Hi Alba. Since I've got no answer to my question I think that my strange behavior is a very deep problem. I let you know when I have news. Bye. Luca. On Aug 29, 2011, at 8:35 AM, Alba Marchisio wrote:

Yes, I'm interested too in this problem. Is any body able to answer to the Luca's question?

Alba

2011/8/25 Luca Ciciriello I found a strange behavior in my code

Here a (very) simplified version of my code:

{-# LANGUAGE PArr #-} {-# OPTIONS -fdph-seq #-}

module Main where

import Control.Parallel import GHC.PArr

parArr :: [:String:] parArr = [: "1", "2", "3", "4" :]

isElement :: String -> Bool isElement x = x `elemP` parArr

main :: IO () main = do putStrLn $ (show . isElement) "5"

The aspected result is "False" ("5" isn't a member of parArr), but the result I get is "True".

I've compiled this code with the command: ghc --make -Odph -threaded Test.hs -rtsopts -o test

and launched with: test +RTS -N2

My system is MacOS X 10.7.1 with Xcode 4.1 and GHC 7.0.4 (I've found the same behavior on Windows).

Its obvious I'm missing something. Could someone help me?

Thanks.

Luca. _______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners

Ok, Thanks. This works fine. I found out that also mapP and sumP are working as expected, but elemP and foldP do not behave correctly. Luca. On Aug 29, 2011, at 11:14 AM, Daniel Fischer wrote:

On Monday 29 August 2011, 08:41:35, Luca Ciciriello wrote:

...

Hi Alba. Since I've got no answer to my question I think that my strange behavior is a very deep problem.

Well, perhaps it's not very deep, just not enough people on this list are familiar with dph. In such cases, ask on haskell-cafe or #haskell to reach a wider audience.

Anyway, I'm pretty sure it's a bug in dph. If you change your code to avoid elemP:

isElement :: String -> Bool isElement x = x `elemP1` parArr where elemP1 :: Eq a => a -> [:a:] -> Bool elemP1 v ar = lengthP (filterP (== v) ar) /= 0

you get the expected result(s) [but not with HEAD].

I've opened a ticket: http://hackage.haskell.org/trac/ghc/ticket/5438

...

I let you know when I have news.

Bye.

Luca.

Cheers, Daniel