Effective automatic parallelization of stencil computations
Sriram Krishnamoorthy, Muthu Manikandan Baskaran, Uday Kumar Reddy Bondhugula et al. · Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation · 2007
Performance optimization of stencil computations has been widely studied in the literature, since they occur in many computationally intensive scientific and engineering applications. Compiler frameworks have also been developed that can transform sequential stencil codes for optimization of data locality and parallelism. However, loop skewing is typically required in order to tile stencil codes along the time dimension, resulting in load imbalance in pipelined parallel execution of the tiles. In this paper, we develop an approach for automatic parallelization of stencil codes, that explicitly addresses the issue of load-balanced execution of tiles. Experimental results are provided that demonstrate the effectiveness of the approach.
Cited by 210 works. doi:10.1145/1250734.1250761
Authors
- Sriram Krishnamoorthy
- Muthu Manikandan Baskaran
- Uday Kumar Reddy Bondhugula
- Jagannathan Ramanujam
- Atanas Rountev
- Ponnuswamy Sadayappan
References
- Computational Electrodynamics: The Finite-Difference Time-Domain Method
- Computational electrodynamics, the finite-difference time-domain method
- Computer and Robot Vision
- A data locality optimizing algorithm
- Supernode partitioning
- More iteration space tiling
- Scanning polyhedra with DO loops
- Tile size selection using cache organization and data layout
- Automatic Blocking of Nested Loops
- Implicit and explicit optimizations for stencil computations