High Order Stencil Transformation Suite
These tools utilize associativity and commutativity of operations in regular loop computations to enhance stencil computations. By aggressively reordering accumulations to the output of a stencil using multi-dimensional retiming, better register reuse is achieved and the total load/store operations required per pixel is significantly reduced.
- Kevin Stock
- Martin Kong
- Tobias Grosser
- Louis-Noël Pouchet
- Fabrice Rastello
- J. Ramanujam
- P. Sadayappan
A Framework for Enhancing Data Reuse via Associative Reordering
Kevin Stock, Martin Kong, Tobias Grosser, Louis-Noël Pouchet, Fabrice Rastello, J. Ramanujam, and P. Sadayappan
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI'14), June 2014.
The recent version of the code can be downloaded from the following location:
Please see the included README file for installation instructions. For questions or comments, contact Kevin Stock.
This work was supported in part by the U.S. National Science Foundation through awards 0811457, 0904549, 0926127, 1059417 and 1321147, by the U.S. Department of Energy through award DE-SC0008844, and a Google European Fellowship in Efficient Computing.