Re: [gem5-users] Reducing the benchmark execution time

[Tod]

Hi Andreas, Okay, that's fine. I could not get exactly, how to launch multiple copies of gem5? Will it work in a single OS and how? How can i specify the disk-image copy (It should be a separate disk-image file, right?) for each gem5 execution?  So far the disk image file has been specified by the environment variable 'LINUX_IMAGE' and it was only one at a given time.  Best, Tod --------- 원본 메일 --------- 보낸사람: Andreas Hansson <andreas.hans...@arm.com> 받는사람 : gem5 users mailing list <gem5-users@gem5.org> 날짜: 2014년 2월 08일 토요일, 23시 16분 21초 +0900 제목: Re: [gem5-users] Reducing the benchmark execution time Hi Tod, I’m glad to hear it worked out. I see no issues with running your experiments in parallel. Simply launch multiple copes of gem5 through e.g. a shell script. Don’t forget to pass “-d” to gem5 and give each parallel run a unique output path. As a minor note, you might also want to double check that you are disabling all debug ports etc (disable listeners in the gem5 scripts) to make sure you don’t accidentally connect to a port assigned to any of the running copies. Good luck with the experiments. Andreas From: Tod <saril...@hanmail.net> Reply-To: gem5 users mailing list <gem5-users@gem5.org> Date: Saturday, 8 February 2014 02:07 To: gem5 users mailing list <gem5-users@gem5.org> Subject: [gem5-users] Reducing the benchmark execution time Thanks Andreas, It did work :) Cutting down to multiple of tens of minutes from multiple hours! I am yet interested to decrease the overall time of my experiment. If there are tens of cores and tens of GBs of memory is available on the real hardware, would it be wise to consider multiple copies of a virtual machines running in parallel?  Where each of them will run a gem5 binary (that is sequential) on top of it and use a disk image independently. That might divide the execution job, hence the execution time, by the number of copies of VM running in parallel? Unless the real hardware or the host OS does not put a bound on it? Would it work and is it okay? OR, is there any other better way to achieve the similar? Thank you. Best regards, Tod --------- 원본 메일 --------- 보낸사람: Andreas Hansson <andreas.hans...@arm.com> 받는사람 : gem5 users mailing list <gem5-users@gem5.org> 날짜: 2013년 11월 26일 화요일, 20시 33분 30초 +0900 제목: Re: [gem5-users] Reducing the benchmark execution time Hi Tod, There are a few things you can try. First, make sure you run gem5.fast.  Second, take a checkpoint just before the ROI and only re-run the part of the workload you want to measure. Third, be more selective about your experiments (see e.g. "A Structured Approach to the Simulation, Analysis and Characterization of Smartphone Applications”, IISWC’13).  Good luck. Andreas From: tod <saril...@hanmail.net> Reply-To: gem5 users mailing list <gem5-users@gem5.org> Date: Tuesday, 26 November 2013 04:25 To: gem5-users <gem5-users@gem5.org> Subject: [gem5-users] Reducing the benchmark execution time Hello Everyone, I am using detailed architecture with caches in FS mode with boot script.  The execution time for my current simulation is about one hour. since I have to repeat my experiment several times so it seems too much.  Can i reduce this execution time with out changing architecture etc for the complete execution?  Thank you. best, tod -- 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 _______________________________________________ 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

_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users