Re: [PATCH] v3: Use gcc __builtin_stdc_* builtins in stdbit.h if possible

[Jakub Jelinek <jakub@redhat.com>]

On Mon, Jan 29, 2024 at 09:07:08PM +0000, Joseph Myers wrote:
> On Fri, 26 Jan 2024, Jakub Jelinek wrote:
> 
> > So like this then (only stdbit.texi changed since last patch)?
> 
> Yes.
> 
> > +ui
> > +leading_zeros (uc a, us b, ui c, ul d, ull e)
> > +{
> > +  return (stdc_leading_zeros (a)
> > +	  + stdc_leading_zeros (b)
> > +	  + stdc_leading_zeros (c)
> > +	  + stdc_leading_zeros (d)
> > +	  + stdc_leading_zeros (e));
> > +}
> 
> I don't see what these functions are for.  They don't seem to be used at 
> all, and the existing tests thoroughly cover the type-generic macros as 
> applied to non-_BitInt types, so additional tests that only verify that 
> such calls can be compiled for non-_BitInt types don't seem to add much 
> value.

Most of the other tests (except the tests for side-effects in arguments)
have constant arguments and so can be folded very early during compilation,
even the side-effects in arguments tests can be folded soon when optimizing,
so these functions were just an attempt to make sure the macros can be
compiled/linked when the arguments are certainly not known.
But if you think it isn't worth checking that or if it is already tested by
some other test, I can remove that.
Another option is to duplicate the test and instead of using the macros
directly use it wrapped into small [[gnu::noinline]] functions.

	Jakub