Re: [RFC PATCH bpf-next] bpf, xdp: per-map bpf_redirect_map functions for XDP

On Fri, 29 Jan 2021 16:35:47 +0100
Björn Töpel [off-list ref] wrote:

On Fri, 29 Jan 2021 at 16:32, Björn Töpel [off-list ref] wrote:

quoted

From: Björn Töpel <redacted>
 

[...]

quoted

For AF_XDP rxdrop this yields +600Mpps. I'll do CPU/DEVMAP
measurements for the patch proper.
 

Kpps, not Mpps. :-P

+600Kpps from 24Mpps to 24.6Mpps I assume.  This corresponds to approx
1 ns ((1/24-1/24.6)*1000 = 1.01626 ns).

This also correlate with saving one function call, which is basically
what the patch does.

Fresh measurement "Intel(R) Xeon(R) CPU E5-1650 v4 @ 3.60GHz" with [1]:
 time_bench: Type:funcion_call_cost Per elem: 3 cycles(tsc) 1.053 ns
 time_bench: Type:func_ptr_call_cost Per elem: 4 cycles(tsc) 1.317 ns

[1] https://github.com/netoptimizer/prototype-kernel/blob/master/kernel/lib/time_bench_sample.c
-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer