Re: [PATCH] PPC64: First in the series of patches implementing POWER8 vector math.

[GT <tnggil at protonmail dot com>]

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Tuesday, February 19, 2019 10:54 PM, Joseph Myers <joseph@codesourcery.com> wrote:

> On Tue, 19 Feb 2019, GT wrote:
>
> > GCC fails if an attempt is made to compile code that uses Altivec
> > functionality without passing the -maltivec flag. Additionally it also
> > fails on attempts to build for pre-POWER7 hardware even when
> > -maltivec flag is given to the compiler. Is there need to add a
> > check in configure for attempts to build libmvec on non-Altivec, non-VSX
> > systems; or let the failure happen when make calls GCC?
>
> It's best to detect the issue at configure time. That's definitely needed
> if you want to enable libmvec by default as on x86_64.
>

I've had trouble modifying configure.ac to detect attempts at building
libmvec on machines without Altivec/VSX. But perhaps the problem isn't
with changes I've made:

After making changes to sysdeps/powerpc/powerpc64/configure.ac I changed
directory to the root of the glibc source. Then I ran autoconf with no
arguments, as instructed in the manual. Checking the timestamps,
configure was not regenerated. I forced regeneration by running
'autoconf -f' and this time configure was regenerated at the root of
glibc source. But sysdeps/powerpc/powerpc64/configure was not
re-created. Running autoconf -v showed lines that begin with
'autom4te: forbidden tokens: .....'.

Just to be sure it isn't something I introduced, I cloned a new glibc
source tree and the only command run there thus far was autoconf -f -v.
The same lines appear noting forbidden tokens.

Is there an issue generating the configure script or are those messages
about forbidden tokens harmless?

And running autoconf at the glibc source-tree root directory should
regenerate configure scripts in all subdirectories which have
configure.ac in them, right?


> I'm not sure why it would fail for pre-POWER7, if you pass -maltivec for
> the AltiVec versions and -mvsx for the versions requiring VSX.
>

Turns out POWER7 is the earliest POWERn that implements v2.06 of the ISA.
v2.06 is when type vector double was introduced. So the behavior is
as expected.