Graphite phone call from 2009_02_25
Attendees: Tobias, Konrad, Kenny, Razya, David, Christophe, Albert, Harsha, Sebastian
- data dependence analysis: first we have to fix the data reference analysis that still fails during bootstrap
- cleanups: reviewed the changes that split graphite.c into several files
- PCP integration: discussed the plan, first step is Clast to PCP, and creation of a testsuite from benchmarks
- PCP will ask CLooG to generate parallel loops, CLooG should annotate loops with a parallel flag, finally code generation should translate 1-to-1 parallel loops without failing.
- another approach would be PCP to Cloog to PCP to Poly to analyze and validate the parallelized loops: big overhead that is not worth
- sese cleanups: Kenny will work on cleaning up the patch on sese regions creation