Thanks, Timothy! This looks like a great addition and I really appreciate your effort to contribute your changes back.
I did several comments on reviewboard, but rather than flood the list with a bunch of "ship it!" comments, I'll just say here that everything I didn't comment on looks good to me. I'm sure Nate will have something to say as well, particularly about the stats backup/restore code. Steve On Fri, Jul 9, 2010 at 7:10 PM, Timothy M Jones <[email protected]> wrote: > Hi everyone, > > I've just posted to the review board 12 patches that enable me to perform > SMARTS-like sampling simulation. Most are finished and I'd be happy to > commit them if everyone is ok with them. A couple are not quite there yet > and I'd like input on whether they are correct and whether I'm doing things > in the best way. > > To summarise what I've done: > > 1) Taken the branch predictor out of O3CPU and made it a stand-alone > SimObject that is accessible from python; > > 2) Added functionality to the simple CPUs to allow them to warm the branch > predictor; > > 3) Altered the takeOverFrom functions so that CPU memory is always connected > up if required. Also provided functions to mark all cache data as ready > which is required after functional warming; > > 4) Provided functions available in python to get the number of instructions > executed by each CPU and to set the maximum number to execute for the next > phase; > > 5) Provided the ability to backup and restore stats so that they are not > counted during the warming up phase; > > 6) Altered the example simulation scripts so that folk can see how I am > simulating like this and do it themselves; > > 7) Made various bug fixes and alterations to correct corner cases I found > when implementing this stuff. > > I'm not so sure about 1) and it would be good to get Kevin's input on that. > I hope to extend it to make one generic branch predictor so that the > in-order core can make use of it too, instead of having it's own, duplicated > code. > > I'm also not so sure about the backing up and restoring of stats. I've not > implemented this in all of the classes and I'm not 100% sure what's going > on, so this is a kind of stab-in-the-dark. Nate, if you could give me > feedback on what is right and wrong here, or if there's a better approach, > that would be much appreciated. > > In general, if there are any glaring errors or things I haven't thought > about, please let me know. > > All these patches should apply with no problems (in the order they're posted > on the review board) and you can try out the sampling for yourselves too :-) > > Cheers > Tim > > -- > Timothy M. Jones > http://homepages.inf.ed.ac.uk/tjones1 > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > _______________________________________________ > m5-dev mailing list > [email protected] > http://m5sim.org/mailman/listinfo/m5-dev > _______________________________________________ m5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/m5-dev
