charm AT lists.cs.illinois.edu

Subject: Charm++ parallel programming system

List archive

[charm] bug in CkReductionMgr::reduceMessages() with CkReduction::set

From: Nicolas Bock <nicolasbock AT gmail.com>
To: "charm AT cs.uiuc.edu" <charm AT cs.uiuc.edu>
Subject: [charm] bug in CkReductionMgr::reduceMessages() with CkReduction::set
Date: Thu, 5 Sep 2013 15:19:37 -0600
List-archive: <http://lists.cs.uiuc.edu/pipermail/charm/>
List-id: CHARM parallel programming system <charm.cs.uiuc.edu>

Hi,

I think I found a bug in CkReductionMgr::reduceMessages() when using the CkReduction::set reducer. The problem is that when reducing up a tree, the first contributions are not wrapped into a setElement struct and subsequent reductions end up sizing the contribution incorrectly and corrupting the final reduction message. I have opened ticket #287 and attached a patch that in my very limited testing fixes the problem. Please have a look and let me know what you think.

Thanks,

nick

[charm] bug in CkReductionMgr::reduceMessages() with CkReduction::set, Nicolas Bock, 09/05/2013