Attendees: Razya, Sebastian, Li, Jan, David, Konrad, Tobias, Harsha
- Reductions patch (Sebastian)
- 8 spec benchmarks do not pass
- Several test cases committed
- Interchange heuristic patch (Sebastian)
- Pranav takes care of this
- Interchange-8.c fails
- Konrad will look at it, incompatibility between dimensions (strip mining?)
- Dependency checks (Konrad)
- Two test cases crashed. Memory problems. Will build 127 polyhedrons.
- Loop Blocking
- Just scheduling blocking/interchange.
Switch NNC -> C was successful. Improved dependency runtime a lot. (Konrad)
- Exact PPL integer calulations might be useful. But not for gcc 4.5 (Tobi)
- IBM scripts show optimal transformations for the test cases (David)
- Alias sets (Li, Tobi)
- Quadratic is not a problem
- Automatic parallelization
- Outer loop parallelization, first to merge (Razya)
- Force parallel flag to loop_p is difficult
- Merge patch without autopar and parallel flag for next week.
- Graphite merge
- Merge the patches outside of graphite from now on.
- Later work on the final graphite patch
- PCP (Jan)
- in and out of the polyhedral model
- Integration is an issue because of the time constrainst
- Build/Configuration problems.
- Finish testsuite / After this merge to trunk
- Cloog isl (Tobi)
- Work on the interface with Sven and Cedric