[GHC] #9221: (super!) linear slowdown of parallel builds on 40 core machine

[GHC]

#9221: (super!) linear slowdown of parallel builds on 40 core machine
-------------------------------------+-------------------------------------
        Reporter:  carter            |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.2.1
       Component:  Compiler          |              Version:  7.8.2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #910, #8224       |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by slyfox):

 I've ran '''synth.bash''' benchmark ('''ghc make -j +RTS -A256M''')
 on current '''ghc-HEAD''' on the following machines:

 - 4x2 HT desktop, speedup is '''3.9x''' compared to '''-j1''' case (best
 at -j8)
 - 12x2 HT haswell VM, speedup is '''12.7x''' compared to '''-j1''' case
 (best at -j24)
 - 8x4 HT Power Power7, speedup is '''8.1x''' compared to '''-j1''' case
 (best at -j12)

 Very similar speedup is achieved in multiprocess mode (via Makefile in
 comment:#65).

 A few facts:
 - Multiprocess run is ~10% faster than multithreaded.
 - Specifying N more than logical CPUs available (say, -j100) in '''ghc
 make -j<N>''' does not lead to severe performance degradation at least on
 up to 12 core hardware.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9221#comment:71>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler