next-gen mapping

[Michael Hohmuth]

"Adam 'WeirdArms' Wiggins" <awiggins at cse.unsw.edu.au> writes:

> 	Would an assembler based L4 benifit from this new mapping tree
> data base too? It certainly sounds easier to handle memory for. 

I think that the binary-tree representation is easier to implement
than the trees-in-fixed-size-arrays one.  One factor which complicates
the latter is that the fixed-size arrays have to be efficiently
allocated, deallocated, and moved around in memory.  Lukas'
implementation includes a generic slab allocator for tree allocation.

> I'm trying to work out my dynamic data structures now. So far i have
> TCB's (512 bytes), 1st level page table entries 16K, 2nd level page
> table entries 1K, mapping tree nodes ??? no idea yet. I got Lukas's
> paper but i failed german in my high school years :( I'm trying to
> get it translated. It there any code somewhere i can look at which
> would illistraight it? Or does lukas have the time to give me an
> english run down?

I think Gernot is trying to translate the paper using Babelfish; Lukas 
has sent him his paper's sources.

Other than that, I'm positive that we can help you understanding it.
(If you're patient enough, you might even just wait until I've
finished my technical report about Fiasco's design -- but don't hold
your breath, please. :)

Michael
-- 
hohmuth at innocent.com, hohmuth at sax.de
http://www.sax.de/~hohmuth/