Re: [patch] "HT scheduler", sched-2.5.63-B3

[Linus Torvalds]

On Thu, 6 Mar 2003, Ingo Molnar wrote:
> 
> another thing. What really happens in the 'recompile job' thing is not
> that X gets non-interactive. Most of the time it _is_ interactive.

This is not what I've seen.

When X is interactive, and is competing against other interactive jobs, 
you don't get multi-second slowdowns. X still gets 10% of the CPU, it just 
gets it in smaller chunks.

The multi-second "freezes" are the thing that bothered me, and those were
definitely due to the fact that X was competing as a _non_interactive
member against other non-interactive members, causing it to still get 10%
of the CPU, but only every few seconds. So you'd get a very bursty
behaviour with very visible pauses.

It's ok to slow X down. Nobody in their right mind would expect X to track 
the mouse 100% when scrolling and the machine load is 15+. I certainly 
don't.

But having X just _pause_ for several seconds gets to me. I can't easily 
make it happen any more thanks to having ridiculous hardware, and I think 
X itself has gotten better thanks to more optimizations in both clients 
and X itself (ie if the CPU requirements of X go down from 5% to 3%, it 
gets a _lot_ harder to trigger).

But it was definitely there. 3-5 second _pauses_. Not slowdowns.

		Linus