Efficiency and Predictability
The multicore platforms contain a large number of shared resources. At the on-chip level, there are multiple processor cores and caches integrated on a single chip, and at the cross-chip level, there are processor chips interconnected together with memory hierarchies, buses, and networks of various capabilities and speeds. An application will consist of a large number of threads, that interact with each other through the shared resources. These shared resources are not only the performance bottleneck, but also the obsticle to predict the system behaviours such as timing for safety-critical applications. As the number of shared resources increases, resource management will become increasingly challenging. In this research track, we investigate how the wide variety of resources can be abstracted, modelled, managed and analyzed. Our research aims at
- Techniques to develop high-performance software on multicores through resource modeling, code transformation, and self-adapting run-time resource allocation.
- Techniques to develop deterministic software on multicores and also techniques to predict and estimate bounds on timing and resource usage of multicore software.