charm AT lists.siebelschool.illinois.edu
Subject: Charm++ parallel programming system
List archive
- From: Tom Quinn <trq AT astro.washington.edu>
- To: charm AT cs.illinois.edu
- Subject: [charm] ibverbs smp dies in IsomallocInit()
- Date: Wed, 11 Jul 2012 10:41:10 -0700 (PDT)
- List-archive: <http://lists.cs.uiuc.edu/pipermail/charm>
- List-id: CHARM parallel programming system <charm.cs.uiuc.edu>
When trying to run SMP ibverbs on a 12 node x 16 core/node partition (192 cores total), I get the dreaded:
Charmrun> started all node programs in 4.059 seconds.
Charmrun> IBVERBS version of charmrun
Converse/Charm++ Commit ID: v6.4.0-505-g2c8dc81
Warning> Randomization of stack pointer is turned on in kernel, thread migration may not work! Run 'echo 0 > /proc/sys/kernel/randomize_va_space' as root to disable it, or try run with '+isomalloc_sync'.
Charm++> scheduler running in netpoll mode.
Charmrun: error on request socket--
Socket closed before recv.
After putting a few debugging statements in convcore.c, it appears that it is dying during IsomallocInit() on one of the nodes. As a work-around, I can use the "+noisomalloc" flag, and the program proceeds as normal.
Tom Quinn Astronomy, University of Washington
Internet:
trq AT astro.washington.edu
Phone: 206-685-9009
- [charm] ibverbs smp dies in IsomallocInit(), Tom Quinn, 07/11/2012
Archive powered by MHonArc 2.6.16.