No subject

faster than Mach, but (and the L4KA guys will dissent) significantly
slower than Jochen's assembly-language L4/x86 (whose IPC path is 20
times faster than Mach's).

> Does L4's performane advantage come from optimisation of the
> implementation or from its architecture?  The literature suggests
> the latter however I note that many L4 implementations are written
> in assembler and that there are performance concerns about Hazelnut.

L4's advantage mainly stems from its architecture (see Jochen's paper
<URL:http://os.inf.tu-dresden.de/papers_ps/jochen/Mikern.ps>).
However, the IPC path can benefit overproportionately from some clever
optimization: It can be so short in theory that it is easy to waste
cycles if functionally is added to it without care.

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