[GHC] #12545: Compilation time/space regression in GHC 8.0/8.1 (search in type-level lists and -O)

[GHC]

#12545: Compilation time/space regression in GHC 8.0/8.1 (search in type-level
lists and -O)
-------------------------------------+-------------------------------------
        Reporter:                    |                Owner:
  mikhail.vorozhtsov                 |
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:  Inlining
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Several points here

 * We don't yet know why 8.0 is worse than 7.10.  That would e helpful.

 * The coercions record a proof of all the list manipulations involved.
 It's not surprising that they can get big.  What is surprising is why they
 are bigger with 8.0

 * GHC includes a "coercion optimiser" `OptCoercion.hs`, which tries to
 shrink a big coercion to a smaller but equivalent coercion. It's possible
 that 8.0 is somehow missing a crucial optimisation that fires in 7.10.

 * We also have a plan, recorded in Trac #8095, to discard coercions
 altogether.  But neither Richard or I have found time to execute on that
 plan.  See also #11598

 Bottom line: more insight into what is going on would be really helpful.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12545#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler