[Haskell-cafe] Optimizing a hash function

[isto]

su, 2006-11-26 kello 15:12 +1100, Ivan Tomac kirjoitti:
> The first version I came up with ran 20 times slower than C.
> Thanks to Don Stewart's suggestions on IRC, I managed to improve the  
> ...

> Options used to compile the version using the hash function written  
> in Haskell:
> ghc -O -funbox-strict-fields -fglasgow-exts -fbang-patterns -cpp -o  
> test hashByteString.hs test.hs
> 
> Options used to compile the version using the hash function in C:
> ghc -O -fglasgow-exts -ffi -fbang-patterns -cpp -DCHASH -o ctest  
> ctest.c test.hs

Hi Ivan Tomac,

Have you tried -O3 -optc-O3  -funfolding-use-threshold=16 
compile flags?  Don, Lemmih, Lennart and Bulat helped me to sort
out a similar problem a couple of weeks ago. More hints can be found at
http://haskell.org/haskellwiki/Performance/GHC
Especially, to check generated code by taking a look of core
-ddump-simpl > core.txt
and to check memory leaks, you could run with
+RTS -sstderr

br, Isto