#7258: Compiling DynFlags is jolly slow -------------------------------------+------------------------------------- Reporter: simonpj | Owner: simonpj Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.6.1 Resolution: | Keywords: deriving-perf Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by alexbiehl): Replying to [comment:77 tdammers]:
It seems that for the misbehaving examples, the offensive cost centre is Stg2Stg. As code size increases, CPU time for this cost centre grows disproportionally in the `read` and `show` examples, while it remains below 40% for the well-behaved ones.
Since you don't compile the examples with profiling on the only "heavy" transformation in Stg2Stg is StgCse. You can try `-fno-stg-cse` to see if it gets better. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/7258#comment:80 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler