OSU Paper Receives Most Influential PLDI Paper Award
The paper entitled "A practical automatic polyhedral parallelizer and locality optimizer" received the ACM SIGPLAN Most Influential PLDI Paper Award at this year's PLDI conference. The paper describes key developments in the doctoral work of Uday Bondhugula, who was supervised by Professor Sadayappan, CSE. The other authors of the paper, Albert Hartono and J. Ramanujan, are also OSU/CSE Ph.D. alumni advised by Dr. Sadayappan. Details of the award citation may be found below.
The PLDI 2008 paper “A Practical Automatic Polyhedral Parallelizer and Locality Optimizer” significantly advanced the power and practical automatic application of polyhedral analysis to automatically determine parallelism and locality improving affine transforms of regular programs expressed as sequences of possibly imperfect loops. The work presented in this paper produced practical implementations that could automatically generate OpenMP parallel programs from sequential nested loops written in C. This inspired a large number of other researchers to further develop polyhedral optimizations based on this approach, like automatic code generation for GPUs and FPGA circuits by high level synthesis from nested sequential loops. The ideas in this paper have proved to be enduring and widely applicable and even to this day keep finding new applications, e.g., for mapping machine learning algorithms to specialized accelerators.
Pictured below: Professor Sadayappan, Uday Bondhugula and J. Ramanujan