#16122: `round :: Double -> Int64` much slower than `fromIntegral @Int @Int64 . round` -------------------------------------+------------------------------------- Reporter: Fuuzetsu | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.6.3 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 simonpj): Here is what is happening * There is a built-in rule for `round` that rewrites `round (RFDict _ _ r _ _) = r`; that is, `round` applied to a suitable dictionary just selects the method. * Currently, built-in rules are always active; they can't be delayed to a later phase. This could be changed relatively easily, but that's the way it is right now. * Rules are applied to the RHS of other rules, if the phases allow; this seems reasonable. After all, that's what will happen whenever the rule is applied. * In particular, the built-in rule is applied to the RHS of the rule in `GHC.Int`, as comment:3 says. * I think the reason things go wrong is that in the RHS of hte rule in `GHC.Int` you really want to apply the rule `"round/Double->Int"` in `GHC.Float`, but perhaps it is not imported by `GHC.Int`? Anyway, probably the best solution here is to rewrite the rule in `GHC.Int` to {{{ "round/Double->Int64" round = (fromIntegral :: Int -> Int64) . GHC.Float.roundDoubleInt }}} (And maybe the same for `fromIntegral`?) -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/16122#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler