Re: [Ptolemy] [Info] needed regarding execution under SDF model
It sounds like what you want is to put a downsampler between HT1 and HT2 with a rate of N1. With the correct setting of the phase parameter, the downsampler will discard all but the last token out of N1. This token will be passed to HT2. similar for the connection between HT2 and HT3. With the token rates of each port of HT1, 2, 3 set to one, you will get the dataflow you want. Setting the iterationCount parameter of the SDF director to 1 will cause the model to stop executing after HT3 executes once. Unfortunately, it means that each firing HT1 produces an dummy value. To be a technically valid SDF model it must... From the point of view of the interface of HT1 alone, I'm a bit at a loss as to why it must execute N1 times? In seems to me like the N1 firings of HT1 and the N2 firings of HT2 could be hidden inside the actor, but without more information about where the desired number of firings is coming from, I couldn't hazard a guess as to the right way to represent it. Steve From: John Doe <[EMAIL PROTECTED]> Subject: Help needed regarding execution under SDF model Date: Tue, 01 Jun 2004 18:07:31 +0200 Hello, I need some help to be able to simulate a new langage in ptolemy. I have a graph with several hierchical tasks (HT) : HT1 --- HT2 --- HT3 - --- I need to execute HT1 N1 times before producing one result, this result is consumed by HT2. I don't need to execute HT1 anymore. After i need to execute HT2 N2 times before producing one result this result is consumed by HT3. And i don't need to execute HT2 anymore. and so on ... N1 is different from N2, N2 from N3 ... Is there a way to do it properly using an SDF director or any other directors ? Thanking you in advance Philippe --- End of Forwarded Message ___ Ptolemy maillist - [EMAIL PROTECTED] http://chess.eecs.berkeley.edu/ptolemy/listinfo/ptolemy Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
Re: [Ptolemy] [Info] needed regarding execution under SDF model
Stephen Andrew Neuendorffer wrote: I'm a bit at a loss as to why it must execute N1 times? In seems to me like the N1 firings of HT1 and the N2 firings of HT2 could be hidden inside the actor, but without more information about where the desired number of firings is coming from, I couldn't hazard a guess as to the right way to represent it. "The N1 firings of HT1 and the N2 firings of HT2 could be hidden inside" I do agree with you, but I don't know how to proceed... In fact in our model, there are two levels : The first is the one I described in my last mail. Each HT (Hierarchical Task), contains the second level : When I take an input token, I cut it in N parts, I do a computation on each part. At the end I take the N results of the computation and I produce the result token. So I use N "firings" to do the N computations. I don't have the choice because I need to increment a counter in the firing method of the director. The director is an SDF director. I try to use the "iterate" method but i don't know how to use it properly. I can't use the "preinitialize" or "initialize" method because all the methods "preinitialize" or "initialize" of all the HT must be executed before a fire. I would like to call it between "initialize" and "prefire". I think it's the best solution. But I do not know if it is possible ? Philippe Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
Re: [Ptolemy] [Info] needed regarding execution under SDF model
At 05:53 AM 6/3/2004, Philippe.Dumont wrote: Stephen Andrew Neuendorffer wrote: I'm a bit at a loss as to why it must execute N1 times? In seems to me like the N1 firings of HT1 and the N2 firings of HT2 could be hidden inside the actor, but without more information about where the desired number of firings is coming from, I couldn't hazard a guess as to the right way to represent it. "The N1 firings of HT1 and the N2 firings of HT2 could be hidden inside" I do agree with you, but I don't know how to proceed... In fact in our model, there are two levels : The first is the one I described in my last mail. Each HT (Hierarchical Task), contains the second level : When I take an input token, I cut it in N parts, I do a computation on each part. At the end I take the N results of the computation and I produce the result token. So I use N "firings" to do the N computations. I don't have the choice because I need to increment a counter in the firing method of the director. OK, so I see two ways to approach this: 1) represent the extraction and composition of the N parts explicitly. 2) have it happen implicitly. I'm not sure what you are interested in: 1) is pretty straightforward (use an ArrayToSequence to extract the N tokens, send them through an actor, and then use a Downsampler to discard all but the last output) 2) I could picture doing a couple of ways, depending on how you are representing the primitive operation and what is important about it. The Director.transferInputs and transferOutputs methods could be overridden to do what you want to do. The primary downside to this is that you end up with an extra layer of hieararchy. The IterateOverArray actor (actor.lib.hoc) implements one way of hiding the extra hierarchy in the user interface. It overrides the director to behave as a "ArrayToSequence", followed by an actor, followed by "SequenceToArray" It sounds like you want something similar, but with a DownSampler instead of a SequenceToArray. The director is an SDF director. I try to use the "iterate" method but i don't know how to use it properly. I can't use the "preinitialize" or "initialize" method because all the methods "preinitialize" or "initialize" of all the HT must be executed before a fire. I would like to call it between "initialize" and "prefire". I think it's the best solution. But I do not know if it is possible ? I don't understand what you are trying to do here... Steve Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
