Re: [PATCH 15/33] target/ppc: Implement Vector Insert from GPR using GPR index insns

[BALATON Zoltan]

On Fri, 22 Oct 2021, Richard Henderson wrote:
> On 10/21/21 12:45 PM, matheus.ferst@eldorado.org.br wrote:
> > +#if defined(TARGET_PPC64)
> > +    return do_vinsx(ctx, a->vrt, size, right, cpu_gpr[a->vra], 
> > cpu_gpr[a->vrb],
> > +                    gen_helper);
> > +#else
> > +    bool ok;
> > +    TCGv_i64 val;
> > +
> > +    val = tcg_temp_new_i64();
> > +    tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]);
> > +
> > +    ok = do_vinsx(ctx, a->vrt, size, right, cpu_gpr[a->vra], val, 
> > gen_helper);
> > +
> > +    tcg_temp_free_i64(val);
> > +    return ok;
> > +#endif
>
> Oh, and what's all this?
>
> Either this isn't defined for !PPC64 at all, or you should just use 
> target_ulong and not do any ifdeffing at all.

You mentioning target_ulong reminded me a question I had. Currently we 
have qemu-system-ppc and qemu-system-ppc64 but the latter includes all 
machines of the former too so you could run for example sam460ex with 
qemu-system-ppc64 (except mac99 which behaves differently based on which 
executable it's part of but you could use mac99 -cpu G4 with 
qemu-system-ppc64 as well). But isn't target_ulong different in these 
executables and could that cause a problem with this? I've always used 
qemu-system-ppc for 32 bit machines but we could have one just executable 
for all machines if there's no need for both.

Regards,
BALATON Zoltan