On Mon, Apr 27, 2009 at 5:23 PM, nathan binkert <[email protected]> wrote:

> > It seems fine to me, however I'm surprised that there is a functional
> > request that ends up in the PhysicalMemory that doesn't need a
> > response. The only thing I could imagine falling into this category is
> > a Functional Writeback if you were attempting to flush the caches
> > before a checkpoint or something similar.
>
> Does a functional write from the debugger or something require a response?
>

Actually devices sinking functional writes should not convert the packet to
a response as the packet needs to be propagated to any other device that
might have a copy of the data (e.g., a cache).  It shouldn't hurt to
generate a response on a functional read though.  Overall I don't see
anything wrong with Cliff's patch, but I'd have to understand the bug it's
addressing a little better to know whether it's really the right fix for
that bug or not.

Steve
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to