[e-lang] Non-local Exits vs Defensive Consistency - David Hopwood

[Mark Miller]

From: David Hopwood <david.nospam.hopwood at blueyonder.co.uk>
Date: Jan 24, 2007 6:10 PM
Subject: Re: Speculative Concurrency for E?

Incidentally, the transactions I'm proposing are not at all speculative.
They only roll back if an exception occurs.

Speculative (a.k.a. "optimistic") STM has a worst-case performance penalty
of O(n), where n is the maximum number of processes that can contend in
parallel for access to any resource. Most of the STM papers I've seen tend
to gloss over this, assuming that it only occurs in pathological cases, but
I'm not convinced. The benefits of transactions come from the ACI[D]
properties, which are independent of whether the implementation is
optimistic or pessimistic.

--
David Hopwood <david.nospam.hopwood at blueyonder.co.uk>