Machine Abstractions and Locality Issues in Studying Parallel Systems
نویسندگان
چکیده
We define a set of overhead functions that capture the salient artifacts representing the interaction between parallel application characteristics and architectural features. An execution-driven simulation testbed is used to separate these overheads in a parallel system. Using this testbed and a set of applications, we address two important issues. The first concerns the use of machine abstractions for performance studies of parallel systems. The second deals with quantifying the impact of locality on the performance of applications. The key conclusions from this study are that the newly proposed model LogP is an effective one for abstracting the network, and that ignoring locality can significantly affect the application performance.
منابع مشابه
Massively Parallel Computing: Mathematics and Communications Libraries Massively Parallel Computing: Mathematics and Communications Libraries
Massively parallel computing holds the promise of extreme performance. The utility of these systems will depend heavily upon the availability of libraries until compilation and run{ time system technology is developed to a level comparable to what today is common on most uniprocessor systems. Critical for performance is the ability to exploit locality of reference and eeective management of the...
متن کاملOn Current Chip Multi-processor Systems
The last few years have seen multicore architectures emerge as the defining technology shaping the future of high-performance computing. Although multicore architectures present tremendous performance potential, to realize the true potential of these systems, software needs to play a key role. In particular, high-level language abstractions and the compiler and the operating system should be ab...
متن کاملLat E X Documents for the Machine Intelligence and Pattern Recognition Series Defon Supporting Rule-based Image Interpretation Using a Distributed Memory Multicomputer
High-level tasks in image understanding are similar to many AI problems in terms of complexity and solution techniques. In particular, rule based systems have been employed for image interpretation using domain knowledge. Past research eeorts on the hardware support of rule-based systems typically assume that little data access locality can be explored. Thus, the Rete network is widely used, an...
متن کاملLat E X Documents for the Machine Intelligence and Pattern Recognition Series on Supporting Rule-based Image Interpretation Using a Distributed Memory Multicomputer
High-level tasks in image understanding are similar to many AI problems in terms of complexity and solution techniques. In particular, rule based systems have been employed for image interpretation using domain knowledge. Past research eeorts on the hardware support of rule-based systems typically assume that little data access locality can be explored. Thus, the Rete network is widely used, an...
متن کاملTwo-stage fuzzy-stochastic programming for parallel machine scheduling problem with machine deterioration and operator learning effect
This paper deals with the determination of machine numbers and production schedules in manufacturing environments. In this line, a two-stage fuzzy stochastic programming model is discussed with fuzzy processing times where both deterioration and learning effects are evaluated simultaneously. The first stage focuses on the type and number of machines in order to minimize the total costs associat...
متن کامل