#haskell - Mon 19 Feb 2007 between 21:36 and 21:40

NY Lost Funds



xicgvdm_other: there is lots of research going on
donsconcurrency and parallelism is very very hot in ghc right now.
Pseudonymdons: I disagree with that. It's usually not bug-ridden.
donsin particular, the data parallel arrays library. thanks MS for funding that
gvdm_otherI was thinking that the big win would be that when programming in haskell you don't even have to think about concurrency and the compiler figures it all for you
Pseudonymdons: The rest is true, though.
donsgvdm_other: well known, intractible problem....
soreargvdm_other: you still have to think about concurrency ...
DariusPseudonym: At the rate they go, they're usually not informally-specified or ad-hoc either.
PseudonymNot true.
DariusHalf-completed, yes.
donsyou can give hints though, that the compiler will figure out
xicgvdm_other: lots of research right now trying to make that a reality, but it's harder then it seems
soreargvdm_other: but you don't have to think about concurrency *bugs*
PseudonymGenerally speaking, you just need a couple of dependent type thingies.
gvdm_otherwould make haskell well suited to the future (once again, the concentration on multi core stuff)
donsgvdm_other: right. have you seen the multicore ghc paper?
SamBPseudonym: how do you informally specify types in Haskell?
gvdm_othernope
link me?
PseudonymSamB: As a matter of fact, I removed one such example from Edison.
gvdm_other?where multicore
lambdabotI know nothing about multicore.
soreargvdm_other: `par` never makes your program nondeterministic or buggy. It only affects performance.
PseudonymThat was the first thing I did.
soreargvdm_other: unlike threading, which introduces bugs.
PseudonymPatricia tries only work on integer types.
donsgvdm_other: tart here, http://haskell.org/haskellwiki/Libraries_and_tools/Concurrency_and_parallelism
lambdabotTitle: Libraries and tools/Concurrency and parallelism - HaskellWiki, http://tinyurl.com/wvlsw
Dariussorear, gvdm_other: Hence the typical distinction between concurrency and parallelism (as is likely discussed on that page).
donsthe paper I referred to, for the new magical parallel arrays, is , http://www.cse.unsw.edu.au/~chak/papers/CLPKM06.html
lambdabotTitle: Research Papers of Manuel Chakravarty
PseudonymBecause H98 doesn't even have fundeps, Chris Okasaki typed this by adding a polymorphic key as a phantom type, created one typeclass instance (for Int), and added a comment that you really shouldn't add any other instances.
That's pretty informal.
dons> let nfib 0 = 1; nfib 1 = 1; nfib x = a `par` b `seq` (a+b) where a = nfib (x - 2); b = nfib (x - 1) in nfib 20
lambdabot10946
dons^^ parallel, pure code!
gvdm_other: note the `par` annotation to fork a thread..
gvdm_otheri see it
xicgvdm_other: another interesting thing about haskell is STM

Page: 5 12 19 26 33 40 47 54 61 68 75 82 89 

IrcArchive

NY Lost Funds