Re: [PATCH 9/9] target/s390x: Use Int128 for passing float128

[Richard Henderson]

On 10/25/22 04:01, Philippe Mathieu-Daudé wrote:
> On 21/10/22 09:30, Richard Henderson wrote:
> > Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> > ---
> >   target/s390x/helper.h          | 32 ++++++-------
> >   target/s390x/tcg/fpu_helper.c  | 88 ++++++++++++++--------------------
> >   target/s390x/tcg/translate.c   | 76 ++++++++++++++++++++---------
> >   target/s390x/tcg/insn-data.def | 30 ++++++------
> >   4 files changed, 121 insertions(+), 105 deletions(-)
>
> > diff --git a/target/s390x/tcg/translate.c b/target/s390x/tcg/translate.c
> > index d1ffbb8710..8023bbab2f 100644
> > --- a/target/s390x/tcg/translate.c
> > +++ b/target/s390x/tcg/translate.c
> > @@ -305,6 +305,18 @@ static TCGv_i64 load_freg32_i64(int reg)
> >       return r;
> >   }
> > +static TCGv_i128 load_freg_128(int reg)
> > +{
> > +    TCGv_i64 h = load_freg(reg);
> > +    TCGv_i64 l = load_freg(reg + 2);
> > +    TCGv_i128 r = tcg_temp_new_i128();
>
> Maybe rename as load_freg_new_128() to make emphasis on the returned
> TCGv need to be freed?

It's no different from the other load_freg* functions just above.  As with those, the 
result is assigned to one of the DisasOps slots, and all of those slots are freed at the 
end of each instruction.

r~