tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: revivesa status 2008/07/09



On Fri, Jul 11, 2008 at 07:35:54PM +0000, Dennis den Brok wrote:
> Bill Stouder-Studenmund <wrstuden%netbsd.org@localhost> schrieb:
> > I think a better question is if you've seen 1:1 threading to be noticably
> > slower with SA compiled in. :-)
> >
> 
> This, and whether "emulating" SA using the new way, which is AFAIU
> what revivesa does, is noticably slower than as it was done before,
> would quite interest me. What are your thoughts about the latter?

It's hard to say. "Emulating" isn't quite the right word to use here. 
revivesa is a reimplimentation of the old system calls in terms of the 
-current thread primitives. There is no support for the system calls 
without it. I see "emulating" as meaning we have an old and a new way of 
doing things and we can map the old to the new. Without revivesa, we have 
no way of supporting scheduler activations. Thus "emulation" doesn't quite 
fit. ;-)

I think that the revivesa implementation is as efficient as the fixsa 
(NetBSD 4.X) one. The major blocks of code are the same, and the life 
cycle of lwps is the same.

A direct comparison will be difficult as the NetBSD 5.x kernel is much 
more concurrent and I believe more efficient.

Take care,

Bill

Attachment: pgpVhqzttHPbZ.pgp
Description: PGP signature



Home | Main Index | Thread Index | Old Index