/* Part 1: Initialization */ Seqrt=0.0 Seq Eventlist=fg for Each Lp I Lp I .parrt=0.0 Lp I .safetime=0.0 Lp I .lastsafetime=0.0 Lp I .eventlist=fg Endfor /* Keep List of Events for the Super-step Waiting to Be Simulated *

LP.EventList if it has already been executed by the sequential portion (so that e.execTime 6 = ?1:0.) So if there is any event not yet simulated, it must be in either SuperStepList or LP.EventList. When the analyzer nds that: SuperStepList = fg and LP i .EventList = fg 8LP i the sequential event list must be empty. And since the set of events in the sequential simulation is the sum of every LP's set of events in the parallel simulation, when the sequential event list becomes empty, in the corresponding parallel simulation, we get MinTime = 1 8LP i , and therefore GST = 1. In a similar way, we can argue that when the parallel simulation terminates, the analyzer will also terminate. For condition 2b, we note that the initial LP safe-times in the analyzer are the same as what would have been in the parallel protocol. The subsequent safe-times simply depend on the previous safetimes and xed lookahead values 6. The safetimes at every subsequent step are therefore the same as in the parallel simulation. >From condition 1, the sequential simulation keeps the ordering of executed events within any LP, the same as what will occur in the parallel super-step protocol. We also note that the time-stamp of all events e must satisfy the constraint: denotes the safetime of LP e at super-step i. Every event falls into a unique simulation time interval, because the safetime values partition the simulation period into disjoint intervals. Since the safe-times are computed in the same way as in the parallel super-step protocol, the sequence of events within an LP are clustered into super-steps in the same way as what would happen in the parallel simulation. 5 Conclusion In this concluding section, we brieey discuss issues to investigate further for this analyzer algorithm. The rst issue is the non-zero lookahead assumption. If this assumption is removed, an implication is that simultaneous events may arrive at the same LP in two consecutive super-steps. Such events will be implicitly ordered by the super-step protocol. The sequential simulation has to be made aware of this or it may simulate the events in a wrong order from that in the parallel protocol. 6 Note that the GST calculation in the protocol in Fig. 1 is no longer necessary with our non-zero lookahead assumption. The assumption removal requires the sequential simulation to be further modiied. The idea …