begin  quoting Darren New as of Thu, Apr 05, 2007 at 09:43:02AM -0700:
> Stewart Stremler wrote:
> >But it pretty much comes down to having an atomic comare and set
> >instruction *somewhere*.
> 
> You can actually do locking between asynchronous processes without any 
> atomic instructions. At least, assuming that (say) reading a memory 
> address while another CPU is writing it doesn't actually break the write 
> or read bit patterns that weren't written at all.

You need a test-and-set or equivalent.

Hm.

Would delays work? I don't think so... 

> >I wasn't trying to describe the concept, but rather the syntax I'd
> >like to see.  The concept is pretty old, I should think.
> 
> It depends. Does "atomic" block out all other execution? Or does it 
> block out only other "atomic" blocks, in which case lots of languages 
> have that and it's pretty much one of the older synchronization 
> primitives invented.

The former.

"Once you start this, you will finish it, and nobody will interrupt you."

-- 
On a single-processor machine, new timeslice, disable interrupts, run, restore.
Stewart Stremler

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

Reply via email to