Skip to contentSkip to conversation

cpuidle: menu governor: reduce latency on exit

PATCH In review Corrado Zoccolo 2 participants 5 messages Jul 19, 2009 – Jul 24, 2009 CPU idle time management framework

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cpuidle: menu governor: reduce latency on exit

This patch moves the state residency accounting and statistics computation off the hot exit path.
On exit, the need to recompute statistics is recorded, and new statistics will be computed when menu_select is called again.

---

drivers/cpuidle/governors/menu.c +19 -1
index f1df59f..5e905e0 100644
@@ -19,6 +19,7 @@
 
 struct menu_device {
 	int		last_state_idx;
+	int             needs_update;
 
 	unsigned int	expected_us;
 	unsigned int	predicted_us;
@@ -29,6 +30,8 @@ struct menu_device {
 
 static DEFINE_PER_CPU(struct menu_device, menu_devices);
 
+static void menu_update(struct cpuidle_device *dev);
+
 /**
  * menu_select - selects the next idle state to enter
  * @dev: the CPU
@@ -39,6 +42,11 @@ static int menu_select(struct cpuidle_device *dev)
 	int latency_req = pm_qos_requirement(PM_QOS_CPU_DMA_LATENCY);
 	int i;
 
+	if (data->needs_update) {
+		menu_update(dev);
+		data->needs_update = 0;
+	}
+
 	/* Special case when user has set very strict latency requirement */
 	if (unlikely(latency_req == 0)) {
 		data->last_state_idx = 0;
@@ -72,7 +80,7 @@ static int menu_select(struct cpuidle_device *dev)
 }
 
 /**
- * menu_reflect - attempts to guess what happened after entry
+ * menu_reflect - records that data structures need update
  * @dev: the CPU
  *
  * NOTE: it's important to be fast here because this operation will add to
@@ -81,6 +89,16 @@ static int menu_select(struct cpuidle_device *dev)
 static void menu_reflect(struct cpuidle_device *dev)
 {
 	struct menu_device *data = &__get_cpu_var(menu_devices);
+	data->needs_update = 1;
+}
+
+/**
+ * menu_update - attempts to guess what happened after entry
+ * @dev: the CPU
+ */
+static void menu_update(struct cpuidle_device *dev)
+{
+	struct menu_device *data = &__get_cpu_var(menu_devices);
 	int last_idx = data->last_state_idx;
 	unsigned int last_idle_us = cpuidle_get_last_residency(dev);
 	struct cpuidle_state *target = &dev->states[last_idx];
#1 lore raw
Quoted 5 lines · On Sun, 19 Jul 2009 16:59:46 +0200 Corrado Zoccolo <[email protected]> wrote:

This patch moves the state residency accounting and statistics computation off the hot exit path.
On exit, the need to recompute statistics is recorded, and new statistics will be computed when menu_select is called again.

Was there any observed and/or measured effect as a result of this patch?

If so, please send a description of that effect, thanks.

#2 lore raw
In reply to Corrado Zoccolo
Quoted 5 lines · On Sun, 19 Jul 2009 16:59:46 +0200 Corrado Zoccolo <[email protected]> wrote:

This patch moves the state residency accounting and statistics computation off the hot exit path.
On exit, the need to recompute statistics is recorded, and new statistics will be computed when menu_select is called again.

Was there any observed and/or measured effect as a result of this patch?

If so, please send a description of that effect, thanks.

#3 lore raw

Hi Andrew,
the expected effect is to reduce processor wakeup latency from sleep (C-states).

We are speaking of few hundreds of cycles reduction out of a several microseconds latency (determined by the hardware transition), so it is difficult to measure.

Corrado

Quoted 14 lines · Andrew Morton
Quoted 5 lines · On Sun, 19 Jul 2009 16:59:46 +0200 Corrado Zoccolo <[email protected]> wrote:

This patch moves the state residency accounting and statistics computation off the hot exit path.
On exit, the need to recompute statistics is recorded, and new statistics will be computed when menu_select is called again.

Was there any observed and/or measured effect as a result of this patch?

If so, please send a description of that effect, thanks.

Signature · 5 lines
__________________________________________________________________________

dott. Corrado Zoccolo                          mailto:[email protected]
PhD - Department of Computer Science - University of Pisa, Italy
--------------------------------------------------------------------------
#4 lore raw
In reply to Andrew Morton

Hi Andrew,
the expected effect is to reduce processor wakeup latency from sleep (C-states).

We are speaking of few hundreds of cycles reduction out of a several microseconds latency (determined by the hardware transition), so it is difficult to measure.

Corrado

Quoted 14 lines · Andrew Morton
Quoted 5 lines · On Sun, 19 Jul 2009 16:59:46 +0200 Corrado Zoccolo <[email protected]> wrote:

This patch moves the state residency accounting and statistics computation off the hot exit path.
On exit, the need to recompute statistics is recorded, and new statistics will be computed when menu_select is called again.

Was there any observed and/or measured effect as a result of this patch?

If so, please send a description of that effect, thanks.

Signature · 5 lines
__________________________________________________________________________

dott. Corrado Zoccolo                          mailto:[email protected]
PhD - Department of Computer Science - University of Pisa, Italy
--------------------------------------------------------------------------