HPCA Workshop on Productivity and Performance in High-End Computing (P-PHEC 2004)

(To be held in conjunction with HPCA 2004, Madrid, Spain)



The goal of this workshop is to examine productivity issues and the metrics used to measure them, in the context of high-end computing (HEC) systems, and to provide a forum for research addressing productivity in high-end computing systems.

The peak performance of processors has been increasing at an exponential rate, typically doubling every eighteen months. However, this increase in processing capability has not resulted in a corresponding improvement in system productivity for high-end applications and systems. In addition to traditional performance metrics, the productivity of a system is governed by other metrics such as application development time, time to solution, performance portability, robustness, and ease of both correctness validation and performance tuning. With both the range and the complexity of high-end applications continuing to increase, factors other than raw performance are becoming more important to the HEC system user.


Following is a partial list of topics upon which papers are solicited:
  1. Methods, approaches and tools for increasing productivity and/or reducing usage complexity of high-end applications and systems.
  2. Productivity metrics: measures of productivity in HEC systems and experiments to validate these measures.
  3. Analysis and characterization of HEC workloads and workflows and their distinct requirements.
  4. Impact of HEC system properties on time-to-solution or other productivity metrics.
  5. Case studies of productivity measurements and experiments on HEC applications and systems.

Dates and Submissions

Submissions due: December 5, 2003
Author Notifications: December 20, 2003
Final Papers Due: Jan 9, 2004
Workshop Date: Saturday, Feb 14, 2004

Papers should not exceed 15 pages, with double-spaced 11pt size fonts in Adobe PDF format. Please e-mail papers to rajamony@us.ibm.com with the subject: "HPCA Productivity Workshop"

Program Committee

Workshop Organizers: Ram Rajamony and Karthick Rajamani.


Breakfast is on your own

8:55 to 9:00

Opening Remarks

9:00 to 10:00

Morning Keynote: The Coming Crisis in Computational Science

Douglass E. Post, Los Alamos National Laboratory


Application Development Productivity Challenges for High-End Computing

Vivek Sarkar, Clay Williams, and Kemal Ebcioglu, IBM T. J. Watson Research Center



10:45 12:15

Comparing Network Processor Programming Environments: A Case Study

Niraj Shah, William Plishker, and Kurt Keutzer, University of California, Berkeley


Templating Transformations for Bitstream Programs

Armando Solar-Lezama, and Rastislav Bodik, University of California, Berkeley


Performance and Productivity in Parallel Programming via Processor Virtualization

Laxmikant V. Kale, University of Illinois at Urbana-Champaign




Afternoon Keynote: Building High Performance Scientific Applications for Parallel and Distributed Systems using a Software Component Architecture

Dennis Gannon, University of Indiana


Introducing the "Application Kernel Matrix"

Brad Chamberlain, John Feo, and David Mizell, Cray Inc.




Managing Complexity in Modern High End Scientific Computing through Component-Based Software Engineering

David E. Bernholdt, Oak Ridge National Laboratory; Robert C. Armstrong, and Benjamin A. Allan, Sandia National Laboratories


The High-Level Parallel Language ZPL Improves Productivity and Performance Bradford L. Chamberlain, University of Washington and Cray Inc.; Sung-Eun Choi, Los Alamos National Laboratory; Steven J. Deitz, and Lawrence Snyder, University of Washington


Raising the Level of Programming Abstraction in Scalable Programming Models 

David E. Bernholdt, Oak Ridge National Laboratory; Jarek Nieplocha, Pacific Northwest National Laboratory; and P. Sadayappan, Ohio State University






If you have any questions, e-mail rajamony@us.ibm.com