5/26/2005 Update - PowerPoint PPT Presentation

1 / 4
About This Presentation
Title:

5/26/2005 Update

Description:

Post report (5/31/2005) summary (6/2/2005) ... a [conference/journal] paper, although probably not enough publishable content. Outline ... – PowerPoint PPT presentation

Number of Views:12
Avg rating:3.0/5.0
Slides: 5
Provided by: dral60
Category:

less

Transcript and Presenter's Notes

Title: 5/26/2005 Update


1
5/26/2005 Update
  • Adam Leko
  • HCS Research Laboratory
  • University of Florida

2
Tool Evaluation Report Overview
  • Evaluation report is coming along
  • Course of action
  • Write report of tools evaluated thus far
    (including HPC toolkit)
  • Post report (5/31/2005) summary (6/2/2005)
  • Update report summary after new tool
    evaluations as needed
  • Originally summary presentation first, but want
    to push back
  • Make sure the presentation is an accurate
    reflection of report

3
Tool Evaluation Report Outline
  • Writing in the form of a conference/journal
    paper, although probably not enough publishable
    content
  • Outline
  • Abstract
  • Introduction brief overview of project,
    motivation behind evaluating tools
  • Evaluation methodology quick overview of
    evaluation strategy given by Hung-Hsun in report
    and refined by Adam Hans last semester
  • Tool evaluations section follow general report
    outlines
  • Installation usage overview
  • Measurement overhead results
  • Bottleneck analysis results
  • Describe good/bad points of tool
  • Discuss tool extendibility
  • Give standard scores for each evaluation category
  • Recommendations
  • Which tool, if any, we should use and how
  • Generate list of candidate tools we could/should
    possibly work with
  • From list of candidate tools (possibilities that
    we should extend), rank on the following
    categories TBD
  • A slightly edited version will go in our final
    report to the NSA
  • Will probably need to do some extra work for TAU
    and Paradyn for overhead measurements and
    bottleneck analysis

4
Prototype Plan
  • Emailed Sameer (TAU contact), found no support
    for GPSHMEM
  • After tool report done, going to work with him to
    add GPSHMEM support to TAU
  • Accomplishes two goals
  • Strengthens relationship with TAU developers if
    we decide to use parts or all of TAU
  • Demonstrates we are able to back up our design
    with code
  • Plan of action
  • Create PSHMEM wrappers for GPSHMEM
  • Link PSHMEM wrappers with TAU and existing
    GPSHMEM code
  • Work with Sameer for getting TAU ready to work
    with GPSHMEM
  • Update ./configure script
  • Add appropriate other code
  • Note that TAU GPSHMEM will provide a basic but
    not very user-friendly or advanced way of
    analyzing SHMEM programs
Write a Comment
User Comments (0)
About PowerShow.com