<br><tt><font size=2>Benjamin Herrenschmidt <benh@kernel.crashing.org>
wrote on 11/07/2006 02:38:12 PM:<br>
<br>
> On Tue, 2006-11-07 at 11:14 -0600, James K Lewis wrote:<br>
> > <br>
> > Ben, <br>
> > <br>
> > This is very interesting. If it will increase Spidernet
performance<br>
> > without causing more bugs then we should investigate. Before<br>
> > attempting any of this though I would like more information.
For<br>
> > example, are we POSITIVE we will see a performance increase by<br>
> > implementing these changes? Any idea how much? Our current driver
is<br>
> > at about 700 Mbps on TX with 1500 byte packets at approx. 30%
CPU<br>
> > usage. On RX, about 720 Mbps at 100% CPU (and thousands of interrupts.<br>
> > NAPI does not work on this thing because of interrupt problems).
<br>
> <br>
> What about Linas patches to do interrupt mitigation with NAPI polling
?<br>
> That didn't end up working ?<br>
> Ben.</font></tt>
<br><tt><font size=2> </font></tt>
<br><tt><font size=2>NAPI did not work. Linas and I found, working independently,
that the turning off and on of the interrupts on Cell does not work properly.
Specifically, you could mask a particular interrupt and it would fire anyway.
I asked repeatedly for help with this from Sony/Toshiba. Their only response
was they needed "more info". </font></tt>
<br><tt><font size=2><br>
I still think your latest idea is something to pursue but not at this time.
Maybe after I deal with my Infiniband issues we can look into this more
closely.<br>
</font></tt>
<br>
<br><font size=2 face="sans-serif"><br>
Jim Lewis<br>
Advisory Software Engineer<br>
IBM Linux Technology Center<br>
512-838-7754<br>
<br>
</font>
<br>