> | It would be nice to make Ben's register allocator available for > testing on > | the 6.8.x series, as it will give better results in some case (albeit > only > | a few at the moment). Since most of the code is already in the > branch, I > | don't think it's a problem to keep pushing patches in. > > The reason for caution is that Ben and Norman are fiddling with the > same code. So they'll be ok on the head, but Ben's patches (without > Norman's) might not apply smoothly to the branch
About half of the new register allocator code is a generic graph colouring library that doesn't ever see the Cmm IR, or even knows what a register is.. It looks like Norman's mega-patch touched some of the interface between Cmm and graph land but not the generic graph stuff itself, which is fairly separate. I should be able to push improvements to the graph algorithms into 6.8 which are separate from what Norman is doing. I'll make sure to validate on 6.8 as well as the head - sounds ok? Ben. _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
