[Cbe-oss-dev] Spider DMA wrongness

[James K Lewis]

Benjamin Herrenschmidt <benh at kernel.crashing.org> wrote on 11/07/2006 
02:38:12 PM:

> On Tue, 2006-11-07 at 11:14 -0600, James K Lewis wrote:
> > 
> > Ben, 
> > 
> >   This is very interesting. If it will increase Spidernet performance
> > without causing more bugs then we should investigate. Before
> > attempting any of this though I would like more information. For
> > example, are we POSITIVE we will see a performance increase by
> > implementing these changes? Any idea how much? Our current driver is
> > at about 700 Mbps on TX with 1500 byte packets at approx. 30% CPU
> > usage. On RX, about 720 Mbps at 100% CPU (and thousands of interrupts.
> > NAPI does not work on this thing because of interrupt problems). 
> 
> What about Linas patches to do interrupt mitigation with NAPI polling ?
> That didn't end up working ?
> Ben.
 
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". 

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.



Jim Lewis
Advisory Software Engineer
IBM Linux Technology Center
512-838-7754


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/cbe-oss-dev/attachments/20061107/692a658a/attachment.htm>