Re: [PATCH v8 16/35] Hexagon (target/hexagon/conv_emu.[ch]) utility functions

[Richard Henderson]

On 3/17/21 9:57 PM, Taylor Simpson wrote:
> > -----Original Message-----
> > From: Richard Henderson <richard.henderson@linaro.org>
> > Sent: Sunday, February 14, 2021 2:57 PM
> > To: Taylor Simpson <tsimpson@quicinc.com>; qemu-devel@nongnu.org
> > Cc: philmd@redhat.com; alex.bennee@linaro.org; laurent@vivier.eu;
> > ale@rev.ng; Brian Cain <bcain@quicinc.com>
> > Subject: Re: [PATCH v8 16/35] Hexagon (target/hexagon/conv_emu.[ch])
> > utility functions
> >
> > On 2/7/21 9:46 PM, Taylor Simpson wrote:
> > > +uint64_t conv_sf_to_8u(float32 in, float_status *fp_status);
> > > +uint32_t conv_sf_to_4u(float32 in, float_status *fp_status);
> > > +int64_t conv_sf_to_8s(float32 in, float_status *fp_status);
> > > +int32_t conv_sf_to_4s(float32 in, float_status *fp_status);
> > > +
> > > +uint64_t conv_df_to_8u(float64 in, float_status *fp_status);
> > > +uint32_t conv_df_to_4u(float64 in, float_status *fp_status);
> > > +int64_t conv_df_to_8s(float64 in, float_status *fp_status);
> > > +int32_t conv_df_to_4s(float64 in, float_status *fp_status);
> >
> > You need to either use the normal float conversion routines, or document
> > what the differences are.
>
> There are some differences in floating point flags raised, so I could write 
> something like this:
>      if (float32_is_infinity(RsV)) {
>          float_raise(float_flag_invalid, &env->fp_status);
>          if (float32_is_neg(RsV)) {
>              RddV = 0ULL;
>          } else {
>              RddV = ~0ULL;
>          }

This isn't different from softfloat.c.

>      } else if (float32_is_any_nan(RsV)) {
>          float_raise(float_flag_invalid, &env->fp_status);
>          RddV = ~0ULL;

Nor is this.

>      } else if (float32_is_zero(RsV)) {
>          RddV = 0;

Not exactly a special case.

>      } else if (float32_is_neg(RsV)) {
>          float_raise(float_flag_invalid, &env->fp_status);
>          RddV = 0;

Not different.

>      } else {
>          RddV = float32_to_uint64_round_to_zero(RsV, &env->fp_status);
>      }
>
> Does that work?

This is 100% identical with round_to_uint_and_pack as used by 
float32_to_uint64_round_to_zero.  It's all straight IEEE 754.


r~