From: Dr Sabine Thuermel
Subject: Debugging of Parallel Lisp
Date:
Message-ID: <1244@ztivax.UUCP>
We are currently developping debugging tools for
a parallel Lisp dialect based on futures.
Does anybody know of existing systems or work in
progress on symbolic or performance debugging -
e.g. for Spur Lisp, Qlisp, MUL-T/Multilisp?
Any pointer would be wellcome!
Sabine Thuermel
e-mail: ········@ztivax.siemens.com
snail-mail: ZFE IS SOF 22
Siemens AG
Otto-Hahn Ring 6
8000 Muenchen 83, FRG
In article <····@ztivax.UUCP>, ········@ztivax.UUCP (Dr Sabine Thuermel)
writes:
|>
|> We are currently developping debugging tools for
|> a parallel Lisp dialect based on futures.
|> Does anybody know of existing systems or work in
|> progress on symbolic or performance debugging -
|> e.g. for Spur Lisp, Qlisp, MUL-T/Multilisp?
|> Any pointer would be wellcome!
|>
Laura Bagnall implemented a neat system for her MS at MIT called ParVis: A
Program Visualization Tool for Multilisp. Her system collected runtime event
information (future creation, touches, etc), analyzed them port-mortem, and
displayed task status and intertask dependencies on a Symbolics 3600. I found
her system extremely useful for performance debugging, though limited in the
size of an application it could deal with. I have a limited number of copies
(2?) of Laura's thesis that I can send interested people.
You should also contact Bert Halstead (········@crl.dec.com) about his current
work.
Randy Osborne
···@crl.dec.com
From: Ken Dickey
Subject: Re: Debugging of Parallel Lisp
Date:
Message-ID: <2835@wrgate.WR.TEK.COM>
In article <····@ztivax.UUCP> ········@ztivax.UUCP (Dr Sabine Thuermel) writes:
>
>We are currently developping debugging tools for
>a parallel Lisp dialect based on futures.
>Does anybody know of existing systems or work in
>progress on symbolic or performance debugging -
>e.g. for Spur Lisp, Qlisp, MUL-T/Multilisp?
>Any pointer would be wellcome!
The Gambit compiler has a variety of futures [lazy, spawning, ...] and a
good profiling environment which includes counts of execution of basic
blocks. I don't believe that it has been released yet. Here is a
reference to the latest paper:
%A Marc Feeley
%A James S. Miller
%T A Parallel Virtual Machine for Efficient Scheme Compilation
%J Proceedings of the 1990 ACM Conference on Lisp
and Functional Programming
%C Nice, France
%D June 1990
You can contact Marc Feeley at <······@cs.brandeis.edu>. I believe that
James Miller is at <·······@cs.brandeis.edu>.
Gambit has options for IEEE Scheme, R4RS Scheme, and MultiScheme. You
might also be interested in Miller's thesis:
%A James S. Miller
%T A Parallel Processing System Based on MIT Scheme
%I MIT LCS Technical Report #402 (Ph.D. Dissertation)
%C Cambridge, Mass.
%D August 1987
David Kranz, et. al. reported on MulT in a paper given in the last
[Portland] conference on the Design and Implementation of Programming
Languages [ACM--sorry, I don't have the reference here].
You should probably contact David Kranz <·····@wheaties.ai.mit.edu> for
information on whether and what hooks have been added to crawl [T's
debugger].
-Ken ····@mrloog.WR.TEK.COM
In article <····@wrgate.WR.TEK.COM>, ····@mrloog.WR.TEK.COM (Ken Dickey)
writes:
|> In article <····@ztivax.UUCP> ········@ztivax.UUCP (Dr Sabine
Thuermel) writes:
|> >
|> >We are currently developping debugging tools for
|> >a parallel Lisp dialect based on futures.
|> >Does anybody know of existing systems or work in
|> >progress on symbolic or performance debugging -
|> >e.g. for Spur Lisp, Qlisp, MUL-T/Multilisp?
|> >Any pointer would be wellcome!
|>
|> David Kranz, et. al. reported on MulT in a paper given in the last
|> [Portland] conference on the Design and Implementation of Programming
|> Languages [ACM--sorry, I don't have the reference here].
|>
|> You should probably contact David Kranz <·····@wheaties.ai.mit.edu> for
|> information on whether and what hooks have been added to crawl [T's
|> debugger].
crawl in MulT is almost the same as in T. MulT's innovation in error handling
for parallel systems is the notion of Unix-like groups (see section 2.3
of the paper quoted above).
Randy Osborne
···@crl.dec.com