cpuidle: Use nanoseconds as the unit of time

Currently, the cpuidle subsystem uses microseconds as the unit of
time which (among other things) causes the idle loop to incur some
integer division overhead for no clear benefit.

In order to allow cpuidle to measure time in nanoseconds, add two
new fields, exit_latency_ns and target_residency_ns, to represent the
exit latency and target residency of an idle state in nanoseconds,
respectively, to struct cpuidle_state and initialize them with the
help of the corresponding values in microseconds provided by drivers.
Additionally, change cpuidle_governor_latency_req() to return the
idle state exit latency constraint in nanoseconds.

Also meeasure idle state residency (last_residency_ns in struct
cpuidle_device and time_ns in struct cpuidle_driver) in nanoseconds
and update the cpuidle core and governors accordingly.

However, the menu governor still computes typical intervals in
microseconds to avoid integer overflows.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Doug Smythies <dsmythies@telus.net>
Tested-by: Doug Smythies <dsmythies@telus.net>

1 files changed, 36 insertions, 40 deletions

diff --git a/drivers/cpuidle/governors/teo.c b/drivers/cpuidle/governors/teo.c
index 702d560eb347..ecbcfaefb0cd 100644
--- a/drivers/cpuidle/governors/teo.c
+++ b/drivers/cpuidle/governors/teo.c
@@ -104,7 +104,7 @@ struct teo_cpu {
 	u64 sleep_length_ns;
 	struct teo_idle_state states[CPUIDLE_STATE_MAX];
 	int interval_idx;
-	unsigned int intervals[INTERVALS];
+	u64 intervals[INTERVALS];
 };
 
 static DEFINE_PER_CPU(struct teo_cpu, teo_cpus);
@@ -117,9 +117,8 @@ static DEFINE_PER_CPU(struct teo_cpu, teo_cpus);
 static void teo_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 {
 	struct teo_cpu *cpu_data = per_cpu_ptr(&teo_cpus, dev->cpu);
-	unsigned int sleep_length_us = ktime_to_us(cpu_data->sleep_length_ns);
 	int i, idx_hit = -1, idx_timer = -1;
-	unsigned int measured_us;
+	u64 measured_ns;
 
 	if (cpu_data->time_span_ns >= cpu_data->sleep_length_ns) {
 		/*
@@ -127,23 +126,21 @@ static void teo_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 		 * enough to the closest timer event expected at the idle state
 		 * selection time to be discarded.
 		 */
-		measured_us = UINT_MAX;
+		measured_ns = U64_MAX;
 	} else {
-		unsigned int lat;
+		u64 lat_ns = drv->states[dev->last_state_idx].exit_latency_ns;
 
-		lat = drv->states[dev->last_state_idx].exit_latency;
-
-		measured_us = ktime_to_us(cpu_data->time_span_ns);
+		measured_ns = cpu_data->time_span_ns;
 		/*
 		 * The delay between the wakeup and the first instruction
 		 * executed by the CPU is not likely to be worst-case every
 		 * time, so take 1/2 of the exit latency as a very rough
 		 * approximation of the average of it.
 		 */
-		if (measured_us >= lat)
-			measured_us -= lat / 2;
+		if (measured_ns >= lat_ns)
+			measured_ns -= lat_ns / 2;
 		else
-			measured_us /= 2;
+			measured_ns /= 2;
 	}
 
 	/*
@@ -155,9 +152,9 @@ static void teo_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 
 		cpu_data->states[i].early_hits -= early_hits >> DECAY_SHIFT;
 
-		if (drv->states[i].target_residency <= sleep_length_us) {
+		if (drv->states[i].target_residency_ns <= cpu_data->sleep_length_ns) {
 			idx_timer = i;
-			if (drv->states[i].target_residency <= measured_us)
+			if (drv->states[i].target_residency_ns <= measured_ns)
 				idx_hit = i;
 		}
 	}
@@ -193,7 +190,7 @@ static void teo_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 	 * Save idle duration values corresponding to non-timer wakeups for
 	 * pattern detection.
 	 */
-	cpu_data->intervals[cpu_data->interval_idx++] = measured_us;
+	cpu_data->intervals[cpu_data->interval_idx++] = measured_ns;
 	if (cpu_data->interval_idx > INTERVALS)
 		cpu_data->interval_idx = 0;
 }
@@ -203,11 +200,11 @@ static void teo_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
  * @drv: cpuidle driver containing state data.
  * @dev: Target CPU.
  * @state_idx: Index of the capping idle state.
- * @duration_us: Idle duration value to match.
+ * @duration_ns: Idle duration value to match.
  */
 static int teo_find_shallower_state(struct cpuidle_driver *drv,
 				    struct cpuidle_device *dev, int state_idx,
-				    unsigned int duration_us)
+				    u64 duration_ns)
 {
 	int i;
 
@@ -216,7 +213,7 @@ static int teo_find_shallower_state(struct cpuidle_driver *drv,
 			continue;
 
 		state_idx = i;
-		if (drv->states[i].target_residency <= duration_us)
+		if (drv->states[i].target_residency_ns <= duration_ns)
 			break;
 	}
 	return state_idx;
@@ -232,8 +229,9 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 		      bool *stop_tick)
 {
 	struct teo_cpu *cpu_data = per_cpu_ptr(&teo_cpus, dev->cpu);
-	int latency_req = cpuidle_governor_latency_req(dev->cpu);
-	unsigned int duration_us, hits, misses, early_hits;
+	s64 latency_req = cpuidle_governor_latency_req(dev->cpu);
+	u64 duration_ns;
+	unsigned int hits, misses, early_hits;
 	int max_early_idx, constraint_idx, idx, i;
 	ktime_t delta_tick;
 
@@ -244,8 +242,8 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 
 	cpu_data->time_span_ns = local_clock();
 
-	cpu_data->sleep_length_ns = tick_nohz_get_sleep_length(&delta_tick);
-	duration_us = ktime_to_us(cpu_data->sleep_length_ns);
+	duration_ns = tick_nohz_get_sleep_length(&delta_tick);
+	cpu_data->sleep_length_ns = duration_ns;
 
 	hits = 0;
 	misses = 0;
@@ -262,7 +260,7 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 			 * Ignore disabled states with target residencies beyond
 			 * the anticipated idle duration.
 			 */
-			if (s->target_residency > duration_us)
+			if (s->target_residency_ns > duration_ns)
 				continue;
 
 			/*
@@ -301,7 +299,7 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 			 * shallow for that role.
 			 */
 			if (!(tick_nohz_tick_stopped() &&
-			      drv->states[idx].target_residency < TICK_USEC)) {
+			      drv->states[idx].target_residency_ns < TICK_NSEC)) {
 				early_hits = cpu_data->states[i].early_hits;
 				max_early_idx = idx;
 			}
@@ -315,10 +313,10 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 			misses = cpu_data->states[i].misses;
 		}
 
-		if (s->target_residency > duration_us)
+		if (s->target_residency_ns > duration_ns)
 			break;
 
-		if (s->exit_latency > latency_req && constraint_idx > i)
+		if (s->exit_latency_ns > latency_req && constraint_idx > i)
 			constraint_idx = i;
 
 		idx = i;
@@ -327,7 +325,7 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 
 		if (early_hits < cpu_data->states[i].early_hits &&
 		    !(tick_nohz_tick_stopped() &&
-		      drv->states[i].target_residency < TICK_USEC)) {
+		      drv->states[i].target_residency_ns < TICK_NSEC)) {
 			early_hits = cpu_data->states[i].early_hits;
 			max_early_idx = i;
 		}
@@ -343,7 +341,7 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 	 */
 	if (hits <= misses && max_early_idx >= 0) {
 		idx = max_early_idx;
-		duration_us = drv->states[idx].target_residency;
+		duration_ns = drv->states[idx].target_residency_ns;
 	}
 
 	/*
@@ -364,9 +362,9 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 		 * the current expected idle duration value.
 		 */
 		for (i = 0; i < INTERVALS; i++) {
-			unsigned int val = cpu_data->intervals[i];
+			u64 val = cpu_data->intervals[i];
 
-			if (val >= duration_us)
+			if (val >= duration_ns)
 				continue;
 
 			count++;
@@ -378,17 +376,17 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 		 * values are in the interesting range.
 		 */
 		if (count > INTERVALS / 2) {
-			unsigned int avg_us = div64_u64(sum, count);
+			u64 avg_ns = div64_u64(sum, count);
 
 			/*
 			 * Avoid spending too much time in an idle state that
 			 * would be too shallow.
 			 */
-			if (!(tick_nohz_tick_stopped() && avg_us < TICK_USEC)) {
-				duration_us = avg_us;
-				if (drv->states[idx].target_residency > avg_us)
+			if (!(tick_nohz_tick_stopped() && avg_ns < TICK_NSEC)) {
+				duration_ns = avg_ns;
+				if (drv->states[idx].target_residency_ns > avg_ns)
 					idx = teo_find_shallower_state(drv, dev,
-								       idx, avg_us);
+								       idx, avg_ns);
 			}
 		}
 	}
@@ -398,9 +396,7 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 	 * expected idle duration is shorter than the tick period length.
 	 */
 	if (((drv->states[idx].flags & CPUIDLE_FLAG_POLLING) ||
-	    duration_us < TICK_USEC) && !tick_nohz_tick_stopped()) {
-		unsigned int delta_tick_us = ktime_to_us(delta_tick);
-
+	    duration_ns < TICK_NSEC) && !tick_nohz_tick_stopped()) {
 		*stop_tick = false;
 
 		/*
@@ -409,8 +405,8 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
 		 * till the closest timer including the tick, try to correct
 		 * that.
 		 */
-		if (idx > 0 && drv->states[idx].target_residency > delta_tick_us)
-			idx = teo_find_shallower_state(drv, dev, idx, delta_tick_us);
+		if (idx > 0 && drv->states[idx].target_residency_ns > delta_tick)
+			idx = teo_find_shallower_state(drv, dev, idx, delta_tick);
 	}
 
 	return idx;
@@ -454,7 +450,7 @@ static int teo_enable_device(struct cpuidle_driver *drv,
 	memset(cpu_data, 0, sizeof(*cpu_data));
 
 	for (i = 0; i < INTERVALS; i++)
-		cpu_data->intervals[i] = UINT_MAX;
+		cpu_data->intervals[i] = U64_MAX;
 
 	return 0;
 }