Optimization Opportunities Based on the Polyhedral Model in GRAPHITE

Tobias Grosser

Research output: Contribution to conferencePaperpeer-review

Abstract / Description of output

The polytope model is used since many years to describe standard loop optimizations like blocking, interchange or fusion, but also advanced memory access optimizations and automatic parallelization. Its exact mathematical description of memory accesses and loop iterations allows to concentrate on the optimization problem and to take advantage of professional problem solving tools developed for operational research. Up to today the polytope model was limited to research compilers or source to source transformations. Graphite generates a polytope description of all programs compiled by the GCC. Therefore polytope optimization techniques are not limited anymore to hand selected code pieces, but can actually be applied in large scale on realworld programs. By showing the impact of GRAPHITE on important benchmarks - “How much runtime is actually spent in code, that can be optimized by polytope optimization techniques?" - we invite people to base their current polytope research on GRAPHITE to make these optimizations available to the large set of GCC compiled applications.
Original languageEnglish
Number of pages13
Publication statusPublished - 10 Jun 2009
EventGCC Developers’ Summit 2009 - Montréal, Canada
Duration: 8 Jun 200910 Jun 2009

Other

OtherGCC Developers’ Summit 2009
Country/TerritoryCanada
CityMontréal
Period8/06/0910/06/09

Fingerprint

Dive into the research topics of 'Optimization Opportunities Based on the Polyhedral Model in GRAPHITE'. Together they form a unique fingerprint.

Cite this