This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH v2 4/6] powerpc64le: Require at least GCC 6.2 for powerpc64le


Joseph Myers <joseph@codesourcery.com> writes:

> On Thu, 22 Jun 2017, Steven Munroe wrote:
>
>> On Wed, 2017-06-21 at 23:27 +0000, Joseph Myers wrote:
>> > I see no problem with this, but as far as I'm concerned it's for powerpc 
>> > maintainers to review.
>> > 
>> I am Ok with this as it is specific to powerpc64le and will not impact
>> the other PPC platforms.
>> 
>> But I would prefer a compile test for __float128 and the require 'q'
>> suffix builtins over s GCC version test. This would allow for clang/llvm
>> compiles (once they complete their float128 support).
>
> The proposed test is for 6.2.  I presume that's because of some fix that 
> wasn't in 6.1 - is it straightforward to make the feature test cover 
> whatever's missing and so reject attempts to build with 6.1?

The current code depend on __builtin_*q that first appeared in GCC 6.2.

I'm changing this patch to check for these built-ins instead and the __float128
type, i.e.:

__float128 a, b, c, d, e;
int i;

__float128
foobar (__float128 x)
{
  a = __builtin_nansq ("0");
  b = __builtin_huge_valq ();
  c = __builtin_infq ();
  d = __builtin_fabsq (x);
  e = __builtin_nanq ("0");
  i = __builtin_signbit (x);
  return __builtin_copysignq (x, x);
}

-- 
Tulio Magno


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]