[PATCH v4 3/5] ARM: topology: Update cpu_power according to DT information
Shilimkar, Santosh
santosh.shilimkar at ti.com
Mon Jul 9 10:37:14 EDT 2012
On Mon, Jul 9, 2012 at 8:06 PM, Vincent Guittot
<vincent.guittot at linaro.org> wrote:
> On 9 July 2012 15:00, Shilimkar, Santosh <santosh.shilimkar at ti.com> wrote:
>> On Mon, Jul 9, 2012 at 6:02 PM, Vincent Guittot
>> <vincent.guittot at linaro.org> wrote:
>>> On 9 July 2012 12:55, Shilimkar, Santosh <santosh.shilimkar at ti.com> wrote:
>>>> Vincent,
>>>> On Mon, Jul 9, 2012 at 2:57 PM, Vincent Guittot
>>>> <vincent.guittot at linaro.org> wrote:
>>>>> Use cpu compatibility field and clock-frequency field of DT to
>>>>> estimate the capacity of each core of the system and to update
>>>>> the cpu_power field accordingly.
>>>>> This patch enables to put more running tasks on big cores than
>>>>> on LITTLE ones. But this patch doesn't ensure that long running
>>>>> tasks will run on big cores and short ones on LITTLE cores.
>>>>>
>>>>> Signed-off-by: Vincent Guittot <vincent.guittot at linaro.org>
>>>>> Reviewed-by: Namhyung Kim <namhyung at kernel.org>
>>>>> ---
>>>>> arch/arm/kernel/topology.c | 153 ++++++++++++++++++++++++++++++++++++++++++++
>>>>> 1 file changed, 153 insertions(+)
>>>>>
>>>> Sorry for not giving this comment on previous version but we should also
>>>> have a way to provide the big.LITTLE information without Device Tree.
>>>> May be a platform device/data.
>>>
>>> Hi Santosh,
>>>
>>> I had thought of adding such additional way to set cpu_power of
>>> big.LITTLE but my conclusion was
>>> -it's a new platform so it should come with DT
>>> -DT is already required by other patches linked to big.LITTLE
>>> (http://lists.infradead.org/pipermail/linux-arm-kernel/2012-January/080873.html)
>>> -There is no device that can be easily used to get such information at
>>> this early boot stage.
>>>
>> I see. Its new processor but it is just 1 one of the IP in an entire SOC.
>> As mentioned below I was talking about full SOC support including
>> all the driver subsystem with DT.
>>
>>>>
>>>> I know we are moving DT way, but remember apart from core kernel
>>>> infrastructure, to have a complete product build with DT means all the
>>>> drivers must be already supporting DT which is not the case with
>>>> many huge driver sub-systems like USB, display subsystem, Audio etc.
>>>>
>>>> Having that support would greatly help for the SOC's which have not yet
>>>> reached to stage where entire SOC is DT compliant and want to use
>>>> big.LITTLE infrastructure.
>>>
>>> Can't you support both type of devices on your platform ? You can move
>>> your device to DT mode when it is supported ?
>>>
>> That is what eventually people end up doing who don't have
>> DT ready for entire SOC. I was trying to ask whether at least some
>> method is proposed(need not be merged in mainline) to have
>> the big.LITTLE information parsing without DT.
>
> Ok, IIUC, you need a temporary methods, which doesn't need to be
> merged in mainline, to set the cpu_scale field ?
>
Yep.
Regards
Santosh
More information about the linux-arm-kernel
mailing list