ARSC T3E Users' Newsletter 190, March 3, 2000

Cray Acquired from SGI by Tera Computer Company

For information, you might visit:

http://www.tera.com/ , or http://www.sgi.com/newsroom/press_releases/2000/march/cray.html

VAMPIR 2.5, VAMPIRtrace 2.0, and ARSC VAMPIR Tutorial

VAMPIR is a graphical tool to help MPI and SHMEM programmers examine and optimize the run-time communication patterns of their codes. We encourage you to make a run with this powerful tool: it can be revealing.

We have updated ARSC's VAMPIR Tutorial to match our recent VAMPIR software upgrades (on yukon, see "news VAMPIR_upgrade"). It is available at:

http://www.arsc.edu/support/howtos/usingvampir.html

The tutorial contains 11 images from VAMPIR 2.5 and covers many features, including:
  • Tracing of SHMEM programs
  • Visual designation of "global operations" (such as collective operations and barriers) in timeline displays.
  • Detailing of individual display components (individual messages, global operations, and subroutine calls).
  • Listing pending (unreceived) messages
  • Displaying run statistics
  • Reducing the size of trace files
  • Zooming

UNICOS/mk Upgrade and Another OS PE for yukon

Next Wednesday night during downtime, we plan two changes to yukon:
  • Well, one new OS PE means one less CMD PE. This will leave yukon with 4 OS PEs and 8 CMD PEs, and is expected to improve file system and network performance.
  • UNICOS/mk will be upgraded from 2.0.4 to 2.0.5. This version should be more resilient, and offers new capabilities to the system administration folks.

Registration for CUG SUMMIT 2000, the 42nd CUG conference

For details on this year's CUG, see:

http://cug2000.sara.nl/

For information on past conferences, see:

http://www.cug.org/

Quick-Tip Q & A


A: {{ I love the T3E, but sometimes I relax with me olde Cray PVP.
   {{ How can I estimate my job's memory utilization so I can make an
   {{ accurate NQS request (and get my job to start sooner)?


    (This answer taken from:
       
http://www.arsc.edu/support/howtos/usingnqs.html


    Use the ja (or job accounting) tool provides a wealth of basic
    information about your run--including its maximum memory
    utilization.

    ja may be used interactively or in qsub scripts. You call it once
    without flags to start accounting; then run one or more programs of
    interest; then call it again with flags to dump reports in various
    formats and to terminate accounting.

    Example qsub script using ja 

      #QSUB -q batch               # Pipe queue, "batch" required
      #QSUB -lM 50Mw               # limit the request to 50 Mw 
      #QSUB -lT 10000              # limit the overall request to 10000 CPU seconds
                  
      cd /u1/uaf/morris/progs
      
      ja                           # Start job accounting
      ./prog data1
      ./prog data2
      ja -sto                      # Dump reports and terminate accounting

    The reports are channeled to stdout and thus to the request's .o
    output file.  In this example, ja -s produces a summary report
    which will contain (among other things) data like this:

      Elapsed Time                :           72      Seconds
      User CPU Time               :            7.6363 [      7.6341] Seconds
      Maximum memory used         :            9.7637 MWords

    The Maximum memory used is the maximum memory that was actually
    used at any single point between the initial invocation of ja and
    the invocation of ja -s.

    The script reserved 50 MWords, but the job only needed about 10
    MWords. If future datasets and runs will be similar, the script
    could be modified to request, say, 15 MWords.

    (ja is also available on the T3E.)




Q: I want to share some directories with members of my group, giving
   them write access. This command:

     chmod -R g+rwX ~

   would work, but I don't want anyone messing with my dot files.  How
   can I share a directory with my group?

[ Answers, questions, and tips graciously accepted. ]


Current Editors:
Ed Kornkven ARSC HPC Specialist ph: 907-450-8669
Kate Hedstrom ARSC Oceanographic Specialist ph: 907-450-8678
Arctic Region Supercomputing Center
University of Alaska Fairbanks
PO Box 756020
Fairbanks AK 99775-6020
E-mail Subscriptions: Archives:
    Back issues of the ASCII e-mail edition of the ARSC T3D/T3E/HPC Users' Newsletter are available by request. Please contact the editors.
Back to Top