On Tue, Apr 28, 2009 at 4:43 PM, Gabe Black <gbl...@eecs.umich.edu> wrote:
> I'm not sure if it will be hugely better. I'd have to just do it and see > how it comes out, I suppose. Wouldn't scheduling an event be similar to > what I have, just using the event queue instead of the new packet > pointer? Yea, at a certain level, the effect is similar--just deferring the callback until later--which is not bad, since that's what addresses the problem. I think that using the existing event queue mechanism (which is explicitly defined for deferring stuff until later) is less convoluted than adding a handful of new variables and control code to cover just this one specific situation. > I could imagine it being better, actually, so I'm tentatively > convinced. There shouldn't be a need to move any x86 code into Alpha to > make that work, but I wouldn't swear to that. > Thanks for giving it a try. Steve
_______________________________________________ m5-dev mailing list m5-dev@m5sim.org http://m5sim.org/mailman/listinfo/m5-dev