Re: [RFC/PATCH] Remove the workaround for _Float128 precision [PR107299]

[Segher Boessenkool <segher@kernel.crashing.org>]

Hi!

On Wed, Dec 21, 2022 at 09:40:24PM +0000, Joseph Myers wrote:
> On Wed, 21 Dec 2022, Segher Boessenkool wrote:
> > > --- a/gcc/tree.cc
> > > +++ b/gcc/tree.cc
> > > @@ -9442,15 +9442,6 @@ build_common_tree_nodes (bool signed_char)
> > >        if (!targetm.floatn_mode (n, extended).exists (&mode))
> > >  	continue;
> > >        int precision = GET_MODE_PRECISION (mode);
> > > -      /* Work around the rs6000 KFmode having precision 113 not
> > > -	 128.  */
> > 
> > It has precision 126 now fwiw.
> > 
> > Joseph: what do you think about this patch?  Is the workaround it
> > removes still useful in any way, do we need to do that some other way if
> > we remove this?

You didn't address these questions.  We don't see negative effects from
removing this workaround, but it isn't clear (to me) what problems were
there that caused you to do this workaround.  Do you remember maybe?  Or
can we just delete it and try to forget such worries :-)

> I think it's best for the TYPE_PRECISION, for any type with the binary128 
> format, to be 128 (not 126).

Well, but why?  Of course it looks nicer, and it is a gross workaround
to have different precisions for the different 128-bit FP modes, more so
if two modes are really the same, but in none of the ways floating point
precision is defined would it be 128 for any 128-bit mode.

> It's necessary that _Float128, _Float64x and long double all have the same 
> TYPE_PRECISION when they have the same (binary128) format,

Yes, agreed.  Or even if it would be not necessary it is the only sane
thing to do.


Segher