Efficient simulation of synthesis-oriented system level designs
Proceedings of the 15th international symposium on System Synthesis
Proceedings of the 14th ACM Great Lakes symposium on VLSI
Optimizing system models for simulation efficiency
Formal methods and models for system design
A fast heuristic scheduling algorithm for periodic ConcurrenC models
Proceedings of the 2010 Asia and South Pacific Design Automation Conference
Accelerating UNISIM-Based Cycle-Level Microarchitectural Simulations on Multicore Platforms
ACM Transactions on Design Automation of Electronic Systems (TODAES)
A conservative approach to systemc parallelization
ICCS'06 Proceedings of the 6th international conference on Computational Science - Volume Part IV
Hi-index | 0.00 |
Simple and powerful modeling of concurrency and reactivityalong with their efficient implementation in the simulationkernel are crucial to the overall usefulness of systemlevel models using the C++-based modeling frame-works.However, the concurrency alignment in most modelingframeworks is naturally expressed along hardwareunits, being supported by the various language constructs,and the system designers express concurrency in their systemmodels by providing threads for some modules/unitsof the model. Our experimental analysis shows that thisconcurrency model leads to inefficient simulation performance,and a concurrency alignment along dataflow givesmuch better simulation performance, but changes the conceptualmodel of hardware structures. As a result, we proposean algorithmic transformation of designs written inthese C++-based environments with concurrency alignmentalong units/modules. This transformation, provided as acompiler front-end, will re-assign the concurrency alongthe dataflow, as opposed to threading along concurrenthardware/software modules, keeping the functionality of themodel unchanged. Such a front-end transformation strategywill relieve hardware system designers from concerns aboutsoftware engineering issues such as, threading architecture,and simulation performance, while allowing them to designin the most natural manner, whereas, the simulation performancecan be enhanced upto almost two times as shown inour experiments.