If the given type has fraction smaller than max_frac/FPROP_FRAC_BASE, the code could be modified to call __fprop_inc_percpu() directly and easier to understand. After this patch, fprop_reflect_period_percpu() will be called twice, and quicky return on pl->period == p->period test, so it would not result to significant downside of performance.
Thanks for Jan's guidance. Signed-off-by: Tan Hu <tan...@zte.com.cn> Reviewed-by: Jan Kara <j...@suse.cz> --- lib/flex_proportions.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c index 7852bfff5..451543937 100644 --- a/lib/flex_proportions.c +++ b/lib/flex_proportions.c @@ -266,8 +266,7 @@ void __fprop_inc_percpu_max(struct fprop_global *p, if (numerator > (((u64)denominator) * max_frac) >> FPROP_FRAC_SHIFT) return; - } else - fprop_reflect_period_percpu(p, pl); - percpu_counter_add_batch(&pl->events, 1, PROP_BATCH); - percpu_counter_add(&p->events, 1); + } + + __fprop_inc_percpu(p, pl); } -- 2.19.1