[Ns-bugs] [Bug 189] CSMA device receives all packets it sends

[bugzilla-daemon@nsnam-www.ece.gatech.edu]

http://www.nsnam.org/bugzilla/show_bug.cgi?id=189





------- Comment #10 from mathieu.lacage at sophia.inria.fr  2008-05-30 20:46 -------
(In reply to comment #9)
> >> Well, you have to increment a reference count and then decrement it.
> 
> > no, you have to call Packet::Copy which performs a heap allocation.
> 
> This is descending into esoterica a little, but okay, you allocate a smart
> pointer with reference count of 1 to hold the reference to the underlying
> buffer, the count of which which is incremented and then eventually decremented
> when the smart pointer reference is decremented and finally the smart pointer
> is deleted.  Right?  I was trying to make the point that this isn't a
> copy-the-whole-packet operation, it's logically a reference counting operation.

That 'smart pointer' is, however, much bigger than a single count integer. I
have not measured it but I expect it to be in the 100 bytes size. I am probably
wrong about the exact figure, but, well.

> Anyway, this operation currently happens for all N of the N devices on the
> channel; putting the source filtering in the channel would allow us to do this
> (N-1) times at the cost of putting the filtering in a strange (IMO) place.
> 
> I lean toward implementing the filtering in the place where it would normally
> (in the real world) go and not in a place that normally has nothing but copper
> atoms ;-)
> 
> If you can make a case that this has a big performance impact I might be
> convinced to do what I think is unusual thing in furtherance of speed.  I might
> also suggest that adding filtering to the device will have negligable
> performance impact since the copy already happens and I'm just adding a
> compare.  Perhaps more impact could be made performance-wise by speeding up all
> N packet operations here and in all other devices with a packet cache to remove
> the new/delete if we are concerned about the issue.

I don't think it is a big deal but I did not want to leave uncorrected an
incorrect assumption about Packet::Copy :)


-- 
Configure bugmail: http://www.nsnam.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.