Ptolemy Classic has a start at using a network of workstations to run
a SDF simulation on multiple workstations using C Code generation.
Unfortunately, we never saw a performance improvement, partly because
we were running on 100base-T network, which has a fairly high latency.
In general, partitioning simulations to run on multiple processors is
tricky, and there are lots of implementation details that can get in
the way.
For details, check out
P. Warner, ``Network of Workstations Active Messages Target for
Ptolemy C Code Generation,'' MS Report, ERL Technical Report UCB/ERL
No. 97/8, University of California, Berkeley, CA 94720, January 24, 1997
at
http://ptolemy.eecs.berkeley.edu/publications/papers/97/now/
See also
http://ptolemy.eecs.berkeley.edu/ptolemyclassic/almagest/docs/user/html/cgc.doc.html
At this time, development of the Ptolemy Classic CGC NOW facilities is
at a standstill.
Ptolemy II, which is implemented using Java was built from the ground
up with threading in mind, so it should be able to take advantage of
multiple processors in one machine and native threads. However, we
have not spent time looking at the performance in such an environment.
The Process Network domain would probably be the first place
to look for performance improvements on multiprocessor machines.
Java is much more network aware, so using CORBA and RMI to run
simulations on multiple machines is easier than using C. We have
been working with CORBA in house, but don't have anything shippable
For some details, see
http://ptolemy.eecs.berkeley.edu/projects/hmad/reports/99/7
or search the Ptolemy website for CORBA
-Christopher
--------
Hello,
We are planning to use Ptolemy for simulating distributed computing
environments/ algorithms. I would like to know if anyone (or group) has
already used Ptolemy for the same purpose. Getting to know about any such
exercise would indeed help our research.
Secondly we would also like to know if Ptolemy has the feature of running
on a NOWs for faster and more efficient simulations. If so then where can
one find the details / documentaions.
Thanking you,
Sayan
---------------------------------------------------------------------------
---
Sayan Mitra Hostel G51
Computer Science And Automation Phone : 080 309 2620
Indian Institute of Science Home :
Bangalore 560 012 7 Clarke Street
email - [EMAIL PROTECTED] Calcutta 700026
[EMAIL PROTECTED] www.angelfire.com/sc/sayan
___________________________________________________________________________
___
---------------------------------------------------------------------------
-
Posted to the ptolemy-hackers mailing list. Please send administrative
mail for this list to: [EMAIL PROTECTED]
--------
----------------------------------------------------------------------------
Posted to the ptolemy-hackers mailing list. Please send administrative
mail for this list to: [EMAIL PROTECTED]
