The prediction for future is difficult and when the cpuidle governor prediction
fails and govenor possibly choose the shallower C-state than it should. How to
quickly notice and find the failure becomes important for power saving.
The patch extends to general case that prediction logic get a small predicted
residency, so it choose a shallow C-state though the expected residency is large
. Once the prediction will be fail, the CPU will keep staying at shallow C-state
for a long time. Acutally, the CPU has change enter into deep C-state.
So when the expected residency is long enough but governor choose a shallow
C-state, an timer will be added in order to monitor if the prediction failure.
When C-state is waken up prior to the adding timer, the timer will be cancelled
initiatively. When the timer is triggered and menu governor will quickly notice
prediction failure and re-evaluates deeper C-states possibility.
Signed-off-by: Youquan Song <youquan.song@xxxxxxxxx>
Signed-off-by: Rik van Riel <riel@xxxxxxxxxx>
---
drivers/cpuidle/governors/menu.c | 22 ++++++++++++++++++++++
1 files changed, 22 insertions(+), 0 deletions(-)
diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c
index beeab6a..b34bf11 100644
--- a/drivers/cpuidle/governors/menu.c
+++ b/drivers/cpuidle/governors/menu.c
@@ -114,6 +114,13 @@ static DEFINE_PER_CPU(int, hrtimer_started);
*
*/
+/*
+ * The C-state residency is so long that is is worthwhile to exit
+ * from the shallow C-state and re-enter into a deeper C-state.
+ */
+static unsigned int perfect_cstate_ms __read_mostly = 30;
+module_param(perfect_cstate_ms, uint, 0000);
+
struct menu_device {
int last_state_idx;
int needs_update;
@@ -351,6 +358,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
/* not deepest C-state chosen for low predicted residency */
if (low_predicted) {
unsigned int timer_us = 0;
+ unsigned int perfect_us = 0;
/*
* Set a timer to detect whether this sleep is much
@@ -361,12 +369,26 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
*/
timer_us = 2 * (data->predicted_us + MAX_DEVIATION);
+ perfect_us = perfect_cstate_ms * 1000;
+
if (repeat && (4 * timer_us < data->expected_us)) {
hrtimer_start(hrtmr, ns_to_ktime(1000 * timer_us),
HRTIMER_MODE_REL_PINNED);
/* menu hrtimer is started */
per_cpu(hrtimer_started, cpu) = 1;
+ } else if (perfect_us < data->expected_us) {
+ /*
+ * The next timer is long. This could be because
+ * we did not make a useful prediction.
+ * In that case, it makes sense to re-enter
+ * into a deeper C-state after some time.
+ */
+ hrtimer_start(hrtmr, ns_to_ktime(1000 * timer_us),
+ HRTIMER_MODE_REL_PINNED);
+ /* menu hrtimer is started */
+ per_cpu(hrtimer_started, cpu) = 1;
}
+
}
return data->last_state_idx;
--
1.7.7.4
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
