[Bug target/67278] [4.9/5/6 Regression] ICE: verify_gimple failed on darwin 14.5 x86_64

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67278

--- Comment #9 from rguenther at suse dot de <rguenther at suse dot de> ---
On Wed, 2 Mar 2016, ubizjak at gmail dot com wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67278
> 
> --- Comment #8 from Uroš Bizjak <ubizjak at gmail dot com> ---
> (In reply to rguenther@suse.de from comment #6)
> > On Tue, 1 Mar 2016, ubizjak at gmail dot com wrote:
> > 
> > > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67278
> > > 
> > > --- Comment #5 from Uroš Bizjak <ubizjak at gmail dot com> ---
> > > (In reply to Richard Biener from comment #4)
> > > > CCing Uros for the followup ICE.  Apply
> > > 
> > > Following patch should fix the ICE:
> > > 
> > > --cut here--
> > > Index: config/i386/i386.c
> > > ===================================================================
> > > --- config/i386/i386.c  (revision 233864)
> > > +++ config/i386/i386.c  (working copy)
> > > @@ -7794,6 +7794,10 @@ type_natural_mode (const_tree type, const CUMULATI
> > >         {
> > >           machine_mode innermode = TYPE_MODE (TREE_TYPE (type));
> > > 
> > > +         /* There are no XFmode vector modes.  */
> > > +         if (innermode == XFmode)
> > > +           return mode;
> > > +
> > 
> > Are there TFmode vector modes?  Yes, insn-modes.h says so at least.
> > No V1<float>modes there though, just V1<int>modes.
> 
> Yes, vector TFmodes are present, and they work OK.
> 
> > Will you test this or shall I?
> 
> The patch was bootstrapped and regression tested, together with your patch. So,
> I'd say it is ready to be applied.

Ok, I'll post/commit it with a testcase then.

Richard.