A while back I think I made totalInstructions() a virtual function because it was impl'ed by all the cpus anyway, don't know if that's basically what you're looking for, functionality-wise.
Lisa On Mon, May 24, 2010 at 9:20 PM, nathan binkert <[email protected]> wrote: > > There seem to be a few issues with the comInstEventQueue. Although the > queue counts instructions, it is still subject to the assert(when >= > curTick); assertion. Additionally, there doesn't seem to be any way to find > out what the current instruction is to make sure that the scheduling always > happens in the future. > > > > Nate, the only thing I can think to do is have a parameter for if the > assert should be checked (either a variable or a template parameter), > thoughts? > In general, this needs to get fixed anyway if we're going to have > multiple event queues. Really, we should get rid of curTick and put > it inside the event queue class itself. The problem is that there's a > hell of a lot of things that use curTick. Perhaps we can have both > for a while and phase out curTick. Initially, curTick could become an > inline function in the EventManager class (a base class of SimObject) > that gets the curTick of the event queue. > > > Anyone mind if I get rid of the instCount() function/variable that was > used for legion validation and create a virtual function in BaseCpu that > returns numInst? > No feelings one way or the other, but it seems reasonable. Does it > get called frequently? > > Nate > _______________________________________________ > 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
