On Tuesday 13 February 2001 12:53 am, Reinoud Zandijk wrote:
> Hi Richard,
>
> On Tue, 13 Feb 2001, Richard Earnshaw wrote:
> > imago@kabel065011.kabel.utwente.nl said:
> > > Wow... so there is allready a kind of support for that build in ...
> > > hmm... have to look into that! That would save lots of work!
> >
> > Yep, look for pmap_isa_dma_* in arm32/pmap.c and then in the shark
> > support dir.
>
> Ah silly me.... I saw that ... and wondered !!!! Maybe we can make it more
> generic for a start ?
>
> 've been secretly thinking of maybe rewriting pmap.c for the arm32 ... a
> more efficient one .... thats why i looked at it :)

There's lots of work still to be done on the pmap.  Especially if any of the 
companies come true on producing multi processor boxen (assuming that they do 
it SMP style, and not hydra style)  Exactly what happens here depends on the 
implementation of the MP, however the basics should be similair.

There's also a PR outstanding on our pmap not doing scheduler locking 
properly.  Also adding dynamically allocated L1's, currently it's a boot time 
parameter or compile time parameter for the number of L1's which limits the 
max number of processes.  There's also the possibility of using pools to hold 
L1 and L2 tables when not in use.  Which may speed up reallocation etc.

I've actually been examining other platform's pmap to see which might be best 
to look at.  Currently alpha looks interesting, (however alpha's have 3 page 
table level, and some sort of unique id (the ASN))  There's also the i386 mp 
branch which might be of interest to see how it differs for supporting mp 
from the mainline branch.

Cheers,
Chris