Program

Location

WHIST 2011 will be held in Salon D of the Loews Ventana Canyon Resort.
Please see this map for the exact location.

Schedule

  Saturday, June 4
8:30 Breakfast
9:00 Keynote
9:30
10:00 Break
10:30 Morning Paper Session
11:00
11:30
12:00 Lunch
12:30
1:00
1:30 Afternoon Paper Session
2:00
2:30
3:00 Break
3:30 Panel Discussion
4:30
5:00  

Saturday, June 4

8:30-9:00
Breakfast

9:00-10:00

Keynote: Jeffrey Hollingsworth

From Paradyn to Dyninst to *API and beyond
Chair: Nathan Tallent
10:00-10:30
Break
10:30-12:00
Chair: Dorian Arnold

Morning Paper Session

Software Analysis Techniques to Approximate Data Centric Direct Measurements.
Nick Rutar and Jeffrey Hollingsworth

Parallelizing Heavyweight Debugging Tools with MPIecho.
Barry Rountree, Guy Cobb, Todd Gamblin, Martin Schulz and Bronis R. de Supinski

A scalable replay-based infrastructure for the performance analysis of one-sided communication.
Marc-Andre Hermanns, Sriram Krishnamoorthy and Felix Wolf
12:00-1:30
Lunch
1:30-3:00
Chair: Sriram Krishnamoorthy

Afternoon Paper Session

Libmonitor: A Tool for First-Party Monitoring.
Mark Krentel

Dynamic Floating-Point Cancellation Detection.
Michael Lam, Jeffrey K. Hollingsworth and G. W. Stewart

A Framework for Bootstrapping Extreme Scale Software Systems.
Joshua Goehner, Dorian Arnold, Dong Ahn, Gregory Lee, Bronis R. de Supinski, Matthew Legendre, Martin Schulz and Barton Miller
3:00-3:30
Break
3:30-4:30
Chair: Todd Gamblin

Panel Discussion

Topic: Future directions for performance tools

Panelists:

  • Martin Schulz, Lawrence Livermore National Laboratory
  • Dorian Arnold, University of New Mexico
  • David Richards, Lawrence Livermore National Laboratory
  • Nathan Tallent, Rice University

Questions:

  1. What are tools already doing well?

    What lessons are there to learn from existing tools, and what techniques should now be reduced to practice? What are the barriers keeping tool developers from reusing each others' tools and techniques?

  2. What are current tools doing poorly, and how will tools provide insight into exascale applications?

    Insight:
    Is the data provided by current performance tools relevant or understandable to developers? How can we tie performance data to application models, and how can we present fewer measurements and more insight?

    Scalability:
    With millions/billions of hierarchical tasks, what will have to change? We already have problems with hundreds of thousands of tasks. What can we learn from the applications, and do exascale tools need co-design too?