Hi Dave, --- David Greene <greened at obbligato.org> wrote:> Roman Levenstein wrote: > > > Yes, the SUIF allocator (or at least one of them) implements the > > Smith/Ramsey/Holloway algorithm. This is what I understand by > looking > > at the coloring reggister allocator code in the SUIF distribution. > The > > SUIF code of this allocator is not very well commented, but I'm > pretty > > sure. > > Do you have a pointer? I'm interested in this algorithm as well.I downloaded the MachSUIF sources from here: http://www.eecs.harvard.edu/hube/software/software.html More precisely, I took this version: 2.02.07.15 Have a look at the color.cpp in the raga subdirectory. This is the implementation of Smith/Ramsey/Holloway algorithm, as far as I understand. -Roman ____________________________________________________________________________________ Never miss an email again! Yahoo! Toolbar alerts you the instant new Mail arrives. http://tools.search.yahoo.com/toolbar/features/mail/
Hi David, I'm currently working on an implementation of the Smith/Ramsey/Holloway algorithm for LLVM 1.9, to use for comparison with a PBQP based allocator I'm also writing. The allocator is almost complete, however it's a little messy (since it was only for personal use) and only supports X86 at the moment. It might me handy as a reference for you though. Let me know if you want a copy of the code. Cheers, Lang.
Hi, I'm not very familiar with the code generation, but this seems like a good opportunity to ask... Has anyone tried optimal graph coloring algorithms? Yeah, it's NP, but the chances are that you are not looking at that many variables at the same time. I'm wondering how much could one improve the register allocation by using optimal algorithms, and how much impact would that have on the speed of the produced native code? Thx, Domagoj Babic