linux/drivers/cpuidle
Daniel Lezcano e93e59ce5b cpuidle: Replace ktime_get() with local_clock()
The ktime_get() can have a non negligeable overhead, use local_clock()
instead.

In order to test the difference between ktime_get() and local_clock(),
a quick hack has been added to trigger, via debugfs, 10000 times a
call to ktime_get() and local_clock() and measure the elapsed time.

Then the average value, the min and max is computed for each call.

From userspace, the test above was called 100 times every 2 seconds.

So, ktime_get() and local_clock() have been called 1000000 times in
total.

The results are:

ktime_get():
============
 * average: 101 ns (stddev: 27.4)
 * maximum: 38313 ns
 * minimum: 65 ns

local_clock():
==============
 * average: 60 ns (stddev: 9.8)
 * maximum: 13487 ns
 * minimum: 46 ns

The local_clock() is faster and more stable.

Even if it is a drop in the ocean, changing the ktime_get() by the
local_clock() allows to save 80ns at idle time (entry + exit). And
in some circumstances, especially when there are several CPUs racing
for the clock access, we save tens of microseconds.

The idle duration resulting from a diff is converted from nanosec to
microsec. This could be done with integer division (div 1000) - which is
an expensive operation or by 10 bits shifting (div 1024) - which is fast
but unprecise.

The following table gives some results at the limits.

 ------------------------------------------
|   nsec   |   div(1000)   |   div(1024)   |
 ------------------------------------------
|   1e3    |        1 usec |      976 nsec |
 ------------------------------------------
|   1e6    |     1000 usec |      976 usec |
 ------------------------------------------
|   1e9    |  1000000 usec |   976562 usec |
 ------------------------------------------

There is a linear deviation of 2.34%. This loss of precision is acceptable
in the context of the resulting diff which is used for statistics. These
ones are processed to guess estimate an approximation of the duration of the
next idle period which ends up into an idle state selection. The selection
criteria takes into account the next duration based on large intervals,
represented by the idle state's target residency.

The 2^10 division is enough because the approximation regarding the 1e3
division is lost in all the approximations done for the next idle duration
computation.

Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
[ rjw: Subject ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-26 02:38:49 +02:00
..
governors cpuidle: menu: Fall back to polling if next timer event is near 2016-03-21 15:50:28 +01:00
coupled.c cpuidle: coupled: remove unused define cpuidle_coupled_lock 2016-01-27 23:08:46 +01:00
cpuidle-arm.c ARM: cpuidle: Register per cpuidle device 2015-03-24 14:46:25 +01:00
cpuidle-at91.c drivers/cpuidle: Convert non-modular drivers to use builtin_platform_driver 2015-06-16 14:12:38 -04:00
cpuidle-big_little.c ARM: MCPM: remove residency argument from mcpm_cpu_suspend() 2015-05-06 11:47:10 -04:00
cpuidle-calxeda.c ARM: migrate to common PSCI client code 2015-08-03 15:38:39 +01:00
cpuidle-clps711x.c drivers/cpuidle: make cpuidle-clps711x.c explicitly non-modular 2015-12-15 00:22:21 +01:00
cpuidle-cps.c cpuidle: Invert CPUIDLE_FLAG_TIME_VALID logic 2014-11-12 21:17:27 +01:00
cpuidle-exynos.c drivers/cpuidle: make cpuidle-exynos.c explicitly non-modular 2015-12-15 00:22:22 +01:00
cpuidle-kirkwood.c ARM: cpuidle: Remove duplicate header inclusion 2015-03-23 18:03:11 +01:00
cpuidle-mvebu-v7.c cpuidle: mvebu: disable the bind/unbind attributes and use builtin_platform_driver 2015-10-23 12:40:48 +02:00
cpuidle-powernv.c powerpc/powernv: remove FW_FEATURE_OPALv3 and just use FW_FEATURE_OPAL 2015-12-17 22:40:54 +11:00
cpuidle-pseries.c cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state 2015-06-22 15:15:15 +02:00
cpuidle-ux500.c drivers/cpuidle: make cpuidle-ux500.c explicitly non-modular 2015-12-15 00:22:22 +01:00
cpuidle-zynq.c drivers/cpuidle: Convert non-modular drivers to use builtin_platform_driver 2015-06-16 14:12:38 -04:00
cpuidle.c cpuidle: Replace ktime_get() with local_clock() 2016-04-26 02:38:49 +02:00
cpuidle.h cpuidle/coupled: Add sanity check for safe_state_index 2015-09-03 03:05:47 +02:00
driver.c cpuidle/coupled: Add sanity check for safe_state_index 2015-09-03 03:05:47 +02:00
dt_idle_states.c Merge back earlier cpuidle material for 3.19-rc1. 2014-11-21 16:31:42 +01:00
dt_idle_states.h drivers: cpuidle: implement DT based idle states infrastructure 2014-09-25 10:52:20 +02:00
governor.c cpuidle: Replace strnicmp with strncasecmp 2014-09-25 01:17:10 +02:00
Kconfig cpuidle: Don't enable all governors by default 2016-01-15 22:39:58 +01:00
Kconfig.arm ARM: cpuidle: Enable the ARM64 driver for both ARM32/ARM64 2015-03-24 10:16:11 +01:00
Kconfig.mips MIPS: Kconfig: Add missing MIPS_CPS dependencies to PM and cpuidle 2014-10-23 19:58:05 +02:00
Kconfig.powerpc powerpc/powernv/cpuidle: Back-end cpuidle driver for powernv platform. 2014-01-29 17:02:24 +11:00
Makefile ARM: cpuidle: Enable the ARM64 driver for both ARM32/ARM64 2015-03-24 10:16:11 +01:00
sysfs.c cpuidle: remove state_count field from struct cpuidle_device 2015-04-03 13:15:50 +02:00