in debian testing.
>
Great news! I think we would keep the CI workaround enabled until debian
has official release of qemu-7.
regards,
Mamone
Regards,
> /Niels
>
> --
> Niels Möller. PGP key CB4962D070D77D7FCB8BA36271D8F1FF368C6677.
> Internet email is subject to wholesale government surveillance.
>
_______________________________________________
nettle-bugs mailing list -- nettle-bugs@lists.lysator.liu.se
To unsubscribe send an email to nettle-bugs-le...@lists.lysator.liu.se
