Efficient parallel timing simulation of synchronous models on networks of workstations
โ Scribed by Edwin Naroska; Feipei Lai; Rung-Ji Shang; Uwe Schwiegelshohn
- Publisher
- Elsevier Science
- Year
- 2001
- Tongue
- English
- Weight
- 353 KB
- Volume
- 47
- Category
- Article
- ISSN
- 1383-7621
No coin nor oath required. For personal study only.
โฆ Synopsis
In this paper we address the parallel timing simulation of synchronous VLSI designs on a network of workstations (NOWs). We suggest combining cycle based and conventional timing simulation techniques to achieve fast timing simulation even on NOWs which are typically characterized by low bandwidth and high communication latency. In particular we execute a timing simulator on each node of the NOW and use cycle based simulation to produce synchronization information required by the timing simulators. As synchronization information is generated exclusively by the cycle based simulator there is no need for any communication between the timing simulators. To verify the feasibility and performance of our approach we simulated several circuits using our approach. The results show that a signiยฎcant speedup can be achieved even for very small circuits.
๐ SIMILAR VOLUMES
The network of workstations (NOW) we consider for scheduling is heterogeneous and nondedicated, where computing power varies among the workstations and local and parallel jobs may interact with each other in execution. An effective NOW scheduling scheme needs sufficient information about system hete
Parallel applications can be executed using the idle computing capacity of workstation clusters. However, it remains unclear how to schedule the processors among different applications most effectively. Processor scheduling algorithms that were successful for shared-memory machines have proven to be
Performing run-time parallelization on general networks of workstations (NOWs) without special hardware or system software supports is very dicult, especially for DOACROSS loops. Parallelizing DOACROSS loops at run time implies complex operations for dependence detection, data accesses, and computat
Cluster algorithms have application in diverse areas, including statistical mechanics of polymer solutions, spin models in physics, and the study of ecological systems. Most parallel cluster labeling algorithms are designed for SIMD and MIMD multiprocessors and based on relaxation methods. We presen