Hi Timo, Note that even the ‘timing’ core is not representative of anything realistic. If all you do is dependent load/store (lat_mem_rd with thrashing) you should be ok, but in general, for any performance numbers use minor or arm_detailed.
Concerning the low bandwidth, you are running without caches. Have a look at m5out/config.dot.pdf or preferably m5out/config.dot.svg for an illustration of the system. If you do not see these files, make sure you have py-dot installed. Andreas On 16/07/2015 22:06, "gem5-users on behalf of Timo Schneider" <gem5-users-boun...@gem5.org on behalf of ti...@inf.ethz.ch> wrote: >On Thu, 2015-07-16 at 21:33 +0100, Andreas Hansson wrote: > >Hi Andreas, > >> As a general rule, never use any performance number from atomic mode. >> Atomic mode is for fast forwarding and warming (and for anything >> non-temporal). The only notion of time is ‘enough to not confuse the >>OS’. >> >> I would recommend to re-run your experiment with a realistic timing core >> model (e.g. minor or arm_detailed), and a realistic DRAM controller >>(e.g. >> DDR3_1600_x64). > >Thanks! That helped! I am using --cpu-type=timing >--mem-type=DDR3_2133_x64 now and it seems to work --- I get 27.9 MB/s >for one and 17.5 MB/s for two procs. > >I am surprised that it is that low, but I am sure that the total memory >bandwidth can be configured somewhere. > >Regards, >Timo > >_______________________________________________ >gem5-users mailing list >gem5-users@gem5.org >http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users -- IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you. ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2557590 ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2548782 _______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
