linux/kernel
Giovanni Gherdovich 6075620b05 sched/cputime: Mitigate performance regression in times()/clock_gettime()
Commit:

  6e998916df ("sched/cputime: Fix clock_nanosleep()/clock_gettime() inconsistency")

fixed a problem whereby clock_nanosleep() followed by clock_gettime() could
allow a task to wake early. It addressed the problem by calling the scheduling
classes update_curr() when the cputimer starts.

Said change induced a considerable performance regression on the syscalls
times() and clock_gettimes(CLOCK_PROCESS_CPUTIME_ID). There are some
debuggers and applications that monitor their own performance that
accidentally depend on the performance of these specific calls.

This patch mitigates the performace loss by prefetching data in the CPU
cache, as stalls due to cache misses appear to be where most time is spent
in our benchmarks.

Here are the performance gain of this patch over v4.7-rc7 on a Sandy Bridge
box with 32 logical cores and 2 NUMA nodes. The test is repeated with a
variable number of threads, from 2 to 4*num_cpus; the results are in
seconds and correspond to the average of 10 runs; the percentage gain is
computed with (before-after)/before so a positive value is an improvement
(it's faster). The improvement varies between a few percents for 5-20
threads and more than 10% for 2 or >20 threads.

pound_clock_gettime:

    threads       4.7-rc7     patched 4.7-rc7
    [num]         [secs]      [secs (percent)]
      2           3.48        3.06 ( 11.83%)
      5           3.33        3.25 (  2.40%)
      8           3.37        3.26 (  3.30%)
     12           3.32        3.37 ( -1.60%)
     21           4.01        3.90 (  2.74%)
     30           3.63        3.36 (  7.41%)
     48           3.71        3.11 ( 16.27%)
     79           3.75        3.16 ( 15.74%)
    110           3.81        3.25 ( 14.80%)
    128           3.88        3.31 ( 14.76%)

pound_times:

    threads       4.7-rc7     patched 4.7-rc7
    [num]         [secs]      [secs (percent)]
      2           3.65        3.25 ( 11.03%)
      5           3.45        3.17 (  7.92%)
      8           3.52        3.22 (  8.69%)
     12           3.29        3.36 ( -2.04%)
     21           4.07        3.92 (  3.78%)
     30           3.87        3.40 ( 12.17%)
     48           3.79        3.16 ( 16.61%)
     79           3.88        3.28 ( 15.42%)
    110           3.90        3.38 ( 13.35%)
    128           4.00        3.38 ( 15.45%)

pound_clock_gettime and pound_clock_gettime are two benchmarks included in
the MMTests framework. They launch a given number of threads which
repeatedly call times() or clock_gettimes(). The results above can be
reproduced with cloning MMTests from github.com and running the "poundtime"
workload:

  $ git clone https://github.com/gormanm/mmtests.git
  $ cd mmtests
  $ cp configs/config-global-dhp__workload_poundtime config
  $ ./run-mmtests.sh --run-monitor $(uname -r)

The above will run "poundtime" measuring the kernel currently running on
the machine; Once a new kernel is installed and the machine rebooted,
running again

  $ cd mmtests
  $ ./run-mmtests.sh --run-monitor $(uname -r)

will produce results to compare with. A comparison table will be output
with:

  $ cd mmtests/work/log
  $ ../../compare-kernels.sh

the table will contain a lot of entries; grepping for "Amean" (as in
"arithmetic mean") will give the tables presented above. The source code
for the two benchmarks is reported at the end of this changelog for
clairity.

The cache misses addressed by this patch were found using a combination of
`perf top`, `perf record` and `perf annotate`. The incriminated lines were
found to be

    struct sched_entity *curr = cfs_rq->curr;

and

    delta_exec = now - curr->exec_start;

in the function update_curr() from kernel/sched/fair.c. This patch
prefetches the data from memory just before update_curr is called in the
interested execution path.

A comparison of the total number of cycles before and after the patch
follows; the data is obtained using `perf stat -r 10 -ddd <program>`
running over the same sequence of number of threads used above (a positive
gain is an improvement):

  threads   cycles before                 cycles after                gain

    2      19,699,563,964  +-1.19%      17,358,917,517  +-1.85%      11.88%
    5      47,401,089,566  +-2.96%      45,103,730,829  +-0.97%       4.85%
    8      80,923,501,004  +-3.01%      71,419,385,977  +-0.77%      11.74%
   12     112,326,485,473  +-0.47%     110,371,524,403  +-0.47%       1.74%
   21     193,455,574,299  +-0.72%     180,120,667,904  +-0.36%       6.89%
   30     315,073,519,013  +-1.64%     271,222,225,950  +-1.29%      13.92%
   48     321,969,515,332  +-1.48%     273,353,977,321  +-1.16%      15.10%
   79     337,866,003,422  +-0.97%     289,462,481,538  +-1.05%      14.33%
  110     338,712,691,920  +-0.78%     290,574,233,170  +-0.77%      14.21%
  128     348,384,794,006  +-0.50%     292,691,648,206  +-0.66%      15.99%

A comparison of cache miss vs total cache loads ratios, before and after
the patch (again from the `perf stat -r 10 -ddd <program>` tables):

  threads   L1 misses/total*100     L1 misses/total*100            gain
		         before                   after
      2           7.43  +-4.90%           7.36  +-4.70%           0.94%
      5          13.09  +-4.74%          13.52  +-3.73%          -3.28%
      8          13.79  +-5.61%          12.90  +-3.27%           6.45%
     12          11.57  +-2.44%           8.71  +-1.40%          24.72%
     21          12.39  +-3.92%           9.97  +-1.84%          19.53%
     30          13.91  +-2.53%          11.73  +-2.28%          15.67%
     48          13.71  +-1.59%          12.32  +-1.97%          10.14%
     79          14.44  +-0.66%          13.40  +-1.06%           7.20%
    110          15.86  +-0.50%          14.46  +-0.59%           8.83%
    128          16.51  +-0.32%          15.06  +-0.78%           8.78%

As a final note, the following shows the evolution of performance figures
in the "poundtime" benchmark and pinpoints commit 6e998916df
("sched/cputime: Fix clock_nanosleep()/clock_gettime() inconsistency") as a
major source of degradation, mostly unaddressed to this day (figures
expressed in seconds).

pound_clock_gettime:

  threads   parent of         6e998916df        4.7-rc7
	    6e998916df            itself
    2        2.23          3.68 ( -64.56%)        3.48 (-55.48%)
    5        2.83          3.78 ( -33.42%)        3.33 (-17.43%)
    8        2.84          4.31 ( -52.12%)        3.37 (-18.76%)
    12       3.09          3.61 ( -16.74%)        3.32 ( -7.17%)
    21       3.14          4.63 ( -47.36%)        4.01 (-27.71%)
    30       3.28          5.75 ( -75.37%)        3.63 (-10.80%)
    48       3.02          6.05 (-100.56%)        3.71 (-22.99%)
    79       2.88          6.30 (-118.90%)        3.75 (-30.26%)
    110      2.95          6.46 (-119.00%)        3.81 (-29.24%)
    128      3.05          6.42 (-110.08%)        3.88 (-27.04%)

pound_times:

  threads   parent of         6e998916df        4.7-rc7
	    6e998916df            itself
    2        2.27          3.73 ( -64.71%)        3.65 (-61.14%)
    5        2.78          3.77 ( -35.56%)        3.45 (-23.98%)
    8        2.79          4.41 ( -57.71%)        3.52 (-26.05%)
    12       3.02          3.56 ( -17.94%)        3.29 ( -9.08%)
    21       3.10          4.61 ( -48.74%)        4.07 (-31.34%)
    30       3.33          5.75 ( -72.53%)        3.87 (-16.01%)
    48       2.96          6.06 (-105.04%)        3.79 (-28.10%)
    79       2.88          6.24 (-116.83%)        3.88 (-34.81%)
    110      2.98          6.37 (-114.08%)        3.90 (-31.12%)
    128      3.10          6.35 (-104.61%)        4.00 (-28.87%)

The source code of the two benchmarks follows. To compile the two:

  NR_THREADS=42
  for FILE in pound_times pound_clock_gettime; do
      gcc -lrt -O2 -lpthread -DNUM_THREADS=$NR_THREADS $FILE.c -o $FILE
  done

==== BEGIN pound_times.c ====

struct tms start;

void *pound (void *threadid)
{
  struct tms end;
  int oldutime = 0;
  int utime;
  int i;
  for (i = 0; i < 5000000 / NUM_THREADS; i++) {
          times(&end);
          utime = ((int)end.tms_utime - (int)start.tms_utime);
          if (oldutime > utime) {
            printf("utime decreased, was %d, now %d!\n", oldutime, utime);
          }
          oldutime = utime;
  }
  pthread_exit(NULL);
}

int main()
{
  pthread_t th[NUM_THREADS];
  long i;
  times(&start);
  for (i = 0; i < NUM_THREADS; i++) {
    pthread_create (&th[i], NULL, pound, (void *)i);
  }
  pthread_exit(NULL);
  return 0;
}
==== END pound_times.c ====

==== BEGIN pound_clock_gettime.c ====

void *pound (void *threadid)
{
	struct timespec ts;
	int rc, i;
	unsigned long prev = 0, this = 0;

	for (i = 0; i < 5000000 / NUM_THREADS; i++) {
		rc = clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &ts);
		if (rc < 0)
			perror("clock_gettime");
		this = (ts.tv_sec * 1000000000) + ts.tv_nsec;
		if (0 && this < prev)
			printf("%lu ns timewarp at iteration %d\n", prev - this, i);
		prev = this;
	}
	pthread_exit(NULL);
}

int main()
{
	pthread_t th[NUM_THREADS];
	long rc, i;
	pid_t pgid;

	for (i = 0; i < NUM_THREADS; i++) {
		rc = pthread_create(&th[i], NULL, pound, (void *)i);
		if (rc < 0)
			perror("pthread_create");
	}

	pthread_exit(NULL);
	return 0;
}
==== END pound_clock_gettime.c ====

Suggested-by: Mike Galbraith <mgalbraith@suse.de>
Signed-off-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1470385316-15027-2-git-send-email-ggherdovich@suse.cz
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-08-10 13:32:56 +02:00
..
bpf Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next 2016-07-27 12:03:20 -07:00
configs config: add android config fragments 2016-08-02 19:35:42 -04:00
debug mm/init: Add 'rodata=off' boot cmdline parameter to disable read-only kernel mappings 2016-02-22 08:51:37 +01:00
events perf/core: Change log level for duration warning to KERN_INFO 2016-08-02 10:23:57 +02:00
gcov gcov: add support for gcc version >= 6 2016-07-15 14:54:27 +09:00
irq genirq: Fix missing irq allocation affinity hint 2016-07-19 10:49:47 +02:00
livepatch modules: add ro_after_init support 2016-08-04 10:16:55 +09:30
locking Merge branch 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-07-25 12:41:29 -07:00
power mm, vmscan: move LRU lists to node 2016-07-28 16:07:41 -07:00
printk Revert "printk: create pr_<level> functions" 2016-08-09 10:48:18 -07:00
rcu Merge branch 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-07-29 13:55:30 -07:00
sched sched/cputime: Mitigate performance regression in times()/clock_gettime() 2016-08-10 13:32:56 +02:00
time Merge branch 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-07-29 13:55:30 -07:00
trace block: rename bio bi_rw to bi_opf 2016-08-07 14:41:02 -06:00
.gitignore
acct.c
async.c
audit_fsnotify.c wrappers for ->i_mutex access 2016-01-22 18:04:28 -05:00
audit_tree.c audit: cleanup prune_tree_thread 2016-04-04 09:46:47 -04:00
audit_watch.c don't bother with ->d_inode->i_sb - it's always equal to ->d_sb 2016-04-10 17:11:51 -04:00
audit.c Merge branch 'stable-4.8' of git://git.infradead.org/users/pcmoore/audit 2016-07-29 17:54:17 -07:00
audit.h Merge branch 'stable-4.8' of git://git.infradead.org/users/pcmoore/audit 2016-07-29 17:54:17 -07:00
auditfilter.c audit: add fields to exclude filter by reusing user filter 2016-06-27 11:01:00 -04:00
auditsc.c Merge branch 'stable-4.8' of git://git.infradead.org/users/pcmoore/audit 2016-07-29 17:54:17 -07:00
backtracetest.c
bounds.c
capability.c kernel: Add noaudit variant of ns_capable() 2016-06-06 20:16:18 +10:00
cgroup_freezer.c
cgroup_pids.c cgroup: Use lld instead of ld when printing pids controller events_limit 2016-06-21 15:03:36 -04:00
cgroup.c Merge branch 'for-4.7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup 2016-07-29 14:29:04 -07:00
compat.c
configs.c
context_tracking.c
cpu_pm.c
cpu.c timers/core: Correct callback order during CPU hot plug 2016-07-28 18:56:22 +02:00
cpuset.c cpuset, mm: fix TIF_MEMDIE check in cpuset_change_task_nodemask 2016-07-28 16:07:41 -07:00
crash_dump.c
cred.c cred: Reject inodes with invalid ids in set_create_file_as() 2016-06-30 18:05:09 -05:00
delayacct.c kmemcg: account certain kmem allocations to memcg 2016-01-14 16:00:49 -08:00
dma.c
elfcore.c
exec_domain.c
exit.c kernel/exit.c: quieten greatest stack depth printk 2016-08-02 19:35:23 -04:00
extable.c
fork.c mm: fix memcg stack accounting for sub-page stacks 2016-07-28 16:07:41 -07:00
freezer.c freezer, oom: check TIF_MEMDIE on the correct task 2016-07-28 16:07:41 -07:00
futex_compat.c ptrace: use fsuid, fsgid, effective creds for fs access checks 2016-01-20 17:09:18 -08:00
futex.c futex: Calculate the futex key based on a tail page for file-based futexes 2016-06-08 19:23:54 +02:00
groups.c
hung_task.c kernel/hung_task.c: use timeout diff when timeout is updated 2016-03-22 15:36:02 -07:00
irq_work.c
jump_label.c powerpc updates for 4.8 #2 2016-08-05 09:00:54 -04:00
kallsyms.c kallsyms: add support for relative offsets in kallsyms address table 2016-03-15 16:55:16 -07:00
kcmp.c ptrace: use fsuid, fsgid, effective creds for fs access checks 2016-01-20 17:09:18 -08:00
Kconfig.freezer
Kconfig.hz
Kconfig.locks
Kconfig.preempt
kcov.c kernel/kcov: unproxify debugfs file's fops 2016-06-15 04:56:35 -07:00
kexec_core.c kexec: add restriction on kexec_load() segment sizes 2016-08-02 19:35:31 -04:00
kexec_file.c kexec: introduce a protection mechanism for the crashkernel reserved memory 2016-05-23 17:04:14 -07:00
kexec_internal.h kexec: move some memembers and definitions within the scope of CONFIG_KEXEC_FILE 2016-01-20 17:09:18 -08:00
kexec.c kexec: allow architectures to override boot mapping 2016-08-02 19:35:27 -04:00
kmod.c
kprobes.c
ksysfs.c kexec: add a kexec_crash_loaded() function 2016-08-02 19:35:30 -04:00
kthread.c
latencytop.c sched/debug: Make schedstats a runtime tunable that is disabled by default 2016-02-09 11:54:23 +01:00
Makefile ELF/MIPS build fix 2016-05-23 17:04:14 -07:00
membarrier.c
memremap.c libnvdimm for 4.8 2016-07-28 17:38:16 -07:00
module_signing.c KEYS: Move the point of trust determination to __key_link() 2016-04-11 22:43:43 +01:00
module-internal.h
module.c Removed the MODULE_SIG_FORCE-means-no-MODULE_FORCE_LOAD patch. 2016-08-04 09:14:38 -04:00
notifier.c
nsproxy.c cgroup: introduce cgroup namespaces 2016-02-16 13:04:58 -05:00
padata.c kernel/padata.c: hide unused functions 2016-05-19 19:12:14 -07:00
panic.c kexec: use core_param for crash_kexec_post_notifiers boot option 2016-08-02 19:35:29 -04:00
params.c
pid_namespace.c
pid.c remove lots of IS_ERR_VALUE abuses 2016-05-27 15:26:11 -07:00
profile.c profile: Convert to hotplug state machine 2016-07-15 10:41:42 +02:00
ptrace.c tree-wide: replace config_enabled() with IS_ENABLED() 2016-08-04 08:50:07 -04:00
range.c
reboot.c
relay.c relay: add global mode support for buffer-only channels 2016-08-02 19:35:41 -04:00
resource.c /proc/iomem: only expose physical resource addresses to privileged users 2016-04-14 12:56:09 -07:00
seccomp.c tree-wide: replace config_enabled() with IS_ENABLED() 2016-08-04 08:50:07 -04:00
signal.c signals: Use hrtimer for sigtimedwait() 2016-07-07 10:35:07 +02:00
smp.c Merge branch 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-07-29 13:55:30 -07:00
smpboot.c cpu/hotplug: Unpark smpboot threads from the state machine 2016-03-01 20:36:56 +01:00
smpboot.h cpu/hotplug: Create hotplug threads 2016-03-01 20:36:56 +01:00
softirq.c arch, ftrace: for KASAN put hard/soft IRQ entries into separate sections 2016-03-25 16:37:42 -07:00
stacktrace.c
stop_machine.c stop_machine: Touch_nmi_watchdog() after MULTI_STOP_PREPARE 2016-07-27 11:12:11 +02:00
sys_ni.c
sys.c prctl: make PR_SET_THP_DISABLE wait for mmap_sem killable 2016-05-23 17:04:14 -07:00
sysctl_binary.c kernel/sysctl_binary.c: use generic UUID library 2016-05-20 17:58:30 -07:00
sysctl.c printk: add kernel parameter to control writes to /dev/kmsg 2016-08-02 19:35:06 -04:00
task_work.c task_work: use READ_ONCE/lockless_dereference, avoid pi_lock if !task_works 2016-08-02 19:35:02 -04:00
taskstats.c taskstats: use the libnl API to align nlattr on 64-bit 2016-04-23 20:13:25 -04:00
test_kprobes.c
torture.c torture: Stop onoff task if there is only one cpu 2016-06-14 16:03:28 -07:00
tracepoint.c kernel/...: convert pr_warning to pr_warn 2016-03-22 15:36:02 -07:00
tsacct.c time, acct: Drop irq save & restore from __acct_update_integrals() 2016-02-29 09:53:09 +01:00
uid16.c
up.c
user_namespace.c fs: Limit file caps to the user namespace of the super block 2016-06-24 10:40:31 -05:00
user-return-notifier.c
user.c
utsname_sysctl.c
utsname.c
watchdog.c Revert "perf/x86/intel, watchdog: Switch NMI watchdog to ref cycles on x86" 2016-07-10 20:58:36 +02:00
workqueue_internal.h sched/core: Get rid of 'cpu' argument in wq_worker_sleeping() 2016-03-02 10:28:47 -05:00
workqueue.c Merge branch 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-07-29 13:55:30 -07:00