On Tuesdayen den 21 August 2001 01:42, Marcelo Tosatti wrote:

On Tue, 21 Aug 2001, Roger Larsson wrote:

quoted

It runs, lets ship it...

First version of a patch that tries to USE a memory_clock to determine
when to run kswapd...

Limits needs tuning... but it runs with almost identical performace as
the original.
Note: that the rubberband is only for debug use...

I will update it for latest kernel... but it might be a week away...

Roger,

Why are you using memory_clock_interval (plus pages_high, of course) as
the global inactive target ?

That makes the inactive target not dynamic anymore.

It is still dymanic due the fact that kswapd will be run not depending on a
wall clock, but on problematic allocations done.
(i.e. inactive_target looses its meaning for the VM since it measures
pages/second but second is no more a base for kswapd runs...
both mean - I want to have this amount of reclaimable pages until the next 
kswapd run...)

/RogerL

-- 
Roger Larsson
Skelleftea
Sweden
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/