[Openmcl-devel] Float & double consing & performance
tfb at tfeb.org
Tue Aug 18 15:21:42 UTC 2009
Hi, does anyone have any good hints for getting good float performance
out of CCL? I couldn't find anything in the Wiki. My (fairly
limited, to put it mildly) tests show that the 64bit version will do
cons-free single-float operations, but conses for doubles, while the
32bit one conses for both. However I've not done much more than
telling it to optimize speed and declaring all the types. I've also
not checked whether it's actually open-coding the operations yet.
So, basically I'm hoping someone else has already worked out what you
need to do :-)
(I'm using the trunk on an x86 mac).
More information about the Openmcl-devel