[PATCH v3] PM / OPP: optimize dev_pm_opp_set_rate() performance a bit
From: Jisheng Zhang <hidden>
Date: 2016-07-25 06:16:01
Also in:
linux-pm, lkml
Subsystem:
driver core, kobjects, debugfs and sysfs, hibernation (aka software suspend, aka swsusp), power management core, suspend to ram, the rest · Maintainers:
Greg Kroah-Hartman, "Rafael J. Wysocki", Danilo Krummrich, Linus Torvalds
In dev_pm_opp_set_rate(), _find_opp_table() is called 4 times: once by _get_opp_clk(), once by dev_pm_opp_set_rate() itself, and twice by dev_pm_opp_find_freq_ceil(). If there are several opp_tables in the system, three times of opp table finding is a big waste. This patch reduced the call of _find_opp_table() to twice. Signed-off-by: Jisheng Zhang <redacted> --- Since v2: - fix gcc warnings Since v1: - avoid opp_table pointer usage after the rcu_read_unlock() as pointed out by Viresh, thus we can only reduced the call of _find_opp_table to twice. drivers/base/power/opp/core.c | 31 +++++++++++++++++++------------ 1 file changed, 19 insertions(+), 12 deletions(-)
diff --git a/drivers/base/power/opp/core.c b/drivers/base/power/opp/core.c
index 7c04c87..df0c709 100644
--- a/drivers/base/power/opp/core.c
+++ b/drivers/base/power/opp/core.c@@ -402,6 +402,22 @@ struct dev_pm_opp *dev_pm_opp_find_freq_exact(struct device *dev, } EXPORT_SYMBOL_GPL(dev_pm_opp_find_freq_exact); +static noinline struct dev_pm_opp *_find_freq_ceil(struct opp_table *opp_table, + unsigned long *freq) +{ + struct dev_pm_opp *temp_opp, *opp = ERR_PTR(-ERANGE); + + list_for_each_entry_rcu(temp_opp, &opp_table->opp_list, node) { + if (temp_opp->available && temp_opp->rate >= *freq) { + opp = temp_opp; + *freq = opp->rate; + break; + } + } + + return opp; +} + /** * dev_pm_opp_find_freq_ceil() - Search for an rounded ceil freq * @dev: device for which we do this operation
@@ -427,7 +443,6 @@ struct dev_pm_opp *dev_pm_opp_find_freq_ceil(struct device *dev, unsigned long *freq) { struct opp_table *opp_table; - struct dev_pm_opp *temp_opp, *opp = ERR_PTR(-ERANGE); opp_rcu_lockdep_assert();
@@ -440,15 +455,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_ceil(struct device *dev, if (IS_ERR(opp_table)) return ERR_CAST(opp_table); - list_for_each_entry_rcu(temp_opp, &opp_table->opp_list, node) { - if (temp_opp->available && temp_opp->rate >= *freq) { - opp = temp_opp; - *freq = opp->rate; - break; - } - } - - return opp; + return _find_freq_ceil(opp_table, freq); } EXPORT_SYMBOL_GPL(dev_pm_opp_find_freq_ceil);
@@ -612,7 +619,7 @@ int dev_pm_opp_set_rate(struct device *dev, unsigned long target_freq) return PTR_ERR(opp_table); } - old_opp = dev_pm_opp_find_freq_ceil(dev, &old_freq); + old_opp = _find_freq_ceil(opp_table, &old_freq); if (!IS_ERR(old_opp)) { ou_volt = old_opp->u_volt; ou_volt_min = old_opp->u_volt_min;
@@ -622,7 +629,7 @@ int dev_pm_opp_set_rate(struct device *dev, unsigned long target_freq) __func__, old_freq, PTR_ERR(old_opp)); } - opp = dev_pm_opp_find_freq_ceil(dev, &freq); + opp = _find_freq_ceil(opp_table, &freq); if (IS_ERR(opp)) { ret = PTR_ERR(opp); dev_err(dev, "%s: failed to find OPP for freq %lu (%d)\n",
--
2.8.1