[GHC] #3758: Huge regression in concurrent app performance and reliability under threaded runtime

[GHC]

#3758: Huge regression in concurrent app performance and reliability under
threaded runtime
-----------------------------+----------------------------------------------
  Reporter:  bos             |          Owner:  simonmar      
      Type:  bug             |         Status:  new           
  Priority:  high            |      Milestone:  6.12.2        
 Component:  Runtime System  |        Version:  6.12.1        
Resolution:                  |       Keywords:                
Difficulty:                  |             Os:  Linux         
  Testcase:                  |   Architecture:  x86_64 (amd64)
   Failure:  Runtime crash   |  
-----------------------------+----------------------------------------------
Comment (by bos):

 Simon, thanks for looking into this.

 I tried the same benchmarks, only this time using a Mac running ab as the
 client over a gigabit ethernet connection.

 With 6.12.1, {{{-N2 -qg}}}, I get good numbers: just below 15,000 requests
 per second.

 With plain {{{-N2}}} (hence using parallel GC), the numbers don't take as
 big a dive as when I was running client and server on a single box, but
 they still drop to about 7,500 requests per second, a 50% reduction in
 throughput.

 Since the server has nothing else running on it in this case, I really do
 have 2 CPUs available. I think :-)

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/3758#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler