Last updated: October 21, 2009 jh/mc/jr/mvw/cp
(To check for possible updates to this document, please see http://www.spec.org/mpi2007/Docs/ )
SPEC MPI2007 is a source code benchmark suite that provides a comparative measure of MPI-parallel, floating point, compute intensive performance. Portability of that source code is one of the chief goals of SPEC MPI2007. SPEC has invested substantial effort to make the benchmarks portable across a wide variety of hardware architectures, operating systems, and compilers. During the development of SPEC MPI2007 testing was done on over many different hardware architectures and operating systems (mainly many variations of Unix and Linux). Most testing was done on 64 bit systems, and limited testing was done on 32 bit systems.
To run and install SPEC MPI2007, you will need:
A computer system running Windows, UNIX or Linux. The benchmark suite includes a toolset. Pre-compiled versions of the toolset are provided. The benchmark suite has been tested on the following systems:
For systems not listed in above, such as earlier or later versions of the above systems, you may find that the tools also work, but SPEC has not tested them. Please see the Portability Notes below.
In addition to the above list, there are several toolsets that are provided as a courtesy. These toolsets are expected to work, but have been tested much less extensively. More importantly, the benchmarks themselves have not been ported to these systems, so if you use one of them, you should be prepared to spend time on the porting effort. Please see the Portability Notes below.
Windows systems that are not based on NT, such as Windows 95, Windows 98, and Windows ME, will definitely NOT work.
A DVD drive.
Alternately, you can install MPI2007 from the compressed tar-file you can get from SPEC/HPG. You will need to verify that it is the Version 2.0 delivery and not some intermediate test kit. Also, the the file space calculations below assume that this file has been deleted once the MPI2007 directory tree is installed.
Memory: typically 1 GB per rank for the medium suite and 2 GB per rank for the large suite. This does not consider OS overhead so more may be required:
The MPI2007 medium benchmarks (code + workload) have been designed to fit within about 1 GB of physical memory per rank. The minimum officially supported memory size is 16 GB. This means that an 8 rank run could require more than 1 GB per rank.
The MPI2007 large benchmarks (code + workload) have been designed to fit within about 2 GB of physical memory per rank. The minimum oficially supported rank count is 64, implying a minimum limit of 128 GB of memory.
The 1 or 2 GB per rank of RAM for the benchmarks does not include space needed for the operating system and other non-SPEC tasks on the system under test.
Attempting to run the suite with insufficient memory is strongly discouraged, as you will end up measuring the speed of your paging file, not the speed of your system running compute intensive workloads.
The disk space requirement will depend on the suite you are running and can vary greatly depending upon your usage and the system you run on:
These estimates are based on the following:
Minimum requirement: It is possible to run with less disk space if: you delete the build directories after the build is done; and you clean run directories between tests. See the discussion of disk space in runspec.html for more information about managing disk space.
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:
Since SPEC supplies only source code for the benchmarks, you will need either:
Please notice that you cannot generate a valid MPI2007 result unless you meet all of requirement 5.a or 5.b. For example, if you are attempting to build the suite but lack a Fortran 95 compiler, you will not be able to measure a SPECmpiM_base2007 result.
Despite SPEC's testing efforts, certain portability problems are likely to arise from time to time. For example:
Some platforms may not have a Fortran 95 compiler available.
Some older compilers may not include all the features needed to run the entire suite.
Sometimes, a new release of a compiler or operating system may introduce new behavior that is incompatible with one of the benchmarks.
If you visit http://www.spec.org/mpi2007/ and look up results for SPEC MPI2007, you will find combinations of OS and compiler versions that are known to work. For example, if a vendor reports a SPECmpiM_base2007 result on the SuperHero 4 using SuperHero Unix V4.0 with SuperHero Fortran v4.0, SuperHero C V4.0 and SuperHero C++ V4.0, you may take that as an assertion by the vendor that the listed versions of Unix, Fortran, C, and C++ will successfully compile and run the SPEC MPI2007 suite on the listed machine.
For systems that have not (yet) been reported by vendors, SPEC can provide limited technical support to resolve portability issues. See techsupport.html for information.
Copyright © 2007-2010 Standard Performance Evaluation Corporation
All Rights Reserved