PARKBENCH 99 - Projects

Home | Projects | Results Database | Publications | Related Links


  This page  contains links to benchmarking and other performance related projects at the University of Southampton. These include the new Genesis codes, a parallel I/O benchmark suite and new tools that have been developed for post-processing PARKBENCH 99 result files.
 

GENESIS - New Low-level Benchmarks

During 1998 new Lowlevel Genesis benchmarks were developed by David Lancaster and the results collected into a new DataBase. The new benchmarks take only timing measurements and do not calculate any performance parameters. It is left to the users of the database to do what analysis of the results that they wish. To aid in this Roger Hockney has written Java Applets that calculate and plot the two- and three-parameter curve fits to the selected benchmarks and display the associated performance parameters. A version  of the interactive PICT tool (see below) is also provided for refining the fits.
 

PICT2: Parkbench Interactive Curve-Fitting Tool

PICT2: Parkbench Interactive Curve-Fitting Tool
 

Parallel Input/Output, MPI-2

A Collaborative Project Involving Fujitsu European Centre for Information Technology (FECIT), the European Centre for Medium Range Weather Forecasting (ECMWF) and the Department of Electronics and Computer Science at the University of Southampton. This project ran from Feb 1997 until Feb 1998.

The technology of parallel processing has matured to the stage where for many computationally intensive problems it is the method of choice. Techniques of parallel I/O have not however kept in step with these developments and applications may soon become I/O bound. An important development in this field was the publication of the MPI-2 standard which contains a chapter on I/O (referred to as MPI-I/O) and presently this is the best way of portably implementing parallel I/O at the application programming level. MPI-1 has been widely taken up in the community and provided MPI-2 is similarly adopted it will form an important basis for parallel I/O.

With this situation in mind, and for the purposes of this project, the ECMWF IFS weather code was ported to use MPI-I/O on the Fujitsu VP700. In order to test the resulting I/O performance Southampton and FECIT developed a suite of tools that provide timing and instrumentation measurements. The aim of the project was to use these tools to analyse the I/O performance in detail: at the system, MPI implementation and application levels. These tools also make it possible, for the first time, to make comparisons of parallel I/O between widely dissimilar systems.

Download

The benchmark suite along with specification and user guide are available for download.

  • Benchmark Suite (as a gzipped tar file, 40kbytes).
  • User Guide (as a gzipped postscript file, 120kbytes).
  • Specification (as a gzipped postscript file, 110kbytes).

  •  

    The HPC Standards Project (Esprit project 21111 )

    The HPC Standards Project has existed to provide funding for a fixed set of European  representatives, drawn from academia and industry, to attend meetings in the U.S.A. convened to address the standardisation of three major High Performance Computing standards: High Performance Fortran (HPF-2), the standard Message-Passing Interface (MPI-2), and a standard set of parallel kernels and benchmarks (Parkbench). It has also provided for the  holding of a European open workshop in each subject area.

    HPC-Standards funding has allowed significant European-US discussions in the area of the PARKBENCH low-level benchmarks. European contribution was made in the following areas; the new "COMMS" benchmarks, a Java applet tool (PICT) for postprocessing and parameterisation of PARKBENCH results, run and reporting rules, Java benchmarks and parallel I/O benchmarks. PARKBENCH awareness and dissemination of results has been increased through the Performance Evaluation and Modelling for Computer Systems (PEMCS) electronic journal which was initiated in early 1997 at the suggestion of the HPCnet Network of Excellence. HPCnet provided some start-up funding for this activity which promotes responsible performance reporting. Therefore, although not funded directly by HPC-Standards or the PARKBENCH group, the journal has active involvement of EU and US experts and this can be seen to be a result of EU support for PARKBENCH in HPC-Standards.

    The HPC-Standards project final management report (February 1998) is available in HTML or postscript format. This details the activities funded by the project from 1st June 1996 until its cessation on November 30th, 1997.
     

    The Journal of Performance Evaluation and Modelling for Computer Systems (PEMCS)

    The Journal of Performance Evaluation and Modelling for Computer Systems (PEMCS)
     

    PERFORM

    PERFORM (Performance EstimatoR For RISC Microprocessors) is a  performance estimating system developed at the University of Southampton, by Dr Alistair Dunlop, for estimating the execution time of SPMD Fortran programs on distributed memory parallel systems. The system is an implementation of a new performance estimating technique based on the concept of execution-driven simulation. Whereas straightforward execution-driven simulation is slow, memory intensive and machine specific, our development of this technique has avoided these traditional problems associated with simulation methods. Specifically, our development of this method uses a feedback loop to reduce the amount of code executed during the simulation, while code sections that do not alter the control flow of the program are removed from the simulation code. Variables used in removed code sections are also removed from the simulation code, significantly reducing the memory requirements. This method is also architecturally independent, in that the properties and performance characteristics of the target platform can be specified in a machine description file. Our results show that PERFORM is a substantial improvement over existing performance estimating methods and that this tool could be used by programmers for code analysis, or for incorporation into an automatic/semi-automatic parallel compilation system. PERFORM is available with either a text-based interface or X-Windows/Motif GUI.
     

    GENESIS

    The GENESIS distributed-memory benchmark suite was assembled to fulfill a need for evaluating the performance of distributed memory MIMD systems on scientific and engineering applications. The suite originated within ESPRIT project P2702 (GENESIS) with the codes being written by a number of the project partners. It was originally prepared for public release by Southampton Novel Architecture Research Centre. A new version, including a database of results is available here .


    Home | Projects | Results Database | Publications | Related Links


    Site last updated 2 March 1999. Comments to Mark Papiani  or David Lancaster.

    Parallel and Distributed Computing Research Group
    Department of Electronics & Computer Science
    University of Southampton
    Southampton
    SO17 1BJ