#5926: Add strict versions of modifyIORef and atomicModifyIORef
---------------------------------+------------------------------------------
Reporter: joeyadams | Owner: simonmar
Type: feature request | Status: new
Priority: normal | Milestone: 7.6.1
Component: libraries/base | Version: 7.4.1
Keywords: | Os: Unknown/Multiple
Architecture: Unknown/Multiple | Failure: Runtime performance bug
Difficulty: Unknown | Testcase:
Blockedby: | Blocking:
Related: |
---------------------------------+------------------------------------------
Changes (by simonmar):
* owner: => simonmar
* difficulty: => Unknown
* milestone: => 7.6.1
Comment:
Ok, I'll apply your patch as is. There's another variant of strict
`atomicModifyIORef` that I have been wanting to experiment with:
{{{
atomicModifyIORef' :: IORef a -> (a -> (a,b)) -> IO b
atomicModifyIORef' (IORef (STRef r#)) f = IO $ \s0 -> loop s0
where
loop s0 =
case readMutVar# r# s0 of { (# s1, a #) ->
let !(a',b) = f a in
case casMutVar# r# a a' s1 of { (# s2, ok, _ #) ->
case ok of { 0# -> (# s2, b #); _ -> loop s2 }}}
}}}
This is like a mini-STM transaction that evaluates `f a` and then attempts
to compare-and-swap the result in. Since evaluating `f a` might take an
arbitrary amount of time, the CAS might fail, and we have to loop. On the
other hand, using this we can avoid ever writing thunks into the `IORef`,
which avoids any problems with black holes.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5926#comment:1>
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
