On Tue, 25 Aug 1998, S.J. Borrill wrote:

> What's the state on:
> 
> a) the vastly speeded up pmap stuff that Neil kept talking about

A lot of Neils changes have been merged in. There is another trick that
has been proven but needs a bit of code added to various functions before
it is in a state to be committed.

> b) the swapping out of L1 page tables so that we can do away with the >
maxproc= stuff. Is this limited to arm32 only?

THis will happen as soon as I finish tracking reference counts on L2 page
tables. When that works L2's will be freed when the last mapping is
removed and similarly the L1 will be removed when the mappings go thus
they will be freed when a process is swapped out etc.
Something is in development on this as I noted in Jason's PR on the
subject for L2 reference counts..

Just thinking about it a quick hack does come to mind that I might give a
go ...

> How large is each L1 page table? 16K seems to ring a bell.

Yep.

Cheers,
				Mark