home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!gatech!hubcap!news
- From: " (Stephan Gutzwiller)" <gutzi@ifi.unibas.ch>
- Subject: parallel algorithm classification
- Message-ID: <1992Dec15.134523.7559@hubcap.clemson.edu>
- Apparently-To: comp-parallel@cernvax.cern.ch
- Sender: fpst@hubcap.clemson.edu (Steve Stevenson)
- Nntp-Posting-Host: linus.ifi.unibas.ch
- Reply-To: baks@ifi.unibas.ch
- Organization: Institut fuer Informatik
- Date: 15 Dec 92 10:06
- Approved: parallel@hubcap.clemson.edu
- Lines: 58
-
- ==================
- Hello netters,
-
- we would like to announce the availability of a technical
- report from the university of Basel (Switzerland) entitled:
-
- 'BAKS: Basler Algorithmen Klassifikations-Schema'
-
- at the moment there is only a german version of this report
- available - we are working on the english translation - via
- anonymous ftp from:
-
- lucy.ifi.unibas.ch (131.152.81.1)
-
- in the directory 'baks', where a compressed postscript file
- can be found (434 kB).
-
- The report introduces a systematic approach to classifications
- of parallel algorithms.
-
- chapter 1 is a general justification for the need of a classi-
- fication of algorithms and consequences of such an
- approach to parallel programming.
- chapter 2 introduces the terminology which is later used for
- different classifications of parallel algorithms and
- explains its use by means of a simple example.
- chapter 3 shows different possibilities for the classification
- of parallel algorithms emerging from the terminology
- introduced in chapter 2; thus is it possible to focus
- on certain aspects of parallel algorithms.
- We also show how our approach can be used to program
- and evaluate existing machines (i.e. existing
- commercially or public-domain available (virtual) machines
- like express or pvm). On the one hand we briefly
- introduce a tool that generates from a given specification of
- an algorithm a program skeleton for a given machine. Thus
- algorithm portability across different machines is achieved.
- On the other hand a system for creation of synthetic
- benchmarks according to user specifications is sketched.
- Thus a machine can be (synthetically) tested on those
- classes of algorithms which seem relevant to the users.
- chapter 4 gives a future outlook.
-
- This technical-report emerged from research projects currently done
- at the university of Basel and we are very interested in feedback
- from the parallel processing community. To us 'BAKS' seems to be
- unique so far, if you know about any other calssification scheme for
- parallel algorithms please let us know. Reply any results,
- questions, problems etc. to
-
- baks@ifi.unibas.ch
-
- ----------------------------------------------------------------------------
- stephan gutzwiller phone: 0041 (0)61 3219991
- institut fuer informatik fax: 0041 (0)61 3219915
- mittlere strasse 142 email: gutzi@ifi.unibas.ch (NeXT mail accepted)
- ch-4054 basel
-
-