Hi Uwe,
> -----Original Message-----
> From: Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx>
> Sent: Thursday, March 14, 2024 10:33 PM
> Subject: Re: [PATCH v18 3/4] pwm: Add support for RZ/G2L GPT
>
> Hello,
>
> On Thu, Mar 14, 2024 at 06:10:50PM +0000, Biju Das wrote:
> > > On Tue, Feb 20, 2024 at 07:43:17PM +0000, Biju Das wrote:
> > > > +
> > > > +static inline u64 rzg2l_gpt_mul_u64_u64_div_u64(u64 a, u64 b, u64
> > > > +c) {
> > > > + u64 retval;
> > > > +
> > > > + if (a > b)
> > > > + retval = mul_u64_u64_div_u64(b, a, c);
> > > > + else
> > > > + retval = mul_u64_u64_div_u64(a, b, c);
> > >
> > > With
> > > https://lore.kernel.org/lkml/20240303092408.662449-2-u.kleine-koenig
> > > @pengutronix.de this function can be replaced by a direct call to
> > > mul_u64_u64_div_u64().
> > > I expect this patch to go into v6.9-rc1 as akpm picked it up before the merge window opened.
> >
> > Ok, I will hold next version until v6.9-rc1 as for-pwm-nexxt doesn't have this patch??
>
> I will rebase the stuff for the v6.10-rc1 merge window on v6.9-rc1, so (assuming my guess is right) you
> can profit of the improved
> mul_u64_u64_div_u64() call. (And even if the patch will go in later, we can live with the inexact
> configuration for that period.)
OK.
>
> > > > +static u32 rzg2l_gpt_calculate_pv_or_dc(u64 period_or_duty_cycle,
> > > > +u8
> > > > +prescale) {
> > > > + return min_t(u64, (period_or_duty_cycle + (1 << (2 * prescale)) - 1) >> (2 * prescale),
> > > > + U32_MAX);
> > >
> > > Can the addition overflow? Is the addition even right? This function
> > > is used in .apply() where it's usually right to round down.
> >
> > No, It won't overflow. The logic is proposed by you in v17 for
> > DIV64_U64_ROUND_UP and it is passing all tests with PWM_DEBUG=y.
>
> Then believe my former self, I didn't redo all the maths in this cycle.
>
> > > > + pm_runtime_enable(&pdev->dev);
> > > > + ret = pm_runtime_resume_and_get(&pdev->dev);
> > > > + if (ret)
> > > > + goto err_reset;
> > > > +
> > > > + ret = clk_rate_exclusive_get(rzg2l_gpt->clk);
> > >
> > > There is a devm variant of this function in the mean time.
> >
> > OK, currently for testing I picked it from next.
>
> For the next submission round make sure to properly use the --base parameter to not annoy the build
> bots. Or feel free to base your patch on next.
OK, I will rebase to next and send v19.
Note:
6.9-rc1 on for-nexxt is still missing a patch[1] for cpu performance on ARM64
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v6.8&id=98323e9d70172f1b46d1cadb20d6c54abf62870d
Cheers,
Biju
