#13390: Strict literal float-out during desugaring regresses T1969 at -O0 -------------------------------------+------------------------------------- Reporter: bgamari | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: 8.4.1 Component: Compiler | Version: 8.0.1 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by dfeuer): [[Commentary/Compiler/Core2CorePipeline]] indicates that in desugaring we only inline "non-recursive bindings that are used only once or where the RHS is trivial". I don't see how to understand what Ben showed based on that. But if a little more inlining is happening for some reason, here's one potential story. Without floating, imagine that the (default-derived) method definitions are somehow inlined into the dictionary, and then `$dme` and `$dmd` are inlined. Then everything squashes down nicely. With floating, something else inexplicable seems to be happening: the default definitions inline into `$ce_a1ik` and `$cd_a1ib`, and somehow `$cc_a1i7` (which is now really small, but surprisingly non-trivial) inlines into them. Glancing at the notes and code for the gentle simplification, I do not understand why any of these inlinings are happening, but maybe someone else does. Or maybe I'm completely misinterpreting. I need to build both commits and dump inlinings. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13390#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler