Messages in this thread | | | Date | Wed, 1 Dec 2021 20:42:43 +0000 | Subject | Re: [RFC 00/12] io_uring zerocopy send | From | Pavel Begunkov <> |
| |
On 12/1/21 17:57, David Ahern wrote: > On 12/1/21 8:32 AM, Pavel Begunkov wrote: [...] >>> mileage varies quite a bit. >> >> Interesting, any brief notes on the setup and the results? Dummy > > VM on Chromebook. I just cloned your repos, built, install and test. As > mentioned above, the skb_orphan_frags_rx change is missing from your > repo and that is the key to your reported performance gains.
Just to clear misunderstandings if any, all the numbers in the cover-letter were measured on the same kernel and during the same boot, and it doesn't include the skb_orphan_frags_rx() change. All double checked by looking at the traces.
When it's routed through the loopback paths (e.g. -D 127.0.0.1), it's indeed slow for both msg_zerocopy and send-zc, and both "benefit" in raw throughput from the hack.
-- Pavel Begunkov
| |