Optimizing parallel simulation of multi-core system
MetadataShow full item record
Multi-core design for CPU is the recent trend and we believe the trend will continue in near future. Researchers and industry architects utilize simulation to evaluate their designs and gain a certain level of confidence before manufacturing the actual products. Due to the fact that modern multi-core systems are complex, traditional sequential simulation can hit the bottlenecks in terms of execution time. To handle the complexity, Parallel Discrete Event Simulation (PDES) programs are employed. PDES program with well-designed partitioning schemes, synchronization algorithm and other optimizations can take advantage of the parallel hardware and achieve scalability for the simulation of multi-core systems. The objective of this dissertation is to design, develop, test and evaluate a variety of technologies to improve the performance and efficiency of parallel simulation of multi-core systems. The technologies include a general guide for partitioning schemes, an efficient front-end for timing-directed simulation, and a new conservative synchronization algorithm.