Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit b4e82b5c authored by Girish Mahadevan's avatar Girish Mahadevan Committed by Stephen Boyd
Browse files

msm: lpm_levels: Adjust latency for Multi core cases 



In multi core cases for several cores going in and out of power collapse
the warmboot is serialized at Trust Zone, which means the exit latency
will scale based on the number of cores exiting power collapse.

Adjust for this when chosing a low power mode based on the number of online
cpus.

Change-Id: I4fe137a1bf8596b7da7990845b18ceb8f3d529da
Signed-off-by: default avatarGirish Mahadevan <girishm@codeaurora.org>
parent 00a6691d

arch/arm/mach-msm/lpm_levels.c

+31 −13
Original line number Diff line number Diff line
@@ -71,6 +71,10 @@ static ssize_t msm_lpm_levels_attr_show( 
static ssize_t msm_lpm_levels_attr_store(struct kobject *kobj,

	struct kobj_attribute *attr, const char *buf, size_t count);



#define ADJUST_LATENCY(x)	\

	((x == MSM_PM_SLEEP_MODE_POWER_COLLAPSE_STANDALONE) ?\

		(num_online_cpus()) / 2 : 0)



static int msm_lpm_lvl_dbg_msk;



module_param_named(
@@ -343,6 +347,9 @@ static void *msm_lpm_lowest_limits(bool from_idle, 
	int i;

	bool modify_event_timer;

	uint32_t next_wakeup_us = time_param->sleep_us;

	uint32_t lvl_latency_us = 0;

	uint32_t lvl_overhead_us = 0;

	uint32_t lvl_overhead_energy = 0;



	if (!msm_lpm_levels)

		return NULL;
@@ -358,24 +365,36 @@ static void *msm_lpm_lowest_limits(bool from_idle, 
		if (sleep_mode != level->sleep_mode)

			continue;



		if (time_param->latency_us < level->latency_us)

		lvl_latency_us =

			level->latency_us + (level->latency_us *

						ADJUST_LATENCY(sleep_mode));



		lvl_overhead_us =

			level->time_overhead_us + (level->time_overhead_us *

						ADJUST_LATENCY(sleep_mode));



		lvl_overhead_energy =

			level->energy_overhead + level->energy_overhead *

						ADJUST_LATENCY(sleep_mode);



		if (time_param->latency_us < lvl_latency_us)

			continue;



		if (time_param->next_event_us &&

			time_param->next_event_us < level->latency_us)

			time_param->next_event_us < lvl_latency_us)

			continue;



		if (time_param->next_event_us) {

			if ((time_param->next_event_us < time_param->sleep_us)

			|| ((time_param->next_event_us - level->latency_us) <

			|| ((time_param->next_event_us - lvl_latency_us) <

				time_param->sleep_us)) {

				modify_event_timer = true;

				next_wakeup_us = time_param->next_event_us -

						level->latency_us;

						lvl_latency_us;

			}

		}



		if (next_wakeup_us <= level->time_overhead_us)

		if (next_wakeup_us <= lvl_overhead_us)

			continue;



		if ((MSM_PM_SLEEP_MODE_POWER_COLLAPSE_STANDALONE == sleep_mode)
@@ -384,22 +403,21 @@ static void *msm_lpm_lowest_limits(bool from_idle, 
					break;



		if (next_wakeup_us <= 1) {

			pwr = level->energy_overhead;

		} else if (next_wakeup_us <= level->time_overhead_us) {

			pwr = level->energy_overhead / next_wakeup_us;

			pwr = lvl_overhead_energy;

		} else if (next_wakeup_us <= lvl_overhead_us) {

			pwr = lvl_overhead_energy / next_wakeup_us;

		} else if ((next_wakeup_us >> 10)

				> level->time_overhead_us) {

				> lvl_overhead_us) {

			pwr = level->steady_state_power;

		} else {

			pwr = level->steady_state_power;

			pwr -= (level->time_overhead_us *

			pwr -= (lvl_overhead_us *

				level->steady_state_power) /

						next_wakeup_us;

			pwr += level->energy_overhead / next_wakeup_us;

			pwr += lvl_overhead_energy / next_wakeup_us;

		}



		if (!best_level || (best_level_pwr >= pwr)) {



			best_level = level;

			best_level_pwr = pwr;

			if (power)
@@ -409,7 +427,7 @@ static void *msm_lpm_lowest_limits(bool from_idle, 
					MSM_PM_SLEEP_MODE_WAIT_FOR_INTERRUPT))

				time_param->modified_time_us =

					time_param->next_event_us -

						best_level->latency_us;

						lvl_latency_us;

			else

				time_param->modified_time_us = 0;

		}