Subject: Re: [boost] performance of a linear algebra/matrix library
From: DE (satan66613_at_[hidden])
Date: 2010-05-06 14:08:25
on 06.05.2010 at 21:54
joel falcou wrote :
> Sound slike you just trash cache ...
> doing matrix product in i,j,k order is like, well, bad as all grad
> student should know ...
i believe for small matrices i,j,k is as efficient as any other
for relatively large matrices i go i,k,j (to minimize cache issues)
but still that may be the case
i haven't tried tiling yet, maybe it will save the day
-- Pavel