#6166: Performance regression in mwc-random since 7.0.x -------------------------------------+------------------------------------- Reporter: bos | Owner: Type: bug | Status: closed Priority: high | Milestone: 8.2.1 Component: Compiler | Version: 7.4.2 Resolution: fixed | Keywords: Operating System: Unknown/Multiple | Architecture: x86_64 Type of failure: Runtime | (amd64) performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Changes (by rwbarton): * status: new => closed * resolution: => fixed Comment: The original version of mwc-random and the minimized test cases in the comments are a bit fragile, in that they rely crucially on `blocks` being floated out of the IO action which contains it, ideally to top level. I'm not sure why, but GHC 7.8.4 isn't doing this floating out in the versions marked SLOW. For better or worse, GHC 7.10.1 does float out `blocks` to top level in the SLOW versions, and I also checked that it produces efficient code for the original version mwc-random-0.13.1.0 that prompted this report. Without a way to reproduce this in a recent version, I'm going to assert that the underlying issue was probably fixed between 7.8 and 7.10. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/6166#comment:21 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler