[PATCHv4 2/2] powerpc64le: ifunc (almost) all *f128 routines in multiarch mode

Joseph Myers joseph@codesourcery.com
Thu Jun 25 18:35:12 GMT 2020

On Thu, 25 Jun 2020, Paul E Murphy via Libc-alpha wrote:

> On 6/24/20 7:00 PM, Joseph Myers wrote:
> > On Wed, 24 Jun 2020, Paul E Murphy via Libc-alpha wrote:
> > 
> > > The design intentionally keeps all of the complexity in one place
> > > hidden, without changes to common code.  Doing each individually is
> > 
> > It's not really hidden, when it means that many new libm functions need
> > architecture-specific changes.  The normal expectation is that any new
> > function (whether using type-generic templates or type-specific
> > implementations, whether real or complex or narrowing-real, and whatever
> > internal functions might be used in the implementation) does not need
> > architecture-specific changes beyond updating the ABI baselines.
> > 
> Can I interpret that as saying an opt-in approach to generating ifunc'ed
> math functions would make a patch like this acceptable?

I don't know what that opt-in approach might look like, or whether it 
would address concerns other people have expressed in this discussion.

Joseph S. Myers

More information about the Libc-alpha mailing list