[ntp:questions] PPS jitter in Linux

Chris Albertson albertson.chris at gmail.com
Tue Apr 3 01:33:58 UTC 2012


On Mon, Apr 2, 2012 at 6:25 PM, Mike S <mikes at flatsurface.com> wrote:
> On 4/2/2012 9:13 PM, Chris Albertson wrote:
>>
>> On Mon, Apr 2, 2012 at 4:28 PM, Mike S<mikes at flatsurface.com>  wrote:
>
> I've played around with different cpufreq setting, thinking it might be
>>>
>>> related to the processor speed during an IRQ, but that seems to have
>>> minimal
>>> impact (performance vs. conservative vs. ondemand).
>>>
>>
>> I think your CPU goes into some power saving mode when there is no
>> load and takes some number of micro seconds to wake up after an
>> interrupt.   Keeping a load on it prevents the sleep or power save
>> mode and hence the need to "wake up."
>
>
> Then explain why setting the cpufreq governor to "performance" (which locks
> it to the highest frequency (
> http://www.mjmwired.net/kernel/Documentation/cpu-freq/governors.txt ) has no
> significant effect.

The mode that your CPU falls into might not be controlled by the above
software.   For example can your CPU shut down a core completely?
How are interrupts assigned to cores?    Which exact model CPU are you
using?

You can look at the code for handling PPS in the linux kernel.  It is
so short and simple that any jitter has to be in the interrupt latency
not in the interrupt handler.



-- 

Chris Albertson
Redondo Beach, California


More information about the questions mailing list