#11075: Confusing parallel spark behaviour with safe FFI calls -------------------------------------+------------------------------------- Reporter: duncan | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Runtime System | Version: 7.10.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by duncan): So thoughtpolice has a very sensible suggestion: annotate this info in the FFI call. That is, rather than just safe and unsafe, we can add an extra bit to say it's safe but cpu intensive. Then the scheduler could keep note of this when it makes the new thread for the FFI call. When there are outstanding cpu intensive C calls that this cap is waiting on, then don't consider the cap idle for the purposes of grabbing and converting new sparks. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11075#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler