Hi, Thank you very much. When I run parsec benchmarks, I see that every time that I use the same parameters, I get a little bit different results while the parameters are not changed. Is this correct and if it is, what could the reason be? Few times I have got the same results but mostly they are different. The range of difference is sometimes large. I set -stopinsns 10000000 but it seems that in different runs, most of times different number of instructions have been run as the number of hits, misses, etc are different. Furthermore, as I don't use ./start_sim, the number of instructions which were run is not written on the terminal. What is the reason?
On Thu, Sep 20, 2012 at 5:22 PM, avadh patel <[email protected]> wrote: > > > On Fri, Sep 14, 2012 at 8:19 PM, Mehrtash Manoochehri <[email protected]>wrote: > >> Hi all >> I have two questions. >> I cannot run parsec. At first i want to run canneal. I do the following, >> it runs but the results are not shown in the stat file and it is empty. >> What should I do? Please see the sequence of my commands. >> scons -c // to clean it >> scons c=4 dramsim=/home/mmanooch/DRAMSim2 //it compiles >> qemu/qemu-system-x86_64 -m 1G -hda /marsVM/parsecROI.img //it boots the >> disc >> > > Here when you invoke qemu, add -snapshot command to make sure disk image > is run under read-only mode. Otherwise there is high probability that your > disk image will get corrupted. > > alt ctrl 2//to set configurations >> simconfig -machine shared_l2 -stopinsns 1000000 -stats m.stat >> > > In simconfig, add -kill-after-run to kill the simulation instance after it > completes and it will flush the stats to specified m.stat file. > > >> alt ctrl 1// >> while I am at root I run the following comamnd: >> parsecmgmt -a run -x roi -c gcc-hooks -p canneal -i simlarge -n 4 >> > > Check out 'create_checkpoints.py' script where we execute couple of other > commands where we export PARSEC_NUM_CPUS and other variables to set per > thread affinity. This is recommended to reduce run-by-run variations. > Don't export checkpoint name variable if you are not creating a checkpoint. > > >> It runs but stat file is not updated. >> >> The other issue is that if I want to run facesim I use the following >> command while the other commands are as shown above. >> >> This is interesting. It might be due to disk image corruption. We are > able to run facesim in ROI region, are you running it with ROI regions? > > - Avadh > > The benchmark runs for a while, but the simulator crashes after a while >> with: >> >> qemu-system-x86_64: ptlsim/build/core/ooo-core/ooo.cpp:870: virtual bool >> ooo::OooCore::runcycle(): Assertion `0' failed. >> Aborted >> >> what is the reason? >> Thank you very much. >> >> _______________________________________________ >> http://www.marss86.org >> Marss86-Devel mailing list >> [email protected] >> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >> >> > -- Regards, Mehrtash Manoochehri Ph.D. Student of Electrical Engineering Department, USC http://www-scf.usc.edu/~mmanooch/
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
