[ofw][patch][WinVerbs tests] Re: fix IPv6related connectionproblem

Sean Hefty sean.hefty at intel.com
Thu Feb 19 02:50:10 PST 2009

>>Possible solutions:
>>1. Exclude T from the result by some artificial trick. (as in the
>>discussed patch)

I don't like this solution...  although I finally understand that you're trying
to overlap the response wait time with getting the time stamp.  The only real
drawback is that the timing of the first ping-pong is too short.  We could
consider excluding the first exchange from the performance data.

>>2. Measure T before the loop and subtract after.

This might be doable.  I think you'll need to subtract T *
current_iteration_count from each timestamp taken inside the loop.

>>3. Make stampings once in N loops decreasing N times it's influence.

This would have been a nice solution, but it loses the histogram.  It should
give the most accurate average.

- Sean

More information about the ofw mailing list