[GHC] #15999: Stabilise nofib runtime measurements

[GHC]

#15999: Stabilise nofib runtime measurements
-------------------------------------+-------------------------------------
        Reporter:  sgraf             |                Owner:  (none)
            Type:  task              |               Status:  new
        Priority:  normal            |            Milestone:  ⊥
       Component:  NoFib benchmark   |              Version:  8.6.2
  suite                              |
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #5793 #9476       |  Differential Rev(s):
  #15333 #15357                      |
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by sgraf):

 I figured out a more useful metric to identify GC wibbles: Watching
 productivity rate change while increasing the nursery.

 Here's the current state of affairs for `paraffins`:

 [[Image(https://i.imgur.com/XxdNJQ8.png)]]

 While there are some wibbles due to short running time, it's clear that
 the function is highly discontinuous. We want something more like this:

 [[Image(https://i.imgur.com/fA4S38S.png)]]

 That's a much smoother curve (which of course is subject to measuring
 inaccuracy). I got this just by iterating `main` 100 times.
 Obviously, this has implications on running time of the benchmark, so the
 `*MODE`s have to be adjusted accordingly.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15999#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler