| Message ID | 20230130093229.27489-3-nylon.chen@sifive.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | Change PWM-controlled LED pin active mode and algorithm | expand |
| Context | Check | Description |
|---|---|---|
| conchuod/cover_letter | success | Series has a cover letter |
| conchuod/tree_selection | success | Guessed tree name to be for-next |
| conchuod/fixes_present | success | Fixes tag not required for -next series |
| conchuod/maintainers_pattern | success | MAINTAINERS pattern errors before the patch: 13 and now 13 |
| conchuod/verify_signedoff | success | Signed-off-by tag matches author and committer |
| conchuod/kdoc | success | Errors and warnings before: 0 this patch: 0 |
| conchuod/module_param | success | Was 0 now: 0 |
| conchuod/build_rv64_gcc_allmodconfig | success | Errors and warnings before: 0 this patch: 0 |
| conchuod/alphanumeric_selects | success | Out of order selects before the patch: 57 and now 57 |
| conchuod/build_rv32_defconfig | success | Build OK |
| conchuod/dtb_warn_rv64 | success | Errors and warnings before: 2 this patch: 2 |
| conchuod/header_inline | success | No static functions without inline keyword in header files |
| conchuod/checkpatch | success | total: 0 errors, 0 warnings, 0 checks, 7 lines checked |
| conchuod/source_inline | success | Was 0 now: 0 |
| conchuod/build_rv64_nommu_k210_defconfig | success | Build OK |
| conchuod/verify_fixes | success | No Fixes tag |
| conchuod/build_rv64_nommu_virt_defconfig | success | Build OK |
Hi Nylon, On Mon, Jan 30, 2023 at 10:32 AM Nylon Chen <nylon.chen@sifive.com> wrote: > The `frac` variable represents the pulse inactive time, and the result of > this algorithm is the pulse active time. Therefore, we must reverse the > result. > > The reference is SiFive FU740-C000 Manual[0]. > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> Thanks for your patch! > --- a/drivers/pwm/pwm-sifive.c > +++ b/drivers/pwm/pwm-sifive.c > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > /* The hardware cannot generate a 100% duty cycle */ > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; Shouldn't the inversion be done before the hardware limitation fixup? > > mutex_lock(&ddata->lock); > if (state->period != ddata->approx_period) { Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > The `frac` variable represents the pulse inactive time, and the result of > this algorithm is the pulse active time. Therefore, we must reverse the > result. > > The reference is SiFive FU740-C000 Manual[0]. > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > --- > drivers/pwm/pwm-sifive.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > index 62b6acc6373d..a5eda165d071 100644 > --- a/drivers/pwm/pwm-sifive.c > +++ b/drivers/pwm/pwm-sifive.c > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > /* The hardware cannot generate a 100% duty cycle */ > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; The same problem exists in pwm_sifive_get_state(), doesn't it? As fixing this is an interruptive change anyhow, this is the opportunity to align the driver to the rules tested by PWM_DEBUG. The problems I see in the driver (only checked quickly, so I might be wrong): - state->period != ddata->approx_period isn't necessarily a problem. If state->period > ddata->real_period that's fine and the driver should continue - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); is wrong for two reasons: it should round down and use the real period. Best regards Uwe
Hi Geert, Thanks for your reply. Geert Uytterhoeven <geert@linux-m68k.org> 於 2023年1月30日 週一 下午5:53寫道： > > Hi Nylon, > > On Mon, Jan 30, 2023 at 10:32 AM Nylon Chen <nylon.chen@sifive.com> wrote: > > The `frac` variable represents the pulse inactive time, and the result of > > this algorithm is the pulse active time. Therefore, we must reverse the > > result. > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > Thanks for your patch! > > > --- a/drivers/pwm/pwm-sifive.c > > +++ b/drivers/pwm/pwm-sifive.c > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > /* The hardware cannot generate a 100% duty cycle */ > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > Shouldn't the inversion be done before the hardware limitation fixup? I think your inference is correct, I will use it. thanks a lot. > > > > > mutex_lock(&ddata->lock); > > if (state->period != ddata->approx_period) { > > Gr{oetje,eeting}s, > > Geert > > -- > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org > > In personal conversations with technical people, I call myself a hacker. But > when I'm talking to journalists I just say "programmer" or something like that. > -- Linus Torvalds
Hi Uwe, Thanks for your reply. Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > The `frac` variable represents the pulse inactive time, and the result of > > this algorithm is the pulse active time. Therefore, we must reverse the > > result. > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > --- > > drivers/pwm/pwm-sifive.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > index 62b6acc6373d..a5eda165d071 100644 > > --- a/drivers/pwm/pwm-sifive.c > > +++ b/drivers/pwm/pwm-sifive.c > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > /* The hardware cannot generate a 100% duty cycle */ > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > As fixing this is an interruptive change anyhow, this is the opportunity > to align the driver to the rules tested by PWM_DEBUG. > > The problems I see in the driver (only checked quickly, so I might be > wrong): > > - state->period != ddata->approx_period isn't necessarily a problem. If > state->period > ddata->real_period that's fine and the driver should > continue > > - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > is wrong for two reasons: > it should round down and use the real period. > I need a little time to clarify your assumptions. If possible, I will make similar changes. e.g. rounddown(num, state->period); if (state->period < ddata->approx_period) ... thanks a lot. > Best regards > Uwe > > -- > Pengutronix e.K. | Uwe Kleine-König | > Industrial Linux Solutions | https://www.pengutronix.de/ |
Hi Uwe, Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > The `frac` variable represents the pulse inactive time, and the result of > > this algorithm is the pulse active time. Therefore, we must reverse the > > result. > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > --- > > drivers/pwm/pwm-sifive.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > index 62b6acc6373d..a5eda165d071 100644 > > --- a/drivers/pwm/pwm-sifive.c > > +++ b/drivers/pwm/pwm-sifive.c > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > /* The hardware cannot generate a 100% duty cycle */ > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > As fixing this is an interruptive change anyhow, this is the opportunity > to align the driver to the rules tested by PWM_DEBUG. > > The problems I see in the driver (only checked quickly, so I might be > wrong): > > - state->period != ddata->approx_period isn't necessarily a problem. If > state->period > ddata->real_period that's fine and the driver should > continue > > - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > is wrong for two reasons: > it should round down and use the real period. are you mean state->period is a redundancy variable so we can use ddata->real_period directly? it seems reasonable, but I don't get your point, why do we need to change the algorithm to DIV_ROUND_DOWN_ULL() and change the if-else condition. frac = DIV_ROUND_DOWN_ULL(num, ddata->real_period); if (state->period < ddata->approx_period) { ... } > > Best regards > Uwe > > -- > Pengutronix e.K. | Uwe Kleine-König | > Industrial Linux Solutions | https://www.pengutronix.de/ |
Hi Uwe, Nylon Chen <nylon.chen@sifive.com> 於 2023年2月3日 週五 下午4:06寫道： > > Hi Uwe, > > Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > > The `frac` variable represents the pulse inactive time, and the result of > > > this algorithm is the pulse active time. Therefore, we must reverse the > > > result. > > > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > > --- > > > drivers/pwm/pwm-sifive.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > > index 62b6acc6373d..a5eda165d071 100644 > > > --- a/drivers/pwm/pwm-sifive.c > > > +++ b/drivers/pwm/pwm-sifive.c > > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > > /* The hardware cannot generate a 100% duty cycle */ > > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > > > As fixing this is an interruptive change anyhow, this is the opportunity > > to align the driver to the rules tested by PWM_DEBUG. > > > > The problems I see in the driver (only checked quickly, so I might be > > wrong): > > > > > - state->period != ddata->approx_period isn't necessarily a problem. If > > state->period > ddata->real_period that's fine and the driver should > > continue > > > > - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > is wrong for two reasons: > > it should round down and use the real period. I have some results from my observations regarding the questions you raised. I don't know if what we are thinking is the same thing. If my assumptions are different from yours, please let me know. Thanks. > are you mean state->period is a redundancy variable so we can use > ddata->real_period directly? > > it seems reasonable, but I don't get your point, why do we need to > change the algorithm to DIV_ROUND_DOWN_ULL() and change the if-else > condition. > > frac = DIV_ROUND_DOWN_ULL(num, ddata->real_period); > if (state->period < ddata->approx_period) { > ... > } > > > > > Best regards > > Uwe > > > > -- > > Pengutronix e.K. | Uwe Kleine-König | > > Industrial Linux Solutions | https://www.pengutronix.de/ |
Hello Nylon, On Wed, Feb 01, 2023 at 04:56:42PM +0800, Nylon Chen wrote: > Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > > The `frac` variable represents the pulse inactive time, and the result of > > > this algorithm is the pulse active time. Therefore, we must reverse the > > > result. > > > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > > --- > > > drivers/pwm/pwm-sifive.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > > index 62b6acc6373d..a5eda165d071 100644 > > > --- a/drivers/pwm/pwm-sifive.c > > > +++ b/drivers/pwm/pwm-sifive.c > > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > > /* The hardware cannot generate a 100% duty cycle */ > > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > > > As fixing this is an interruptive change anyhow, this is the opportunity > > to align the driver to the rules tested by PWM_DEBUG. > > > > The problems I see in the driver (only checked quickly, so I might be > > wrong): > > > > - state->period != ddata->approx_period isn't necessarily a problem. If > > state->period > ddata->real_period that's fine and the driver should > > continue > > > > - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > is wrong for two reasons: > > it should round down and use the real period. > > > I need a little time to clarify your assumptions. If possible, I will > make similar changes. > > e.g. > rounddown(num, state->period); > if (state->period < ddata->approx_period) > ... the idea is that for a given request apply should do the following to select the hardware setting: - Check polarity, if the hardware doesn't support it, return -EINVAL. (A period always starts with the active phase for the duration of duty_cycle. For normal polarity active = high.) - Pick the biggest period length possible that is not bigger than the requested period. - For the picked period, select the biggest duty_cycle possible that is not bigger than the requested duty_cycle. Then if possible switch to the selected setting in an atomic step. Does this clearify your doubts? Best regards Uwe
Hi Uwe Thanks for your reply. Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年3月1日 週三 下午5:21寫道： > > Hello Nylon, > > On Wed, Feb 01, 2023 at 04:56:42PM +0800, Nylon Chen wrote: > > Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > > > The `frac` variable represents the pulse inactive time, and the result of > > > > this algorithm is the pulse active time. Therefore, we must reverse the > > > > result. > > > > > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > > > --- > > > > drivers/pwm/pwm-sifive.c | 1 + > > > > 1 file changed, 1 insertion(+) > > > > > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > > > index 62b6acc6373d..a5eda165d071 100644 > > > > --- a/drivers/pwm/pwm-sifive.c > > > > +++ b/drivers/pwm/pwm-sifive.c > > > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > > > /* The hardware cannot generate a 100% duty cycle */ > > > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > > > > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > > > > > As fixing this is an interruptive change anyhow, this is the opportunity > > > to align the driver to the rules tested by PWM_DEBUG. > > > > > > The problems I see in the driver (only checked quickly, so I might be > > > wrong): > > > > > > - state->period != ddata->approx_period isn't necessarily a problem. If > > > state->period > ddata->real_period that's fine and the driver should > > > continue > > > > > > - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > > is wrong for two reasons: > > > it should round down and use the real period. > > > > > I need a little time to clarify your assumptions. If possible, I will > > make similar changes. > > > > e.g. > > rounddown(num, state->period); > > if (state->period < ddata->approx_period) > > ... > > the idea is that for a given request apply should do the following to > select the hardware setting: > > - Check polarity, if the hardware doesn't support it, return -EINVAL. > (A period always starts with the active phase for the duration of > duty_cycle. For normal polarity active = high.) > - Pick the biggest period length possible that is not bigger than the > requested period. > - For the picked period, select the biggest duty_cycle possible that is > not bigger than the requested duty_cycle. > > Then if possible switch to the selected setting in an atomic step. > > Does this clearify your doubts? I need a little time to clarify your assumptions. Thanks again. > > Best regards > Uwe > > -- > Pengutronix e.K. | Uwe Kleine-König | > Industrial Linux Solutions | https://www.pengutronix.de/ |
Hi Uwe, Sorry it's so long ago. I have completed the implementation of the new version, but there is one thing about this letter that I still don't quite understand. Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > The `frac` variable represents the pulse inactive time, and the result of > > this algorithm is the pulse active time. Therefore, we must reverse the > > result. > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > --- > > drivers/pwm/pwm-sifive.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > index 62b6acc6373d..a5eda165d071 100644 > > --- a/drivers/pwm/pwm-sifive.c > > +++ b/drivers/pwm/pwm-sifive.c > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > /* The hardware cannot generate a 100% duty cycle */ > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > As fixing this is an interruptive change anyhow, this is the opportunity > to align the driver to the rules tested by PWM_DEBUG. > > The problems I see in the driver (only checked quickly, so I might be > wrong): > > - state->period != ddata->approx_period isn't necessarily a problem. If > state->period > ddata->real_period that's fine and the driver should > continue > I still don’t quite understand the description of this paragraph. state->period != ddate->approx_period seems to be used to compare the results of the previous and next two times. I'm unsure what to do if I replace the conditional expression with something else. In addition, I don't understand the meaning of this. "if state->period > ddata->real_period that's fine, and the driver should continue" At present, my new version of the implementation has passed the test of the pwm_apply_state_debug() function. Would you suggest I send the new implementation version before continuing the discussion? Thank you again for everything you’ve done. > - frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > is wrong for two reasons: > it should round down and use the real period. > > Best regards > Uwe > > -- > Pengutronix e.K. | Uwe Kleine-König | > Industrial Linux Solutions | https://www.pengutronix.de/ |
Hello, On Fri, Sep 08, 2023 at 06:41:00PM +0800, Nylon Chen wrote: > Sorry it's so long ago. > > I have completed the implementation of the new version, but there is > one thing about this letter that I still don't quite understand. > > Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > > The `frac` variable represents the pulse inactive time, and the result of > > > this algorithm is the pulse active time. Therefore, we must reverse the > > > result. > > > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > > --- > > > drivers/pwm/pwm-sifive.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > > index 62b6acc6373d..a5eda165d071 100644 > > > --- a/drivers/pwm/pwm-sifive.c > > > +++ b/drivers/pwm/pwm-sifive.c > > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > > /* The hardware cannot generate a 100% duty cycle */ > > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > > > As fixing this is an interruptive change anyhow, this is the opportunity > > to align the driver to the rules tested by PWM_DEBUG. > > > > The problems I see in the driver (only checked quickly, so I might be > > wrong): > > > > - state->period != ddata->approx_period isn't necessarily a problem. If > > state->period > ddata->real_period that's fine and the driver should > > continue > > I still don’t quite understand the description of this paragraph. > > state->period != ddate->approx_period seems to be used to compare the > results of the previous and next two times. There are two things to consider: - usually the hardware doesn't support all requestable states because the hardware's quantum is > 1 ns. That is, it might for example support periods in the form (160 ns * i / 3) for i in 1 .. 1023. If this hardware runs with i = 500 (that is period ~= 26666.66 ns) because the first channel is configured to run with period = 26667, and .request is called for the 2nd channel with .period = 26700 ns, there is no need to refuse that, because 26666.66 is the best possible approximation for 26700 ns anyhow. - .apply is supposed to implement the highest possible period that isn't bigger than the requested period. So in the above case even if the hardware runs at 26666.66 ns without the possibility to change that, a request to configure for period = 30000 ns could succeed (and keep 26666.66 ns). > Would you suggest I send the new implementation version before > continuing the discussion? Note that the above implements the optimal behaviour for a driver. (For some definition of "optimal" that admittedly also yields strange behaviour at times. The reasoning for this to the be thing to implement is, that's the corner cases are easier to implement, idempotency is possible and it's easier to work with than rounding to the nearest possible value.) While I'd like to see the sifive driver to implement this optimal behaviour, it's not mandatory that you convert the driver to that behaviour. Just make sure to not make it worse. So to answer your question: If you understood what I wrote above and are motivated to improve the driver, it would be great to do that before the next review round. Best regards Uwe
Hi Uwe I'm glad the example was helpful to you. I will double-check and verify my implementation once again. Thank you for your assistance. Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年9月8日 週五 下午10:50寫道： > > Hello, > > On Fri, Sep 08, 2023 at 06:41:00PM +0800, Nylon Chen wrote: > > Sorry it's so long ago. > > > > I have completed the implementation of the new version, but there is > > one thing about this letter that I still don't quite understand. > > > > Uwe Kleine-König <u.kleine-koenig@pengutronix.de> 於 2023年1月30日 週一 下午6:17寫道： > > > > > > On Mon, Jan 30, 2023 at 05:32:29PM +0800, Nylon Chen wrote: > > > > The `frac` variable represents the pulse inactive time, and the result of > > > > this algorithm is the pulse active time. Therefore, we must reverse the > > > > result. > > > > > > > > The reference is SiFive FU740-C000 Manual[0]. > > > > > > > > [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf > > > > > > > > Signed-off-by: Nylon Chen <nylon.chen@sifive.com> > > > > --- > > > > drivers/pwm/pwm-sifive.c | 1 + > > > > 1 file changed, 1 insertion(+) > > > > > > > > diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c > > > > index 62b6acc6373d..a5eda165d071 100644 > > > > --- a/drivers/pwm/pwm-sifive.c > > > > +++ b/drivers/pwm/pwm-sifive.c > > > > @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, > > > > frac = DIV64_U64_ROUND_CLOSEST(num, state->period); > > > > /* The hardware cannot generate a 100% duty cycle */ > > > > frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); > > > > + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; > > > > > > The same problem exists in pwm_sifive_get_state(), doesn't it? > > > > > > As fixing this is an interruptive change anyhow, this is the opportunity > > > to align the driver to the rules tested by PWM_DEBUG. > > > > > > The problems I see in the driver (only checked quickly, so I might be > > > wrong): > > > > > > - state->period != ddata->approx_period isn't necessarily a problem. If > > > state->period > ddata->real_period that's fine and the driver should > > > continue > > > > I still don’t quite understand the description of this paragraph. > > > > state->period != ddate->approx_period seems to be used to compare the > > results of the previous and next two times. > > There are two things to consider: > > - usually the hardware doesn't support all requestable states because > the hardware's quantum is > 1 ns. That is, it might for example > support periods in the form (160 ns * i / 3) for i in 1 .. 1023. > > If this hardware runs with i = 500 (that is period ~= 26666.66 > ns) because the first channel is configured to run with period = > 26667, and .request is called for the 2nd channel with .period = > 26700 ns, there is no need to refuse that, because 26666.66 is the > best possible approximation for 26700 ns anyhow. > > - .apply is supposed to implement the highest possible period that > isn't bigger than the requested period. So in the above case even if > the hardware runs at 26666.66 ns without the possibility to change > that, a request to configure for period = 30000 ns could succeed (and > keep 26666.66 ns). > > > Would you suggest I send the new implementation version before > > continuing the discussion? > > Note that the above implements the optimal behaviour for a driver. > (For some definition of "optimal" that admittedly also yields strange > behaviour at times. The reasoning for this to the be thing to implement > is, that's the corner cases are easier to implement, idempotency is > possible and it's easier to work with than rounding to the nearest > possible value.) > > While I'd like to see the sifive driver to implement this optimal > behaviour, it's not mandatory that you convert the driver to that > behaviour. Just make sure to not make it worse. > > So to answer your question: If you understood what I wrote above and are > motivated to improve the driver, it would be great to do that before the > next review round. > > Best regards > Uwe > > -- > Pengutronix e.K. | Uwe Kleine-König | > Industrial Linux Solutions | https://www.pengutronix.de/ |
diff --git a/drivers/pwm/pwm-sifive.c b/drivers/pwm/pwm-sifive.c index 62b6acc6373d..a5eda165d071 100644 --- a/drivers/pwm/pwm-sifive.c +++ b/drivers/pwm/pwm-sifive.c @@ -158,6 +158,7 @@ static int pwm_sifive_apply(struct pwm_chip *chip, struct pwm_device *pwm, frac = DIV64_U64_ROUND_CLOSEST(num, state->period); /* The hardware cannot generate a 100% duty cycle */ frac = min(frac, (1U << PWM_SIFIVE_CMPWIDTH) - 1); + frac = (1U << PWM_SIFIVE_CMPWIDTH) - 1 - frac; mutex_lock(&ddata->lock); if (state->period != ddata->approx_period) {
The `frac` variable represents the pulse inactive time, and the result of this algorithm is the pulse active time. Therefore, we must reverse the result. The reference is SiFive FU740-C000 Manual[0]. [0]: https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf Signed-off-by: Nylon Chen <nylon.chen@sifive.com> --- drivers/pwm/pwm-sifive.c | 1 + 1 file changed, 1 insertion(+)