From: James Rogers (jamesr@best.com)
Date: Wed Jan 16 2002 - 10:25:21 MST
On 1/16/02 1:13 AM, "Eugene Leitl" <Eugene.Leitl@lrz.uni-muenchen.de> wrote:
> On Tue, 15 Jan 2002, James Rogers wrote:
>> What is an example of an algorithm that is qualitatively different on
>> a parallel system versus a serial one?
>
> Realtime control in iterated predator-prey context. It doesn't help that
> your hardware will eventually finish, if it is being eaten before it
> finishes.
That's actually independent of the algorithm though. A large pile of 286
computers in parallel will be slower than a P4 for solving a given problem.
For the types of stuff you are talking about, system level performance only
has to meet a certain spec, whether it is done on one processor or millions.
If the spec is "as fast as possible" then maspar is the way to go, but there
is typically an economic component that puts a damper on ambitions with
these kinds of specs.
Cheers,
-James Rogers
jamesr@best.com
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:11:42 MST