Putting Polyhedral Loop Transformations to Work
Cédric Bastoul, Albert Cohen, Sylvain Girbal, Saurabh Sharma, Olivier Temam
To appear at
16th Workshop on Languages and Compilers for Parallel Computing (LCPC03), College Station, TX, 2-4 October 2003
Full Text, Printable Abstract.
Abstract
We seek to extend the scope and efficiency of iterative compilation
techniques by searching not only for program transformation
parameters but for the most appropriate transformations themselves.
For that purpose, we need to find a generic way to express program
transformations and compositions of transformations. In this
article, we introduce a framework for the polyhedral representation
of a wide range of program transformations in a unified way. We also
show that it is possible to generate efficient code after the
application of polyhedral program transformations. Finally, we
demonstrate an implementation of the polyhedral representation and
code generation techniques in the Open64/ORC compiler.