[Gllug] UKUUG

Nix nix at esperi.demon.co.uk
Thu Jul 5 23:01:22 UTC 2001


On Tue, 3 Jul 2001, Jake Jellinek stated:
> I found some of the finer details of Andrew Arcangeli's (Again from SuSE

s/Andrew/Andrea/ (as others have said). (Finger macro trouble, I guess :) )

> optimisation to the application kernel/userspace interaction. The basic
> concept here is that syscalls such as gettimeofday are very fast internally
> and are called a large number of times, very often by some applications. (In

Hello, X11.

> Moving gettomeofday into userspace entirely isn't an option since it needs
> to get priveleged information about interrupts and chipsets from the
> hardware.

Surely not. The *updating* must be done from kernel space, but
gettimeofday() itself can be entirely implemented in userspace; it was
the first application mentioned for what would become vsyscalls entirely
*because* it can be implemented entirely in userspace.

Most other syscalls, of course, you're quite right about; but
gettimeofday() and the almost totally useless to optimize (except for
benchmarks) getpid() can be implemented in userspace exclusively.

(Unless I'm missing something major, that is.)

-- 
`I'm not sure whether libtool is an existence proof that you _can_
 write a shell script that handles its arguments correctly, or a
 demonstration that you may try but you are doomed to failure.'
                                                       -- Zack Weinberg


-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug




More information about the GLLUG mailing list