Re: [PATCH 22/33] target/mips: Convert MSA 3R instruction format to decodetree (part 1/4)

[Richard Henderson]

On 10/24/21 11:57 AM, Philippe Mathieu-Daudé wrote:
> On 10/24/21 19:52, Richard Henderson wrote:
> > On 10/23/21 2:47 PM, Philippe Mathieu-Daudé wrote:
> > > +static bool trans_msa_3r_df(DisasContext *ctx, arg_msa_r *a,
> > > +                            void (*gen_msa_3r_df)(TCGv_ptr, TCGv_i32,
> > > TCGv_i32,
> > > +                                                  TCGv_i32, TCGv_i32))
> > > +{
> > > +    TCGv_i32 tdf = tcg_constant_i32(a->df);
> > > +    TCGv_i32 twd = tcg_const_i32(a->wd);
> > > +    TCGv_i32 tws = tcg_const_i32(a->ws);
> > > +    TCGv_i32 twt = tcg_const_i32(a->wt);
> > > +
> > > +    gen_msa_3r_df(cpu_env, tdf, twd, tws, twt);
> >
> > Missing check_msa_access.
>
> It is included in TRANS_MSA(). How could I make that clearer?
>
> The style I follow is:
> - trans_UPPER() -> direct implementation
> - trans_lower() -> called via TRANS_xxx() macro

Hmm.  I dunno.  Especially since some of the cases require checking for the null function 
pointer *before* the access check.

How many trans helpers do you wind up with in the end?
Perhaps drop TRANS_MSA entirely?


r~