On Monday 26 May 2008 16:29:53 Damien Doligez wrote:
> On 2008-05-21, at 10:06, Martin Berger wrote:
> > Here I disagree. Shared  memory concurrency is a specific form
> > of message passing: Writing to a memory cell is in fact sending
> > a message to that cell carrying two items, the new value and a
> > return channel that is used to inform the writer that sending
> > has succeeded, and likewise for reading.
>
> This is completely wrong.  A few machines have a simple model like
> that, but they were all built in the last century.  Nowadays, writing
> to memory is more like broadcasting a message and having no idea when
> it will arrive at each destination.  And if you write to another piece
> of memory, you don't know in what order the updates will become
> visible to a given processor.
>
> You are neglecting a very important parameter, which is called the
> "memory model" of your multiprocessor.

The memory model of a multiprocessor is just a specific form of communication 
fabric. That does not disagree with Martin's statement. So he was certainly 
not "completely wrong". At worst it was a simplification. I suspect he simply 
did not aticipate anyone treating his comment as a seminal work on multicore 
computing.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/products/?e