On Wed, Jul 28, 2010 at 11:59 AM, Hz Xes <[email protected]> wrote: > Hi > > > > I have questions about “Cannot get snapshot 'final'”. I found a related > post a couple months ago. But I can not get solution. > > > > (1) > > > > When I use the “ –stopinsns “ option to make simulation stop after a mount > of instructions, I could not get statistics convert to a readable file. > > > > It looks that simulation could stop at the specified limit, but I could not > convert the statistics file. > > > > At end of the logfile: > > …… > > Completed 10087000 cycles, 9971993 commits: 94698 Hz, 85831 > insns/sec: rip 00007f6b784ed646 > > Completed 10103000 cycles, 9991262 commits: 79624 Hz, 95892 > insns/sec: rip 00007f6b784cea34 > > Stopping simulation loop at specified limits (10110227 iterations, 10000000 > commits) > > update mode count for ctx 0 > > update mode count for ctx 0 > > update mode count for ctx 0 > > > > Stopped after 10110227 cycles, 10000000 instructions and 108 seconds of sim > time (cycle/sec: 93613 Hz, insns/sec: 92592, insns/cyc: 0.9890974752594576) > > > > > > the simulation configuration I used in run_bench.py (download from Marss86 > site): > > > > # run_bench.py > > …… > > sim_cmd_generic = 'simconfig -stats %s.stats -logfile %s.log -run > -stopinsns 10m -kill-after-finish' > > …… > > > > And at produced qemu window, it shows > > > > Execution command: > > Generic simulation command: simconfig -stats %s.stats -logfile %s.log -run > -stopinsns 10m -kill-after-finish > > Chekcpoints to run: ['blackscholes'] > > All files will be saved in: ./ > > Running thread with img: ./disks/parsec_image/parsec.qcow2 > > All Threads are started > > Starting Checkpoint: blackscholes > > initing the ooomachine > > char device redirected to /dev/pts/4 > > char device redirected to /dev/pts/5 > > QEMU 0.12.3 monitor - type 'help' for more information > > (qemu) loadvm blackscholes > > (qemu) simconfig -stats ./blackscholes.stats -logfile ./blackscholes.log > -run -stopinsns 10m -kill-after-finish > > simulation options received:-stats ./blackscholes.stats -logfile > ./blackscholes.log -run -stopinsns 10m -kill-after-finish > > > > > > Did I make some mistakes on the configurations or somewhere? Please let me > know, great thanks! > > > Finally I came across this bug yesterday and the issue is that if MARSS is not killed after the simulation is completed, it overwrites the stats file. So right now workaround is to give '-kill-after-run' option in your simconfig and it should work.
> (2) Second question about run_bench.py (download from Marss86 site): > > > > I found that simulations are blocked after first benchmark reaches the > instructions limit. It looks run_bench.py blocks at > > > > # run_bench.py > > …… > > p.wait() > > …… > > > > And at produced qemu window, it shows > > > > Execution command: > > Generic simulation command: simconfig -stats %s.stats -logfile %s.log -run > -stopinsns 10m -kill-after-finish > > Chekcpoints to run: ['blackscholes'] > > All files will be saved in: ./ > > Running thread with img: ./disks/parsec_image/parsec.qcow2 > > All Threads are started > > Starting Checkpoint: blackscholes > > initing the ooomachine > > char device redirected to /dev/pts/4 > > char device redirected to /dev/pts/5 > > QEMU 0.12.3 monitor - type 'help' for more information > > (qemu) loadvm blackscholes > > (qemu) simconfig -stats ./blackscholes.stats -logfile ./blackscholes.log > -run -stopinsns 10m -kill-after-finish > > simulation options received:-stats ./blackscholes.stats -logfile > ./blackscholes.log -run -stopinsns 10m -kill-after-finish > > (qemu) > > ^ infinite wait here. > > > > Did I make some mistakes on running simulation via run_bench.py script? > Please let me know. > Yes, your configuration has '-kill-after-finish' and its an invalid option, so the simulator switch to emulation mode and does not 'kill' itself. So the 'run_bench.py' waits for that process to finish and does not kick off other simulation. Use '-kill-after-run' option as mentioned above. It should fix this issue. > (3) > > Could any share a typical simulation configuration file a multic-core with > cache configuration for Marss86? > Currently for quick simulation runs we use the default configuration. For caches, you can use some standard processor cache sizes for your purpose. You can find more details of pipeline in wiki for latest processors like intel nehalem. - Avadh > > > Thanks a lot, > > HZ > > > _______________________________________________ > 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
