On 28.03.23 04:06, Jakub Kicinski wrote:

On Sat, 25 Mar 2023 06:42:43 +0100 Felix Fietkau wrote:

quoted

quoted

quoted

In my tests it brings down latency (both avg and p99) considerably in
some cases. I posted some numbers here:
https://lore.kernel.org/netdev/e317d5bc-cc26-8b1b-ca4b-66b5328683c4@nbd.name/ (local)  

Could you provide the full configuration for this test?
In non-threaded mode the RPS is enabled to spread over remaining
3 cores? 

In this test I'm using threaded NAPI and backlog_threaded without any 
fixed core assignment.

I was asking about the rps_threaded=0 side of the comparison.
So you're saying on that side you were using threaded NAPI with
no pinning and RPS across all cores?

Yes.

- Felix