Re: Speed of memcpy, csum_partial and csum_partial_copy
Linus Torvalds (torvalds@cs.helsinki.fi)
Sun, 9 Jun 1996 09:11:29 +0300 (EET DST)
On Sat, 8 Jun 1996, Alan Cox wrote:
>
> > In short, the two copies that occur in TCP loopback (first from the
> > sender into the kernel, and then from the kernel into the receiver) alone
> > account for 60% of the TCP stack..
>
> There are three for TCP loopback unless someone has fiddled recently.
Umm.. Why do you think the TCP performance jumped from 3.4MB/s to 12MB/s
on my old P100?
Yes, I "fiddled" with it (the larger MTU did a lot, but if you think the
larger MTU in 2.0 is the only thing that's changed, you're wrong).
> We copy
>
> user->sk_buff [with csum]
> sk_buff to sk_buff
> sk_buff->user [skip csum]
Nope. The sk_buff->sk_buff is just a "skb_clone()", no copy.
Linus