> It seems as if bitsets are only used to determine the # of destinations in a > memory hierarchy (e.g. L2 sharers) > > We dont know that a compile time but we do know that at configuration time > (swig time?). > > That means, once Ruby builds its system, it can pass the max sharers to the > gem5 builder and use that to generate a file with a typedef bitset. > Something like: > typedef std::bitset<MaxSharers> NetDest; > > Would that make sense? Its a least a better option then being limited to 64 > cores of simulation.
The biggest problem with this approach is that some of the operations don't work correctly if there are extra unused bits at the end. For example, there's a count all ones function, if the bits at the end are unused and set at zero, we need to rewrite that function. Nate _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
