Hi Andreas, I agree that determinism, worthwhile performance improvement, and not re-inventing the wheel are very important; I tried to capture that on the wiki page. I would prefer to start working with timing (rather than atomic) mode though, since that's where the performance is worst, and the decoupling between objects in timing mode is better than in atomic mode.
Steve On Tue, Sep 6, 2011 at 12:38 AM, Andreas Hansson <[email protected]>wrote: > Hi all, > > An interesting starting point would also be to focus the parallelisation on > the atomic mode and introduce the notion of time quanta and temporal > decoupling, similar to loosely-timed TLM-2.0. Overall there is probably much > to be learnt from the parallelisation efforts of the SystemC kernel and the > large body of academic papers on optimistic/conservative parallel simulation > of these high-level models. If we're not careful the result will be a > non-deterministic simulator, with a marginal performance improvement when > running in multiple threads and a slow-down when running it single threaded. > > Andreas > > > -----Original Message----- > From: [email protected] [mailto:[email protected]] On > Behalf Of Ali Saidi > Sent: 05 September 2011 17:58 > To: gem5 Developer List > Subject: Re: [gem5-dev] Parallel M5 > > Hi Digant, > > Steve has been looking at parallelization a little more recently than Nate > probably has. As far as that list goes, we're probably at around step 6. The > next thing to do is creating a number of event queues and assigning sim > objects to one queue or another. A good first step would be to support > simulation of two systems in two threads but a single process. A next step > beyond that would be supporting 2 systems communicating over an ethernet > link that was multi-thread aware. > > Thanks, > Ali > > > > On Sep 2, 2011, at 7:55 PM, Digant wrote: > > > Hi Nathan, > > I am thinking to working in the direction of parallel M5, can you brief > me > > whats the current status of http://m5sim.org/wiki/index.php/Parallel_M5? > > > > Thanks > > Digant Desai. > > _______________________________________________ > > 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
