[PATCH] clk: samsung: add set_rate and round_rate callbacks for pll45xx
Yadwinder Singh Brar
yadi.brar01 at gmail.com
Fri Jul 12 01:11:13 EDT 2013
On Thu, Jul 11, 2013 at 8:54 PM, Thomas Abraham
<thomas.abraham at linaro.org> wrote:
> Hi Yadwinder,
>
> Thanks for your comments.
>
> On 11 July 2013 16:14, Yadwinder Singh Brar <yadi.brar01 at gmail.com> wrote:
>> Hi Thomas,
>>
>> On Thu, Jul 11, 2013 at 2:57 PM, Thomas Abraham
>> <thomas.abraham at linaro.org> wrote:
>>> Add support for set_rate and round_rate callbacks for pll45xx pll. This allows
>>> configuring pll45xx to generate a desired clock output.
>>>
>>> Signed-off-by: Thomas Abraham <thomas.abraham at linaro.org>
>>> ---
>>>
>>> The set_rate and round_rate callbacks in this patch for pll45xx are handled
>>> slightly differently from the way it is done in the (not yet merged) patch
>>> series from Yadwinder
>>> (http://www.mail-archive.com/linux-samsung-soc@vger.kernel.org/msg19540.html)
>>>
>>> In this patch, the pll lookup table is kept as part of the pll configuration
>>> code itself, since the pll is just a hardware block which takes an input
>>> frequency and configuration values to genertate a output clock frequency.
>>> Given a set of inputs, a pll of a given type will generate the same output
>>> regardless of the SoC it is used in. So instead of supplying the pll lookup
>>
>> Tomasz also raised similar point earlier in discussion on that series.
>> That time a question which remained unanswered for which we were waiting
>> (as requested from hardware guys also) was that :
>> Whether we need to stick to recommended values only or we can drive a
>> particular output frequency with a given input frequency with any possible
>> set of (P, M, S) values ?
>
> The default lookup table could list all the possible target frequency
> and its associated PMS values. Having a bigger list of PMS lookup
same target frequency can have different associated PMS values recommended
in manual. So the question was in that case whether to stick to
recommended values.
> values should not matter since the CPUFreq driver would have a smaller
> subset of frequency which can be selected on a particular SoC.
>
>>
>> Theoretically it seems possible but from manual its seems to prefer
>> recommended values.
>> Also I am not sure whether same values are always recommended in user manuals
>> for different SoCs using same PLL type.
>>
> [...]
Regards,
Yadwinder
More information about the linux-arm-kernel
mailing list