Summary of Research

I am interested in extracting parallelism from irregular applications.  Specifically, I am investigating compiler and run-time techniques for exploiting data-parallelism in real-world algorithms such as Delaunay mesh generation and refinement.  To this end, I have developed a run-time system, called Galois, which can correctly and efficiently parallelize a wide range of applications.

Publications

Milind Kulkarni, Keshav Pingali and L. Paul Chew. Using Transactions in Delaunay Mesh Generation. In Workshop on Transactional Memory Workloads (WTW), June, 2006

Milind Kulkarni and Keshav Pingali. Scheduling Issues in Optimistic Parallelization. In NSF Next Generation Software (NGS) Workshop, March 2007.

Milind Kulkarni, Keshav Pingali, Bruce Walter, Ganesh Ramanarayanan, Kavita Bala and Paul Chew. Optimistic Parallelization Requires Abstractions. In Programming Languages Design and Implementation (PLDI), June, 2007.