On 18 June 2013 13:54, Lukasz Majewski [off-list ref] wrote:

On Tue, 18 Jun 2013 08:42:13 +0200, Viresh Kumar wrote:

quoted

Its not about how long.. One cpu type can work longer with boost freq
compared to other.

What we probably need is:
- Enabled boost from sysfs if required (now below steps will come into
  picture)
- See how many cpus are running, if only one then start using boost
freqs

You are right here.

I'd like to propose following solution:
1. For acpi (where boost_enable come into play) - do not consider
number of active cpus (this is done in HW anyway)

2. For SW solution evaluate how many CPUs are running. If only one is
running then allow enabling boost from sysfs.

Looks fine.

But following situation is also possible: User enable boost when one
core is only running and then for some reason other core is woken up.
What shall be done then?
Shall we then disable boost immediately when cpufreq detects that
more than one core is running? Or leave this situation to be handled by
thermal subsystem?

Obviously disable boost ASAP. Every SoC might not have a thermal
framework glue to do it.

As a side note:
Logic proposed at point 2, is already implemented at LAB
(enable LAB only when one core is running and disable it when more
than one come into play).

Hmm.. So, eventually that will go away now :)

quoted

- Now thermal should be come into picture to save chip in case a
single cpu running at boost can burn it out.

I will extent v4 to embrace code which switches off boost at thermal.

Gud.