netdev
[Top] [All Lists]

Re: More measurements

To: Andrew Morton <andrewm@xxxxxxxxxx>, Andi Kleen <ak@xxxxxx>
Subject: Re: More measurements
From: Andrey Savochkin <saw@xxxxxxxxxxxxx>
Date: Wed, 31 Jan 2001 11:23:17 +0800
Cc: netdev@xxxxxxxxxxx
In-reply-to: <3A76C5DE.993BB140@xxxxxxxxxx>; from "Andrew Morton" on Wed, Jan 31, 2001 at 12:47:10AM
References: <3A75785A.42B9E7CE@xxxxxxxxxx>, <3A75785A.42B9E7CE@xxxxxxxxxx>; <20010130124839.24151@xxxxxxxxxxxx> <3A76C5DE.993BB140@xxxxxxxxxx>
Sender: owner-netdev@xxxxxxxxxxx
Hi,

On Wed, Jan 31, 2001 at 12:47:10AM +1100, Andrew Morton wrote:
> Andi Kleen wrote:
> > 
> > On Tue, Jan 30, 2001 at 10:08:02AM +0100, Andrew Morton wrote:
> > > Lots of interesting things here.
> > >
> > > - eepro100 generates more interrupts doing TCP Tx, but not
> > >   TCP Rx.  I assume it doesn't do Tx mitigation?
> > 
> > The Intel driver (e100.c) uploads special firmware and does it for RX and 
> > TX.
> > eepro100 doesn't. Perhaps you could measure that driver too?

That's about RX mitigation, where a special hardware assistance is required.
For TX interrupts, the current driver arbitrarily asks for a TX interrupt
every forth packet (to do garbage collection).

> Sure.  Anyone have a URL for intel's driver?

ftp://download.intel.com/support/network/adapter/pro100/ e100-*.tgz

> 
> > >
> > > - Changing eepro100 to use IO operations instead of MMIO slows
> > >   down this dual 500MHz machine by less than one percent at
> > >   100 mbps.  At 12,000 interrupts per second. Why all the fuss
> > >   about MMIO?
> > 
> > iirc Ingo at some point found at some monster machine that the IO operations
> > in the eepro100 interrupt handler dominated some Tux profile.

The same was reported by kumon@xxxxxxxxxxxxxxxxxx
I think that serialization effects of IO instructions lead to them staying at
the top of profile.

Best regards
                Andrey

<Prev in Thread] Current Thread [Next in Thread>