> i have implemented this recently, it's a 'swapin-predictor', completely
> kernel-based, it's similar to CPU branch prediction. It monitors per-vma
> swapin addresses, and maintains an array 'probable next swapin' pointers
> and probability counters. It does improve the speed of linear trashing
> _alot_ ;) (from 80K/sec to disk speed) Not that linear trashing is a good
> thing.
>
> i've got to clean it up. I also have a hack that flips two-page COW pages
> when doing the COW-copy, to get better cache utilization on
> non-write-allocate CPUs... although i still fail to see any measureable
> advantage.
Maybe we should form a developer-group and integrate all
mm patches into one large 'ready-to-apply' patch (I'm willing
to volunteer for this one).
Once we have 'enough' features we could work on debugging
it, and once it's stable we can send it to Linus...
I am also willing to keep the patch up-to-date against
_every_ kernel version to come (well, development-kernels).
grtz,
Rik.
----------
Send Linux memory-management wishes to me: I'm currently looking
for something to hack...