From: Xiaoge Wang (wangx147_at_MSU.EDU)
Date: Tue Jul 28 2015 - 12:50:23 CDT
Hi,
I am trying to run NAMD on Xeon Phi cluster to test for the performance advantages.
The machine has 28 nodes accelerated with two Intel Xen Phi 5110P, two 2.5Ghz 10-core Inel Xero E5-2670v2 and 128GB of ram. I down load the pre-built Linux-x86_64-multicore-MIC <http://www.ks.uiuc.edu/Development/Download/download.cgi?UserID=&AccessCode=&ArchiveID=1323> and run the benchmark configure apoa1.namd.
My questions are:
(1) Is there any documents/instructions on how to run on Xeon Phi cluster?
(2) Which part of the computation is offloaded to the coprocessors? How the load distributed? How to use command line parameters (such as +p, +ppn, pemap, commap etc.) to manage the workload distribution for reasonable performance?
Thank you very much in advance for your help.
Xiaoge
Xiaoge Wang
Research Specialist
Institute for Cyber-Enabled Research
Michigan State University
567 Wilson Road, Room 1440
East Lansing, MI 48824
Phone: 517-884-9472 <tel:517-884-9472>
wangx147_at_msu.edu <mailto:wangx147_at_msu.edu>
iCER.msu.edu <http://icer.msu.edu/>
This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:15 CST