| Message ID | f48f17a7-7f50-c5ef-cc8f-007d0cb302b0@gmail.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v2] pwm: meson: modify and simplify calculation in meson_pwm_get_state | expand |
On Mon, May 01, 2023 at 04:03:16PM +0200, Heiner Kallweit wrote: > I don't see a reason why we should treat the case lo < hi differently > and return 0 as period and duty_cycle. The current logic was added with > c375bcbaabdb ("pwm: meson: Read the full hardware state in > meson_pwm_get_state()"), Martin as original author doesn't remember why > it was implemented this way back then. > So let's handle it as normal use case and also remove the optimization > for lo == 0. I think the improved readability is worth it. > > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> > --- > v2: > - improve commit description > --- > drivers/pwm/pwm-meson.c | 14 ++------------ > 1 file changed, 2 insertions(+), 12 deletions(-) > > diff --git a/drivers/pwm/pwm-meson.c b/drivers/pwm/pwm-meson.c > index 5732300eb..3865538dd 100644 > --- a/drivers/pwm/pwm-meson.c > +++ b/drivers/pwm/pwm-meson.c > @@ -351,18 +351,8 @@ static int meson_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm, > channel->lo = FIELD_GET(PWM_LOW_MASK, value); > channel->hi = FIELD_GET(PWM_HIGH_MASK, value); > > - if (channel->lo == 0) { > - state->period = meson_pwm_cnt_to_ns(chip, pwm, channel->hi); > - state->duty_cycle = state->period; > - } else if (channel->lo >= channel->hi) { > - state->period = meson_pwm_cnt_to_ns(chip, pwm, > - channel->lo + channel->hi); > - state->duty_cycle = meson_pwm_cnt_to_ns(chip, pwm, > - channel->hi); > - } else { > - state->period = 0; > - state->duty_cycle = 0; > - } The last else branch is even wrong, isn't it? .apply() can for a greater than 50% relative duty cycle well have lo < hi, right? So this is not a mere optimisation but a fix?! > + state->period = meson_pwm_cnt_to_ns(chip, pwm, channel->lo + channel->hi); > + state->duty_cycle = meson_pwm_cnt_to_ns(chip, pwm, channel->hi); Note that meson_pwm_calc() has a similar construct that can be simplified in a similar way. All three variants have channel->pre_div = pre_div; and the last else branch is universal and can replace the others. Another issue I just spotted is that duty = state->duty_cycle is wrong for state->duty_cycle > UINT_MAX. (Ditto the assignment to period.) Making both duty and period u64 shoudl fix that. After that duty_cnt > 0xffff cannot happen as the core ensures that duty_cycle <= period. Having said that, the proposed change here is an improvement, so: Reviewed-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> I also suggest to add a Fixes line, i.e. Fixes: c375bcbaabdb ("pwm: meson: Read the full hardware state in meson_pwm_get_state()") Best regards Uwe
On 01.05.2023 20:07, Uwe Kleine-König wrote: > On Mon, May 01, 2023 at 04:03:16PM +0200, Heiner Kallweit wrote: >> I don't see a reason why we should treat the case lo < hi differently >> and return 0 as period and duty_cycle. The current logic was added with >> c375bcbaabdb ("pwm: meson: Read the full hardware state in >> meson_pwm_get_state()"), Martin as original author doesn't remember why >> it was implemented this way back then. >> So let's handle it as normal use case and also remove the optimization >> for lo == 0. I think the improved readability is worth it. >> >> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> >> --- >> v2: >> - improve commit description >> --- >> drivers/pwm/pwm-meson.c | 14 ++------------ >> 1 file changed, 2 insertions(+), 12 deletions(-) >> >> diff --git a/drivers/pwm/pwm-meson.c b/drivers/pwm/pwm-meson.c >> index 5732300eb..3865538dd 100644 >> --- a/drivers/pwm/pwm-meson.c >> +++ b/drivers/pwm/pwm-meson.c >> @@ -351,18 +351,8 @@ static int meson_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm, >> channel->lo = FIELD_GET(PWM_LOW_MASK, value); >> channel->hi = FIELD_GET(PWM_HIGH_MASK, value); >> >> - if (channel->lo == 0) { >> - state->period = meson_pwm_cnt_to_ns(chip, pwm, channel->hi); >> - state->duty_cycle = state->period; >> - } else if (channel->lo >= channel->hi) { >> - state->period = meson_pwm_cnt_to_ns(chip, pwm, >> - channel->lo + channel->hi); >> - state->duty_cycle = meson_pwm_cnt_to_ns(chip, pwm, >> - channel->hi); >> - } else { >> - state->period = 0; >> - state->duty_cycle = 0; >> - } > > The last else branch is even wrong, isn't it? .apply() can for a greater > than 50% relative duty cycle well have lo < hi, right? So this is not a > mere optimisation but a fix?! > I *think* too that it's wrong. However I have no test hw and I'm not aware of any problem caused by the current code. Therefore I was reluctant to make the patch a fix. >> + state->period = meson_pwm_cnt_to_ns(chip, pwm, channel->lo + channel->hi); >> + state->duty_cycle = meson_pwm_cnt_to_ns(chip, pwm, channel->hi); > > Note that meson_pwm_calc() has a similar construct that can be > simplified in a similar way. All three variants have > > channel->pre_div = pre_div; > The pre_div member will be gone anyway with a patch series that is in discussion currently ("make full use of CCF"). > and the last else branch is universal and can replace the others. > > Another issue I just spotted is that > > duty = state->duty_cycle > > is wrong for state->duty_cycle > UINT_MAX. (Ditto the assignment to > period.) Making both duty and period u64 shoudl fix that. After that > duty_cnt > 0xffff cannot happen as the core ensures that duty_cycle <= > period. > I saw that one too. It's something for a follow-up patch. > Having said that, the proposed change here is an improvement, so: > > Reviewed-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > I also suggest to add a Fixes line, i.e. > > Fixes: c375bcbaabdb ("pwm: meson: Read the full hardware state in meson_pwm_get_state()") > OK > Best regards > Uwe > Heiner
diff --git a/drivers/pwm/pwm-meson.c b/drivers/pwm/pwm-meson.c index 5732300eb..3865538dd 100644 --- a/drivers/pwm/pwm-meson.c +++ b/drivers/pwm/pwm-meson.c @@ -351,18 +351,8 @@ static int meson_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm, channel->lo = FIELD_GET(PWM_LOW_MASK, value); channel->hi = FIELD_GET(PWM_HIGH_MASK, value); - if (channel->lo == 0) { - state->period = meson_pwm_cnt_to_ns(chip, pwm, channel->hi); - state->duty_cycle = state->period; - } else if (channel->lo >= channel->hi) { - state->period = meson_pwm_cnt_to_ns(chip, pwm, - channel->lo + channel->hi); - state->duty_cycle = meson_pwm_cnt_to_ns(chip, pwm, - channel->hi); - } else { - state->period = 0; - state->duty_cycle = 0; - } + state->period = meson_pwm_cnt_to_ns(chip, pwm, channel->lo + channel->hi); + state->duty_cycle = meson_pwm_cnt_to_ns(chip, pwm, channel->hi); state->polarity = PWM_POLARITY_NORMAL;
I don't see a reason why we should treat the case lo < hi differently and return 0 as period and duty_cycle. The current logic was added with c375bcbaabdb ("pwm: meson: Read the full hardware state in meson_pwm_get_state()"), Martin as original author doesn't remember why it was implemented this way back then. So let's handle it as normal use case and also remove the optimization for lo == 0. I think the improved readability is worth it. Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> --- v2: - improve commit description --- drivers/pwm/pwm-meson.c | 14 ++------------ 1 file changed, 2 insertions(+), 12 deletions(-)