Hi Christopher,
I could see the early stop only once with your workflow, in which case only 9 from 10 tokens arrived at the end. Since then it seems to be warmed up, as never more behaves incorrectly.

Opening user preferences PtolemyPreferences.xml...
247 ms. Memory: 4812K Free: 1882K (39%)
173 ms. Memory: 4812K Free: 1511K (31%)
72 ms. Memory: 4812K Free: 1198K (25%)
80 ms. Memory: 4812K Free: 193K (4%)
58 ms. Memory: 4812K Free: 1917K (40%)
56 ms. Memory: 4812K Free: 1377K (29%)
Warning: '.Branching-switch-PN-ptolemy.Test' The test produced only 9 tokens, yet the correctValues parameter was expecting 10 tokens.
66 ms. Memory: 4812K Free: 1038K (22%)
88 ms. Memory: 4812K Free: 457K (9%)
55 ms. Memory: 4812K Free: 1708K (35%)
65 ms. Memory: 4812K Free: 1121K (23%)
51 ms. Memory: 4812K Free: 328K (7%)
...

When I replaced the Test for Display, I got bad results more frequently (1 from 4-5 runs), with 5,6,7 tokens out of 10, resp.

Test env:
  P4 2.80 D, 1GB RAM
  Linux, Fedora Core 4
  Java 1.4.2_10-b03

Norbert



On Mon, 31 Jul 2006, Christopher Brooks wrote:

Hi Norbert,

I modified your example slightly and added it as a test as
ptII/ptolemy/domains/pn/test/auto/Branching-switch-PN-ptolemy.xml

I can't get this to fail though.
I'm running under Java 1.5.0_06 under Windows with the Ptolemy CVS
tree.
I changed the Sleep actor to 0L and I always get 10 outputs.

I've left a copy of what I checked in at:
http://ptolemy.eecs.berkeley.edu/~cxh/models/Branching-switch-PN-ptolemy.xml

If you have a chance, could you download that version and try
replicating the error?
Also, what version of Java and what OS (Linux? Windows?) are you
running.

If you let me know via email, I'll summarize to the list once we get
to the bottom of the problem.

_Christopher


--------

   Hi,

   I have created a small workflow to demonstrate the Switch actor. A Ramp
   produces numbers 1..10, an Expression produces 0..2 (input mod 3), and the
   Switch has three channels, each with just one Expression to produce a Strin
  g.
   Finally, the three branches are brough together with NondeterministicMerge
  and
   the strings are displayed.

   So there I expect 10 strings as output, which is mostly the case if I run i
  t.
   However, sometimes - quite frequently - less strings are displayed (tokens
   somewhere are lost) when the workflow stops.

   The workflow works nicely if I do one of the followings:
     - replace NondeterministicActor for Select
     - put a Sleep with 1ms to slow down the execution

   So it seems to me that PN senses termination condition wrongly in the origi
  nal
   case. Or what else?

   I attach the workflow, already with the Sleep, so it works as expected.
   If you delete the Sleep actor (or even just set to 0) and run it a couple o
  f
   times, you will see shorter outputs occasionally.

   I would like to understand what is happening and why (except if it is a bug
  ).
   Thanks in advance for explanation.
   Norbert

         Norbert Podhorszki
       ------------------------------------
         University of California, Davis
         Department of Computer Science
         1 Shields Ave, 2236 Kemper Hall
         Davis, CA 95616
         (530) 754-8188
         [EMAIL PROTECTED]
         ----------------------------------


----------------------------------------------------------------------------
Posted to the ptolemy-hackers mailing list.  Please send administrative
mail for this list to: [EMAIL PROTECTED]

Reply via email to