#367: Infinite loops can hang Concurrent Haskell
------------------------------------------+---------------------------------
Reporter: simonpj | Owner: ezyang
Type: bug | Status: new
Priority: lowest | Milestone: _|_
Component: Compiler | Version: 6.4.1
Resolution: None | Keywords: scheduler
allocation
Os: Unknown/Multiple | Architecture: Unknown/Multiple
Failure: Incorrect result at runtime | Difficulty: Unknown
Testcase: | Blockedby:
Blocking: | Related:
------------------------------------------+---------------------------------
Comment(by ezyang):
Here are the aggregate stats for the stack check version (overall, it
seems that it produces more bloated executables, but is a little bit
faster):
{{{
--------------------------------------------------------------------------------
Program Size Allocs Runtime Elapsed TotalMem
----------------------------------------------------------------------------
Min +6.1% -0.1% -7.2% -6.7% -50.0%
Max +6.7% +5.8% +5.9% +6.1% +0.8%
Geometric Mean +6.6% +0.1% +0.1% +0.2% -0.8%
}}}
Switching to stack checks has some interesting interactions with some of
the optimizations; for example, we can no longer optimize away Sp - 0 <
SpLim as true, since SpLim may have been twiddled with. If we run nofib
without adding the extra yields but with those optimizations turned off,
we see:
{{{
Min +5.9% -0.1% -7.3% -7.0% -50.0%
Max +6.7% +5.8% +3.9% +4.1% +0.0%
Geometric Mean +6.6% +0.1% -0.1% -0.1% -0.8%
}}}
So we already pay most of the cost from having to ditch the optimizations.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/367#comment:25>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
