Hi Andreas, I wrote my own script to configure a new system using "learning gem-5" scripts. But I couldn't run a bench mark with that. So I tried to edit the fs.py file. But it has some problems; 1) I got an error like "fatal: Unable to find destination for addr 0x100000000 on system.membus". Then I tried to increase the addrrange for system.mem_range, and it went upto 5GB.
2) So whenever I tried to simulate parsec benchmark the system hangs up after a few seconds. 3) I noticed in stats.txt file like the workload is not allocating to all the 4 cores, only first cpu have all the workload. Can you give me some suggestions to avoid above problems regards, Rohith Mathew On Fri, Nov 6, 2015 at 1:48 PM, Andreas Hansson <[email protected]> wrote: > Hi Rohith, > > The classic memory system is using a MOESI-based coherency protocol, and > supports very flexible topologies built around crossbars. For anything up > to 8-16 cores it is a very good starting point in my view. > > For the second question the scope is truly huge. You may want to have a > look at the caches, as well as the memory controller. Also, these days > gem5 supports “Buffer on Board” style memory systems using an on-chip > controller as well as an off-chip memory controller, so you may even > consider both of these. I suspect the answer to the question “where should > this functionality go” would be enough for a couple of PhD theses... > > Andreas > > On 06/11/2015, 05:07, "gem5-dev on behalf of rohith mathew" > <[email protected] on behalf of [email protected]> wrote: > > >Hi Andreas, > >Thanks for the support. I have some doubt like, > > > >1)if we want to use coherency protocols, is classic memory system enough > >or > >do we need to go for ruby? > >2)If I want to occupy some database operations in main memory for my HTM > >implementation, which all files I should work with? > > > >Thanks in advance. > > > >regards, > >Rohith > > > >On Thu, Nov 5, 2015 at 2:00 PM, Andreas Hansson <[email protected]> > >wrote: > > > >> Hi Rohith, > >> > >> I am afraid I cannot help you with the HTM bit (there was some previous > >> work posted on the review board years ago...), but when it comes to the > >> system topology, the gem5 classic memory system is very flexible. > >> > >> I’d suggest to write your own config script, possibly based on > >> config/fs.py, or even the “learning-gem5” scripts. Simply instantiate > >>the > >> components, and CPU clusters one by one, connect them up, and off you > >>go. > >> It really is not that complex. The biggest problem here is that the > >> existing example scripts try to accommodate a lot of various topologies, > >> and do so by means of if-statements. > >> > >> Andreas > >> > >> On 05/11/2015, 05:59, "gem5-dev on behalf of rohith mathew" > >> <[email protected] on behalf of [email protected]> wrote: > >> > >> >Hi, > >> > > >> >I am trying to implement a HTM(Hardware Transactional Memory) in Gem5 > >>with > >> >x86. For that I want to configure a system with 4 cores, each having 2 > >> >levels of private caches and a third level of shared cache. can anyone > >> >help > >> >me out with this, because I am really new to this Gem5. > >> > > >> >regards, > >> >Rohith > >> >_______________________________________________ > >> >gem5-dev mailing list > >> >[email protected] > >> >http://m5sim.org/mailman/listinfo/gem5-dev > >> > >> > >> ________________________________ > >> > >> -- 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. > >> _______________________________________________ > >> gem5-dev mailing list > >> [email protected] > >> http://m5sim.org/mailman/listinfo/gem5-dev > >> > >_______________________________________________ > >gem5-dev mailing list > >[email protected] > >http://m5sim.org/mailman/listinfo/gem5-dev > > > ________________________________ > > -- 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. > _______________________________________________ > gem5-dev mailing list > [email protected] > http://m5sim.org/mailman/listinfo/gem5-dev > _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
