Who wants to adopt a polyhedral compiler?

                     Albert Cohen

     Institut national de recherche en informatique
                et en automatique (INRIA)
                         and
             Ecole Normale Superieure (ENS)
                    Paris, France


The polyhedral model is a powerful framework for automatic
optimization and parallelization.  Programs are represented using
systems of affine (linear) inequalities, allowing to construct and
search for advanced loop optimizations.

In this talk, we will survey a few challenges for the adoption of
polyhedral techniques in production compilers. We will study ongoing
work to extend the reach of the framework. From dynamic,
data-dependent control flow, to interprocedural analysis and
optimization across function boundaries. We will also discuss new
applications of polyhedral compilation, including just-in-time
specialization, code generation for heterogeneous architectures, and
how these trends may reinforce the push for polyhedral techniques in
modern compilers.