Newsgroups: comp.parallel
From: dionne@IRO.UMontreal.CA (Carl Dionne)
Subject: Taxonomy of parallels debuggers?
Organization: Universite de Montreal
Date: Tue, 21 Feb 1995 17:35:21 GMT
Message-ID: <D427sA.J00@IRO.UMontreal.CA>

For my master's dissertation I'm describing an implantation of a replay
algorithm for an actor langage.  I describes various types of
parallels debuggers, and I would like to include a taxonomy of replay
debuggers.

Has anyone seen information on the subject?  I know of several
implantations of replay algorithms but I'm not sure how
to classify them.

I'm planning to use three axes:

type of replay (synchronisation based / data based): 
in data based replay, the data
needed to replay the program are recorder, while in synchronisation
based replay, only the order in which certain events occurs is recorded.

architecture on which the algorithm is implanted: shared memory / 
distributed memory

scope of the replay: This axe is concerned with the limits of the replay
algorithm: whether it replays all the tasks or a subset of those tasks,
whether it always replays from the start of the program, or can start
at a check point after the start of the program.

Any information on an existing texonomy, or any comments
regarding the proposed classification are welcomed!

Thanks,

Carl


