Re: [Gems-users] simulating a 4 cpu CMP with dnuca protocol with gems

[Bradford Beckmann <beckmann@xxxxxxxxxxx>]

Alessandro,

The biggest change you'll need to make is correlating the L2 bank mapping
functions to your new file specified network.  In particular what are the
local and remote banks for each processor.  I recommended trying to fully
understand how these mapping function work with the 8 processor DNUCA
network before trying to implement your 4 processor mapping functions and
network design.

Other than that, I don't suspect you'll run into any major problems.
However, I haven't done it myself.  So there may be other subtle issues as
well.

Let me know if you have any more questions.

Brad



On Thu, 23 Mar 2006, Alessandro Bardine wrote:

> I would like to make some simulations that
> involve a 4 CPU CMP using the DNUCA coherence protocol.
> The wiki documentation for the NUCACOL says that at the moment only
> 8 CPU based architectures are supported.
> To get a 4 CPU architecture I must, of course, rewrite the topology
> description file but
> is there some other changes to be done in the coherence protocol or in ruby
> parameters?
> In other words, are the slicc files for DNUCA scalable versus the number of
> the CPU or
> no?
>
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>