This problem has been existing for quite a few years.. I believe. As far as I can remember, specify "garnet-network=flexible" fix the issue. Also, if your work does not require running the whole benchmark, you can specify the number of instructions to run, for example, 500M. That may not incur such deadlock problem for some reason that I don't know...
On Sat, Jun 21, 2014 at 7:58 AM, Steve Reinhardt via gem5-users < [email protected]> wrote: > If it used to work, and has stopped working, then 'hg bisect' is very > useful to identify exactly where it broke. > > Steve > > > > On Sat, Jun 21, 2014 at 1:30 AM, Choi, Wonje via gem5-users < > [email protected]> wrote: > >> Hi Castillo, >> >> After the simulation was terminated with deadlock message, I have started >> the simulation from checkpoint. However, it still ran into the same error. >> >> Thanks. >> >> Best, >> Wonje >> ________________________________________ >> From: Castillo Villar, Emilio <[email protected]> >> Sent: Saturday, June 21, 2014 1:15 AM >> To: Choi, Wonje; gem5 users mailing list; nihar.b.rathod >> Subject: Re: [gem5-users] Deadlock problem with Ruby in the newest Gem5 >> >> Hello >> >> Ruby deadlocks right after taking a checkpoint. Despite the checkpoint >> being OK. >> Once the simulation is checkpointed you have to stop the simulation and >> start it again from the checkpoint. >> >> ________________________________________ >> De: gem5-users [[email protected]] en nombre de Choi, Wonje >> via gem5-users [[email protected]] >> Enviado: sábado, 21 de junio de 2014 6:42 >> Para: nihar.b.rathod; gem5 users mailing list >> Asunto: Re: [gem5-users] Deadlock problem with Ruby in the newest Gem5 >> >> Nihar, >> >> I'm not sure exactly which version I'm working on, but I think it is late >> 2012 version. Before they convert topology class to regular class. >> >> Best, >> -Wonje >> >> ________________________________________ >> From: gem5-users <[email protected]> on behalf of >> nihar.b.rathod via gem5-users <[email protected]> >> Sent: Friday, June 20, 2014 9:26 PM >> To: [email protected] >> Subject: Re: [gem5-users] Deadlock problem with Ruby in the newest Gem5 >> >> Choi, Wonje via gem5-users <gem5-users <at> gem5.org> writes: >> >> > >> > >> > >> > Hi all, >> > >> > I was trying to run PARSEC and SPLASH-2 benchmarks in ALPHA FS mode with >> Ruby in the newest Gem5. >> > However, I've run into this error message: >> > >> > >> > warn: Prefetch instructions in Alpha do not do anything >> > >> > info: Entering event queue <at> 2618625984500. Starting simulation... >> > Writing checkpoint >> > info: Entering event queue <at> 2618625986000. Starting simulation... >> > panic: Possible Deadlock detected. Aborting! >> > version: 18 request.paddr: 0x[0x330000, line 0x330000] >> m_writeRequestTable: 1 current time: 2618875986000 issue_time: >> 2618625986000 >> difference: 250000000 >> > <at> cycle 2618875986000 >> > [wakeup:build/ALPHA/mem/ruby/system/Sequencer.cc, line 122] >> > Memory Usage: 1739068 KBytes >> > Program aborted at cycle 2618875986000 >> > >> > >> > I ran this simulation with different protocols, such as >> MOESI_CMP_directory, MESI_cmp_directory, and MOESI_hammer. But, all >> simulations ran into same deadlock problem. I used 64 core on mesh >> topology >> with 8 mesh row. I've also ran with both detailed and timing cpu types. >> > I have searched the mailing list, but I could not find a solution. It >> worked with old Gem5 version, I'm not sure why it's doing this in newest >> Gem5. Do you have any suggestion what I should do to solve this problem? >> > >> > Thanks. >> > >> > Best, >> > Wonje Choi >> > >> > Here is the command line that I've used for the simulation. >> > >> > build/ALPHA/gem5.fast --outdir=m5out/network/fft/ >> configs/example/ruby_fs.py --cpu-type=timing --num-cpus=64 --caches >> --l2cache --l1d_size=32kB --l1d_size=32kB --l2_size=8MB --num-dirs=64 >> --num-l2caches=64 --topology=Mesh --mesh-rows=8 --garnet-network=fixed >> --checkpoint-dir=m5out/network/fft/ -b fft >> > >> > >> > >> > >> > >> > >> > >> > _______________________________________________ >> > gem5-users mailing list >> > gem5-users <at> gem5.org >> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> >> >> Hi Wonje Choi, >> >> I also have same problem with parsec and FS mode with latest version of >> gem5. >> Can pls tell me for which version it was working for you? >> >> Thank you, >> >> Best >> Nihar Rathod >> >> >> >> >> _______________________________________________ >> gem5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> _______________________________________________ >> gem5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> _______________________________________________ >> gem5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> > > > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
