| Message ID | 1410303964-6943-1-git-send-email-dianders@chromium.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 09/10/2014 07:06 AM, Doug Anderson wrote: > Let's define the voltatile registers (those that can't be cached) and > enable caching. The rk808 is accessed almost constantly with cpufreq > so this is really nice. > > As measured by ftrace: > before this change: cpu0_set_target() => ~2200us > after this change: cpu0_set_target() => ~500us > > Signed-off-by: Doug Anderson <dianders@chromium.org> > --- > drivers/mfd/rk808.c | 30 ++++++++++++++++++++++++++++++ > 1 file changed, 30 insertions(+) > > diff --git a/drivers/mfd/rk808.c b/drivers/mfd/rk808.c > index 0324422..bd02150 100644 > --- a/drivers/mfd/rk808.c > +++ b/drivers/mfd/rk808.c > @@ -29,10 +29,40 @@ struct rk808_reg_data { > int value; > }; > > +static bool rk808_is_volatile_reg(struct device *dev, unsigned int reg) > +{ > + /* > + * Notes: > + * - Technically the ROUND_30s bit makes RTC_CTRL_REG volatile, but > + * we don't use that feature. It's better to cache. > + * - It's unlikely we care that RK808_DEVCTRL_REG is volatile since > + * bits are cleared in case when we shutoff anyway, but better safe. > + */ > + > + switch (reg) { > + case RK808_SECONDS_REG ... RK808_WEEKS_REG: > + case RK808_RTC_STATUS_REG: > + case RK808_VB_MON_REG: > + case RK808_THERMAL_REG: > + case RK808_DCDC_UV_STS_REG: > + case RK808_LDO_UV_STS_REG: > + case RK808_DCDC_PG_REG: > + case RK808_LDO_PG_REG: > + case RK808_DEVCTRL_REG: > + case RK808_INT_STS_REG1: > + case RK808_INT_STS_REG2: > + return true; > + } > + > + return false; > +} > + > static const struct regmap_config rk808_regmap_config = { > .reg_bits = 8, > .val_bits = 8, > .max_register = RK808_IO_POL_REG, > + .cache_type = REGCACHE_RBTREE, > + .volatile_reg = rk808_is_volatile_reg, > }; > > static struct resource rtc_resources[] = { Reviewed-by: Chris Zhong <zyw@rock-chips.com>
On Tue, 09 Sep 2014, Doug Anderson wrote: > Let's define the voltatile registers (those that can't be cached) and > enable caching. The rk808 is accessed almost constantly with cpufreq > so this is really nice. > > As measured by ftrace: > before this change: cpu0_set_target() => ~2200us > after this change: cpu0_set_target() => ~500us > > Signed-off-by: Doug Anderson <dianders@chromium.org> > --- > drivers/mfd/rk808.c | 30 ++++++++++++++++++++++++++++++ > 1 file changed, 30 insertions(+) Applied with Chris' Reviewed-by. > diff --git a/drivers/mfd/rk808.c b/drivers/mfd/rk808.c > index 0324422..bd02150 100644 > --- a/drivers/mfd/rk808.c > +++ b/drivers/mfd/rk808.c > @@ -29,10 +29,40 @@ struct rk808_reg_data { > int value; > }; > > +static bool rk808_is_volatile_reg(struct device *dev, unsigned int reg) > +{ > + /* > + * Notes: > + * - Technically the ROUND_30s bit makes RTC_CTRL_REG volatile, but > + * we don't use that feature. It's better to cache. > + * - It's unlikely we care that RK808_DEVCTRL_REG is volatile since > + * bits are cleared in case when we shutoff anyway, but better safe. > + */ > + > + switch (reg) { > + case RK808_SECONDS_REG ... RK808_WEEKS_REG: > + case RK808_RTC_STATUS_REG: > + case RK808_VB_MON_REG: > + case RK808_THERMAL_REG: > + case RK808_DCDC_UV_STS_REG: > + case RK808_LDO_UV_STS_REG: > + case RK808_DCDC_PG_REG: > + case RK808_LDO_PG_REG: > + case RK808_DEVCTRL_REG: > + case RK808_INT_STS_REG1: > + case RK808_INT_STS_REG2: > + return true; > + } > + > + return false; > +} > + > static const struct regmap_config rk808_regmap_config = { > .reg_bits = 8, > .val_bits = 8, > .max_register = RK808_IO_POL_REG, > + .cache_type = REGCACHE_RBTREE, > + .volatile_reg = rk808_is_volatile_reg, > }; > > static struct resource rtc_resources[] = {
diff --git a/drivers/mfd/rk808.c b/drivers/mfd/rk808.c index 0324422..bd02150 100644 --- a/drivers/mfd/rk808.c +++ b/drivers/mfd/rk808.c @@ -29,10 +29,40 @@ struct rk808_reg_data { int value; }; +static bool rk808_is_volatile_reg(struct device *dev, unsigned int reg) +{ + /* + * Notes: + * - Technically the ROUND_30s bit makes RTC_CTRL_REG volatile, but + * we don't use that feature. It's better to cache. + * - It's unlikely we care that RK808_DEVCTRL_REG is volatile since + * bits are cleared in case when we shutoff anyway, but better safe. + */ + + switch (reg) { + case RK808_SECONDS_REG ... RK808_WEEKS_REG: + case RK808_RTC_STATUS_REG: + case RK808_VB_MON_REG: + case RK808_THERMAL_REG: + case RK808_DCDC_UV_STS_REG: + case RK808_LDO_UV_STS_REG: + case RK808_DCDC_PG_REG: + case RK808_LDO_PG_REG: + case RK808_DEVCTRL_REG: + case RK808_INT_STS_REG1: + case RK808_INT_STS_REG2: + return true; + } + + return false; +} + static const struct regmap_config rk808_regmap_config = { .reg_bits = 8, .val_bits = 8, .max_register = RK808_IO_POL_REG, + .cache_type = REGCACHE_RBTREE, + .volatile_reg = rk808_is_volatile_reg, }; static struct resource rtc_resources[] = {
Let's define the voltatile registers (those that can't be cached) and enable caching. The rk808 is accessed almost constantly with cpufreq so this is really nice. As measured by ftrace: before this change: cpu0_set_target() => ~2200us after this change: cpu0_set_target() => ~500us Signed-off-by: Doug Anderson <dianders@chromium.org> --- drivers/mfd/rk808.c | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+)