Welcome! We are the Parallel Programming Laboratory.

Our goal is to develop technology that improves performance of parallel applications while also improving programmer productivity. We aim to reach a point where, with our freely distributed software base, complex irregular and dynamic applications can (a) be developed quickly and (b) perform scalably on machines with thousands of processors.

Processor virtualization is one of our core techniques: the programmer divides the computation into a large number of entities, which are mapped to the available processors by an intelligent runtime system. This separation of concerns between programmers and the system is key to attaining both our goals together.

Mitigating Process Variation Through Dynamic Load Balancing

The design and manufacture of present-day CPUs causes inherent variation in supercomputer architectures such as variation in power and temperature of the chips. The variation also manifests itself as frequency differences among processors under Turbo Boost dynamic overclocking. This variation can lead to unpredictable and suboptimal performance in tightly coupled HPC applications. In this study, we use compute-intensive kernels and applications to analyze the variation among processors in four top supercomputers: Edison, Cab, Stampede, and Blue Waters.

As shown in...