If you fast-forward the execution before switching into the simulation mode, your collected statistics includes cold simulation stats (cold cache misses, empty pipe line , etc ). I think the best way is to use ptlcall in the source code of your workload like what Paul said.
On Tue, Sep 11, 2012 at 7:28 PM, Ankita (Garg) Goel <[email protected]>wrote: > Hi, > > Is there a way I could fast-forward the execution of a benchmark for a few > million of instructions before switching into the simulation mode ? For > instance, I create the checkpoint like below: > > # ./create_checkpoint bb; ./bzip2 <param>; ./stop_sim; > > Now, once I start marss again, from the above checkpoint 'bb', I want to > fast-forward about 1million user instructions and then run simulation for > the next few million instructions. Can I do this ? I did look around in the > archives, but did not find the answer there. Any suggestions/thoughts on > this will greatly help me! > > Thanks! > > -- > Regards, > Ankita > > _______________________________________________ > http://www.marss86.org > Marss86-Devel mailing list > [email protected] > https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel > >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
