Re: [PATCH 13/13] Use GCC builtins for copysign functions if desired.

[Stefan Liebler <stli at linux dot ibm dot com>]

On 12/2/19 10:00 PM, Joseph Myers wrote:
> On Mon, 2 Dec 2019, Stefan Liebler wrote:
>
> > This patch is using the corresponding GCC builtin for copysignf, copysign,
> > copysignl and copysignf128 if the USE_FUNCTION_BUILTIN macros are defined to one
> > in math-use-builtins.h.
>
> I believe this is always safe for these implementations (the only case
> where GCC might not expand copysign functions inline is copysignl for IBM
> long double, in the soft-float case).
Thus you mean we can do the following preset in 
sysdeps/generic/math-use-builtins.h?
#define USE_COPYSIGN_BUILTIN 1
#define USE_COPYSIGNF_BUILTIN 1
#define USE_COPYSIGNL_BUILTIN 0
#define USE_COPYSIGNF128_BUILTIN 0

Or even also set USE_COPYSIGNL_BUILTIN to one as IBM long double has its 
own implementation in ./sysdeps/ieee754/ldbl-128ibm/s_copysignl.c.

On the other hand, I'm not able to run tests on the different 
architectures. Thus the safest way would be to leave all those macros 
set to zero and let the architectures decide.

In each case, I would prefer to set those macros to one in a separate 
patch. Then this patch could be reverted in case of failing on one 
architecture.