Last updated: $Date: 2009/12/14 14:48:09
(To check for possible updates to this document, please
see http://www.spec.org/mpi2007/Docs/ )
SPEC MPI2007 V2.0 adds the large data set to the SPEC MPI2007 V1.1. This means SPEC MPI2007 will now have two different metric sets, one for smaller MPI systems (16-128 ranks) and one which is larger (64-2048 ranks). Results generated with V2.0 for the SPEC MPIM2007 (medium) suite are comparable to results from V1.1 and V1.0. Results generated with V2.0 for the SPEC MPIL2007 (large) suite are not comparable to SPEC MPIM2007 (medium) suite results, as the large data set results are different from the medium. V2.0 is intended to add the large suite, as well as improve compatibility, stability, documentation and ease of use of the medium suite. Changes are intended to be useful to several kinds of users:
For users of with bigger systems:
For anyone who reads (or produces) a result:
If you have already used SPEC MPI2007 V1.1 or V1.0 and already have configuration files, it is recommended that you read through this document and also consider the document changes-in-v1.1, so as to avoid surprises during use of V2.0. Among the changes that you are likely to be affected by are the addition of the large benchmark suite, the naming of data sets, and the changes to runspec.
Contents
(This table of contents proceeds in rough order of time for a user of the suite: you acquire a platform, ensure that you are familiar with the rules, build the benchmarks, run them, generate reports, and occasionally use utilities and other features.)
1. Runspec - large
2. Data Sets - name changes
1. New Metrics
2. Report Names
Note: links to SPEC MPI2007 documents on this web page assume that you are reading the page from a directory that also contains the other SPEC MPI2007 documents. If by some chance you are reading this web page from a location where the links do not work, try accessing the referenced documents at one of the following locations:
Version 2.0 of the SPEC MPI2007 benchmark suite sees the addition of the large component. The large suite is targetted for use on systems ranging from 16 to 2048 ranks. The suite introduces the test data set ltest and the train data set ltrain. These are smaller running data sets for making sure things are working. The suite also introduces a new ref data set called lref. This larger data set shows scaling improvement for runs ranging from 64 ranks to 2048 ranks. It has not been studied for larger than 2048 rank counts.
The following benchmark changes were made in V2.0:
121.pop2
122.tachyon
126.lammps
128.GAPgeofem
129.tera_tf
132.zeusmp2
137.lu
The following benchmarks were added in V2.0:
125.RAxML
142.dmilc
143.dleslie
145.lGemsFDTD
147.l2wrf2
Runspec: The target large is available for the running of the large suite from the tool runspec.
Data Sets: There are now multiple data sets for each data set class of test, train and ref. The medium data sets are mtest, mtrain and mref. The new large data sets are ltest, ltrain and lref. For reportable runs, it will automatically select mref or lref as the data set depending upon if you select medium or large as the target benchmark suite.
New Metrics: The addition of the new large suite has added new metrics SPECmpiL_base2007 and SPECmpiL_peak2007. Runs made for large will have the large metric names listed in the reports.
Report Names: Report names have the form
<suite>.<nnn>.<workload>.<type>
for example, MPIM2007.003.mref.ps, MPIM2007.003.mref.txt, MPIM2007.003.mref.pdf for the medium suite and MPIL2007.009.lref.ps, MPIL2007.009.lref.txt, MPIL2007.009.lref.pdf for the large suite.
Testing on Windows systems is still minimal. Things are expected to work, but no test systems have been available.
Documentation has been added for the new features mentioned in this document. Most of the changes are linked from the descriptions above.
Copyright © 2010 Standard Performance Evaluation Corporation
All Rights Reserved