On 2020-09-22 6:30 a.m., Hans de Goede wrote:
Hi,
<snip>
The big question is, do we want to expose that even though the user
configured
a performance-profile of high, the user is only getting medium atm
because of
reasons?
Note I say "because of reasons" specifically, because things become even
more
complicated if we want to spell out the reasons in the sysfs interface too.
I mean high will give different results even when in desk mode, depending
on if there is a cloth on the desk (bad) or if the table is a metal
picknick
table full of round holes to drain the rain (allowing more airflow to the
bottom of the laptop) not to mention that the ambient temperature in which
the laptop is used can probably vary from 15 to 35 degrees celcius.
IOW there can be many factors why high may not really lead to high turbo
clocks; or why it leads to higher turbo clocks then normally expected...
I still have the feeling that it would be best to drop the UI requirement
to show being in a degraded performance mode, because the performance
with modern laptops is just very variable and dependent on many factors.
If we drop that UI requirement; then there also is no need to advertise
configured vs actual performance profile in the sysfs interface.
Users who really want to know what is going on will get much more
detailed and useful information when using something like turbostat
(or a UI for that) anyways.
Regards,
Hans
Thought about this some and I'm in agreement. My vote is to keep the
first version simple and see where we go from there based on user feedback.
I'm going to make the lapmode information available via debugfs for
those users who do want to see it anyway - and this exercise is really
about the user space controller so I can see how fitting in all these
extra pieces just makes it awkward.
I had some other thoughts but I think they tag on better to Elia's email
so I'll save those for there
Mark
