[parisc-linux] Re:[parisc-linux-cvs] linux-2.6 kyle

[Michael S. Zick]

On Mon August 28 2006 11:42, Grant Grundler wrote:
> On Wed, Aug 23, 2006 at 06:45:47PM +0200, Joel Soete wrote:
> > (the idea was so that sync would force the write back to ram but then I read
> > more about memory strongly/weakly ordering and so became less sure?)
> 
> Simple explanation: "sync" instruction enforces an "ordering" (completion)
> of all previous instructions.
> 

Sorry if I am speaking out of turn Joel...

Joel's test of my write-up example (which still has the "sync" in place)
has died...

But it looks like it is related to the preemptive scheduling, not a clear
lock failure as before.

There will probably be a break in our work on this problem while we
decide on our next step of testing (and we both get some rest).

> AFAIK the CPU will respect ordering for memory accesses at all times.
> 
> grant

Mike