c9364636dc
htb_sched structures are big, and source of false sharing on SMP. Every time a packet is queued or dequeue, many cache lines must be touched because structures are not lay out properly. By carefully splitting htb_sched in two parts, and define sub structures to increase data locality, we can improve performance dramatically on SMP. New htb_prio structure can also be used in htb_class to increase data locality. I got 26 % performance increase on a 24 threads machine, with 200 concurrent netperf in TCP_RR mode, using a HTB hierarchy of 4 classes. Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Tom Herbert <therbert@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> |
||
|---|---|---|
| .. | ||
| Kconfig | ||
| Makefile | ||
| act_api.c | ||
| act_csum.c | ||
| act_gact.c | ||
| act_ipt.c | ||
| act_mirred.c | ||
| act_nat.c | ||
| act_pedit.c | ||
| act_police.c | ||
| act_simple.c | ||
| act_skbedit.c | ||
| cls_api.c | ||
| cls_basic.c | ||
| cls_cgroup.c | ||
| cls_flow.c | ||
| cls_fw.c | ||
| cls_route.c | ||
| cls_rsvp.c | ||
| cls_rsvp.h | ||
| cls_rsvp6.c | ||
| cls_tcindex.c | ||
| cls_u32.c | ||
| em_canid.c | ||
| em_cmp.c | ||
| em_ipset.c | ||
| em_meta.c | ||
| em_nbyte.c | ||
| em_text.c | ||
| em_u32.c | ||
| ematch.c | ||
| sch_api.c | ||
| sch_atm.c | ||
| sch_blackhole.c | ||
| sch_cbq.c | ||
| sch_choke.c | ||
| sch_codel.c | ||
| sch_drr.c | ||
| sch_dsmark.c | ||
| sch_fifo.c | ||
| sch_fq_codel.c | ||
| sch_generic.c | ||
| sch_gred.c | ||
| sch_hfsc.c | ||
| sch_htb.c | ||
| sch_ingress.c | ||
| sch_mq.c | ||
| sch_mqprio.c | ||
| sch_multiq.c | ||
| sch_netem.c | ||
| sch_plug.c | ||
| sch_prio.c | ||
| sch_qfq.c | ||
| sch_red.c | ||
| sch_sfb.c | ||
| sch_sfq.c | ||
| sch_tbf.c | ||
| sch_teql.c | ||