[parsec-users] Simulation metrics
huanf at kth.se
Mon Mar 10 18:45:34 EDT 2008
Thanks,I think I have a clearer view now.
PARSEC seems to work well on my x86/Linux. However we are considering to
use PARSEC on our target marchine which is a solaris system simulated by
simics.I wonder if the "native input" is too large for a simulated system
since it's said for real machines.
> Hello Huan,
> PARSEC itself is just the program collection. It is independent from any
> metrics, so you can monitor everything that your processor / simulator
> supports. If you want to add instrumentation source code, you should first
> consider adding it to the PARSEC hooks (that's why they're there). The
> allow you to write instrumentation code only ONCE and it will
> be used by ALL PARSEC programs at the designated code locations.
> The hooks are special function calls that are made by all PARSEC programs
> the defined locations. You can find the hook library and some basic
> documentation in the package "hooks" in the "libs" group. After you
> the package simply add your code and rebuild PARSEC. Your build
> must explicitly enable the hooks, we provide one configuration which does
> that by default.
> - Chris
> On Sunday 09 March 2008 11:07 am, Huan Fang wrote:
>> I am now doing some research on cache coherence of CMPs. Can someone
>> tell me what metrics can I test using PARSEC? For example, L1/L2 miss
>> rates, miss latency, traffics are of my interest. I have run some test
>> like simsmall/simlarge but the only output is program runtime.
>> parsec-users mailing list
>> parsec-users at lists.cs.princeton.edu
> parsec-users mailing list
> parsec-users at lists.cs.princeton.edu
MSc. System-on-Chip Design
Royal Institute of Technology (KTH), Stockholm, Sweden
More information about the parsec-users