提交 af556884 编写于 作者: B Bruno Randolf 提交者: John W. Linville

lib: Improve EWMA efficiency by using bitshifts

Using bitshifts instead of division and multiplication should improve
performance. That requires weight and factor to be powers of two, but i think
this is something we can live with.

Thanks to Peter Zijlstra for the improved formula!
Signed-off-by: NBruno Randolf <br1@einfach.org>

--

v2:	use log2.h functions
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
上级 5dcc03fe
master alk-4.19.24 alk-4.19.30 alk-4.19.34 alk-4.19.36 alk-4.19.43 alk-4.19.48 alk-4.19.57 ck-4.19.67 ck-4.19.81 ck-4.19.91 github/fork/deepanshu1422/fix-typo-in-comment github/fork/haosdent/fix-typo linux-next v4.19.91 v4.19.90 v4.19.89 v4.19.88 v4.19.87 v4.19.86 v4.19.85 v4.19.84 v4.19.83 v4.19.82 v4.19.81 v4.19.80 v4.19.79 v4.19.78 v4.19.77 v4.19.76 v4.19.75 v4.19.74 v4.19.73 v4.19.72 v4.19.71 v4.19.70 v4.19.69 v4.19.68 v4.19.67 v4.19.66 v4.19.65 v4.19.64 v4.19.63 v4.19.62 v4.19.61 v4.19.60 v4.19.59 v4.19.58 v4.19.57 v4.19.56 v4.19.55 v4.19.54 v4.19.53 v4.19.52 v4.19.51 v4.19.50 v4.19.49 v4.19.48 v4.19.47 v4.19.46 v4.19.45 v4.19.44 v4.19.43 v4.19.42 v4.19.41 v4.19.40 v4.19.39 v4.19.38 v4.19.37 v4.19.36 v4.19.35 v4.19.34 v4.19.33 v4.19.32 v4.19.31 v4.19.30 v4.19.29 v4.19.28 v4.19.27 v4.19.26 v4.19.25 v4.19.24 v4.19.23 v4.19.22 v4.19.21 v4.19.20 v4.19.19 v4.19.18 v4.19.17 v4.19.16 v4.19.15 v4.19.14 v4.19.13 v4.19.12 v4.19.11 v4.19.10 v4.19.9 v4.19.8 v4.19.7 v4.19.6 v4.19.5 v4.19.4 v4.19.3 v4.19.2 v4.19.1 v4.19 v4.19-rc8 v4.19-rc7 v4.19-rc6 v4.19-rc5 v4.19-rc4 v4.19-rc3 v4.19-rc2 v4.19-rc1 ck-release-21 ck-release-20 ck-release-19.2 ck-release-19.1 ck-release-19 ck-release-18 ck-release-17.2 ck-release-17.1 ck-release-17 ck-release-16 ck-release-15.1 ck-release-15 ck-release-14 ck-release-13.2 ck-release-13 ck-release-12 ck-release-11 ck-release-10 ck-release-9 ck-release-7 alk-release-15 alk-release-14 alk-release-13.2 alk-release-13 alk-release-12 alk-release-11 alk-release-10 alk-release-9 alk-release-7
无相关合并请求
#ifndef _LINUX_AVERAGE_H
#define _LINUX_AVERAGE_H
#include <linux/kernel.h>
/* Exponentially weighted moving average (EWMA) */
/* For more documentation see lib/average.c */
......@@ -26,7 +24,7 @@ extern struct ewma *ewma_add(struct ewma *avg, unsigned long val);
*/
static inline unsigned long ewma_read(const struct ewma *avg)
{
return DIV_ROUND_CLOSEST(avg->internal, avg->factor);
return avg->internal >> avg->factor;
}
#endif /* _LINUX_AVERAGE_H */
......@@ -8,6 +8,7 @@
#include <linux/module.h>
#include <linux/average.h>
#include <linux/bug.h>
#include <linux/log2.h>
/**
* DOC: Exponentially Weighted Moving Average (EWMA)
......@@ -24,18 +25,21 @@
* ewma_init() - Initialize EWMA parameters
* @avg: Average structure
* @factor: Factor to use for the scaled up internal value. The maximum value
* of averages can be ULONG_MAX/(factor*weight).
* of averages can be ULONG_MAX/(factor*weight). For performance reasons
* factor has to be a power of 2.
* @weight: Exponential weight, or decay rate. This defines how fast the
* influence of older values decreases. Has to be bigger than 1.
* influence of older values decreases. For performance reasons weight has
* to be a power of 2.
*
* Initialize the EWMA parameters for a given struct ewma @avg.
*/
void ewma_init(struct ewma *avg, unsigned long factor, unsigned long weight)
{
WARN_ON(weight <= 1 || factor == 0);
WARN_ON(!is_power_of_2(weight) || !is_power_of_2(factor));
avg->weight = ilog2(weight);
avg->factor = ilog2(factor);
avg->internal = 0;
avg->weight = weight;
avg->factor = factor;
}
EXPORT_SYMBOL(ewma_init);
......@@ -49,9 +53,9 @@ EXPORT_SYMBOL(ewma_init);
struct ewma *ewma_add(struct ewma *avg, unsigned long val)
{
avg->internal = avg->internal ?
(((avg->internal * (avg->weight - 1)) +
(val * avg->factor)) / avg->weight) :
(val * avg->factor);
(((avg->internal << avg->weight) - avg->internal) +
(val << avg->factor)) >> avg->weight :
(val << avg->factor);
return avg;
}
EXPORT_SYMBOL(ewma_add);
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册
反馈
建议
客服 返回
顶部