The efficacy of a computing innovation over a broad range of applications largely defines its success.
Given the vast diversity in application characteristics, it is evident to predict the next generation target architectures to be heterogeneous---comprising variable granularity computing blocks and varied communication mechanisms. Capturing this heterogeneity within an application is an area of research that has a profound impact. In the architectural domain, sequential processors and parallel FPGAs are two disparate processing paradigms. Breaking down the longstanding rigidity between the two processing models can lead to new dimensions of research and innovations in new computing models. Synergistic operation between applications, architectures and their adaptability is only possible with systematic innovations in programming models, compilers, architectures, and systems that strive to strike a balance between flexibility and heterogeneity.
Heterogeneous computing using CPUs, GPUs & FPGAs.
Automating application-to-architecture mapping
Faster access to innovative solutions and their implementations leads to new scientific advancements. Faster computing capabilities not only accelerate existing applications but also identify novel solutions that were previously infeasible on account of high design time. In addition to speeding up applications, the proposed research plan also aims at reducing energy consumption and enhancing resource efficiency of computing systems. Reduced energy consumption makes a direct impact on costs and portability of computing systems.