Hi Mohamed Salem,
I think this would be a nice way to look at it.
It never struck to me!
I can always split this actor into 8 different atomic actors and put them into a 
composite actor.
May be this can solve the matter (My project release is due this month end!!).
Anyway thanks a lot for the suggestion, I will try it and let you know.

Thanks
kiran

----- Original Message -----
From: Mohamed Salem <[EMAIL PROTECTED]>
Date: 04 Dec 2003 16:12:55 +0200
To: Kiran SB <[EMAIL PROTECTED]>
Subject: Re: Need to delay execution by 1 clk...

> Hi Kiran and all,
> 
>       If this actor is constructed as one atomic-actor that models the eight
> stages of execution, I would suggest that you re-build it as a composite
> actor from eight diffent atmoic-actors (your own models and/or from
> library actors).Each atomic actor in this model would be corresponding
> to one of the eight stages of execution.
> 
>       Pipeline your actors; in other words output of (actor1/stage 1) is
> connected to input of (actor2/stage2);meanwhile all eight are
> synchronous;i.e connected to the same main clock.
> 
>       The DE director reacts to each positive edge as an event. The
> running(execution) of the overall composite actor is a sequene of events
> corresponding to each positive edge of the clock.
> 
>       In this criteria, Consider two consecutive actors (one delivering to
> the next) in (execution graph) ; at the positive edge both of them are
> Fired such that at this instance(event) T ; (actorX) delivers its output
> at time (T-1)to (actorX+1) and it produces simultaneously an output that
> to be delivered to (actorX+1)in the next firing event(next positive
> edge)at time (T+1).In other words the output token of actorX is reserved
> (not changing) in the gap between two consecutive events.
> 
> for example :
> assume 
> actorX output at T-1 = m 
> actorX output at T   = n 
> 
> (ActorX reserves-stores(flipflop) a value =m for one clock cycle)
> 
> Then
>  
> actor X+1 processes an input = m at time T 
> actor X+1 processes an input = n at time T+1 
> 
>       Conclusion , if I understand your exact problem , this method could
> do your job without a delay actor in between actorX and actor X+1. 
> Otherwise, insertion of a Timedelay with delay parameter of 1 could
> help.
> 
> I hope that this would be helpful ...
> let me know about results.
> 
> THANKS for your time and interest
> 
> Best Regards,
> 
>       Mohamed A.Salem
> 
> ========================================================================
> 
> On Thu, 2003-11-27 at 13:06, Kiran SB wrote:
> > Hi
> > I am new to ptolemy (6 months).
> > Currently I am using PTII 3.0.2 for building(simulating) custom actors in the
> > DE domain, which are basically executional elements (of a reconfigurable 
> > device).
> > 
> > In one particular situation I have an actor with 8 stages of execution.
> > During initialize I construct a Execution(Data) flow graph depending on the
> > selector settings in the actor.
> > I traverse through the graph and execute each stage in it during fire.
> > Now at one stage I need to create a delay of 1 clock.
> > The executional unit needs to sleep for a clock and proceed to the next stage.
> > Basically I need to forward time by a clock.
> > setCurrentTime (currentTime+1) might not be a good idea as it might cause
> > problems to the token's timestamp.
> > What is to be achieved is that the currentTime after this stage should be
> > t+1(say we started at t).
> > I don't know if I can use DEEventQueue in any way for this purpose.
> > Maybe its a wrong or a repeat question!
> > I have searched all group messages and couldn't find anybody with such a
> > requirement.
> > Can somebody help me out with this problem?
> > 
> > - kiran
> 

-- 
______________________________________________
Check out the latest SMS services @ http://www.linuxmail.org 
This allows you to send and receive SMS through your mailbox.


Powered by Outblaze

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

Reply via email to