#7198: New codegen more than doubles compile time of T3294 -------------------------------------+------------------------------------- Reporter: simonmar | Owner: simonmar Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.4.2 (CodeGen) | Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: #4258 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonmar): The basic problem here is that * The Stg->Cmm code generator generates suboptimal code * The `CmmSink` pass would mostly optimise it away, but it is only run at -O. * The extra code is hurting compile time * The old code generator generated simpler code, and thus was faster It's not obvious what the solution is. We might need several improvements in various places. it's not clear whether the lets should be flattened; if so, this is the job of earlier phases (simplifier or CorePrep). Flattening lets is a tradeoff, and I suspect the simplifier is already doing a reasonable job here, and has just decided not to in this case. In some cases I made the STG->Cmm code generator a bit more clever, so as to generate less code and improve compile time, even though later optimisation phases would have produced the same result. I like this approach because it improves compile times for -O0, relying on later optimisation is brittle, and a few tweaks can have a big effect due to the regular patterns that occur in Cmm. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/7198#comment:9 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler