Re: [PATCH V2 3/4] clk: meson: c3: add support for the C3 SoC PLL clock

[Jerome Brunet]

On Tue 17 Oct 2023 at 22:39, Chuan Liu <chuan.liu@xxxxxxxxxxx> wrote:


>>>>> +
>>>>> +static struct clk_fixed_factor fclk_div2p5_div = {
>>>>> +     .mult = 2,
>>>>> +     .div = 5,
>>>>> +     .hw.init = &(struct clk_init_data){
>>>>> +             .name = "fclk_div2p5_div",
>>>>> +             .ops = &clk_fixed_factor_ops,
>>>>> +             .parent_hws = (const struct clk_hw *[]) {
>>>>> +                     &fixed_pll.hw
>>>>> +             },
>>>>> +             .num_parents = 1,
>>>>> +     },
>>>>> +};
>>>> This one is wrong if I follow the doc.
>>>> It is supposed to be fixed 8 divider taking it's source directly from
>>>> the DCO, skipping the OD post divider ... assuming the doc is up to date.
>>>>
>>> No, C3 SoC div2p5 is not skipping the OD post divider.
>> I a bit surprised there would be a frequency multiplier considering the
>> complexity of it, when skiping a divider is possible HW wise. Are you
>> sure ?
> This part confirms with our chip design engineer that fclk_div2p5 here is
> actually a clock output by a divider with decimal (divider factor is
> 2.5). The divider factor in clk-divider.c is int, so this description is
> used in the software. Or what do you suggest would be a better way to
> describe this type of divider with decimals?

It's alright. keep it that way then.
Consider fixing the doc maybe.

>>>>> +
>>>>> +static struct clk_regmap fclk_div2p5 = {
>>>>> +     .data = &(struct clk_regmap_gate_data){
>>>>> +             .offset = ANACTRL_FIXPLL_CTRL4,
>>>>> +             .bit_idx = 4,
>>>>> +     },
>>>>> +     .hw.init = &(struct clk_init_data) {
>>>>> +             .name = "fclk_div2p5",
>>>>> +             .ops = &clk_regmap_gate_ro_ops,
>>>>> +             .parent_hws = (const struct clk_hw *[]) {
>>>>> +                     &fclk_div2p5_div.hw
>>>>> +             },
>>>>> +             .num_parents = 1,
>>>>> +     },
>>>>> +};
>>>>> +