--IKC1RK7+nRZ8SZaO
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Thu, Jun 09, 2005 at 10:34:05PM -0700, Jonathan Stone wrote:
> >Finally, making smp-safe datastructures doesn't necessarily involve
> >locks;=20
>
> (Mmm, I wonder, did you write that before seeing my preivous message,
> or after?)

Before or after which one? :) Before, I think, but the subthreads are
converging it seems.

> Much of Herlihy's early work recovers from
> contention by copying the entire structure being updated. =20

I had (roughly) in mind:

  The sequent-derived RCU that linux now apparently uses, which by
  demonstration seems to fit the 'general-purpose CPU-portable' bill,
  and which seems like it might be largely this, or even exactly this,
  without checking the name reference;

  The Xen rings I mentioned elsewhere; and

  Various constructions like cpu-local storage and cpu-bound work that
  aim to 'manage' contention by (say) bidding or delegation,
  especially as used in NUMA-type platforms.  Most of these aren't
  much use for your single-fat-fast stream scenario, though.

--
Dan.
--IKC1RK7+nRZ8SZaO
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (NetBSD)

iD8DBQFCqT4VEAVxvV4N66cRAtf1AJ4htlBFvlUKyB5YXh9FYrdkkWFnBgCg+ygv
UHiPAbChM6yv0+UInvDF/KM=
=/RC6
-----END PGP SIGNATURE-----

--IKC1RK7+nRZ8SZaO--