[PATCH] Fix Tsi108 ethernet driver performance

[Alexandre Bounine]

I agree that 20 ms is not the best option. But, as you mentioned as
well, shorter period may bring more wasted CPU cycles.
Using this approach to interrupt moderation I see performance
improvement using frame flow that generates about one interrupt per
frame. Below are CPU usage numbers reported by top for the current and
patched versions of the Tsi108_eth driver.       

Current version:
Cpu(s):  0.0% us, 0.3% sy, 0.0% ni, 14.6% id, 0.0% wa, 14.6% hi, 70.4%
si

Patched version:
Cpu(s):  0.0% us, 0.0% sy, 0.0% ni, 35.2% id, 1.2% wa, 2.4% hi, 61.2% si

- Alex.


-----Original Message-----
From: Scott Wood [mailto:scottwood at freescale.com] 
Sent: Friday, July 13, 2007 4:33 PM
To: Scott Wood
Cc: Alexandre Bounine; linuxppc-dev at ozlabs.org
Subject: Re: [PATCH] Fix Tsi108 ethernet driver performance

Scott Wood wrote:
> Wouldn't this cause latencies of up to a second in responding to 
> received packets?  I'd think that would be considered excessive.
> 
> And shouldn't NAPI be reducing the RX interrupt load anyway?

Sorry, I missed that you reduced the interval to 20 ms.  Still, a bit on

the high side, and a source of wasted CPU cycles when the network isn't 
heavily loaded.  How much of a difference in CPU usage did you see with 
this patch?

-Scott

---

Important Notice: This message is intended for the use of the individual to whom it is addressed and may contain information which is privileged, confidential and/or exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, or is not the employee or agent responsible for delivering the message to the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this communication is strictly prohibited. If you have received this communication in error, please notify the sender immediately by telephone or return e-mail and delete the original message from your systems. Thank you.