This is a McPAT-related question, not a gem5 one.
In McPAT, throughout and latency in xml file are in fact cycle time and
access time in cacti. So, McPAT tries to optimize the cache for the
specified throughput (cycle time) and latency (access time).

Amin


On Mon, Oct 28, 2013 at 7:11 PM, Ivan Stalev <[email protected]> wrote:

> Hi everyone,
>
> I am trying to use the GEM5 stats and config outputs and plug them into
> McPAT in order to get area/power numbers for the architecture I simulated
> in GEM5. I am using the perl script and template XML file found 
> here<https://www.cl.cam.ac.uk/~acr31/sicsa/>.
> However, I am running into a problem; I am getting infinity for the Runtime
> Dynamic power in my McPAT output.
>
> I believe the problem may be coming from the cache configuration (I am
> getting McPAT warnings such as "dcache array structure cannot satisfy
> latency constraint". More specifically, for the icache and dcache config
> parameters for McPAT, McPAT expects "throughput w.r.t. core clock" and
> "latency w.r.t. core clock". In the original template XML file, the values
> are 1 (hardcoded constant) and "ceil(config.system.cpu.icache.latency /
> config.system.cpu.clock)", respectively. However, GEM5 now has hit_latency
> and response_latency as opposed to just latency.
>
> For those who use GEM5 with McPAT, what values do you have for throughput
> and latency w.r.t to core clock? Has anyone else ran into the problem of
> getting infinity and NaN values for the Runtime Dynamic power?
>
> Currently (getting infinity power) I have throughput set to 1 and latency
> set to "ceil((config.system.cpu.icache.hit_latency +
> config.system.cpu.icache.response_latency) / config.system.cpu.clock)"
>
> Thank you in advance,
>
> Ivan
>
> _______________________________________________
> gem5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to