Practical limits to Moore's law have been reached as a result of the significance of problems such as quantum tunnelling and heat dissipation associated with increasing integrated circuit element density. However, a reliance has been established on this unsustainable trend in processor clock speeds and performance.
The consensus is that this developmental crisis will be resolved by adopting parallel computing. However, the adaptation of a few decades worth of acquired knowledge, algorithms, tools and infrastructure to the parallel model and the development of effective applications, which can be staged in a scalable structured fashion, capable of exploiting parallelism on multiple architectural levels, is a formidable technical challenge.
Our research involves the development of intelligent coordination frameworks and adaptive architectural patterns for established and emerging parallel architectures that include dynamic heterogeneous deployments of multi-core, multi-processor, and multi-node platforms with Graphics Processing Units.