The Changing Landscape of Parallel Computing – I


July 17, 2012


David Patterson, Josep Torrellas, and Tim Mattson


University of Southern California, Intel, University of Illinois at Urbana-Champaign


Juan Vargas from Microsoft chairs this session at Faculty Summit 2012.

For decades, the microprocessor industry improved the performance of on-chip processors by increasing the number of operations executed per CPU cycle. By 2007, it was clear that heat and power would limit further increases in clock rates. Today, it is widely accepted that performance gains can only be obtained by putting more than one processor in a single chip.

Multicore parallelism represents a paradigm shift that breaks the 50-year tradition of conventional sequential programming models and software development. Knowing that research in this area would have deep implications, Microsoft and Intel launched in 2008 the Universal Parallel Computing Research Centers at the University of Illinois at Urbana-Champaign (UIUC) and the University of California, Berkeley. Through this unique collaboration, Microsoft, Intel, UIUC, and UC Berkeley have joined forces to conduct research into all aspects of parallelism at the client, including architectures, operating systems, run-time middleware, programming tools and libraries, and applications.

During this session, researchers from Microsoft, Intel, UC Berkeley, and UIUC share the outcomes of their three-year efforts and present their insights about new directions in the changing landscape of parallel computing.