Re: [Gems-users] FILE_SPECIFIED Question?

[hb166307@xxxxxxxx]

Hello GEMS users,

I have been trying to use the FILE_SPECIFIED topology and I can not seem to get
a system created. I am using the files with the prefix NUCA that came in the
ruby/network/simple/Network_Files directory. I set the appropriate ruby
parameters to use a use one of the unmodified network description. The only ones
that seem to work are:

NUCA_Procs-1_ProcsPerChip-1_L2Banks-1_Memories-1.txt
NUCA_Procs-1_ProcsPerChip-1_L2Banks-32_Memories-1.txt

Every other one give either hangs while creating the system of gives the
following error.

Segmentation fault (SIGSEGV) in main thread
The simulation state has been corrupted. Simulation cannot continue.
Please restart Simics.

I really hope that it is something that I am missing. Any ideas? Thank you in
advance. Would really appreciate any help!

---------------------------------------------------------------------------

Quoting Mike Marty <mike.marty@xxxxxxxxx>:

> ------=_Part_35741_18221839.1223215779174
> Content-Type: text/plain; charset=ISO-8859-1
> Content-Transfer-Encoding: 7bit
> Content-Disposition: inline
> 
> On Fri, Oct 3, 2008 at 11:38 PM, Niket <niketa@xxxxxxxxxxxxx> wrote:
> 
> > The configuration you have created is a 8 processor system with 2
> L2
> > banks and 2 Routers connecting various components. It does not
> > necessarily imply a 2-chip configuration. For instance, all these
> > connections are also valid for a single chip system. The
> > NUM_PROCS_PER_CHIP controls that. You might also want to use the
> > appropriate protocol for that.
> >
> > If you want to use a 2-chip configuration with 4 procs/chip, I
> believe
> > you can use the MSI_MOSI_CMP_directory protocol. However, the L2's
> would
> > not be unified between the 2-chips. The 2-chips in that case would
> be
> > connected by the link between Routers 0 and 1.
> >
> > I am not sure whether there is a protocol with the GEMS
> distribution
> > that lets you use a unified L2 across 2 chips. The GEMS team might
> > clarify that.
> >
> 
> Unless things have changed, the PROCS_PER_CHIP parameter essentially
> determines how the L2 is split across "chips".  You could set
> PROCS_PER_CHIP
> to 8, but then configure your interconnect into two separate regions
> of
> higher-bandwidth, lower-latency links.
> 
> ------=_Part_35741_18221839.1223215779174
> Content-Type: text/html; charset=ISO-8859-1
> Content-Transfer-Encoding: 7bit
> Content-Disposition: inline
> 
> <div dir="ltr"><br><br><div class="gmail_quote">On Fri, Oct 3, 2008
> at 11:38 PM, Niket <span dir="ltr">&lt;<a
> href="mailto:niketa@xxxxxxxxxxxxx";>niketa@xxxxxxxxxxxxx</a>&gt;</span>
> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0
> .8ex;border-left:1px #ccc solid;padding-left:1ex;">
> The configuration you have created is a 8 processor system with 2
> L2<br>
> banks and 2 Routers connecting various components. It does not<br>
> necessarily imply a 2-chip configuration. For instance, all
> these<br>
> connections are also valid for a single chip system. The<br>
> NUM_PROCS_PER_CHIP controls that. You might also want to use the<br>
> appropriate protocol for that.<br>
> <br>
> If you want to use a 2-chip configuration with 4 procs/chip, I
> believe<br>
> you can use the MSI_MOSI_CMP_directory protocol. However, the
> L2&#39;s would<br>
> not be unified between the 2-chips. The 2-chips in that case would
> be<br>
> connected by the link between Routers 0 and 1.<br>
> <br>
> I am not sure whether there is a protocol with the GEMS
> distribution<br>
> that lets you use a unified L2 across 2 chips. The GEMS team
> might<br>
> clarify that.<br></blockquote><div><br></div><div>Unless things have
> changed, the PROCS_PER_CHIP parameter essentially determines how the
> L2 is split across &quot;chips&quot;. &nbsp;You could set
> PROCS_PER_CHIP to 8, but then configure your interconnect into two
> separate regions of higher-bandwidth, lower-latency links.</div>
> <div><br></div><div><br></div></div></div>
> 
> ------=_Part_35741_18221839.1223215779174--
>