Re: [PATCH for-8.0 v2] target/ppc: Fix temp usage in gen_op_arith_modw

[Nicholas Piggin]

On Sun Apr 9, 2023 at 7:24 AM AEST, Cédric Le Goater wrote:
> On 4/8/23 09:05, Richard Henderson wrote:
> > Fix a crash writing to 't3', which is now a constant.
> > Instead, write the result of the remu to 't1'.
> > 
> > Fixes: 7058ff5231a ("target/ppc: Avoid tcg_const_* in translate.c")
> > Reported-by: Nicholas Piggin <npiggin@gmail.com>
> > Reviewed-by: Anton Johansson <anjo@rev.ng>
> > Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>
> Looks good:
>
>    https://gitlab.com/legoater/qemu/-/pipelines/831847446
>
> I have a PR ready for this same branch. If you want to me send,
> just tell.

Thanks Richard and Cedric. LGTM.

> I don't think we have tcg tests for the prefix or mma instructions
> introduced in P10. That would be good to have.

I agree, we need to do a bit better on ppc.

I'm trying to get a handle on all the tests we have for these things,
I haven't looked too closely before. kvm-unit-tests actually works
well for TCG and I did find some (system level) prefix issues with it.
I don't know if that's the right place to focus on instruction level
testing though. QEMU's tcg tests sounds like a better place for it,
but is it only for userspace tests? There are also some verification
tests people are using for verifying hardware cores.

Seems like a common upstream project that others can use might be
useful.

Thanks,
Nick