Richard, Thanks for the clarification. So we ought to see some improvements in the compile time of the polyhedron benchmarks by tomorrow...
http://users.physik.fu-berlin.de/~tburnus/gcc-trunk/benchmark/ The graphite merge should provide some performance improvement regarding loop optimizations, no? Jack On Mon, Jul 28, 2008 at 05:04:56PM +0200, Richard Guenther wrote: > On Mon, 28 Jul 2008, Jack Howarth wrote: > > > Richard, > > Is there any information on what sort of performance improvements > > we might see from the tuples merge? Or would those require additional > > merges like graphite? > > The tuples is supposed to reduce the memory footprint of our tree IL > and as a second order effect reduce compile-time. There are no > performance improvements expected - in fact any changes in performance > are unwanted. > > Richard.