RE: [lwip-users] Performance results

[Kieran Mansley]

On Tue, 3 Feb 2004, John Taylor wrote:

> >From: address@hidden
> [mailto:lwip->address@hidden On
> Behalf Of Kieran >Mansley
> >
> >... but it does demonstrate that the stack can run fast (if you
> >rip out all the slow bits!)
>
> What are the "slow bits?"

The mailboxes for passing messages, pbuf allocation, and data copies are
obvious candidates.  In particular if you're using the sockets API you'll
get the worst of all of these.  As I said before I'm using a fairly
radically altered lwIP, and this requires a rather different netif layer
(and some unusual hardware) to allow efficient sockets implementation.

> One last caveat.  My port lwIP is not finished.  I still have bug where
> the TCP "ack" number gets out-of-sync and effectively stalls some of the
> connections.  However, I can flood ping the board and not loose a single
> packet.

This sounds vaguely familiar.  Have you mentioned it before?  If so, that
might be why it sounds familiar.  If not, it might be worth searching the
mailing list to see if it's been brought up.

> Is a 1.3msec ping response time reasonable for lwIP or have I screwed
> something up in my port?

Not sure.  Doesn't sound too bad, but doesn't sound great either.  I don't
have much experience of the performance of "vanilla" lwIP I'm afraid.

Kieran