Hi Conor,

Thank you for patiently giving me advice. I appreciate your help.

Not long ago, I said, "This patch needs to be accompanied by
modifications to the pwm_sifive_apply() function to make sense."

I recently reviewed the v3 version, and after discussing it with Emil,
there are several areas that require modification. I will provide the
necessary changes for each of them:

1. polarity check. (Suggestion from Uwe)
- if (state->polarity != PWM_POLARITY_INVERSED)
+ if (state->polarity != PWM_POLARITY_NORMAL)
2. avoid using old periodperiod, not state->period
- period = max(state->period, ddata->approx_period);
- frac = DIV64_U64_ROUND_CLOSEST(num, state->period);
+ frac = DIV64_U64_ROUND_CLOSEST(num, period);
3. add a conditional check can be added in the code to set
ddata->approx_period to state->period when state->period is smaller
than ddata->approx_period
  if (state->period != ddata->approx_period) {
  ...
+       if (state->period < ddata->approx_period) {
+               ddata->approx_period = state->period;
+       }
-       ddata->approx_period = state->period;
+       period = ddata->approx_period;

I will use 'unmatched' on my end to verify again. If there are any
other errors, feel free to point them out. Thank you.

Nylon Chen [off-list ref] 於 2023年8月4日 週五 上午9:42寫道：

Hi Conor,

Conor Dooley [off-list ref] 於 2023年8月3日 週四 下午5:44寫道：

quoted

Hey Nylon,

(I yoinked the reply to 1/1 to here, as it makes more sense in this
context)

quoted

On Thu, Aug 03, 2023 at 10:15:10AM +0100, Conor Dooley wrote:

quoted

On Thu, Aug 03, 2023 at 04:57:33PM +0800, Nylon Chen wrote:

quoted

According to the circuit diagram of User LEDs - RGB described in themanual hifive-unleashed-a00.pdf[0] and hifive-unmatched-schematics-v3.pdf[1].

The behavior of PWM is acitve-high.

Removed patches: 1
New patches: (none)

Links:
- [0]:  https://sifive.cdn.prismic.io/sifive/c52a8e32-05ce-4aaf-95c8-7bf8453f8698_hifive-unleashed-a00-schematics-1.pdf
- [1]:  https://sifive.cdn.prismic.io/sifive/6a06d6c0-6e66-49b5-8e9e-e68ce76f4192_hifive-unmatched-schematics-v3.pdf
- [2]:  https://sifive.cdn.prismic.io/sifive/1a82e600-1f93-4f41-b2d8-86ed8b16acba_fu740-c000-manual-v1p6.pdf

Changed in v4:
 - Remove previous updates to the PWM algorithm.

Why? I don't recall the conclusion on the previous version being that
that patch was not needed.

I apologize for forgetting about this update earlier. Just now,
I tried to pull rebase master and noticed that other developers seem
to have made some fixes to the algorithm. Upon closer inspection, I
found that they addressed the part we previously discussed with Emil
and Uwe, such as "first pwm_apply_state."

Therefore, my instinct tells me that they have already taken care of
the issues we discussed before.

I didn't see anything in linux-next that would solve this problem of
inversion. The last meaningful change is:
        commit 334c7b13d38321e47d1a51dba0bef9f4c403ec75
        Author:     Emil Renner Berthing [off-list ref]
        AuthorDate: Wed Nov 9 12:37:24 2022 +0100
        Commit:     Thierry Reding [off-list ref]
        CommitDate: Mon Jan 30 16:42:45 2023 +0100

            pwm: sifive: Always let the first pwm_apply_state succeed

which predates your v3 by quite a bit.

quoted

I will review the conflicting parts in the pwm-sifive.c code in my v4
version once again to ensure there are no omissions. If I find any, I
will submit v5 accordingly.

And if this patch is okay in isolation, please reply here explaining
which commit fixed the algorithm, so that I can pick it up.

This patch needs to be accompanied by modifications to the
pwm_sifive_apply() function to make sense.

I will double-check and review the previous discussions to ensure
that. Thank you for your response.

quoted

Thanks,
Conor.