The advent of multi-core processors has brought a new importance to developing parallel software and has reinvigorated research into the design and implementation of programming language features for synchronization and concurrency. Why do we need to revisit this area — after all, researchers have been studying parallel programming for over thirty years? One reason is that, historically, research in parallelism has focused on the requirements of scientific computing, server applications and operating systems. In these settings there is often a ready source of work for parallel hardware — different portions of a computation to perform at the same time, different clients to serve, or different processes or devices to interact with. In contrast, the kinds of client-side application that will run on multi-core desktops and laptops present new challenges — not only must the different threads in the system interact with each other safely, but the application must often also be restructured to find sources of parallelism in the first place.
展开▼
