Currently, if there's a very long run queue, a minor GC won't have to
#7606: Stride scheduling for Haskell threads with priorities ---------------------------------+------------------------------------------ Reporter: ezyang | Owner: ezyang Type: feature request | Status: new Priority: normal | Milestone: 7.8.1 Component: Runtime System | Version: 7.7 Keywords: | Os: Unknown/Multiple Architecture: Unknown/Multiple | Failure: None/Unknown Difficulty: Unknown | Testcase: Blockedby: | Blocking: Related: | ---------------------------------+------------------------------------------ Comment(by ezyang): traverse a lot of it because it will be in the old generation, but here the GC has to look at every element of the priority queue. We should make a benchmark with a long run queue and see if that's an issue. Based on some investigation, this appears to be probably one of the primary causes of slowdown. I checked to see if this was an issue by checking how often markCapability calls occurred versus actual evacuations performed on TSO objects. Although the number of markCapability calls is the same in both cases, TSO objects get evacuated a lot more frequently in the new world. One of the things that I'm not really clear about is how to tell if I can skip evacuating an object. This GC part of the code is pretty opaque to me, and while I feel like there should be something like a log(n) solution I'm not seeing an implementation path. -- Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7606#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler