Re: stress testing and loadavg

Adam McKee (amckee@poboxes.com)
Tue, 5 Aug 1997 17:05:32 -0600 (CST)

I was not trying to argue that it's not a problem -- I realize that user
code could very well do nasty tricks to get more CPU - it's just kind of
sad that this would be a problem. Perhaps it should be modified so that
you must block for a certain length of time before going back to your
min-run-queue? And/or maybe you don't go back to your min-run-queue when
you block, just decrement the run-queue? I will think about this more -
it is a problem I had not even considered.

I would like to see something like this make it into the kernel, though I
realize it may not stand a chance of acceptance in its current form. I
will continue to experiment with a few more ideas I have for it. At the
very least, it has been a nice introduction to the kernel sources for me.
This was my first kernel hack 8)

-- Adam

On Tue, 5 Aug 1997, Alan Cox wrote:

> > its minimum run-queue. I hope people would realize that putting in system
> > calls just to get back to the minimum run-queue is a brain-damaged thing
> > to do. Of course, I'm sure compiler writers realize that optimizing for
>
> Doesn't hold water. If I do stupid tricks to get 99.95% of the CPU time
> by misusing scheduling tricks and kill the machine because I know its
> a brain damaged scheduler ..
>
> Alan
>