Parallel Systems Architecture Laboratory

Parallel Systems Architecture Laboratory. It is a research institution dedicated to advancing the design and implementation of high-performance computing systems. The laboratory focuses on the intersection of hardware and software to solve complex computational problems across scientific and engineering domains. Its work is integral to progress in fields requiring immense processing power, such as computational fluid dynamics, climate modeling, and genomics.

Overview

The laboratory operates as a center for innovation within the broader ecosystem of academic research and industrial R&D. Its mission encompasses exploring novel architectures that move beyond traditional von Neumann architecture limitations. Researchers at the institution often publish findings in premier venues like the International Symposium on Computer Architecture and IEEE conferences. The work conducted there directly contributes to the evolution of exascale computing initiatives supported by agencies like the United States Department of Energy.

Research Focus

Primary investigations center on manycore processor design, heterogeneous computing, and interconnection networks for large-scale systems. A significant emphasis is placed on energy-efficient computing, seeking to reduce the power consumption of data centers and supercomputers. The lab also pioneers research into resilient computing techniques to manage hardware faults in extreme-scale environments. Additional strands of inquiry include hardware acceleration for machine learning workloads and advanced memory hierarchy designs incorporating non-volatile memory.

Key Projects

Historically, the laboratory has been involved in designing prototype systems for the DARPA Ubiquitous High Performance Computing program. It contributed to the development of the Adaptive Supercomputing vision, exploring reconfigurable computing via FPGA technologies. Another major initiative involved creating simulation frameworks for next-generation network-on-chip architectures, used by partners like Intel and AMD. More recent projects include co-design efforts for quantum-classical hybrid computing systems and architectures optimized for graph analytics.

Facilities and Resources

The laboratory hosts a state-of-the-art computational cluster equipped with the latest GPU accelerators from NVIDIA and AMD. It maintains a suite of electronic design automation tools for VLSI and system-on-chip design, supported by licenses from Cadence Design Systems and Synopsys. A dedicated data center module provides a controlled environment for testing thermal management and power delivery solutions. Researchers have access to extensive benchmark suites and performance analysis tools developed in collaboration with the SPEC organization.

Collaborations and Affiliations

The lab maintains strong ties with industry leaders, including IBM Research, Hewlett Packard Enterprise, and Microsoft Research. It is a key participant in multi-university consortia funded by the National Science Foundation and the Semiconductor Research Corporation. International partnerships exist with institutions like the RIKEN Center for Computational Science in Japan and the Jülich Supercomputing Centre in Germany. The laboratory also works closely with national laboratories such as Lawrence Livermore National Laboratory and Sandia National Laboratories on applied problems.

Notable People

The laboratory has been directed by several influential figures in computer architecture, including contributors to the development of multithreading and cache coherence protocols. Alumni have taken prominent roles at companies like Google, Apple Inc., and Tesla, Inc. in their autopilot hardware divisions. Faculty affiliates have received recognition such as the ACM Gordon Bell Prize and IEEE Fellow status for their contributions to parallel processing. Visiting scientists have included pioneers from the ILLIAC IV project and architects of the Cray supercomputing legacy.