New loop flattening pass on tree-ssa

Submitted by Ira Rosen on Oct. 27, 2010, 8:39 p.m.

Details

Message ID OFAFC35A7A.7F05A24D-ONC22577C9.0070D038-C22577C9.00717064@il.ibm.com
State New
Headers show

Commit Message

Ira Rosen Oct. 27, 2010, 8:39 p.m.
Sebastian Pop <sebpop@gmail.com> wrote on 27/10/2010 09:41:43 PM:

> prefetch and iv-opt could be done before loop flattening.
> I don't know if we could move these two passes before SLP-vect.
> Any ideas here Richi? Ira?

I tried to move SLP after iv-opt:


and nothing bad happened to SLP tests. So it seems to be ok from the SLP
point of view.

Ira

>
> Thanks,
> Sebastian
> [attachment "0001-Call-if-conversion-from-loop-flattening.patch"
> deleted by Ira Rosen/Haifa/IBM]

Patch hide | download patch | download mbox

Index: passes.c
===================================================================
--- passes.c    (revision 165969)
+++ passes.c    (working copy)
@@ -913,10 +913,10 @@ 
            }
           NEXT_PASS (pass_predcom);
          NEXT_PASS (pass_complete_unroll);
-         NEXT_PASS (pass_slp_vectorize);
          NEXT_PASS (pass_parallelize_loops);
          NEXT_PASS (pass_loop_prefetch);
          NEXT_PASS (pass_iv_optimize);
+         NEXT_PASS (pass_slp_vectorize);
          NEXT_PASS (pass_tree_loop_done);
        }
       NEXT_PASS (pass_cse_reciprocals);