[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: ethernet.c changes from 1.33 (e100lx_mcm) -> 1.39.2.2 (devboard_82)



>What's meant by a "quiet" network? (one with little or no traffic?)

Yes

>and does this mean that UDP was not working under such a scenario in past 
>drivers? 

Yes

The problem was that UDP waited for the ethernet driver to report 
a packet to be sent before trying to send the next. We had a 
performance tweak that only checked for transmitted packets
when a packet was received. So if no packet were received, no
packets got sent...

/Mikael

-----Original Message-----
From: owner-dev-etrax@xxxxxxx.com]On">mailto:owner-dev-etrax@xxxxxxx.com]On
Behalf Of Dave Chevalier
Sent: Wednesday, July 09, 2003 5:28 PM
To: dev-etrax
Subject: Q: ethernet.c changes from 1.33 (e100lx_mcm) -> 1.39.2.2
(devboard_82)



  I was looking at the change log in the header of ethernet.c that came with 
the devboard_82 release (kernel 2.4.20), ethernet.c rev 1.39.2.2, and I was 
curious about the following entry:

>  * Revision 1.39.2.1  2003/03/31 14:12:46  starvik
>  * Transmit interrupt always enabled. This has two side effects:
>  *   1. UDP (and possibly other protocols) works on quiet networks
>  *   2. Slightly lower transmit performance

I was most interested in item #1 -
  What's meant by a "quiet" network? (one with little or no traffic?)
and does this mean that UDP was not working under such a scenario in past 
drivers? (like rev 1.33 which came with e100lx_mcm, kernel 2.4.19)

Can anyone at Axis elaborate?

Thanks!
  Dave