Skip to content

reduce overhead for tg->shares distribution

Paul Turner lkml 4 versions – Applied

Status from this message: Thanks, applied!

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v4 Latest11 patches 12 repliesApplied
  2. v311 patches 6 repliesSuperseded
  3. v211 patches 4 repliesSuperseded
Earlier versions · 1
VersionPostedPatchesRepliesStatus
v1 RFC12 patches15 repliesSuperseded

Changes from the cover letter

v4
all patches
- fix "-p0" stripping

Patches in v4 Applied

#Subject and filesSizeRepliesApplied
1/11sched: rewrite tg_shares_upinclude/linux/sched.h kernel/sched.c kernel/sched_debug.c +3 more+162 -2131 reply
2/11sched: on-demand (active) cfs_rq listkernel/sched.c kernel/sched_fair.c kernel/sched_rt.c+92 -831 reply
3/11sched: make tg_shares_up() walk on-demandkernel/sched.c kernel/sched_fair.c+58 -671 reply
4/11sched: fix load corruption from update_cfs_shareskernel/sched_fair.c+8 -81 reply
5/11sched: fix update_cfs_load synchronizationkernel/sched.c kernel/sched_fair.c+22 -131 reply
6/11sched: hierarchal order on shares update listkernel/sched_fair.c+16 -101 reply
7/11sched: add sysctl_sched_shares_windowinclude/linux/sched.h kernel/sched_fair.c kernel/sysctl.c+16 -11 reply
8/11sched: update shares on idle_balancekernel/sched_fair.c+1 -01 reply
9/11sched: demand based update_cfs_load()kernel/sched.c kernel/sched_fair.c+20 -11 reply
10/11sched: allow update_cfs_load to update global loadkernel/sched_fair.c+29 -151 reply
11/11sched: update tg->shares after cpu.shares writekernel/sched.c+11 -311 reply

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v4 yet.

Read the v4 thread

Cover letter [tg_shares_up rewrite v4 00/11] reduce overhead for tg->shares distribution

This is v4 of the tg_shares_up re-write previously posted at:
http://lkml.org/lkml/2010/10/16/17

Corrected versus the previous version is the quilt patch-strip level.

The key idea behind this patchset is an idea of Peter Zijlstra'a to compute global loads as a function of local averages, avoiding the cpu-cpu interactions which previously plagued tg_shares_up.

Beyond bug fixes not much material has changed since v1 as results seem favourable/stable.

Changes:
v4
all patches
- fix "-p0" stripping

v3 (http://lkml.org/lkml/2010/11/11/526)
sched-rework-tg_shares_up.patch
- remove lint: {max,min}_sched_shares_ratelimit from sysctl.c

sched-tg-ondemand.patch
- (bug fix) fix synchronization on leaf_cfs_rq_list removal during task_group reclaim.
- fix !CONFIG_FAIR_GROUP_SCHED build
- fix comment on what task_group_lock protects

sched-tg-global_update_load.patch
- add missing arguments to update_cfs_load() in !CONFIG_FAIR_GROUP_SCHED case

v2 (http://lkml.org/lkml/2010/11/10/618)
sched-rework-tg_shares_up.patch
- (bug fix) use owning tg->shares value in effective_load
- remove redundant cfs_rq_of() in {enqueue,dequeue}_entity

sched-tg-add-shares_window_sysctl.patch
- update comment on sysctl interpretation

sched-tg-fix-update_cfs_load.patch
- update comment to reflect periods instead of foldings (more intuitive)

Some motivation again:
24 thread Intel box, 150 active cgroups, multiple threads/group, system load at ~90% (sample size=10s):
tip:

    2.64%  [k] tg_shares_up <!>
    0.15%  [k] __set_se_shares

patched:

    0.02%  [k] update_cfs_load
    0.01%  [k] update_cpu_load
    0.00%  [k] update_cfs_shares

Fairness coverage for the above at: http://rs5.risingnet.net/~pjt/patches/shares_data_v1.txt

Thanks,

- Paul