Rik van Riel
0132c3e177
sched/numa: Examine a task move when examining a task swap ...
Running "perf bench numa mem -0 -m -P 1000 -p 8 -t 20" on a 4
node system results in 160 runnable threads on a system with 80
CPU threads.
Once a process has nearly converged, with 39 threads on one node
and 1 thread on another node, the remaining thread will be unable
to migrate to its preferred node through a task swap.
However, a simple task move would make the workload converge,
witout causing an imbalance.
Test for this unlikely occurrence, and attempt a task move to
the preferred nid when it happens.
# Running main, "perf bench numa mem -p 8 -t 20 -0 -m -P 1000"
###
# 160 tasks will execute (on 4 nodes, 80 CPUs):
# -1x 0MB global shared mem operations
# -1x 1000MB process shared mem operations
# -1x 0MB thread local mem operations
###
###
#
# 0.0% [0.2 mins] 0/0 1/1 36/2 0/0 [36/3 ] l: 0-0 ( 0) {0-2}
# 0.0% [0.3 mins] 43/3 37/2 39/2 41/3 [ 6/10] l: 0-1 ( 1) {1-2}
# 0.0% [0.4 mins] 42/3 38/2 40/2 40/2 [ 4/9 ] l: 1-2 ( 1) [50.0%] {1-2}
# 0.0% [0.6 mins] 41/3 39/2 40/2 40/2 [ 2/9 ] l: 2-4 ( 2) [50.0%] {1-2}
# 0.0% [0.7 mins] 40/2 40/2 40/2 40/2 [ 0/8 ] l: 3-5 ( 2) [40.0%] ( 41.8s converged)
Without this patch, this same perf bench numa mem run had to
rely on the scheduler load balancer to first balance out the
load (moving a random task), before a task swap could complete
the NUMA convergence.
The load balancer does not normally take action unless the load
difference exceeds 25%. Convergence times of over half an hour
have been observed without this patch.
With this patch, the NUMA balancing code will simply migrate the
task, if that does not cause an imbalance.
Also skip examining a CPU in detail if the improvement on that CPU
is no more than the best we already have.
Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: chegu_vinod@hp.com
Cc: mgorman@suse.de
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/n/tip-ggthh0rnh0yua6o5o3p6cr1o@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-07-05 11:17:38 +02:00
kernel/printk: use symbolic defines for console loglevels
2014-06-04 16:54:17 -07:00
Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
2014-06-12 19:18:49 -07:00
gcov: add support for GCC 4.9
2014-06-10 15:34:46 -07:00
genirq: Improve documentation to match current implementation
2014-05-27 10:16:44 +02:00
Merge branch 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
2014-06-12 18:48:15 -07:00
Merge branch 'pm-sleep'
2014-06-12 13:43:08 +02:00
kernel/printk: use symbolic defines for console loglevels
2014-06-04 16:54:17 -07:00
Merge branch 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip into next
2014-06-03 12:57:53 -07:00
sched/numa: Examine a task move when examining a task swap
2014-07-05 11:17:38 +02:00
nohz: Support nohz full remote kick
2014-06-16 16:26:54 +02:00
One bug fix that goes back to 3.10. Accessing a non existent buffer
2014-06-12 21:07:25 -07:00
ipc, kernel: clear whitespace
2014-06-06 16:08:14 -07:00
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
2014-06-12 14:27:40 -07:00
Merge git://git.infradead.org/users/eparis/audit
2014-04-12 12:38:53 -07:00
auditsc: audit_krule mask accesses need bounds checking
2014-06-10 08:44:40 -07:00
kernel/backtracetest.c: replace no level printk by pr_info()
2014-06-04 16:54:14 -07:00
fs,userns: Change inode_capable to capable_wrt_inode_uidgid
2014-06-10 13:57:22 -07:00
Merge branch 'for-3.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
2014-06-09 15:03:33 -07:00
cgroup: remove css_parent()
2014-05-16 13:22:48 -04:00
kernel/compat.c: use sizeof() instead of sizeof
2014-06-04 16:54:19 -07:00
asmlinkage: Add explicit __visible to drivers/*, lib/*, kernel/*
2014-05-05 16:07:46 -07:00
More ACPI and power management updates for 3.16-rc1
2014-06-12 13:14:19 -07:00
Merge branch 'for-3.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
2014-06-09 15:03:33 -07:00
kernel/exec_domain.c: code clean-up
2014-06-04 16:54:15 -07:00
signals: mv {dis,}allow_signal() from sched.h/exit.c to signal.[ch]
2014-06-06 16:08:11 -07:00
ptrace: fix fork event messages across pid namespaces
2014-06-06 16:08:11 -07:00
Merge branch 'next' (accumulated 3.16 merge window patches) into master
2014-06-08 11:31:16 -07:00
Merge branch 'perf/urgent' into perf/core, to resolve conflict and to prepare for new patches
2014-06-06 07:55:06 +02:00
kernel/hung_task.c: convert simple_strtoul to kstrtouint
2014-06-04 16:54:15 -07:00
irq_work: Remove BUG_ON in irq_work_run()
2014-07-05 11:17:26 +02:00
locking/rwlocks: Introduce 'qrwlocks' - fair, queued rwlocks
2014-06-06 07:58:28 +02:00
kernel/kexec.c: convert printk to pr_foo()
2014-06-06 16:08:12 -07:00
signals: change wait_for_helper() to use kernel_sigaction()
2014-06-06 16:08:12 -07:00
kprobes: Show blacklist entries via debugfs
2014-04-24 10:26:41 +02:00
kobject: Make support for uevent_helper optional.
2014-04-25 12:00:49 -07:00
kthread: fix return value of kthread_create() upon SIGKILL.
2014-06-04 16:53:51 -07:00
kernel/latencytop.c: convert seq_printf to seq_puts
2014-06-04 16:54:15 -07:00
Most of this is cleaning up various driver sysfs permissions so we can
2014-06-11 16:09:14 -07:00
kprobes, notifier: Use NOKPROBE_SYMBOL macro in notifier
2014-04-24 10:26:39 +02:00
kernel/panic.c: add "crash_kexec_post_notifiers" option for kdump after panic_notifers
2014-06-06 16:08:12 -07:00
param: hand arguments after -- straight to init
2014-04-28 11:48:34 +09:30
kernel/profile.c: use static const char instead of static char
2014-06-06 16:08:13 -07:00
kernel/reboot.c: convert simple_strtoul to kstrtoint
2014-06-04 16:54:15 -07:00
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
2014-04-12 14:49:50 -07:00
kernel/res_counter.c: replace simple_strtoull by kstrtoull
2014-06-04 16:54:15 -07:00
resources: Clarify sanity check message
2014-05-23 10:47:21 -06:00
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
2014-06-12 14:27:40 -07:00
signals: introduce kernel_sigaction()
2014-06-06 16:08:12 -07:00
irq_work: Implement remote queueing
2014-06-16 16:26:54 +02:00
Merge branch 'rcu/next' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu into core/rcu
2014-05-22 11:36:10 +02:00
kernel/stop_machine.c: kernel-doc warning fix
2014-06-04 16:54:15 -07:00
sched: Consolidate open coded implementations of nice level frobbing into nice_to_rlimit() and rlimit_to_nice()
2014-05-22 11:16:36 +02:00
sys_sgetmask/sys_ssetmask: add CONFIG_SGETMASK_SYSCALL
2014-06-04 16:54:14 -07:00
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
2014-06-12 14:27:40 -07:00
timer: Prevent overflow in apply_slack
2014-04-30 13:46:17 +02:00
torture: Remove __init from torture_init_begin/end
2014-05-14 09:46:30 -07:00
kernel/tracepoint.c: kernel-doc fixes
2014-06-04 16:54:15 -07:00
kernel/user.c: drop unused field 'files' from user_struct
2014-06-04 16:54:16 -07:00
kernel/user_namespace.c: kernel-doc/checkpatch fixes
2014-06-06 16:08:13 -07:00
sysctl: convert use of typedef ctl_table to struct ctl_table
2014-06-06 16:08:16 -07:00
kernel/watchdog.c:touch_softlockup_watchdog(): use raw_cpu_write()
2014-04-18 16:40:08 -07:00
Merge branch 'for-3.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
2014-06-09 14:56:49 -07:00
workqueue: rename manager_mutex to attach_mutex
2014-05-20 10:59:32 -04:00
0132c3e177