This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH 0/6] More vector permutation work
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: uweigand at de dot ibm dot com (Ulrich Weigand)
- Cc: rth at redhat dot com (Richard Henderson), gcc-patches at gcc dot gnu dot org, irar at il dot ibm dot com, dje dot gcc at gmail dot com, davem at davemloft dot net, developer at sandoe-acoustics dot co dot uk
- Date: Thu, 27 Oct 2011 10:05:53 +0200 (CEST)
- Subject: Re: [PATCH 0/6] More vector permutation work
> Richard Henderson wrote:
> > On 10/26/2011 07:30 AM, Ulrich Weigand wrote:
> > > This fails since for u == 4 and mode == V4SFmode it attempts to expand
> > > a V4SFmode shift, which is unsupported.
> > >
> > > Shouldn't this be using the mode of the selector rather than the mode
> > > of the result in any case?
> >
> > Yes, it should use the mode of the selector. And doing just that is
> > enough to fix the bug. But I noticed that the actual results for a
> > constant permutation were much nastier than they ought to be.
> >
> > Try this.
>
> Thanks, this has indeed fixed the build problem. Full test suite is
> still running ...
OK, the test suite has completed with no new regression, except that
every single LTO test is now failing with:
lto1: fatal error: bytecode stream generated with LTO version 2.1 instead of the expected 2.0
This would appear to be some independent problem ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
Ulrich.Weigand@de.ibm.com