Simon Marlow wrote:

The shootout

http://gist.github.com/349196 My machine is way to old to figure out whether switching to a table-based lookup like the C version has would still gain anything, so It'd be nice if someone of you with a recent machine could have a look at it. Other input is of course very welcome, too, as well as possibly a version for dual- and upcores that distributes random number generation/ lookup and output over several cores. (My machine is way too... you know the story) -- (c) this sig last receiving data processing entity. Inspect headers for copyright history. All rights reserved. Copying, hiring, renting, performance and/or quoting of this signature prohibited.

I'm wondering... Since the DPH libraries are shipped with GHC by default are we allowed to use them for the shootout? Roman On 30/03/2010, at 19:25, Simon Marlow wrote:

The shootout (sorry, Computer Language Benchmarks Game) recently updated to GHC 6.12.1, and many of the results got worse. Isaac Gouy has added the +RTS -qg flag to partially fix it, but that turns off the parallel GC completely and we know that in most cases better results can be had by leaving it on. We really need to tune the flags for these benchmarks properly.

http://shootout.alioth.debian.org/u64q/haskell.php

It may be that we have to back off to +RTS -N3 in some cases to avoid the last-core problem (http://hackage.haskell.org/trac/ghc/ticket/3553), at least until 6.12.2.

Any volunteers with a quad-core to take a look at these programs and optimise them for 6.12.1?

Cheers, Simon _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe