netstacklat: Experiment with 4 nth packet sampling

name netstacklat_tcp_recv_timestamp
 - run_time_ns 17510185954 run_cnt 101083454 = 173.23 ns

Sample every 4-nth packet (25%)
 - Overhead: 173.23 ns
 - Compared to nth-2 (187.44 ns) saved 14.21 ns (187.44-173.23)
 - Compared to nth-32 (157.67 ns) cost 15.56 ns more (173.23-157.67)

Signed-off-by: Jesper Dangaard Brouer <hawk@kernel.org>

Author: netoptimizer

examples/netstacklat.bpf.c

@@ -29,7 +29,7 @@ const __s64 TAI_OFFSET = (37LL * NS_PER_S);
 const struct netstacklat_bpf_config user_config = {
 	.network_ns = 0,
 	.filter_min_queue_len = 0, /* zero means filter is inactive */
-	.filter_nth_packet = 2, /* reduce recorded event to every nth packet, use power-of-2 */
+	.filter_nth_packet = 4, /* reduce recorded event to every nth packet, use power-of-2 */
 	.filter_pid = false,
 	.filter_ifindex = true,
 	.filter_cgroup = true,