Subject: Re: ingres and atomicity
To: None <>
From: David Laight <>
List: tech-userlevel
Date: 06/14/2005 07:36:46
On Mon, Jun 13, 2005 at 09:07:04PM -0400, James K. Lowden wrote:
> Does the term "atomic clear" have some generally accepted meaning, or is
> it  Ingrespeak?  

Well, atomic functions are usually used to perform read-modify-write
sequences on small memeory items, and 'clear' is usually a single bus
transaction so is atomic anyway.

The functions given did an atomic clear of a 64bit integer - not immediately
obvious from the name, and not necessarily a single bus transaction.

> Do we support, anywhere, the idea of changing a word (4 bytes) of memory
> atomically, meaning that no other thread can intervene?  

Code will assume that a write of a 32bit (aligned) quantity is atomic.


David Laight: