OK. So do you mean the parameter should be used to set the iteration times? And we will output the last time iteration results? Do we need a global AVG result for all the iterations?
> -----Original Message----- > From: Wisam Jaddo <[email protected]> > Sent: Tuesday, November 18, 2025 8:43 PM > To: Bing Zhao <[email protected]>; Slava Ovsiienko > <[email protected]>; [email protected]; Raslan Darawsheh > <[email protected]> > Cc: Ori Kam <[email protected]>; Dariusz Sosnowski <[email protected]>; > Suanming Mou <[email protected]>; Matan Azrad <[email protected]>; NBU- > Contact-Thomas Monjalon (EXTERNAL) <[email protected]> > Subject: RE: [PATCH] test-flow-perf: support measuring 2nd round rate > > Hi, > > > -----Original Message----- > > From: Bing Zhao <[email protected]> > > Sent: Monday, November 17, 2025 9:30 AM > > To: Slava Ovsiienko <[email protected]>; [email protected]; Raslan > > Darawsheh <[email protected]>; Wisam Jaddo <[email protected]> > > Cc: Ori Kam <[email protected]>; Dariusz Sosnowski > > <[email protected]>; Suanming Mou <[email protected]>; Matan > > Azrad <[email protected]>; NBU-Contact-Thomas Monjalon > > (EXTERNAL) <[email protected]> > > Subject: [PATCH] test-flow-perf: support measuring 2nd round rate > > > > In some driver, the flow table may start from a small size and > > increase based on the rules number. The table resize may not reflact > > the actual rules insertion rate due to the extra overhead. > > > > Keeping 1 rule in the table and measuring the 2nd iteration would be > > more accurate. > > I'm ok with the approach of enabling the measurement of the table warm up > effect over the insertion/deletion, but I tend to believe that we should > have it by iterations were user can set the needed rounds instead of > depending on only 2 rounds > > BRs, > Wisam Jaddo
