Hello Mirko, Friday, May 12, 2006, 4:42:02 PM, you wrote:
PS: I am still curious: does threadDelay use the wall clock or the per-process clock (CPU time)?
I think it uses wall clock time. Proof:
And regardless of the answer - how could one obtain the opposite behaviour? (I don't find this discussed in the visible docs. Or am I missing something?)
use threadDelay in cycle, testing CPU time each time. the following code used by me to measure both wall and CPU time. but i should say that cpu time measurement sometimes gives negative results :) i think it may be because it returns time for CURRENT OS thread, but Haskell runtime sometimes creates new OS thread and continue execute Haskell code in this new created thread (at least under Windows) benchmark h str times action = do prev <- getCPUTime prev2 <- getClockTime action current <- getCPUTime current2 <- getClockTime let secs = fromIntegral (current-prev) / 1e12 secs2 = diffTimes current2 prev2 putStrLn$ str ++ ": " ++ showTime secs2 ++ " (user: " ++ showTime secs ++ ")" showTime secs = showFFloat (Just 3) secs " secs" diffTimes (TOD sa pa) (TOD sb pb) = i(sa - sb) + (i(pa-pb) / 1e12) i x = fromIntegral x -- Best regards, Bulat mailto:Bulat.Ziganshin@gmail.com