[GHC] #13536: Program which terminates instantly in GHC 8.0.2 runs for minutes with 8.2.1

[GHC]

#13536: Program which terminates instantly in GHC 8.0.2 runs for minutes with 8.2.1
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  highest           |            Milestone:  8.2.1
       Component:  Compiler          |              Version:  8.1
      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 rwbarton):

 Well this is quite interesting. The culprit is apparently the new STG CSE
 pass. It's fast with `-fno-stg-cse` and slow with `-fstg-cse`, and the
 only change in the STG is
 {{{#!diff
                                                      case eta_s5Di of {
                                                        (,) a_s5Dn
 [Occ=Once] b_s5Do [Occ=Once] ->
 -                                                          (,) [a_s5Dn
 b_s5Do];
 +                                                          eta_s5Di;
                                                      };
 }}}
 which is the body of
 {{{#!hs
   unmodel (a,b) = (unmodel a, unmodel b)
 }}}

 It certainly makes sense that STG CSE would do that and that it couldn't
 be done in Core-level CSE, because the types `Model Bool` and `Bool` are
 different there; but I don't yet understand why it is ''bad'' in this
 program.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13536#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler