[GHC] #11029: Performance loss due to eta expansion
GHC
ghc-devs at haskell.org
Wed Oct 28 12:46:10 UTC 2015
#11029: Performance loss due to eta expansion
-------------------------------------+-------------------------------------
Reporter: NeilMitchell | Owner:
Type: bug | Status: new
Priority: normal | Milestone:
Component: Compiler | Version: 7.10.2
Resolution: | Keywords:
Operating System: Unknown/Multiple | Architecture:
Type of failure: Runtime | Unknown/Multiple
performance bug | Test Case:
Blocked By: | Blocking:
Related Tickets: | Differential Rev(s):
Wiki Page: |
-------------------------------------+-------------------------------------
Comment (by nomeata):
JFTR, the relevant code is this bit in `CoreArity.lhs`:
{{{#!hs
arityType env (Case scrut _ _ alts)
| exprIsBottom scrut || null alts
= ABot 0 -- Do not eta expand
-- See Note [Dealing with bottom (1)]
| otherwise
= case alts_type of
ABot n | n>0 -> ATop [] -- Don't eta expand
| otherwise -> ABot 0 -- if RHS is bottomming
-- See Note [Dealing with bottom
(2)]
ATop as | not (ae_ped_bot env) -- See Note [Dealing with bottom
(3)]
, ae_cheap_fn env scrut Nothing -> ATop as
| exprOkForSpeculation scrut -> ATop as
| otherwise -> ATop (takeWhile
isOneShotInfo as)
where
alts_type = foldr1 andArityType [arityType env rhs | (_,_,rhs) <-
alts]
}}}
and you can prevent this behaviour with `-fpedantic-bottoms`.
And it looks that I made this eta-expansion more aggressive in
changeset:2931d19e90d2366f2ce308d65a36333336ca6059 when trying to fix
#2915.
--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11029#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
More information about the ghc-tickets
mailing list