Affiliation:
1. Department of Computer Science, University of Pittsburgh, Pittsburgh, Pa
Abstract
The efficient implementation of high level languages which support concurrency is imperative, particularly in such application areas as embedded real-time systems. Various storage management schemes have been proposed for the implementation concurrency. In this work, a methodology is presented for evaluating the performance of these schemes including the standard heap, Berry-heap, cactus stack and quantized heap when used for Ada programs. The methodology involves the generation of workload traces from an executing program and the use of simulators of the alternative storage management schemes to produce measures of the performances of the schemes. The performance evaluation metrics used are space, time and the space-time product.
Publisher
Association for Computing Machinery (ACM)