From: pjt@google.com
To: linux-kernel@vger.kernel.org
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>,
Ingo Molnar <mingo@elte.hu>,
Srivatsa Vaddagiri <vatsa@in.ibm.com>,
Chris Friesen <cfriesen@nortel.com>,
Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>,
Pierre Bourdon <pbourdon@excellency.fr>,
Paul Turner <pjt@google.com>,
Bharata B Rao <bharata@linux.vnet.ibm.com>
Subject: [RFC tg_shares_up improvements - v1 09/12] sched: demand based update_cfs_load()
Date: Fri, 15 Oct 2010 21:43:58 -0700 [thread overview]
Message-ID: <20101016045119.133213772@google.com> (raw)
In-Reply-To: 20101016044349.830426011@google.com
[-- Attachment #1: sched-tg-demand_based_update_cfs_load.patch --]
[-- Type: text/plain, Size: 2211 bytes --]
When the system is busy dilation of rq->next_balance makes lb->update_shares()
insufficiently frequent for threads which don't sleep (no dequeue/enqueue
updates). Adjust for this by making demand based updates based on the
accumulation of execution time sufficient to wrap our averaging window.
Signed-off-by: Paul Turner <pjt@google.com>
---
kernel/sched.c | 9 ++++++++-
kernel/sched_fair.c | 10 ++++++++++
2 files changed, 18 insertions(+), 1 deletion(-)
Index: kernel/sched.c
===================================================================
--- kernel/sched.c.orig
+++ kernel/sched.c
@@ -355,9 +355,16 @@ struct cfs_rq {
*/
unsigned long h_load;
+ /*
+ * Maintaining per-cpu shares distribution for group scheduling
+ *
+ * load_stamp is the last time we updated the load average
+ * load_last is the last time we updated the load average and saw load
+ * load_unacc_exec_time is currently unaccounted execution time
+ */
u64 load_avg;
u64 load_period;
- u64 load_stamp, load_last;
+ u64 load_stamp, load_last, load_unacc_exec_time;
unsigned long load_contribution;
#endif
Index: kernel/sched_fair.c
===================================================================
--- kernel/sched_fair.c.orig
+++ kernel/sched_fair.c
@@ -538,6 +538,9 @@ static u64 sched_vslice(struct cfs_rq *c
return calc_delta_fair(sched_slice(cfs_rq, se), se);
}
+static void update_cfs_load(struct cfs_rq *cfs_rq);
+static void update_cfs_shares(struct cfs_rq *cfs_rq, long weight_delta);
+
/*
* Update the current task's runtime statistics. Skip current tasks that
* are not in our scheduling class.
@@ -557,6 +560,12 @@ __update_curr(struct cfs_rq *cfs_rq, str
curr->vruntime += delta_exec_weighted;
update_min_vruntime(cfs_rq);
+
+ cfs_rq->load_unacc_exec_time += delta_exec;
+ if (cfs_rq->load_unacc_exec_time > sysctl_sched_shares_window) {
+ update_cfs_load(cfs_rq);
+ update_cfs_shares(cfs_rq, 0);
+ }
}
static void update_curr(struct cfs_rq *cfs_rq)
@@ -712,6 +721,7 @@ static void update_cfs_load(struct cfs_r
}
cfs_rq->load_stamp = now;
+ cfs_rq->load_unacc_exec_time = 0;
cfs_rq->load_period += delta;
if (load) {
cfs_rq->load_last = now;
--
next prev parent reply other threads:[~2010-10-16 4:54 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-10-16 4:43 [RFC tg_shares_up improvements - v1 00/12] [RFC tg_shares_up - v1 00/12] Reducing cost of tg->shares distribution pjt
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 01/12] sched: rewrite tg_shares_up pjt
2010-10-21 6:04 ` Bharata B Rao
2010-10-21 6:28 ` Paul Turner
2010-10-21 8:08 ` Bharata B Rao
2010-10-21 8:38 ` Paul Turner
2010-10-21 9:08 ` Peter Zijlstra
[not found] ` <AANLkTi=zYAfb_izD15ROxH=C6+zPzX+XEGw7r5UUoAar@mail.gmail.com>
2010-11-04 21:00 ` Paul Turner
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 02/12] sched: on-demand (active) cfs_rq list pjt
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 03/12] sched: make tg_shares_up() walk on-demand pjt
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 04/12] sched: fix load corruption from update_cfs_shares pjt
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 05/12] sched: fix update_cfs_load synchronization pjt
2010-10-21 9:52 ` Bharata B Rao
2010-10-21 18:25 ` Paul Turner
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 06/12] sched: hierarchal order on shares update list pjt
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 07/12] sched: add sysctl_sched_shares_window pjt
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 08/12] sched: update shares on idle_balance pjt
2010-10-16 4:43 ` pjt [this message]
2010-10-16 4:43 ` [RFC tg_shares_up improvements - v1 10/12] sched: allow update_cfs_load to update global load pjt
2010-10-16 4:44 ` [RFC tg_shares_up improvements - v1 11/12] sched: update tg->shares after cpu.shares write pjt
2010-10-16 4:44 ` [RFC tg_shares_up improvements - v1 12/12] debug: export effective shares for analysis versus specified pjt
2010-10-16 19:46 ` [RFC tg_shares_up improvements - v1 00/12] [RFC tg_shares_up - v1 00/12] Reducing cost of tg->shares distribution Peter Zijlstra
2010-10-21 6:36 ` Paul Turner
2010-10-22 0:14 ` Paul Turner
2010-10-17 5:24 ` Balbir Singh
2010-10-17 9:38 ` Peter Zijlstra
2010-10-17 12:09 ` Balbir Singh
2010-11-03 18:27 ` Karl Rister
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20101016045119.133213772@google.com \
--to=pjt@google.com \
--cc=a.p.zijlstra@chello.nl \
--cc=bharata@linux.vnet.ibm.com \
--cc=cfriesen@nortel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=pbourdon@excellency.fr \
--cc=svaidy@linux.vnet.ibm.com \
--cc=vatsa@in.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox