Subj : Re: Memory visibility and MS Interlocked instructions
To   : comp.programming.threads
From : Joe Seigh
Date : Fri Aug 26 2005 12:50 pm

Alexander Terekhov wrote:
> Joe Seigh wrote:
> [...]
> 
>>Why can't you directly answer specific questions like what is meant
>>by bidirectional and unidirectional?  
> 
> 
> I'm reluctant because you're going to forget it almost immediately and 
> I'll just waste time and bandwidth once again. But ok, 
> 
> http://groups.google.de/group/comp.programming.threads/msg/3f519417b2a619c5
> http://groups.google.de/group/comp.programming.threads/msg/a08095e4e4b61155
> 

One issue you seem to be touching upon is whether memory accesses have to
complete before some instruction "executes", or not initiate until after
some instruction "executes".  I haven't seen any situations, outside of
context switching by the OS, where this is required.  I haven't seen any
examples from you where you think this is required, i.e. it won't work
without it.  AFAIK, all you need is relative ordering of the memory
accesses.  You can't directly observe the instruction "execution" anyway.

-- 
Joe Seigh

When you get lemons, you make lemonade.
When you get hardware, you make software.

.