<br><font size=2><tt>Leonid Arsh <leonida@voltaire.com> wrote on
04/27/2006 01:24:49 AM:<br>
<br>
> Shirley Ma wrote:<br>
> > Without seeing your patch, I coudn't say anything. I guess your
<br>
> > implemention<br>
> > didn't handler multithreads simultanously. If you only have one
<br>
> > interrupt handler,<br>
> > couldn't see any reason you can get better performance number
with<br>
> > splitting CQs.<br>
> Shirley, you are right.<br>
> I just wanted share our experience with you.<br>
> <br>
> All the tests we made on our IPoIB driver, so our NAPI implementation
<br>
> isn't relevant here.<br>
> Unfortunately, we didn't plan to work on the IPoIB performance in
the <br>
> nearest future, so I can't<br>
> implement NAPI on the OpenIB driver right now.<br>
> <br>
> I think it would be very interesting to compare the NAPI performance
<br>
> against the work queue.<br>
> Please let me know if you are planning to do it yourself. <br>
> ><br>
> > Could you please post your NAPI patch here?<br>
> ><br>
> > As I mentioned I will test my patch to see how's the performance.<br>
> <br>
> <br>
> Thank you,<br>
> Leonid</tt></font>
<br>
<br><font size=2><tt>How many percentage throughput you got from your NAPI
implementation?</tt></font>
<br>
<br><font size=2><tt>So far work queue gives very consistent 15% througput
increase in my </tt></font>
<br><font size=2><tt>local test with one dual core cpu over mthca. I am
planning to add one more </tt></font>
<br><font size=2><tt>cpu to see the difference.</tt></font>
<br>
<br><font size=2><tt>Yes, NAPI is in our plan. We can see NAPI vs. work
queue results soon.</tt></font>
<br><font size=2><tt>Actualy if we can combine work queue with NAPI, that
would be more</tt></font>
<br><font size=2><tt>interesting.</tt></font>
<br>
<br><font size=2><tt>Thanks<br>
</tt></font><font size=2 face="sans-serif">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>