Re: [PATCH 0/4] CPUFreq statistics retrieved by drivers
From: Sudeep Holla
Date: Thu Aug 06 2020 - 13:38:07 EST
On Wed, Aug 05, 2020 at 10:33:02AM -0700, Florian Fainelli wrote:
> On 8/5/2020 9:03 AM, Sudeep Holla wrote:
> > On Wed, Aug 05, 2020 at 06:34:36PM +0530, Viresh Kumar wrote:
> >> On 05-08-20, 12:04, Lukasz Luba wrote:
> >>> I know that Viresh is going to develop patches and improve these
> >>> cpufreq stats framework. Maybe he also had this 'aggregation' in mind.
> >>> I will leave it him.
> >> I am only going to look at cpufreq's view of stats independently from
> >> the firmware.
> > +1, I agree with that. Kernel must avoid any logic to aggregate or
> > interpret the data in a generic way. The userspace tools can manage that
> > especially if this tend to be platform specific.
> We can probably standardize on how to expose the firmware maintained
> statistics such that these tools do not have to widely vary from
> platform to platform, right?
Ofcourse. I just don't want any logic that interpret/analyse the stats
comparing with the in-kernel stats or otherwise.