Hello all --

I just realized I created a ConstVectorNode back in the spring but never
sent out the diff of statistics.hh.  Here it is.  Better late than never I
guess...

Note that this diff includes two separate changes to statistics.hh:  the
ConstVectorNode, and the fix to the BinaryNode total() function (that was
also discussed on the list at around the same time).  Also note that this is
a diff from statistics.hh in m5-2.0b2; I know there were some changes in
this file in m5-2.0b3, but I haven't yet made the switch (sorry about
that).  Hopefully that won't be too much of an impediment.

Thanks,

   -Vilas

On 3/6/07, Nathan Binkert <[EMAIL PROTECTED]> wrote:
>
> I guess I missed a little.  You should just create a ConstVectorNode like
> ConstNode, and create a new version of constant that builds a VectorNode
> if an array or vector is passed in.
>
> > You could add a constant vector without too much trouble. If you end up
> doing
> > it, please send us a diff and we'll include it in the next release.
> >
> > Thanks,
> > Ali
> >
> > On Mar 6, 2007, at 5:21 PM, Vilas Sridharan wrote:
> >
> >> Hi Ali,
> >>
> >> Thank you for pointing that out -- I hadn't seen that originally.  Of
> >> course, one of the things that I glossed over in my initial mail was
> that I
> >> actually was hoping to be able to create a constant Stats::Vector in
> the
> >> constructor.  The Stats::constant seems to understandably only support
> >> single values.  (I am essentially trying to compute a weighted sum
> using a
> >> Stats::Formula as the product of two Stats::Vectors.)
> >>
> >> I have come up with a workaround for my own purposes, but it's ugly
> (set
> >> the Stats::Vector values the first time the object is called).  I was
> >> hoping for a slightly more elegant way of doing it, but oh well :-).
> >>
> >> Thanks,
> >>
> >>    -Vilas
> >>
> >> On 3/6/07, Ali Saidi <[EMAIL PROTECTED]> wrote:
> >> Vilias,
> >>
> >> Take a look at the the end of NSGigE::regStats(). I think you just
> >> need to wrap your constant in Stats::constant(myconstant) like is
> >> done with the txBandwidth formula.
> >>
> >> Ali
> >>
> >> On Mar 6, 2007, at 4:14 PM, Vilas Sridharan wrote:
> >>
> >>> Hello again --
> >>>
> >>> Using M5 2.0b2, I have created a class which I instantiate in
> >>> various CPU structures.  This class uses Stats::Formula objects to
> >>> calculate values for me at the end of simulation.  These values
> >>> involve per-structure parameters which are constant over simulation
> >>> but different between each structure.  I'm having trouble getting
> >>> these values into my Stats::Formula variables.  I may just be
> >>> overlooking something silly here...
> >>>
> >>> From what I can tell, it seems that all the Stats get re-
> >>> initialized sometime between the constructor call and the time
> >>> simulation starts.  (I've tried variously putting init code in the
> >>> constructor and in regStats(), after the Stats have been
> >>> registered, with the same results.)  Is there a way to declare a
> >>> 'constant' stat that doesn't get initialized like this at the
> >>> begining of execution?  For some reason, I can't seem to use the
> >>> value that I pass in directly in the Stats::Formula, either (the
> >>> website has a note to this effect.)
> >>>
> >>> Am I missing something?
> >>>
> >>> Thanks,
> >>>
> >>>    -Vilas
> >>> _______________________________________________
> >>> m5-users mailing list
> >>> [email protected]
> >>> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> >>
> >> _______________________________________________
> >> m5-users mailing list
> >> [email protected]
> >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> >>
> >> _______________________________________________
> >> m5-users mailing list
> >> [email protected]
> >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> >
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>

Attachment: statistics.hh.diff
Description: Binary data

_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to