Messages in this thread | | | Subject | Re: Memory hogs | Date | Tue, 20 Jul 1999 21:18:23 +0200 | From | Hermann Schichl <> |
| |
Bernd Paysan wrote: > > With a useful setup, tight memory is so far from ordinary operation, that > all this really should never happen. This should only happen when there is > really a malicious program attacking the machine (or a program that has > gone wild). If it really happens during normal operation, add a new swap > partition, or better buy more RAM. > Aaargh! I hate to explain this over and over.... There seems to be a FUNDAMENTAL problem involved here when computer science guys and numerical mathematics guys talk about this issue.
When you are trying to solve (e.g.) NP-hard problems like finding global optima you generally cannot guess in advance how much memory you will need (if you can do it, the problem is not interesting). There is no useful setup in this situation. Unfortunately, if you design your algorithm in such a way that you try to use as little memory (and other resources) as possible, you have in advance not the slightest idea how much you will need. The dimension of the problem (e.g.) certainly is not a good hint (though you know that the resource requirements grow at least exponentially (asymptotically, for all problems,...)). There are certain 30-dimensional problems which can kill almost every machine around, and there are 1000-dimensional ones which can be solved on every PC. And it is NP-hard to know in advance what is true, and different algorithms yield different results (and again no way to know in advance). So these are probably malicious problems but not malicious programs (except for Matlab, Maple, and Mathematica).
Hermann
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |