RPi and powerd, was: Re: RPI4 clock speeds and serial port ( temperatures idle and -j4 buildworld buildkernel )

[Mark Millard]

[This did not get sent when I wrote it. So sending way late.]

On 2021-Mar-21, at 11:13, bob prohaska <fbsd at www.zefox.net> wrote:

> On Sat, Mar 20, 2021 at 09:20:36PM -0700, Mark Millard wrote:
>> 
>> You are getting the vast majority of the time
>> benefit from powerd use compared to my build
>> context (based on the world build times). Avoiding
>> building to allow compiling non-arm could cut
>> your time noticeably, possibly around 30 minutes
>> for world.
>> 
> 
> Half an hour I can wait 8-) More likely I'll botch the changes.
> 
> After trying powerd on the Pi4 with success it was applied
> to Pi3 and Pi2. Powerd seemed to work well on the Pi3, but


> the Pi2 v1.1 stalled consistently in the building libraries
> phase. The stall was unresponsive to the debugger escape.
> After power-cycling and disabling powerd, the Pi2 picked
> up where it left off. Not sure if this is a property of 
> powerd or an artifact of my setup on the Pi2, being
> 14.0-CURRENT FreeBSD 14.0-CURRENT #13 main-f5e930b369:
> executing
> make -j4 -DWITH_META_MODE  buildworld 

It might be that if you set the speed in /etc/sysctl.conf
instead of using powerd (so the arm running at a constant
rate), that the problem would repeat on the RPi2.

Were you monitoring temperature on the RPi2? Voltage?

In my testing the RPi4's are the first RPi*'s that seemed
to operate well somwhat outside the default configuration's
range. (Just the power connection distinctions tend to
suggest such a status.) The RPi4's have fair sized margin.
There is margin beyond where I operate them from what I've
seen. (Or I'd not be operating them at the settings that
I'm using.)

> There was an effect from powerd on the Pi2, but only
> 600MHz > 900MHz. The Pi3 went from 600 to 1200, the
> Pi4 600 to 1500.

That is consistent with the table entries
for arm_freq and arm_freq_min defaults in:

https://www.raspberrypi.org/documentation/configuration/config-txt/overclocking.md


> There was no obvious serial console
> disruption on any of the three machines. 
> 
> 
> At least on the Pi4 and Pi3 use of -DWITH_META_MODE seems
> notably faster than before, much more like use of -DNO_CLEAN.
> Quite a bit has changed since my earlier gripes about needless
> recompiles so it's hard to fairly credit the improvement. 
> 
>> 
>>> The RPi4 support for powerd (cpufreq) was added
>>> on 2020-Jun-20 but my RPi4B activity started
>>> well before then. That is why I did not go down
>>> the powerd path in how I configured things. It
>>> will be interesting to compare to your time.
>>> 
> 
> I always thought of powerd as a complication inflicted to
> conserve battery life. I had no idea it could be used for
> performance enhancement. Thank you, Klaus!





===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)