[GHC] #13193: Integer (gmp) performance regression?

[GHC]

#13193: Integer (gmp) performance regression?
-------------------------------------+-------------------------------------
        Reporter:  j.waldmann        |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:
Operating System:  Linux             |         Architecture:  x86_64
 Type of failure:  Runtime           |  (amd64)
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by j.waldmann):

 Hi.

 * how to reproduce: add the type declaration that prevents the defaulting
 {{{
 a :: Int -> ST s Int -> ST s Int -> ST s Int -> ST s Int -> ST s Int -> ST
 s Int`
 }}}

 * how big the change is
 {{{
 ghc-6.12.3:  Int:  1.28 sec  Integer: 1.56 sec
 ghc-8.0.2 :  Int:  1.30 sec  Integer: 2.30 sec
 }}}

 * "Integer fit in Int" - when I replace Integer by Int, the program
 produces the exact same result, so I assume there were no overflows. When
 I replace the numerals 1, -1 by something really large (+/-
 1000000000000000000), the execution path should be identical (affine
 transformation). The Int result does no longer agree with the Integer
 result, and the Integer runtime goes up slightly.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13193#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler