Re: [GHC] #12367: Commit adding instances to GHC.Generics regression compiler performance

18 Jul 2016

      #12367: Commit adding instances to GHC.Generics regression compiler performance
-------------------------------------+-------------------------------------
        Reporter:  bgamari           |                Owner:  bgamari
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.0.2
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:  Generics
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:                    |
-------------------------------------+-------------------------------------
Changes (by simonpj):

 * cc: ezyang (added)

Comment:

 Great catch!  This is all Edward's fault :-).
 {{{
 commit 4c834fdddf4d44d12039da4d6a2c63a660975b95
 Author: Edward Z. Yang 
 Date:   Mon Nov 17 21:23:52 2014 -0800
 }}}
 The `is_dfun` field of a `ClsInst` should be lazy, only pulled on if the
 instance is actually needed.  See the `forkM` in `TcIface.tcIfaceInst`.

 But this `instIsVisible` stuff is pulling on the `is_dfun` which forces
 lots of stuff to get loaded that is entirely unnecessary.

 Solution: add an `is_mod :: Module` field to `ClsInst`, which gives the
 `Module` for the `ClsInst`.  Invariant: it's the same Module as that for
 the `is_dfun`.  And use that `is_mod` field in `instIsVisible`.

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12367#comment:15
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler