We document the design and implementation of a "production" incremental garbage collector for GHC 6.02. It builds on our earlier work (Non-stop Haskell) that exploited GHC's dynamic dispatch mechanism to hijack object code pointers so that objects in to-space automatically scavenge themselves when the mutator attempts to "enter" them. This paper details various optimisations based on code specialisation that remove the dynamic space (and associated time) overheads that accompanied our earlier scheme. We detail important implementation issues and provide a detailed evaluation of a range of design alternatives in comparison with Non-stop Haskell and and GHC's current generational collector. We also show how the same code specialisation techniques can be used to eliminate the write barrier in a generational collector.
pubs.doc.ic.ac.uk: built & maintained by Ashok Argent-Katwala.