Michael Neumann wrote:
Matthew Dillon wrote:
I would like to revisit SMP but it isn't going to happen until HAMMER
is well into production. I try to encourage others to take up the SMP
work but so far there's only been some minor poking around.
IMHO a first step would be to write down all the things that needs to be
done to get SMP going.
SMP works, it just doesn't work with a high parallelism in the kernel.
It's just a question of removing the implicity dependence on the mplock
for common code paths. So the only thing to be done is a) identify common
code paths b) in those paths: either aquire the mplock or protect the
data in a different way c) in other paths that deal with the same data,
adjust accordingly. Easy huh? :)
My point is that this stuff is all over the place, starting from syscalls,
going to VFS, VM and drivers.
cheers
simon
--
Serve - BSD +++ RENT this banner advert +++ ASCII Ribbon /"\
Work - Mac +++ space for low €€€ NOW!1 +++ Campaign \ /
Party Enjoy Relax | http://dragonflybsd.org Against HTML \
Dude 2c 2 the max ! http://golden-apple.biz Mail + News / \
