Mitigation and autonegotiation

[Bogdan Costescu]

On Fri, 21 Apr 2000, Andrew Morton wrote:

> What advantage is there to generating extra interrupts?

No extra interrupts, but one interrupt per Tx-ed packet.

> The only advantage I can see is that the tx ring will always be pretty
> much full, rather than filling right up in a big burst then emptying
> down to a single packet.  If your system is experiencing large interrupt
> blockages then you will see better tx performance because the NIC
> hardware has more to be going on with.
> 
> Linux occasionally blocks interrupts for 3 milliseconds in the console
> code, and frequently blocks them for ~500 microseconds in the IDE code
> (unless you use 'hdparm -u'), so maybe this is the difference.

There is a small discussion in the docs about trading CPU usage for
network performance, but I forgot exactly where... As I see now, we are
somehow on opposite positions from this POV: you want to have the network
driver loading as little as possible the host, while I want the max speed
(meaning smallest latency, as the bandwitdh cannot be changed)... Also,
usually, the computers doing parallel computations are not using much the
console (while IDE can be used more often), so these blocks do not occur.
In my case, there is only CPU load produced by floating point calculations
and NIC interrupts (and timer interrupts that occur anyway) and usually
the program is happy to be interrupted by NIC interrupts because it
scatter/gather data for next calculations.

I've finished yesterday reading the 212 doc pages and I have found several
things that (theoretically, at least) can be done to improve the speed.
Some of them might be 905C specific (I haven't read the docs for 905B),
anyway I will come back with these after some thinking.

> Still two problems remain with this driver:
> 
> 1: Why the 3c575 (cardbus) cards usually come up receiving but not
> transmitting (another person is experiencing this exact thing with the
> 574_cs and 589_cs drivers also).
> 
> 2: The "sleepy NIC" problem.
> 
> You don't have a laptop, do you Bogdan? :-)

No and it's very unlikely to have one soon. 

Sincerely,

Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De

-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-vortex-bug-request@beowulf.org