Hey, I'm a CS Master's student at Cal Poly Pomona focusing on programming language theory & implementation -- particularly optimizations. Particularly optimization "frameworks": ways of simplifying compilers' often repetitive, ad-hoc code into a single conceptual model.
This summer, I've been reading the source of Factor's compiler on again / off again. I'm most of the way through compiler/tree (I've read up through escape-analysis in depth), though I've given all of the code many look-overs. I wouldn't dare call myself an expert in Factor, but reading it has become pretty natural and I'm not at that "I hear it's good for pushing stacks, but you can't do variables" phase. :) This upcoming quarter (Fall 2010-11) I'm working on my thesis, and one paper keeps getting stuck in my mind: Composing Dataflow Analyses and Transformations by Lerner, Grove, and Chambers ( http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.11.6534 ). It presents a unified way of writing dataflow optimizations so they automatically compose together in mutually beneficial ways, even though they're defined separately with no special knowledge of each other. It's the stuff behind the GHC guys' Hoopl framework, which is apparently being merged into their new backend: http://www.cs.tufts.edu/~nr/pubs/dfopt-abstract.html (a newer version is linked there, too). With their powers combined, they form! ...an idea my thesis advisor seemed receptive to. Would overhauling the dataflow analyses & optimizations of the Factor compiler be useful or interesting to you guys, even just to see the results? Or am I way off base here? If so, any better ideas? I'm finding polyhedral optimization interesting, but if the guys at GCC and LLVM take such a long time coding their frameworks, I can only imagine how terribly I'd do. :P Thanks for your feedback, -- Alex Vondrak ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ Factor-talk mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/factor-talk
