Re: [PATCH v3 11/11] RFC: powerpc64le: Enable support for IEEE long double

[Paul E Murphy <murphyp at linux dot ibm dot com>]

On 12/13/19 4:41 PM, Gabriel F. T. Gomes wrote:
> On Mon, 09 Dec 2019, Paul E Murphy wrote:

> > Ignoring the individual files for a moment, is it appropriate to split
> > this into one or more patches to introduce the
> > __LONG_DOUBLE_USE_FLOAT128 usage into the various user headers?
>
> Since __LONG_DOUBLE_USES_FLOAT128 is only defined to 1 in
> ldbl-ibm128-compat's bits/long-double.h, I suppose this could work.  It's
> always defined to 0 by all other bits/long-double.h headers (as suggested
> by Florian [1]), including what's currently installed for powerpc64le.
>
> Let me check if I got your suggestion right.  Are you suggesting that all
> the changes to installed headers in this patch be split into a separate
> patch that can be integrated to master?  Then, this patch (the last one)
> would only touch the Implies, .abilist, and ldbl-128ibm-compat-abi.h
> files...
>
> I.e.:
>
> > >    sysdeps/powerpc/powerpc64/le/Implies-before   |   1 +
> > >    .../powerpc64/le/ldbl-128ibm-compat-abi.h     |   8 ++
> > >    .../linux/powerpc/powerpc64/le/libc.abilist   |  93 ++++++++++++++++
> > >    .../linux/powerpc/powerpc64/le/libm.abilist   | 104 ++++++++++++++++++
>
> ^ these?

Exactly.  They can remain inert until the remaining support arrives and 
is enabled, while getting the benefit of any ancillary refactoring 
occuring while the remaining libm work is finalized.  Though, that is my 
opinion and may not represent a consensus.