Making computers work
Henrik Nilsen Omma
henrik at ubuntu.com
Thu May 25 12:13:42 BST 2006
Matthew Garrett wrote:
>
> That's possibly the problem - the assumption that computers can't be
> made to work properly. It's fundamentally wrong, and it holds back
> development.
>
> Computers behave deterministically[1]. Software behaves
> deterministically. The combination of these two things lets us produce
> software that works 100% of the time.
I agree with you on this, just an interesting side-note: I did Fluid
dynamics simulations on Linux clusters for my thesis that were in theory
100% deterministic (no random numbers involved). However, from a
combination of load-balancing and rounding errors I would get slightly
different results each time. When you then add the butterfly effect in
fluid dynamics you can get wildly different morphologies (but the same
global dynamics).
- Henrik
More information about the sounder
mailing list