An evaluation of software multithreading in a conventional distributed memory multiprocessor

Abstract

Multithreading is often seen as a solution to the problem of large memory latencies that occur when remote data is needed for local computation. This paper quantifies the costs and benefits of software multithreading on a distributed memory multiprocessor. We describe the design of a machine-independent software multithreading system as part of a runtime system for a high-level parallel programming language, and present a quantitative analysis of the costs of our multithreading system, as well as its performance on the nCUBE/2 multiprocessor. We show that, in the presence of a sufficient number of remote references to cover the initial costs, our multithreading system provides speedup factors of between 1.27 and 1.65.

Multiprocessor, Multithreading and Memory Optimization for On-Chip Multimedia Applications

“…The performance of cache is studied with a multithreaded virtual processor (MVP) [3]. The performance was improved by tolerating memory latency but also lower cache miss rates due to exploitation of data locality.…”

“…Given their combination of parallel data processing in a multiprocessor system with the high level of integration of System-on-Chip (SoC), they are great candidates for systems such as network processors and complex multimedia platforms [1]. The important amount of data manipulated by these applications requires a large memory size and a significant number of accesses to the external memory for each processor node in the MPSoC architecture [2,3]. Therefore, it is important to optimize, at the applicationlevel, the access to the memory and it is important to observe the effect of multithreading in order to improve processing time and power consumption.…”

“…In these applications, the majority of the area and power cost is due to the global communication and memory interaction [4][5][6]. Indeed, a key area of concentration to handle both real-time and energy/power problems is the memory system [2,3]. The development of new strategies and techniques to decrease memory space, code size and shrink the number of accesses to the memory is necessary.…”

The Nexus Approach to Integrating Multithreading and Communication

“…Multithreading has proven useful for overlapping computation and communication or I/O [7,22], for load balancing [14], and for implementing process abstractions in parallel languages [10,12,25,28,34,45].…”

Evaluating the Performance of Multithreading and Prefetching in Multiprocessors

“…Previous experimental research on multithreading performance shows that multithreading is effective at tolerating memory latencies for some applications [17,33,18,32]. Previous analytical research [2,29,27,14] focuses on The multithreading model shows that 2-4 contexts are sufficient to maximize this technique's potential benefit in the current generation of shared-memory multiprocessors.…”