Parallel Computing Laboratory

Parallel Computing Laboratory
Name	Parallel Computing Laboratory
Established	2008
Research field	Parallel computing, manycore architectures, programming models
Location	University of Illinois Urbana-Champaign
Affiliations	Intel, Microsoft, AMD, Cray

Parallel Computing Laboratory. The Parallel Computing Laboratory is a major academic-industry research consortium established to address the fundamental challenges of the manycore computing era. Founded in 2008 and based at the University of Illinois Urbana-Champaign, its mission is to develop new hardware architectures, software systems, and programming models that can efficiently harness hundreds of processor cores. The lab's work is widely recognized for its influence on the design of future high-performance computing systems and mainstream commercial hardware.

Overview

The laboratory was launched through a significant partnership between the University of Illinois Urbana-Champaign and Intel, with foundational contributions from researchers at the University of California, Berkeley and Microsoft. Its creation was a direct response to the end of Dennard scaling and the shift in the semiconductor industry towards increasing core counts rather than single-core clock speeds, a trend often referred to as the "manycore threshold." The lab operates under the umbrella of the Center for Extreme-Scale Computation and collaborates closely with the National Center for Supercomputing Applications. Its research philosophy emphasizes a holistic, co-design approach, tightly integrating advancements in computer architecture, compilers, and operating systems to overcome performance and programmability barriers.

Research Focus

The primary research focus is on overcoming the limitations of existing parallel programming paradigms for emerging manycore and heterogeneous systems. A core tenet is the "3Cs" research strategy, which targets improvements in **Correctness**, by developing tools to manage concurrency bugs; **Composability**, enabling software components to work efficiently together; and **Continuity**, ensuring software longevity across hardware generations. Key technical challenges include managing memory hierarchy, power and energy efficiency, and data races in highly concurrent environments. The lab's research agenda is heavily informed by the need to support future exascale systems and pervasive parallel computing in devices from smartphones to supercomputers.

Hardware Infrastructure

Researchers have access to a sophisticated array of experimental hardware platforms for prototyping and benchmarking. This infrastructure includes large-scale clusters of Intel Xeon Phi manycore processors and AMD APUs, along with custom research chips designed in collaboration with Intel Labs. The lab utilizes FPGA-based emulation systems for rapid architectural exploration and validation. Furthermore, it has conducted significant experiments on pre-production systems from vendors like Cray and IBM, and has contributed to the design of the Blue Waters supercomputer. This direct access to cutting-edge and prototype silicon is critical for its co-design methodology.

Software and Programming Models

A major output has been the development of innovative software frameworks and programming models designed for productivity and performance. The lab is the birthplace of the **Parallel Research Kernels**, a benchmark suite used to evaluate parallel hardware and software. It has made substantial contributions to the open-source LLVM compiler infrastructure, particularly for parallel code generation and optimization. Researchers have also developed new models for task parallelism, such as the **BOCR** system, and have advanced runtime systems for managing dynamic, fine-grained parallelism on heterogeneous architectures. Work on formal methods and tools like the DPOR algorithm aims to ensure correctness in concurrent software.

Key Projects and Applications

Notable projects include the **Runnemede** manycore architecture project, which explored scalable on-chip networks and memory systems. The **Perfect** suite, funded by the DARPA, focused on power efficiency for embedded systems. The lab's research has directly influenced commercial products, including features in Intel's TBB library and Microsoft's .NET Framework parallel extensions. Application-driven research spans domains such as computational fluid dynamics, molecular dynamics simulations, and machine learning algorithms, demonstrating the real-world impact of its foundational work.

Collaborations and Affiliations

The laboratory maintains a vast network of strategic partnerships with leading technology corporations and government agencies. Its founding and sustained support comes from a consortium including Intel, Microsoft, AMD, and Cray. It also receives research funding from the U.S. Department of Energy and the National Science Foundation. Academic collaborations extend beyond University of Illinois Urbana-Champaign to include the University of California, Berkeley, the University of Texas at Austin, and international partners. These affiliations ensure its research remains grounded in industrial realities while pursuing long-term, transformative academic goals.

Category:Computer science laboratories Category:Parallel computing Category:University of Illinois Urbana-Champaign