More IOC3 performance, was: Re: [Fwd: TTCP data test data] more on ftp c

To:	Ken Pearlstein <pearlste@xxxxxxxxxxxx>
Subject:	More IOC3 performance, was: Re: [Fwd: TTCP data test data] more on ftp curiosity
From:	Ralf Baechle <ralf@xxxxxxxxxxx>
Date:	Fri, 23 Feb 2001 18:00:01 +0100
Cc:	server1-plat@xxxxxxxxxxxx, linux-origin@xxxxxxxxxxx
In-reply-to:	<3A8B249F.F0D3A053@xxxxxxxxxxxx>; from pearlste@xxxxxxxxxxxx on Wed, Feb 14, 2001 at 07:36:47PM -0500
References:	<3A8B249F.F0D3A053@xxxxxxxxxxxx>
Sender:	owner-linux-origin@xxxxxxxxxxx
User-agent:	Mutt/1.2.5i

More on IOC3 performance - I fixed a bug which did result in outgoing
packets that are larger than 104 bytes and cross a 16kb boundary getting
sent corrupted.  This didn't show up as data corruption as it was caught
by the IP checksum mechanism; it just ruined performance.  So for my
test platform (for convenience I use an O200 running Linux to develop the
driver) and a second O200 running IRIX at the remote end I now clearly get
into the leage of sane benchmarks, > 90mbit/s including TCP/IP/802.3 header
overhead, for UPD even close to 99mbit/s.  On TX only, still and still
at the price of burning much more CPU than necessary.

  Ralf

<Prev in Thread]	Current Thread	[Next in Thread>
More IOC3 performance, was: Re: [Fwd: TTCP data test data] more on ftp curiosity, Ralf Baechle <=

Previous by Date:	BUG() oops from skb_put() of len=-12, from tcp_sendmsg(), John Hawkes
Previous by Thread:	BUG() oops from skb_put() of len=-12, from tcp_sendmsg(), John Hawkes
Indexes:	[Date] [Thread] [Top] [All Lists]