Since this one's trivially parallizable, I took a crack at the mandelbrot test case. It was fairly easy to thread it on a per-line basis in Haskell without changing the original too much. It might be more efficient to break the work into larger chunks, but that would require some slightly more complicated changes so I didn't try that yet. Anyway, the code and the results are here. See README for details: http://www.thenewsh.com/~newsham/shootout/ Tim Newsham http://www.thenewsh.com/~newsham/
newsham:
Since this one's trivially parallizable, I took a crack at the mandelbrot test case. It was fairly easy to thread it on a per-line basis in Haskell without changing the original too much. It might be more efficient to break the work into larger chunks, but that would require some slightly more complicated changes so I didn't try that yet.
Anyway, the code and the results are here. See README for details: http://www.thenewsh.com/~newsham/shootout/
a) have you submitted it to the shootout. b) is it faster c) can you put it on the parallel shootout wiki, http://haskell.org/haskellwiki/Shootout/Parallel BTW, I just added parallel spectral-norm, http://haskell.org/haskellwiki/Shootout/Parallel/SpectralNorm
a) have you submitted it to the shootout.
no
b) is it faster
yes
c) can you put it on the parallel shootout wiki, http://haskell.org/haskellwiki/Shootout/Parallel
http://haskell.org/haskellwiki/Shootout/Parallel/Mandelbrot Tim Newsham http://www.thenewsh.com/~newsham/
newsham:
a) have you submitted it to the shootout.
no
b) is it faster
yes
c) can you put it on the parallel shootout wiki, http://haskell.org/haskellwiki/Shootout/Parallel
Nice, on quad core, the old entry, $ time ./Old 6400 > /dev/null ./Old 6400 > /dev/null 6.99s user 0.01s system 99% cpu 7.015 total And Tim's parallel entry, $ time ./A 6400 +RTS -N6 > /dev/null ./A 6400 +RTS -N6 > /dev/null 6.89s user 0.03s system 346% cpu 1.995 total Noice! -- Don
Here's a whack at regex-dna: http://haskell.org/haskellwiki/Shootout/Parallel/RegexDNA only modest speedup (memory bw bound?). A regex engine that could run several machines concurrently in one pass would prob be a big win. Tim Newsham http://www.thenewsh.com/~newsham/
participants (2)
-
Don Stewart -
Tim Newsham