Berkeley ELENG C249A - Hierarchical FSM’s with Multiple Concurrency Models

Unformatted text preview:

1Hierarchical FSM’s with Multiple Concurrency ModelsPresented by Perry Tsao on October 31, 2000Introduction• Reactive Systems• All have concurrency• Includes embedded systems, real-time systems, some software systems• MoC’s that support concurrency:• Comm. sequential processes (CSP)• dataflow process networks (DF)• discrete events (DE)• synchronous/reactive (SR)2• FSM’s with concurrent MoC’s• Statecharts• Argos- FSM’s and SR model• Argos and Lustre• SDL- process networks with FSM’s• CFSM- FSM’s and DE• hybrid- continuous with FSM’s• All intertwine concurrency and automata semantics• Limited compositionalityIntroduction• *charts (“star-charts”) • decouple concurrency model from semantics• can hierarchically combine different concurrency models• Desirable MoC’s properties• compositional- composite modules used as primitives • heterogeneity- composite modules embedded within foreign MoC’sIntroduction3• The Basic FSM• deterministic• at most one enabled transition for each input• reactive• at least one enabled transition for each input• self-transitions assumedFinite State Machines• Multiple Inputs and Outputs• Pure• size of input symbol set and a power of two• ?Σ?52M• size of each signal alphabet = 2• ex. signal consists of “event” that is present or absentFinite State Machines4• Valued• size of a signal alphabet > 2• used in FSM’s that have arithmetic operations• Hierarchical• slave FSM action first, then master FSM action• outputs merged• valued FSM: no two triggered transitions can have same output signalFinite State Machines• Hiererarchy and Valued FSM• makes notations more compact, not fundamentally more expressiveFinite State Machines5• Termination• To refine a FSM, MoC must have a finite iterationMixing FSM’s• Dataflow with FSM• General DF and PN• Finite iteration existence is undecidable • SDF• Decidable if iteration exists• Deadlock existence decidable• Special case: homogenous SDF• 1 input => 1 output for all arcsMixing FSM with Dataflow6• Example:• 2 pure FSM’s in a homogenous SDF• FSM’s fired sequentially within 1 SDF iteration• FSM can fire concurrently across iterationsFSM inside SDFFSM in SDF• {D,C,C,C,E}• Two types of firings:• Type A firing• No transition, only refined subsystems of current state fired• Type B firing• Transition taken• Allows FSM to stay in same state for top level SDF iterationFinite State Machines7SDF in FSM• FSM refined to SDF• SDF graph iteration followed by FSM reaction• Easy if SDF is homogenousSDF in FSM• Non-homogenous SDF’s in FSM• Type signature• FSM that SDF embedded in treated as SDF• Complications• SDF composition not always possible• FSM (with embedded SDF) not always in SDF• Type signature can change in different FSM states8Heterochronous Dataflow• For type signature changes• Two sets of balance equations needed• Type signature constant during HDF iteration• Compromises modularityDynamic Dataflow• FSM in DDF• Fire FSM until input tokens consumed• DDF in FSM• Very complicated, not discussed9Boo!Discrete Events and FSM• DE has global time• Events have time stamps• FSM inside DE• FSM is a zero delay actor• FSM fires whenever DE actor it refines fires• Output has same time stamp as input10DE inside FSM• Analagous to SDF inside FSM• DE properties exported to environment of FSM• Still have problems with zero-delay loopsSynchronous/Reactive with FSM• Quick SR system overview• Everything happens at ticks• Fixed points theorem ensures determinacy• Produce phase• Function evaluated to determine outputs (find fixed points)• Transition phase• Outputs changed to prepare for next tick• Esterel is an example11FSM inside SR• Simple FSM inside SR• Simple => not refined• Firing types• Type C determines output (if possible)• Type D writes the outputs (no calculations)• Type C firing occurs until fixed point resolvedFSM inside SR• Refined FSM inside SR• Current state of FSM refined to SR• No problem• Current state of FSM refined to non-SR• Fire these states after SR states (if possible)• Invoke transition function (if possible)• SR inside FSM• Pretend FSM is an SR, no problem (just as above)12Verification and Synthesis• Synthesis from FSM, SDF, SR done before• Combining synthesis methods should be possible• Verification• FSM’s and SDF’s separately is well studied• Combination unclear, simulation necessaryExample• Reflex game• Heterogenous realization (*charts)• Easier to understand• Esterel• More compact• Does not model environment• VHDL and C• Much longer• Messier• Cell phones• Good candidate application for *charts13Conclusion• *charts • Allow designs to use multiple MoC’s• Pick MoC most suitable for each


View Full Document

Berkeley ELENG C249A - Hierarchical FSM’s with Multiple Concurrency Models

Documents in this Course
Load more
Download Hierarchical FSM’s with Multiple Concurrency Models
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Hierarchical FSM’s with Multiple Concurrency Models and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Hierarchical FSM’s with Multiple Concurrency Models 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?