1 Mar
2014
1 Mar
'14
1:27 p.m.
On Sat, Mar 01, 2014 at 07:49:18AM +0100, user web210p1 wrote:
I'm currently taking a course in University called
"How to write fast numerical
code". We are to do a group project of 3 persons where we choose a numerical
algorithm and optimize it over the course for a particular processor (e.g. apply
vectorization, cache optimization etc...).
Are you supposed to optimize by
1. understanding the algorithm and the context in which it is used,
and eventually modifying it within given limits on performance,
precision, etc., or
2. only by changing the way it is coded (which seems to be suggested
by your message) ?
In my experience there is much more to be gained from (1) than from
(2), so that is always the first thing I work on.
I've got something you work on but it would have to be treated as
confidential.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)