#11068: Make Generic/Generic1 methods inlinable -------------------------------------+------------------------------------- Reporter: glguy | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1447 Wiki Page: | -------------------------------------+------------------------------------- Changes (by glguy): * differential: D1447 => Phab:D1447 -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11068#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11068: Make Generic/Generic1 methods inlinable -------------------------------------+------------------------------------- Reporter: glguy | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1447 Wiki Page: | -------------------------------------+------------------------------------- Comment (by osa1): I remember chatting with some people at ICFP about optimizing Generics code. The problem was that sometimes GHC is able to optimize Generics-based code so that it doesn't use Generics methods anymore. For example, using Generics- based default method of `NFData` sometimes works well, generated code is same as hand-written code. But in some cases it never optimizes enough to eliminate Generics methods. For example, in my tests I was never able to eliminate Generics methods in Generics-based default implementation of `Binary`. I'm wondering if after this patch GHC is able to optimize such instances. @glguy have you tried anything like this? Looks like a good improvement to me, +1. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11068#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11068: Make Generic/Generic1 methods inlinable -------------------------------------+------------------------------------- Reporter: glguy | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1447 Wiki Page: | -------------------------------------+------------------------------------- Comment (by glguy): Hi Simon, The problem I'm working on is indeed related to José's work. I'm solving the problem that GHC doesn't use Appliative and Functor laws to rewrite code using Haskell instead of rewrite rules. I "run" the code at a lifted version of the desired type `f` that uses the Applicative laws to rewrite a term to be left associated in the `<*>`s and using a single `<$>`. Using this technique allows me to derive efficient code that is generic in terms of Functor and Applicative on a case by case basis rather than with global rules. {{{ badTraversal :: Traversal' (Int,Int,Int) Int badTraversal f (x,y,z) = pure (\x' (y',z') -> (x',y',z')) <*> f x <*> ((,) <$> f y <*> f z) goodTraversal :: Traversal' (Int,Int,Int) Int goodTraversal = boggling badTraversal -- generated core is roughly: \f (x,y,z) -> (,,) <$> f x <*> f y <*> f z }}} The code from `badTraversal` could have been easily generated using GHC.Generics, but the efficient version can be recovered using the `boggling` operation defined in repository mentioned above. As far as my request to add the inline pragma, I think I might have misunderstood what GHC was doing. My goal was to ensure that from/to/from1/to1 would be able to inline across module boundaries even when their definitions were large due to the type being large. However after revisiting my test case I was struggling to reproduce that the change fixed anything, so I closed the change request (but forgot to change this ticket) -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11068#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11068: Make Generic/Generic1 methods inlinable -------------------------------------+------------------------------------- Reporter: glguy | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Resolution: | Keywords: Generics, | newcomer, Inlining Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1447 Wiki Page: | -------------------------------------+------------------------------------- Changes (by mpickering): * keywords: Generics => Generics, newcomer, Inlining -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11068#comment:8 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11068: Make Generic/Generic1 methods inlinable -------------------------------------+------------------------------------- Reporter: glguy | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Resolution: | Keywords: Generics, | newcomer, Inlining Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1447 Wiki Page: | -------------------------------------+------------------------------------- Comment (by mpickering): There is at least some investigative work to be done. It seems no one has checked whether the unfoldings of `from`, `to`, `from1` , `to1` were ever making it into the interface files. If they are, then adding an inlinable pragma won't achieve anything but if they aren't then it is probably worth adding the pragma as Pedro suggested. Feel free to remove the tag if you disagree. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11068#comment:10 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler