<br><font size=2><tt>Roland Dreier <rdreier@cisco.com> wrote on 04/18/2006
03:06:33 PM:<br>
<br>
> And actually it argues against splitting the CQ, because having one
CQ<br>
> increases the number of CQ entries that we have a chance to poll at<br>
> any one time, by lumping send and receive completions together...<br>
> <br>
> - R.<br>
</tt></font>
<br><font size=2><tt>The send needs to obtain tx_lock to handle send wc.
Receive doesn't </tt></font>
<br><font size=2><tt>need to. That might explain some of the performance.
</tt></font>
<br><font size=2><tt>I have tried to remove tx_lock in send wc handler
by removeing tx_ring. </tt></font>
<br><font size=2><tt>It did increase the send speed. I hit an ehca problem,
the test couldn't move on.</tt></font>
<br>
<br><font size=2><tt>And splitting CQ is really benefit the duplex throughput.</tt></font>
<br>
<br><font size=2><tt>We are doing more testing now to dig out more info.</tt></font>
<br><font size=2 face="sans-serif"><br>
Thanks<br>
Shirley Ma<br>
IBM Linux Technology Center<br>
15300 SW Koll Parkway<br>
Beaverton, OR 97006-6063<br>
Phone(Fax): (503) 578-7638<br>
<br>
</font>