#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): New performance results are in from [https://perf.haskell.org/ghc/#revision/e05b00ef8e05e58e93e0559e4499a830bc51b... GHC speed] {{{ nofib/time/cryptarithm1 0.421 - 4.04% 0.404 seconds tests/alloc/T9020 723162856 + 17.86% 852298336 bytes tests/alloc/T9872b 4754051528 - 3.24% 4600233488 bytes tests/alloc/T9872c 4452756568 - 3.28% 4306667256 bytes Interestingly, `T9020` (long sequence of `return()` commands) was improved by your previous commit. }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Changes (by nomeata): * cc: nomeata (added) -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): OK, well next thing is to use the profiler. That of course may cause the difference to disappear, in which case ticky is next. 17% can't be too hard to find. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Comment (by nomeata): Ah, the problem seems to triggered by different Core, but not in T9020, but rather in GHC.Base. And indeed, the simplifier phase after float out uses more memory with the patch applied. What are the changes that could have caused this? Likely this: Before {{{ 7af6f512e836b8b3376592bbd63e1ae5 $fMonadIO :: Monad IO DFunId {- Strictness: m, Inline: [ALWAYS] CONLIKE, Unfolding: DFun:. @ IO $fMonadIO_$cp1Monad bindIO thenIO returnIO $fMonadIO_$cfail -} }}} After {{{ d1aceb0a764a371523ef868c174b2704 $fMonadIO :: Monad IO DFunId {- Strictness: m, Inline: [ALWAYS] CONLIKE, Unfolding: DFun:. @ IO $fMonadIO_$cp1Monad bindIO $fApplicativeIO_$c*> $fMonadIO_$creturn $fMonadIO_$cfail -} }}} Note that `(>>)` now goes through Applicative, and not through `thenIO` as before. And that is obviously more expensive when simplifying the `return () >> return () >> ...` chains: {{{ 42258bccfed0ab422b11c0c7ba4ead55 $fApplicativeIO_$c*> :: IO a -> IO b -> IO b {- Arity: 3, HasNoCafRefs, Strictness: , Inline: INLINE (sat-args=2), Unfolding: InlineRule (2, False, False) (\ @ a @ b (m :: IO a) (k :: IO b) (eta :: State# RealWorld)[OneShot] -> (bindIO @ a @ b m (\ (ds :: a) -> k)) `cast` (N:IO[0] <b>_R) eta) `cast` (forall (a :: <*>_N) (b :: <*>_N). <IO a>_R ->_R <IO b>_R ->_R Sym (N:IO[0] <b>_R)) -} }}} Also `return` itself is now implemented via {{{ a2f6e8e629337f5fbb3dfb5d4b7caaa8 $fMonadIO_$creturn :: a -> IO a {- Arity: 2, HasNoCafRefs, Strictness: , Unfolding: InlineRule (0, True, True) returnIO1 `cast` (forall (a :: <*>_N). <a>_R ->_R Sym (N:IO[0] <a>_R)) -} }}} I’m attaching both interface dumps, in case anyone wants to have a closer look. Does this analysis sound plausible? -- Ticket URL: GHC The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Changes (by nomeata): * Attachment "iface-after.txt" added. GHC.Base iface with this patch. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): Aha, good work. `-fddump-simpl-ticks` might have nailed it earlier; should have though of that. It should show more inlining of `$fApplicativeIO_$c*>` etc, during some particular simplifier run; so the result is the same but the journey is different. Does it? It's not the casts. It's caused by `Note [CSE for stable unfoldings]`. Note that the IO instances of both `>>` and `*>` get INLINE pragams in `GHC.Base`, so we are (now) very cautious about inlining them, rightly I think. What is more mysterious to me is why the unfolding for `$fApplicativeIO_$c*>` doesn't have `bindIO`, and then `bindIO` inlined into it, which would allow a cascade of further improvements, which would, I think, produce essentially the same code as `thenIO`. The offending corner is this {{{ active_unfolding_gentle id = isInlinePragma prag -- WHY?? && isEarlyActive (inlinePragmaActivation prag) }}} So I tried removing the `isInlinePragma prag` test, and indeed the code becomes identical. Conclusion: * We could change the code in `GHC.Base` to have {{{ *> = thenIO

...

= thenIO }}} No reason why not; it's simple and direct. Does it fix the perf problem in the compiler?

* I'd love to make the above change to `active_unfolding_gentle` and see if any other performance numbers budge. I doubt that anything will change a lot -- it really only affects whether optimisation happens before or after inlining -- but it should improve compiler performance a bit for that very reason. This could be a separate ticket Incidentally, see #5928 which is somewhat related. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781#comment:9 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: closed Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: fixed | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Changes (by nomeata): * status: new => closed * resolution: => fixed -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781#comment:13 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler

#11781: Improve common sub-expression -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: closed Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Resolution: fixed | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D2074 Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): For the record, the actual commit was {{{ commit 5b986a4de288e2c703c38ee37222a7bf3260cc11 Author: Simon Peyton Jones Date: Fri Apr 1 12:24:50 2016 +0200 CSE code cleanup and improvement Triggered by an observation by Joachim, Simon felt the urge to clean up the CSE code a bit. This is the result. (Code by Simon, commit message and other leg-work by Joachim) Differential Revision: https://phabricator.haskell.org/D2074 }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11781#comment:15 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler