<html><body><div style="color:#000; background-color:#fff; font-family:arial, helvetica, sans-serif;font-size:12pt"><div><div>Oh, my - what an indentation :-)</div><div><br></div><div>New try:</div><div><br></div><div>----- Videresendt melding ----</div><div>Fra: Fixie Fixie <fixie.fixie@rocketmail.com></div><div>Til: "haskell-cafe@haskell.org" <haskell-cafe@haskell.org> </div><div>Kopi: Clark Gaebel <cgaebel@uwaterloo.ca> </div><div>Sendt: Torsdag, 29. november 2012 20.57</div><div>Emne: Vedr: [Haskell-cafe] To my boss: The code is cool, but it is about 100 times slower than the old one...</div><div><br></div><div>Sure, the code is from the url I mentioned, both in C and in Haskell.</div><div><br></div><div>It allready seems like the haskell-version should run fast to me - it uses unboxed arrays and even unsafe functions to make it run faster. Well, have a
look:</div><div><br></div><div>C-version:</div><div><br></div><div>#include <stdint.h></div><div>#include <stdio.h></div><div><br></div><div><br></div><div>#define VDIM 100</div><div>#define VNUM 100000</div><div><br></div><div><br></div><div><br></div><div>uint64_t prng (uint64_t w) {</div><div> w ^= w << 13;</div><div> w ^= w >> 7;</div><div> w ^= w << 17;</div><div> return w;</div><div>};</div><div><br></div><div>void kahanStep (double *s, double *c, double x) {</div><div> double y, t;</div><div> y = x - *c;</div><div> t = *s + y;</div><div> *c = (t - *s) - y;</div><div> *s = t;</div><div>}</div><div><br></div><div>void kahan(double s[], double c[]) {</div><div> for (int i = 1; i <= VNUM; i++) {</div><div> uint64_t w =
i;</div><div> for (int j = 0; j</div><div> < VDIM; j++) {</div><div> kahanStep(&s[j], &c[j], w);</div><div> w = prng(w);</div><div> }</div><div> }</div><div>};</div><div><br></div><div><br></div><div>int main (int argc, char* argv[]) {</div><div> double acc[VDIM], err[VDIM];</div><div> for (int i = 0; i < VDIM; i++) {</div><div> acc[i] = err[i] = 0.0;</div><div> };</div><div> kahan(acc, err);</div><div> printf("[ ");</div><div> for (int i = 0; i < VDIM; i++) {</div><div> printf("%g ", acc[i]);</div><div> };</div><div> printf("]\n");</div><div>};</div><div><br></div><div>And the haskell
version:</div><div><br></div><div>{-# LANGUAGE CPP, BangPatterns #-}</div><div>module Main (main) where</div><div><br></div><div>#define VDIM 100</div><div>#define VNUM 100000</div><div><br></div><div>import Data.Array.Base</div><div>import Data.Array.ST</div><div>import Data.Array.Unboxed</div><div>import Control.Monad.ST</div><div>import GHC.Word</div><div>import Control.Monad</div><div>import Data.Bits</div><div><br></div><div>prng :: Word -> Word</div><div>prng w = w'</div><div> where</div><div> !w1 = w `xor` (w `shiftL` 13)</div><div> !w2 = w1 `xor` (w1 `shiftR` 7)</div><div> !w' = w2 `xor` (w2 `shiftL` 17)</div><div><br></div><div>type Vec s = STUArray s Int Double</div><div><br></div><div>kahan :: Vec s -> Vec s -> ST s ()</div><div>kahan s c = do</div><div> let inner w j</div><div> | j < VDIM = do</div><div>
!cj <- unsafeRead c j</div><div> !sj <- unsafeRead s j</div><div> let !y = fromIntegral w - cj</div><div> !t = sj + y</div><div> !w' = prng w</div><div> unsafeWrite c j ((t-sj)-y)</div><div> unsafeWrite s j t</div><div> inner w' (j+1)</div><div> | otherwise = return ()</div><div> forM_ [1 .. VNUM] $ \i -> inner (fromIntegral i) 0</div><div><br></div><div>calc :: ST s (Vec s)</div><div>calc = do</div><div> s <- newArray (0,VDIM-1)
0</div><div> c <- newArray (0,VDIM-1) 0</div><div> kahan s c</div><div> return s</div><div><br></div><div>main :: IO ()</div><div>main = print . elems $ runSTUArray calc</div><div><br></div><div>Cheers,</div><div><br></div><div>Felix</div></div> </div></body></html>