#9221: (super!) linear slowdown of parallel builds on 40 core machine -------------------------------------+------------------------------------- Reporter: carter | Owner: Type: bug | Status: new Priority: high | Milestone: 7.10.1 Component: Compiler | Version: 7.8.2 Resolution: | Keywords: Operating System: | Architecture: Unknown/Multiple Unknown/Multiple | Difficulty: Unknown Type of failure: Compile- | Blocked By: time performance bug | Related Tickets: #910 Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- Comment (by carter): Yeah, I definitely found that making the nursery larger helped A LOT, i'd done `-A50m` or `-A100m` on a few occasions, though I must admit I didn't think to do that for running GHC itself! Good point about hyperthreads, that would match my experience of only seeing good perf up to about 15-20ish for -N. This does raise the problem that the default -N flag should probably be set to the number of physical cores rather than hyper threaded cores, would that be a reasonable patch for me to write for GHC? (though of course theres the matter of correctly detecting that info in a portable way! ) I unfortunately dont have access anymore to that machine (or presently any SSH access to a mega core sized box) so I can't repro it to dig in deeper, I was using some cli tool who's name escapes me right now -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/9221#comment:13 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler