Re: [PATCH 4/7] target/arm: Handle writeback in VLDR/VSTR sysreg with no memory access

[Peter Maydell]

On Fri, 18 Jun 2021 at 17:15, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 6/18/21 7:10 AM, Peter Maydell wrote:
> > @@ -633,9 +665,11 @@ static void fp_sysreg_to_memory(DisasContext *s, void 
> > *opaque, TCGv_i32 value)
> >           gen_helper_v8m_stackcheck(cpu_env, addr);
> >       }
> >
> > -    gen_aa32_st_i32(s, value, addr, get_mem_index(s),
> > -                    MO_UL | MO_ALIGN | s->be_data);
> > -    tcg_temp_free_i32(value);
> > +    if (do_access) {
> > +        gen_aa32_st_i32(s, value, addr, get_mem_index(s),
> > +                        MO_UL | MO_ALIGN | s->be_data);
> > +        tcg_temp_free_i32(value);
> > +    }
>
> So, this bit looked funny at first glance -- are we missing the free of 
> value?  But of
> course value == NULL when !do_access.  Which made me wonder if it wouldn't be 
> better to
> just use value == NULL and not add the extra argument?

That was how I thought I'd do it initially, but of course you
can't do that for the fp_memory_to_sysreg callbacks, and I
preferred to be consistent across the two.

-- PMM