Re: [PATCH 10/10] x86: correct VFPCLASSP{S,D} operand size handling

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Mon, Aug 12, 2019 at 1:25 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 09.08.2019 19:13,  H.J. Lu  wrote:
> > On Fri, Aug 9, 2019 at 8:58 AM Jan Beulich <jbeulich@suse.com> wrote:
> >> On 09.08.2019 16:57,  H.J. Lu  wrote:
> >>> When I see
> >>>
> >>> 22   vfpclasspd $0, (%eax), %k0
> >>>
> >>> I can't tell what the memory size is.
> >>
> >> Excuse me, but when you go through source code it is assumed
> >> that you know what your source code means and does. No-one
> >
> > The assembly source code can come from anywhere.
> >
> >> requires you to omit the suffix. But equally no-one should be
> >> required to specify a suffix just to meet your taste.
> >
> > These instructions can take different memory sizes.  One should
> > be able to tell what the memory size is by looking at it. Is that too
> > much to ask?
>
> This is not too much to ask, but is a decision of the programmers
> writing assembly code, not something to enforce by gas. Once
> again - given the context things may be entirely unambiguous. I
> can only re-emphasize that as a maintainer you should weigh your
> personal preferences against what others may think, want, or need.

I think a suffix should be required if the memory operand size
isn't fixed by default.  One shouldn't need to know how assembly
codes is assembled to tell the memory operand size.

> >> Let me be frank here: If you continue to refuse to allow this
> >> change in, I'll have to make it work correctly for Intel syntax
> >> mode only (which requires more code for no gain), just to avoid
> >> the need to have you ack the change. I don't think though that
> >> this would be a good course of action.
>
> I find it quite interesting that you didn't even care to comment
> on this part.

That is entirely up to you.


-- 
H.J.