#10034: Regression in mapM_ performance -------------------------------------+------------------------------------- Reporter: dfeuer | Owner: ekmett Type: bug | Status: new Priority: highest | Milestone: 7.10.1 Component: Core Libraries | Version: 7.10.1-rc2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: Phab:D632 -------------------------------------+------------------------------------- Changes (by simonpj): * priority: normal => highest Comment: I'm very unhappy with applying this patch without any comments, and without any understanding of what is going on. Now `mapM_` has an INLINE pragma that may or may not be necessary, and an eta-expanded argument, and no one understands which of the two is important, if either. Maybe there is a simplifier bug underlying this, which is simply being covered up. Please please can we have a reproducible test case that shows the regression. Ideally one that just showed it, but if it's necessary to switch Call Arity off (which you hypothesise is making the test case harder to create) then do that. But please let's NOT just cover up the problem. I'm quite inclined to revert the patch, just to serve as a forcing function to try to elicit this information. For now I'll just increase the priority to try to make sure we notice it. I know it's a nuisance, but this is an example of what people mean by incurring technical debt. Thanks! Simon -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10034#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler