I'd be more for having ClockDomains which the various things that needed a clock got a pointer and a multiplier to. Adjustments to the ClockDomain would then change all the clocks that were based on it. This would make implementing frequency scaling reasonably easy to do, although figuring out the object hierarchy would probably be a bit annoying without MI.
Ali On Apr 1, 2012, at 3:59 AM, Andreas Hansson wrote: > Hi guys, > > As Nate pointed out, I wasn't saying all modules should run on the same > clock. What I hinted at was the ability to use the parent.clock as the > default, and then ensure that there is always at least one clock, on the > system level. > > The important part here, from my point of view, is to not force a module to > have a single clock, but rather push it down to the level of ports. A clock > domain crossing is a module with two clocks, and it would be extremely > valuable (in my opinion) to be able to check that two ports agree on the > clock when they are bound together. > > For the rest the ClockedObject sounds very reasonable. Why not make it > MemObject btw? > > Andreas > ________________________________________ > From: [email protected] [[email protected]] On Behalf Of Gabe > Black [[email protected]] > Sent: Sunday, April 01, 2012 6:57 AM > To: [email protected] > Subject: Re: [gem5-dev] Clock Defaults > > On 03/31/12 21:30, nathan binkert wrote: >>> but I guess that isn't what he meant. Making clocks into simobjects >>> seems reasonable. >> Seems overkill to me. What benefit is there in it being a sim object >> over just it already being a special parameter type. In fact I see >> downsides since there's all of the proxy stuff with math that would >> have to be reimplemented. >> >> Nate >> _______________________________________________ >> gem5-dev mailing list >> [email protected] >> http://m5sim.org/mailman/listinfo/gem5-dev > > The described benefit is that you could have special clocks which have > additional behavior, but thinking about it more, if you made the CPU do > that for instance, there would probably still need to be some > integration between the CPU and the clock so the CPU would know what was > going on, the clock would know when to switch behavior, etc. > > Gabe > _______________________________________________ > 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
