Vortrag: Debugging Parallel Programs using DETOP and OMIS, 11.5.1998

[Maria Cherry]

                          UNIVERSITAET WIEN=20
          INSTITUT FUER SOFTWARETECHNIK UND PARALLELE SYSTEME
                            gemeinsam mit=20
                                VCPC=20
           EUROPEAN CENTRE FOR PARALLEL COMPUTING AT VIENNA=20


      EINLADUNG ZU EINEM VORTRAG IM RAHMEN DES INSTITUTS-KOLLOQUIUMS:
               =20
        =20
              Debugging Parallel Programs using DETOP and OMIS
                  =20
 =09=09=09  =20
                          Dr. Roland Wism=FCller
                               LRR-TUM
                    Technische Universit=E4t M=FCnchen
                                =20

                 ZEIT: Montag, 11. 5. 1998, 15.30 Uhr c.t.
       ORT: Institut fuer Softwaretechnik und Parallele Systeme
                  1090 Wien, Liechtensteinstrasse 22,=20
                         Seminarraum, Mezzanin


Abstract

Debugging is one of the most important aspects of software development. Thi=
s
is equally true for parallel programming. However, since debuggers are one
of the most complex tools to build, research activities in this field are
often limited to user interface considerations. LRR-TUM is one of the few
European research groups that also deals with the system level aspects of
parallel debugging.

The talk will be divided into three parts: The first part will give an
introduction to and a motivation for parallel debuggers. It will present
some of the problems these tools have to handle and gives an overview of
current activities in this field.

In the second part, we will present the parallel debugger DETOP that has
been developed at LRR-TUM. The most advanced version of this debugger
is available for the multithreaded programming model PARIX running
on Parsytec computers, however, a port to the PVM and MPI programming
libraries on NOWs is currently done. The talk will introduce the basic
requirements we tried to meet and presents the most important features of
DETOP.

When we started to port our set of on-line tools to PVM on NOWs, we realize=
d
that having a standard interface for monitoring and manipulating parallel
programs could greatly reduce the cost of porting and would also have other
benefits, such as tool interoperability and uniformity across target platfo=
rms.
This lead to the start of the OMIS (On-line Monitoring Interface Specificat=
ion)
project, which will be the topic of the last part of the talk. We will pres=
ent
the concepts of OMIS, explain how we try to achieve the flexibility and
efficiency required for such an interface, illustrate its use, and present =
an
overview on its implementation.