POEMS/MASC: Peformance-Oriented End-to-end Modeling System/ Model-based Adaptive Software Control

Directed by
M. K. Vernon,
UW-Madison Logo
V. S. Adve,
Univ of Illinois at Urbana Logo
R. Bagrodia,
UCLA Logo
J. C. Browne
Univ. of Texas at Austin Logo
E. N. Houstis,
Purdue Univ. Logo
J. R. Rice,
Purdue Univ. Logo
P. J. Teller,
Univ. of Texas at El Paso Logo
S. Wright
UW-Madison Logo

Overview
People
Publications
Current Topics

Research Topics

If you are a graduate student interested in gaining some research/publishing experience and contributing to the POEMS/MASC project, for example through an independent study project, possible topics and who to contact for further information about each topic are listed below:
  • Generalized/reusable LogGP model
    The goal is to generalize the LogGP model developed and validated for Sweep3D running on an SP system, so that the model can be applied to predict the total execution time for any parallel application that has a parellel pipelined structure and uses message passing communication (e.g., MPI or PVM). The first step is to adapt the model and validate it for another significant application that has a parallel pipelined synchronization structure, such as one of the NAS benchmarks. (Contact vadve@cs.uiuc.edu or vernon@cs.wisc.edu).

  • Parallel Implementation of MOL
    The parallel version has been designed to enable the MOL application to run on multiple processors.The parallel implementation uses the Message Passing Interface (MPI) to communicate data between processes. (Contact jrr@cs.purdue.edu).

  • Parallel application performance modeling
    Model the performance of a large application running on Condor. (Contact vernon@cs.wisc.edu).

  • Graphical User Interface to :
    (1) Experiment with Adaptive Parallel Computations
    (2) Collect experimental data about Parallel Performance of MOL
    The primary objective of the GUI is to enable users to interactively control problem parameters and obtain performance data in the form of tables and/or graphs. Also the performance data can be stored in a database , which can be used for further analysis. (Contact jrr@cs.purdue.edu).

If you are interested in joining the POEMS/MASC project as an RA, please see any of the project directors.


vernon@cs.wisc.edu