Hi Rodrigo, It is likely that instructions involved with synchronization will affect the absolute instruction counts on cores that wait for other cores (and thus, the IPC). A better metric for analyzing the balance of work between cores is the mix of instructions that they commit (e.g. integer, floating point, control, etc.). Depending on which core type (atomic, timing, O3, inorder) you're using in gem5, there are different stats that represent these values. I'd recommend looking for these different stats in your output.
Joel On Mon, Jun 24, 2013 at 11:18 AM, Rodrigo Reynolds Ramírez < rodrigo.r...@hotmail.com> wrote: > Yes, I runned it for the ROI from a checkpoint, I got three sets of > statics, I copy the results for ROI. > > Rodrigo > > ------------------------------ > Date: Mon, 24 Jun 2013 21:44:30 +0530 > > From: biswa....@gmail.com > To: gem5-users@gem5.org > Subject: Re: [gem5-users] Parsec and Alpha > > No that's not the case always. Did you run for the ROI? > > > On Mon, Jun 24, 2013 at 9:35 PM, Rodrigo Reynolds Ramírez < > rodrigo.r...@hotmail.com> wrote: > > I think the OS scheduler tries to balance the cpu's load avoiding that ipc > difference. > > Rodrigo > > ------------------------------ > Date: Mon, 24 Jun 2013 19:48:55 +0530 > From: biswa....@gmail.com > To: gem5-users@gem5.org > Subject: Re: [gem5-users] Parsec and Alpha > > > why would the ipc be same? > > > On Mon, Jun 24, 2013 at 7:33 PM, Rodrigo Reynolds Ramírez < > rodrigo.r...@hotmail.com> wrote: > > Hello everyone > > I am currently working with parsec+alpha, I have executed some tests using > the small inputs using two cpus, is it normal to get a very unbalanced ipc? > These are my results: > > Bench / IPC CPU0 CPU1 blackscholes 1.659 0.338138 bodytrack 0.374274 > 0.319496 canneal 0.108077 0.529619 dedup 0.585347 0.292882 ferret > 0.71402 0.337503 fluidanimate 1.493102 0.297794 freqmine 0.596511 > 0.305602 streamcluster 0.243343 0.303156 swaptions 0.24285 0.331316 > vips 0.280734 0.403565 x264 0.491017 0.314645 > For most benchmarks there is a big difference in the ipc between cpus, I > would think the ipc should be almost the same. > > I am using the alpha image provided here > http://www.cs.utexas.edu/~parsec_m5/, I am using the memory classic > model, checkpoints and the launcher scripts generated with .rcS Run > Script Generator <http://www.cs.utexas.edu/~parsec_m5/writescripts.pl>. > > Thanks in advance, > > Rodrigo > > _______________________________________________ > gem5-users mailing list > gem5-users@gem5.org > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > > > > > -- > > *thanks®ards > * > *BISWABANDAN* > http://www.cse.iitm.ac.in/~biswa/ > > “We might fall down, but we will never lay down. We might not be the best, > but we will beat the best! We might not be at the top, but we will rise.” > > > > _______________________________________________ gem5-users mailing list > gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > > _______________________________________________ > gem5-users mailing list > gem5-users@gem5.org > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > > > > > -- > > *thanks®ards > * > *BISWABANDAN* > http://www.cse.iitm.ac.in/~biswa/ > > “We might fall down, but we will never lay down. We might not be the best, > but we will beat the best! We might not be at the top, but we will rise.” > > > > _______________________________________________ gem5-users mailing list > gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > > _______________________________________________ > gem5-users mailing list > gem5-users@gem5.org > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > -- Joel Hestness PhD Student, Computer Architecture Dept. of Computer Science, University of Wisconsin - Madison http://pages.cs.wisc.edu/~hestness/
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
