A summary of research in system software and concurrency at the University of Malta : multithreading

Abstract

Multithreading has emerged as a leading paradigm for the development of applications with demanding performance requirements. This can be attributed to the benefits that are reaped through the overlapping of I/O with computation and the added bonus of speedup when multiprocessors are employed. However, the use of multithreading brings with it new challenges. Cache utilisation is often very poor in multithreaded applications, due to the loss of data access locality incurred by frequent context switching. This problem is compounded on shared memory multiprocessors when dynamic load balancing is introduced, as thread migration also disrupts cache content. Moreover, contention for shared data within a thread scheduler for shared memory multiprocessors has an adverse effect on efficiency when handling fine grained threads. Over the past few years, the System Software Research Group at the University of Malta has conducted research into the effective design of user-level thread schedulers, identifying several weaknesses in conventional designs and subsequently proposing a radical overhaul of the status quo to overcome these deficiencies. Various results have been published in academic conferences and journals [1–4]; this brief report highlights the principal findings. The related problem of communication and I/O bottlenecks in multithreaded systems and contemporary computer systems in general is discussed elsewhere in these proceedings [5].