I think I've done the tests before and DRAMSim imposes, on average, 30% overhead, so if it's just supposed to run for a few seconds, maybe the benchmark isn't actually running at all.
On Wed, May 22, 2013 at 10:44 PM, Brendan Fitzgerald < [email protected]> wrote: > FFT is my go-to test benchmark. Normally it'll take a few seconds using > just MARSS. I'm not sure how long the MARSS-DRAMSim2 integration takes. > > Plus, I'm pretty sure that FFT doesn't take 100m instructions, but I can't > remember right now. > > > On Wed, May 22, 2013 at 10:39 PM, Paul Rosenfeld <[email protected]>wrote: > >> I'm not entirely sure what the FFT benchmark does, but I just want to >> point out that it's not uncommon for benchmarks to run for that long in >> full system simulators. I will routinely have simulations running for a >> week or longer. >> >> Are you able to see any sign of progress that the simulator is actually >> running (for example by running tail -f ptlsim.log) >> >> >> On Wed, May 22, 2013 at 6:23 PM, Qiuling Zhu <[email protected]>wrote: >> >>> Hi Zoran, >>> >>> Thank you so much! Yes, I used the command "qemu-img info /MARSS/ >>> full_system_image/ubuntu-natty.qcow2" and I saw the checkpoint was >>> created shown as below: >>> >>> ------------------------- >>> system-x86_64 >>> >>> 3 FFT_checkpoint 73M 2013-05-21 19:10:07 00:00:57.521 >>> >>> ------------------------- >>> >>> For your second suggestion, how do I check the memory size for the >>> image? I increased the memory = 4G in the util.cfg, but it seems not work >>> as well. I wonder usually how long the simulation would take if everything >>> goes well? Also, if I don't run simulations in batch mode, do I still have >>> to create the checkpoint? Or can I just run "./start_sim; ./FFT; >>> ./stop_sim" without creating the checkpoint? Many thanks. >>> >>> >>> Regards, >>> >>> Jolin >>> >>> >>> >>> >>> On Wed, May 22, 2013 at 5:15 PM, Zoran Jaksic <[email protected]> wrote: >>> >>>> Maybe my answer is stupid but have you check if the checkpoint is >>>> created? >>>> >>>> If they are, maybe the problem is that memory size for the image is >>>> different when you run run_bench.py script. Try with 4G. >>>> >>>> Those are the problems i had when started with that script so. >>>> >>>> Regards >>>> Zoran >>>> On May 22, 2013 11:07 PM, "Qiuling Zhu" <[email protected]> >>>> wrote: >>>> >>>>> Hi, >>>>> >>>>> I tried to run run_bench.py and it seems to take forever to finish: it >>>>> has been running for 24 hours and it is still running. I wonder if there >>>>> is something wrong with my approach? Really appreciate it if someone can >>>>> help. >>>>> >>>>> Below is how I run the simulation (By the way, I use the ‘dramsim’ >>>>> branch of MARSSx86 which I downloaded from >>>>> https://github.com/dramninjasUMD/DRAMSim2/wiki/MARSSx86-with-DRAMSim2): >>>>> >>>>> >>>>> 1. First of all, I used the ubuntu-natty.qcow2 diskimage and tried to >>>>> simulate its own FFT benchmark >>>>> >>>>> 2. In the disk image create the checkpoint FFT as below and named it >>>>> as FFT_checkpoint >>>>> >>>>> Ubuntu:~# ./create_checkpoin FFT_checkpoint; ./FFT; ./stop_sim >>>>> >>>>> 3. Edit the util.cfg which I attached in the end; >>>>> >>>>> 4. running simulation with run_bench.py as below: >>>>> >>>>> ./util/run_bench.py -d ./marss_results/ -e FFT >>>>> >>>>> Then I see the following stuffs printed on the screen but nothing >>>>> more happens (the result FFT_checkpoint.out is also always empty): >>>>> >>>>> -------------------------------------------------------------------------- >>>>> Reading config file /MARSS/marss.dramsim/util/util.cfg >>>>> Checkpoints: ['FFT_checkpoint'] >>>>> simconfig: -stopinsns 100m -logfile bench.log.txt -stats >>>>> bench.stats.yml.txt -machine single_core %(default_simconfig)s >>>>> Total run configurations: 1 >>>>> 1 parallel simulation instances will be run. >>>>> All files will be saved in: /MARSS/marss_results/ >>>>> Checkpoint FFT_checkpoint >>>>> All Threads are started >>>>> simconfig: -stopinsns 100m -logfile bench.log.txt -stats >>>>> bench.stats.yml.txt -machine single_core -kill-after-run -quiet >>>>> Config file written >>>>> Starting Checkpoint: FFT_checkpoint >>>>> Command: /MARSS/marss.dramsim/qemu/qemu-system-x86_64 -m 2G -serial >>>>> pty -nographic -drive >>>>> cache=unsafe,file=/MARSS/full_system_image/ubuntu-natty.qcow2 -simconfig >>>>> bench.simcfg.txt -loadvm FFT_checkpoint -snapshot >>>>> char device redirected to /dev/pts/0 >>>>> >>>>> (The screen just stays here) >>>>> >>>>> -------------------------------------------------------------------------- >>>>> >>>>> -------------------------------------------------------------------------- >>>>> >>>>> Here is my util.cfg: >>>>> ----------------------- >>>>> [DEFAULT] >>>>> marss_dir = /MARSS/marss.dramsim >>>>> util_dir = /MARSS/marss.dramsim/util >>>>> img_dir = /MARSS/full_system_image >>>>> qemu_bin = %(marss_dir)s/qemu/qemu-system-x86_64 >>>>> default_simconfig = -kill-after-run -quiet >>>>> [suite FFT] >>>>> checkpoints = FFT_checkpoint >>>>> [run FFT] >>>>> suite = FFT >>>>> images = /MARSS/full_system_image/ubuntu-natty.qcow2 >>>>> memory = 2G >>>>> simconfig = -stopinsns 100m -logfile bench.log.txt -stats >>>>> bench.stats.yml.txt -machine single_core %(default_simconfig)s >>>>> >>>>> -------------------------------------------------------------------------- >>>>> >>>>> -------------------------------------------------------------------------- >>>>> >>>>> >>>>> Below is "bench.log.txt" >>>>> ----------------------------- >>>>> // Copyright 1999-2007 Matt T. Yourst <[email protected]> >>>>> // Copyright 2009-2011 Avadh Patel <[email protected]> >>>>> // >>>>> // Git branch 'dramsim' on date Mon Sep 24 13:37:22 2012 (HEAD: >>>>> a01f9c) >>>>> // Built May 21 2013 19:27:22 on BUILDHOST using gcc-4.6 >>>>> // Running on marinara. >>>>> // >>>>> >>>>> Active parameters: >>>>> -help disabled >>>>> -run disabled >>>>> -stop disabled >>>>> -kill disabled >>>>> -flush disabled >>>>> -quiet enabled >>>>> -logfile bench.log.txt >>>>> -loglevel 0 >>>>> -startlog 0 >>>>> -startlogrip 18446744073709551615 >>>>> -consolelog disabled >>>>> -logbufsize 524288 >>>>> -logfilesize 67108864 >>>>> -dump-state-now disabled >>>>> -screenshot >>>>> -log-user-only disabled >>>>> -dump-config-file >>>>> -stats bench.stats.yml.txt >>>>> -yamlstats bench.stats.yml.txt >>>>> -stats-format yaml >>>>> -snapshot-cycles infinity >>>>> -snapshot-now >>>>> -time-stats-logfile >>>>> -time-stats-period 10000 >>>>> -startrip 18446744073709551615 >>>>> -fast-fwd-insns 0 >>>>> -fast-fwd-user-insns 0 >>>>> -fast-fwd-checkpoint >>>>> -stopinsns 100 M >>>>> -stopcycle infinity >>>>> -stopiter infinity >>>>> -stoprip 18446744073709551615 >>>>> -stop-at-marker infinity >>>>> -stop-at-marker-hits infinity >>>>> -stopinsns-rel infinity >>>>> -bbinsns 65536 >>>>> -flushevery infinity >>>>> -kill-after-run enabled >>>>> -event-record >>>>> -event-record-stop disabled >>>>> -event-replay >>>>> -corefreq 0 >>>>> -enable-checker disabled >>>>> -checker-startrip 18446744073709551615 >>>>> -perfect-cache disabled >>>>> -dumpcode test.dat >>>>> -dump-at-end disabled >>>>> -bbdump >>>>> -verify-cache disabled >>>>> -machine single_core >>>>> -enable-mongo disabled >>>>> -mongo-server 127.0.0.1 >>>>> -mongo-port 27017 >>>>> -bench-name >>>>> -tags >>>>> -run-tests disabled >>>>> -execute-after-kill >>>>> -sync 0 >>>>> -simpoint >>>>> -simpoint-interval 10 M >>>>> -simpoint-chk-name simpoint >>>>> -dramsim-device-ini-file ini/DDR3_micron_8M_8B_x16_sg15.ini >>>>> -dramsim-pwd ../DRAMSim2 >>>>> -dramsim-system-ini-file system.ini >>>>> -dramsim-results-dir-name MARSS >>>>> ~ >>>>> >>>>> >>>>> >>>>> Thanks, >>>>> Jolin >>>>> >>>>> _______________________________________________ >>>>> http://www.marss86.org >>>>> Marss86-Devel mailing list >>>>> [email protected] >>>>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >>>>> >>>>> >>> >>> >>> -- >>> There will be miracles when you believe! >>> >>> _______________________________________________ >>> 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 >> >> >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
