From: Peter Zijlstra <a.p.zijlstra@chello.nl>
To: linux-kernel@vger.kernel.org
Cc: Ingo Molnar <mingo@elte.hu>,
Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>,
Mike Galbraith <efault@gmx.de>,
Peter Zijlstra <a.p.zijlstra@chello.nl>
Subject: [PATCH 03/30] sched: fix calc_delta_asym()
Date: Fri, 27 Jun 2008 13:41:12 +0200 [thread overview]
Message-ID: <20080627115211.107047525@chello.nl> (raw)
In-Reply-To: 20080627114109.724249622@chello.nl
[-- Attachment #1: sched-asyn_gran.patch --]
[-- Type: text/plain, Size: 2358 bytes --]
calc_delta_asym() is supposed to do the same as calc_delta_fair() except
linearly shrink the result for negative nice processes - this causes them
to have a smaller preemption threshold so that they are more easily preempted.
The problem is that for task groups se->load.weight is the per cpu share of
the actual task group weight; take that into account.
Also provide a debug switch to disable the asymmetry (which I still don't
like - but it does greatly benefit some workloads)
This would explain the interactivity issues reported against group scheduling.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
kernel/sched_fair.c | 28 +++++++++++++++++++++++++++-
kernel/sched_features.h | 1 +
2 files changed, 28 insertions(+), 1 deletion(-)
Index: linux-2.6/kernel/sched_fair.c
===================================================================
--- linux-2.6.orig/kernel/sched_fair.c
+++ linux-2.6/kernel/sched_fair.c
@@ -430,6 +430,29 @@ calc_delta_asym(unsigned long delta, str
for_each_sched_entity(se) {
struct load_weight *se_lw = &se->load;
+#ifdef CONFIG_FAIR_SCHED_GROUP
+ struct cfs_rq *cfs_rq = se->my_q;
+ struct task_group *tg = NULL
+
+ if (cfs_rq)
+ tg = cfs_rq->tg;
+
+ if (tg && tg->shares < NICE_0_LOAD) {
+ /*
+ * scale shares to what it would have been had
+ * tg->weight been NICE_0_LOAD:
+ *
+ * weight = 1024 * shares / tg->weight
+ */
+ lw.weight *= se->load.weight;
+ lw.weight /= tg->shares;
+
+ lw.inv_weight = 0;
+
+ se_lw = &lw;
+ } else
+#endif
+
if (se->load.weight < NICE_0_LOAD)
se_lw = &lw;
@@ -1154,7 +1177,10 @@ static unsigned long wakeup_gran(struct
* More easily preempt - nice tasks, while not making it harder for
* + nice tasks.
*/
- gran = calc_delta_asym(sysctl_sched_wakeup_granularity, se);
+ if (sched_feat(ASYM_GRAN))
+ gran = calc_delta_asym(sysctl_sched_wakeup_granularity, se);
+ else
+ gran = calc_delta_fair(sysctl_sched_wakeup_granularity, se);
return gran;
}
Index: linux-2.6/kernel/sched_features.h
===================================================================
--- linux-2.6.orig/kernel/sched_features.h
+++ linux-2.6/kernel/sched_features.h
@@ -6,3 +6,4 @@ SCHED_FEAT(CACHE_HOT_BUDDY, 1)
SCHED_FEAT(SYNC_WAKEUPS, 1)
SCHED_FEAT(HRTICK, 1)
SCHED_FEAT(DOUBLE_TICK, 0)
+SCHED_FEAT(ASYM_GRAN, 1)
--
next prev parent reply other threads:[~2008-06-27 11:59 UTC|newest]
Thread overview: 39+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-06-27 11:41 [PATCH 00/30] SMP-group balancer - take 3 Peter Zijlstra
2008-06-27 11:41 ` [PATCH 01/30] sched: clean up some unused variables Peter Zijlstra
2008-06-27 11:41 ` [PATCH 02/30] sched: revert the revert of: weight calculations Peter Zijlstra
2008-06-30 18:07 ` Balbir Singh
2008-07-15 20:16 ` Peter Zijlstra
2008-06-27 11:41 ` Peter Zijlstra [this message]
2008-06-27 11:41 ` [PATCH 04/30] sched: fix calc_delta_asym Peter Zijlstra
2008-06-27 11:41 ` [PATCH 05/30] sched: revert revert of: fair-group: SMP-nice for group scheduling Peter Zijlstra
2008-06-27 11:41 ` [PATCH 06/30] sched: sched_clock_cpu() based cpu_clock() Peter Zijlstra
2008-06-27 11:41 ` [PATCH 07/30] sched: fix wakeup granularity and buddy granularity Peter Zijlstra
2008-06-27 11:41 ` [PATCH 08/30] sched: add full schedstats to /proc/sched_debug Peter Zijlstra
2008-06-27 11:41 ` [PATCH 09/30] sched: fix sched_domain aggregation Peter Zijlstra
2008-06-27 11:41 ` [PATCH 10/30] sched: update aggregate when holding the RQs Peter Zijlstra
2008-06-27 11:41 ` [PATCH 11/30] sched: kill task_group balancing Peter Zijlstra
2008-06-27 11:41 ` [PATCH 12/30] sched: dont micro manage share losses Peter Zijlstra
2008-06-27 11:41 ` [PATCH 13/30] sched: no need to aggregate task_weight Peter Zijlstra
2008-06-27 11:41 ` [PATCH 14/30] sched: simplify the group load balancer Peter Zijlstra
2008-06-27 11:41 ` [PATCH 15/30] sched: fix newidle smp group balancing Peter Zijlstra
2008-06-27 11:41 ` [PATCH 16/30] sched: fix sched_balance_self() " Peter Zijlstra
2008-06-27 11:41 ` [PATCH 17/30] sched: persistent average load per task Peter Zijlstra
2008-06-27 11:41 ` [PATCH 18/30] sched: hierarchical load vs affine wakeups Peter Zijlstra
2008-06-27 11:41 ` [PATCH 19/30] sched: hierarchical load vs find_busiest_group Peter Zijlstra
2008-06-27 11:41 ` [PATCH 20/30] sched: fix load scaling in group balancing Peter Zijlstra
2008-06-27 11:41 ` [PATCH 21/30] sched: fix task_h_load() Peter Zijlstra
2008-06-27 11:41 ` [PATCH 22/30] sched: remove prio preference from balance decisions Peter Zijlstra
2008-06-27 11:41 ` [PATCH 23/30] sched: optimize effective_load() Peter Zijlstra
2008-06-27 11:41 ` [PATCH 24/30] sched: disable source/target_load bias Peter Zijlstra
2008-06-27 11:41 ` [PATCH 25/30] sched: fix shares boost logic Peter Zijlstra
2008-06-27 11:41 ` [PATCH 26/30] sched: update shares on wakeup Peter Zijlstra
2008-06-27 11:41 ` [PATCH 27/30] sched: fix mult overflow Peter Zijlstra
2008-06-27 11:41 ` [PATCH 28/30] sched: correct wakeup weight calculations Peter Zijlstra
2008-06-27 11:41 ` [PATCH 29/30] sched: incremental effective_load() Peter Zijlstra
2008-06-27 11:41 ` [PATCH 30/30] sched: bias effective_load() error towards failing wake_affine() Peter Zijlstra
2008-06-27 12:46 ` [PATCH 00/30] SMP-group balancer - take 3 Ingo Molnar
2008-06-27 17:33 ` Dhaval Giani
2008-06-28 17:08 ` Dhaval Giani
2008-06-30 12:59 ` Ingo Molnar
2008-06-30 14:53 ` Dhaval Giani
2008-07-01 10:57 ` Dhaval Giani
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=20080627115211.107047525@chello.nl \
--to=a.p.zijlstra@chello.nl \
--cc=efault@gmx.de \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=vatsa@linux.vnet.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