[GHC] #427: Random.StdGen slowness
GHC
cvs-ghc at haskell.org
Wed Apr 3 19:19:06 CEST 2013
#427: Random.StdGen slowness
--------------------------------------+-------------------------------------
Reporter: remit | Owner: rrnewton
Type: bug | Status: new
Priority: normal | Milestone: _|_
Component: libraries/random | Version:
Resolution: None | Keywords:
Os: Unknown/Multiple | Architecture: Unknown/Multiple
Failure: Runtime performance bug | Difficulty: Unknown
Testcase: | Blockedby:
Blocking: | Related:
--------------------------------------+-------------------------------------
Comment(by joell):
This ticket seems to be almost describing two separate issues: the
performance of the random package and the "thunk-bomb" problem in
`getStdRandom`. While I admit I haven't looked into the performance
aspect, I am now well acquainted with the "thunk-bomb" aspect.
Turns out a metrics library we were using at my company made a `randomRIO`
call that in some cases (all of them in our configuration) wouldn't
actually use the generated random value. This led to me spending a few
days trying to figure out why our services were consuming and retaining
gigabytes of memory. (For those interested in case studies, the
discussion with the metrics library folks is here:
https://github.com/brendanhay/network-metrics/issues/4 .)
So I would like to petition that the change Remi suggested in comment:17
to make `getStdRandom` strict (or something along those lines) be made.
To my mind (heavily biased by the experience of the last few days), I
wouldn't really consider the "thunk-bomb" aspect a performance problem so
much as a stability one.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/427#comment:20>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
More information about the ghc-tickets
mailing list