Re: musings on performance

[Jason Stover]

On Sun, May 14, 2006 at 08:37:20PM -0700, Ben Pfaff wrote:
> Ben Pfaff <address@hidden> writes:
> 
> > That said, there are numerous optimizations that are likely to
> > have real benefit, that we should consider implementing as time
> > goes on.  The ones that I've come up with so far are below.
> > They're roughly in order of increasing difficulty and benefit,
> > and each one tends to depend on the previous.
> 
> It's pretty clear that my wilder ideas on optimization just
> exposed my ignorance of the structure of statistical
> computations.  Oh well.  I'll work on the optimizations that
> definitely make sense and we can think about whether anything
> more ambitious is worthwhile.

I think what you said about the map/reduce model made a lot of sense,
though it would be a lot of work and results from such an approach
would not typically be identical to those from an analogous procedure
executed without the map/reduce method.

-Jason