[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[bgl-discuss] allreduce performance bug in coprocessor mode

To: discuss@xxxxxxxxxxxxxxx
Subject: [bgl-discuss] allreduce performance bug in coprocessor mode
From: Pete Beckman <beckman@xxxxxxxxxxx>
Date: Thu, 27 Oct 2005 07:50:39 -0500
Delivered-to: discuss-outgoing@bgl.mcs.anl.gov
Delivered-to: discuss@bgl.mcs.anl.gov
Sender: owner-discuss@xxxxxxxxxxxxxxx

While running some benchmarks, we found out that MPI_Allreduce has a serious performance bug in coprocessor mode on our machine (driver 202). Allreduce for virtual node mode is 100x faster (for twice the number of processors). Allreduce for co-processor mode is slow.

The bug is fixed in the next IBM driver release, but in the mean time, realize allreduce will be very slow in coprocessor mode.

Also, shows our need for performance regression testing of MPI...

-Pete

- --------------------------------------------------------------------
To add or remove yourself from this mailing list, use the 'notifyme'
command on any BGL machine. To remove: notifyme -n, to add: notifyme -y.

Prev by Date: Re: [bgl-discuss] building flash i/o kernel on BGL
Next by Date: [bgl-discuss] allreduce performance bug in coprocessor mode
Previous by thread: Re: [bgl-discuss] building flash i/o kernel on BGL
Next by thread: [bgl-discuss] allreduce performance bug in coprocessor mode
Index(es):
- Date
- Thread