Re: New Performance (+Memory saver) patch

Mark Hemment (markhe@nextd.demon.co.uk)
Fri, 11 Apr 1997 07:15:13 +0100 (BST)

On Thu, 10 Apr 1997, Alan Cox wrote:
> I think that code is a false path. Amongst other things it prevents
> get_fd() running without locks on the SMP kernel - thats makes some paths
> far more expensive and prone to be scaling problems. Perhaps blocks of 256
> so that most programs fit under 256 and can run unlocked ?

Good point.
I'm currently looking at some basic threading of the SLAB allocator (an
irq spin-lock per cache, plus one for the cache chain). That should allow
some code (including get_fd()?) to mirgate outside of the global kernel
lock.

Regards,

markhe

------------------------------------------------------------------
Mark Hemment, Unix/C Software Engineer (Contractor)
markhe@nextd.demon.co.uk http://www.nextd.demon.co.uk/
"Success has many fathers, failure is a B**TARD!" - anon
------------------------------------------------------------------