On Mon, 11 Aug 2003, Kris Van Hees wrote:

> What you describe is a very common problem with OSes that implement virtual
> memory. It's typically pretty much OK when your program's data space fits in
> real memory, but once you run beyond that, performance will most definitely
> be much worse when your inner loop runs through all the pages.  After all, you
> are hitting a page fault at a lot of pages, causing major paging activity.
> And when the working set itself is generally larger than available real memory,
> you end up paging out earlier touched pages in favour of new pages, only to
> reverse that again when the next offset loop starts.  You just keep cycling
> through the pages being swapped in and out.
>
> As far as I know, no OS with virtual memory has a real solution to this because
> there is no way for the OS to know how to solve the problem.  It's a bad design
> on the programmer's part :)  Some performance issues can only be solved by

Or the machine is simply under spec for the job. Who's at fault that
Taroon runs like a dog on my Pentium II with 128 Mbytes? Red Hat for a
poor package, or me for not adding RAM?

The hardware vendor of course!

Actually, there's a grain of truth there coz the box does get shirty
about mixed SIMM modules.


--


Cheers
John.

Join the "Linux Support by Small Businesses" list at
http://mail.computerdatasafe.com.au/mailman/listinfo/lssb
Copyright John Summerfield. Reproduction prohibited.

Reply via email to