#5859: unsafeInterleaveIO duplicates computation when evaluated by multiple
threads
---------------------------------+------------------------------------------
Reporter: joeyadams | Owner:
Type: bug | Status: new
Priority: high | Milestone: 7.4.2
Component: libraries/base | Version: 7.2.2
Keywords: | Os: Unknown/Multiple
Architecture: Unknown/Multiple | Failure: Incorrect result at runtime
Difficulty: Unknown | Testcase:
Blockedby: | Blocking:
Related: |
---------------------------------+------------------------------------------
Comment(by simonpj):
This does indeed look bogus. I've done a little investigation, which I'm
going to record here so I don't forget it.
Fundamentally the problem is the "state hack". (Compile with `-fno-state-
hack` and the problem goes away.) Consider
{{{
let x = factorial 1000 in replicate 10 (print x)
}}}
After a bit of inlining we get something like
{{{
let x = factorial 1000 in replicate 10 (\s. wprint x s)
}}}
where `s :: State# RealWorld` is a state token. Now the float-in pass
pushes the let-binding inwards, to give
{{{
replicate 10 (\s. wprint (factorial 1000) s)
}}}
and we are dead: we evaluate `(factorial 1000)` each time round the loop.
Why does float-in do this? Because the state hack says that the type of
`s` means it is only applied once, which is patently false.
I think the solution is the narrow (once more) teh scope of the state
hack, so that it's only used in arity expansion. I know how to do this
but have to set up nofib runs to compare before and after.
Thanks for the report. Hacks usually bit you in the end.
Simon
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5859#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
