begin  quoting Andrew Lentvorski as of Tue, Apr 03, 2007 at 08:29:21PM -0700:
[atomic syntax]
> Congratulations, you just described "Composable Memory Transactions".
> 
> http://research.microsoft.com/~simonpj/papers/stm/stm.pdf

Okay, I've finally gotten around to reading that paper.

Clever.

> This is one of the cornerstones of Shared Transactional Memory.

While an impressive piece of work[1], the problem (as I see it) is that it
does not remove the need for locks (but then, nothing will, I think),
and it accomplishes the "atomicity" by rollback-and-retry.

The second problem would presumably have the realtime guys throwing
a fit... given a fast process changing memory and a slow process
changing memory, you're going to have a likely case of starvation.

Of course, that nailed down the problem of *my* viewpoint in thinking of
an "atomic" sequence of instructions -- it simply doesn't work on a
multiprocessor or multicore system without introducing locks. So the
locks are still there, but they're now hidden behind some syntactic
sugar.

I'm not sure this would be an improvement.

[1] Honest.

-- 
Thoughts on the other paper will be forthcoming, unless my brain melts.
Stewart Stremler

-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg

Reply via email to