#7450: Regression in optimisation time of functions with many patterns (6.12 to 7.4)? -------------------------------------+------------------------------------- Reporter: iustin | Owner: bgamari Type: bug | Status: patch Priority: normal | Milestone: 7.12.1 Component: Compiler | Version: 7.6.1 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: Phab:D1012 -------------------------------------+------------------------------------- Comment (by bgamari): I finally had a chance to look further into this. It seems that after the fix `CmmPipeline.doRTs` may still be non-linear. With 2048 constructors, {{{ COST CENTRE MODULE %time %alloc lintAnnots CoreLint 12.5 20.4 doSRTs CmmPipeline 11.6 16.9 FloatOutwards SimplCore 6.3 2.0 simplLetUnfolding Simplify 4.5 2.7 pprNativeCode AsmCodeGen 4.3 5.0 completeBind Simplify 4.1 2.0 StgCmm HscMain 3.3 2.7 ... }}} With 4096 constructors, {{{ COST CENTRE MODULE %time %alloc doSRTs CmmPipeline 23.0 33.0 lintAnnots CoreLint 16.9 26.5 FloatOutwards SimplCore 7.8 1.6 completeBind Simplify 4.0 1.2 simplLetUnfolding Simplify 3.9 1.6 pprNativeCode AsmCodeGen 3.3 3.3 ... }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/7450#comment:22 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler