Re: [gomp4] Some progress on #pragma omp simd

[Aldy Hernandez <aldyh at redhat dot com>]

On 06/12/13 16:36, Jakub Jelinek wrote:
> On Wed, Jun 12, 2013 at 10:38:00AM -0700, Richard Henderson wrote:
> > On 06/12/2013 10:30 AM, Jakub Jelinek wrote:
> > > So the built-ins would take address of this decl, something else?
> >
> > Perhaps address, perhaps just referenced uninitialized?
>
> True, assuming no pass would actually want to change that SSA_NAME of the
> magic decl just because it is undefined (coalesce with some other undefined
> SSA_NAME or something similar).  I hope nothing does that, it would be
> problematic for the uninitialized warning pass too I bet.

Gentlemen:

Attached is a lightly tested patch rewriting the aforementioned UID to 
use the address of a dummy DECL (see the lower_rec_input_clauses fragment).

I first tried an uninitialized reference but into-SSA happens after 
lower_rec_input_clauses, and the DECL gets rewritten into SSA, while 
loop->simduid maintains the original DECL.  This causes all the equality 
checks throughout to fail.  I thought of changing the SSA pass to update 
loop->simduid as well, but that seemed kludgy.  I didn't think about it 
too hard, perhaps there's a more elegant way.

Anyway...this at least works with my contrived reduction tests for Cilk 
Plus pragma simd.

If this is what you want, I can start looking at how this behaves with 
LTO and inlining.

Aldy

Attachment: curr
Description: Text document