According to "Modelling with Generalized Stochastic
Petri Nets" (M. Ajome Marsan et al., Wiley, 1995),
immediate transitions are supposed to have precedence
over timed transitions.
In the simulation module of PIPE2, immediate
transitions do not have this precedence, hence the
behaviour of a GSPN may differ from the expected behaviour.
This should be fixed or documented. (Or did I miss it?)
Logged In: YES
user_id=1609637
Originator: NO
The code in the pipe.dataLayer.DataLayer class in fireRandomTransition method was changed to give precedence to immediate transitions over timed transitions. This may result in a dramatic change in how a network behaves. For example, the GSPN3 sample network now will never fire the T1 transitions except if it is selected as the first transition to fire. Since it can only fire when two places have tokens. Once in operation, placing a token in either of these places (P0 and P3) will cause an immediate transition (either T0 or T2) te be enabled and it will fire (along with T3 or T4) before "the other" place can receive a token from the "timed" transition. So, T1 is initially enabled, but will never be enabled again.