Distinguished Lecture Series
Keshav Pingali
University of Texas at Austin
Friday October 21, 2011
2230 WEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Title: Towards a Science of Parallel Programming
Keshav Pingali
University of Texas at Austin
Friday October 21, 2011
2230 WEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Title: Towards a Science of Parallel Programming
Abstract
When parallel programming started in the 70's and 80's, it was mostly art: languages such as functional and logic programming languages were designed and appreciated mainly for their elegance and beauty. More recently, parallel programming has become engineering: imperative languages like FORTRAN and C++ have been extended with parallel constructs, and we now spend our time benchmarking and tweaking large programs no one understands to obtain performance improvements of 5-10%. In spite of all this activity, we have few insights into how to write parallel programs to exploit the performance potential of multicore processors.
In this talk, I will argue that these problems arise largely from the limitations of the program-centric abstractions like dependence graphs that we currently use to think about parallelism. I will then propose a novel data-centric abstraction called the operator formulation of algorithms, which reveals that a generalized form of data-parallelism called amorphous data-parallelism is ubiquitous in diverse applications ranging from mesh generation/refinement/partitioning to SAT solvers, maxflow algorithms, stencil computations and event-driven simulation. I will also show that the operator formulation can be used to perform a structural analysis of algorithms that can be exploited for efficient implementations of these algorithms. Finally, I will describe a system based on these ideas called Galois for programming multicore processors.
BIO
Keshav Pingali is a Professor in the Department of Computer Science at the University of Texas at Austin, and he holds the W.A."Tex" Moncrief Chair of Computing in the Institute for Computational Engineering and Sciences (ICES) at UT Austin. He was on the faculty of the Department of Computer Science at Cornell University from 1986 to 2006, where he held the India Chair of Computer Science.
Pingali's research has focused on programming languages and compiler technology for program understanding, restructuring, and optimization. His group is known for its contributions to memory-hierarchy optimization; some of these have been patented and are in use in industry compilers. His current research is focused on programming languages and tools for multicore processors.
Pingali is a Fellow of the IEEE and a Fellow of the American Association for the Advancement of Science. He was the Editor-in-chief of the ACM Transactions on Programming Languages and Systems. He also serves on the NSF CISE Advisory Committee.
When parallel programming started in the 70's and 80's, it was mostly art: languages such as functional and logic programming languages were designed and appreciated mainly for their elegance and beauty. More recently, parallel programming has become engineering: imperative languages like FORTRAN and C++ have been extended with parallel constructs, and we now spend our time benchmarking and tweaking large programs no one understands to obtain performance improvements of 5-10%. In spite of all this activity, we have few insights into how to write parallel programs to exploit the performance potential of multicore processors.
In this talk, I will argue that these problems arise largely from the limitations of the program-centric abstractions like dependence graphs that we currently use to think about parallelism. I will then propose a novel data-centric abstraction called the operator formulation of algorithms, which reveals that a generalized form of data-parallelism called amorphous data-parallelism is ubiquitous in diverse applications ranging from mesh generation/refinement/partitioning to SAT solvers, maxflow algorithms, stencil computations and event-driven simulation. I will also show that the operator formulation can be used to perform a structural analysis of algorithms that can be exploited for efficient implementations of these algorithms. Finally, I will describe a system based on these ideas called Galois for programming multicore processors.
BIO
Keshav Pingali is a Professor in the Department of Computer Science at the University of Texas at Austin, and he holds the W.A."Tex" Moncrief Chair of Computing in the Institute for Computational Engineering and Sciences (ICES) at UT Austin. He was on the faculty of the Department of Computer Science at Cornell University from 1986 to 2006, where he held the India Chair of Computer Science.
Pingali's research has focused on programming languages and compiler technology for program understanding, restructuring, and optimization. His group is known for its contributions to memory-hierarchy optimization; some of these have been patented and are in use in industry compilers. His current research is focused on programming languages and tools for multicore processors.
Pingali is a Fellow of the IEEE and a Fellow of the American Association for the Advancement of Science. He was the Editor-in-chief of the ACM Transactions on Programming Languages and Systems. He also serves on the NSF CISE Advisory Committee.