Good dotProduct

Andrei Alexandrescu SeeWebsiteForEmail at erdani.org
Tue Jun 29 08:21:37 PDT 2010


bearophile wrote:
> This is the faster I've done, further loop unrolling is not useful (for arrays of about 1400 doubles it's about 2.9 times faster than std.numeric.dotProduct on a recent Celeron CPU):
[snip]

That's pretty cool! Do we have your permission to replace dotProduct 
with yours?

Andrei


More information about the Digitalmars-d mailing list