Next update maybe?
Here there is about 100 options about power governators:
But those available are here listed
[root@sfos ~]# cat /sys/devices/system/cpu/cpufreq/policy0/scaling_available_governors
conservative powersave interactive performance schedutil
[root@sfos ~]# cat /sys/devices/system/cpu/cpufreq/policy4/scaling_available_governors
conservative powersave interactive performance schedutil
It would be possible to keep 4 cpu in a conservative mode and 4 in schedutil or any combination. I am not aware if it is better to run at full throttle few CPUs or keep balanced. My gut feeling indicates the second but …
I am trying this one configuration:
for i in /sys/devices/system/cpu/cpu[0-3]/cpufreq/scaling_governor;
do echo "interactive" >$i; done
for i in /sys/devices/system/cpu/cpu[4-7]/cpufreq/scaling_governor;
do echo "conservative" >$i; done`
mcetool \
--set-power-saving-mode=enabled \
--set-low-power-mode=enabled \
--set-psm-threshold=100 \
--set-forced-psm=disabled \
--set-ps-on-demand=enabled
The 10 III doesn’t have the interactive governor, and its CPU’s big.LITTLE core architecture is 6:2 and not 4:4. Please stop messing up in yet another thread.
Different hardware have different configuration but most the commands are the same.
-
Xperia 10 III : Octa-core (2x2.0 GHz Kryo 560 Gold & 6x1.7 GHz Kryo 560 Silver)
-
Xperia 10 II : Octa-core (4x2.0 GHz Kryo 260 Gold & 4x1.8 GHz Kryo 260 Silver)
Therefore it makes sense to have two different governor policies in the two CPUs sets:
for i in /sys/devices/system/cpu/cpu[0-1]/cpufreq/scaling_governor;
do echo “ondemand” >$i; donefor i in /sys/devices/system/cpu/cpu[2-7]/cpufreq/scaling_governor;
do echo “conservative” >$i; done`
The differences for Xperia 10 III in the code has been highlighted, very little. Anyway, thanks for the suggestion and information. Because now, I can write a script that can works for both. 
But if you post in a thread strictly related to the 10 III then the information you post should apply to the 10 III, or else it can only confuse others. It is not a thread about governors or CPU cores in general, but a 10 III specific one, so the information should be valid for the 10 III.
I did not question that it is possible (or maybe even useful) to use two different governors for the big and little cores (in fact, that’s how the 10 III used to be configured by default in previous OS releases, with schedutil for the little cores and performance governor for the big ones), but only that your information was not applicable to the 10 III and thus potentially misleading to others.
Anyway, now that Jolla switched to using schedutil for 10 III’s all 8 cores, and that several people tested using lower minimum frequencies (e.g. 300 MHz) for all cores and also other governors (e.g. ondemand), none of which gave any noticeable power drain reduction, neither during use nor in standby, we already came to the conclusion that it is all in vain because the actual reason of 10 III’s high power drain lays elsewhere, most probably in the AOSP layer.
1 Like
Maybe, but there are area of improvements:
because as far as I see the power management policy is disrupted by deamons that should sleep in theory but instead they are alive for too long time in CPUs and there are significant amount of CPUs used by kworkers and rcu_preempt. I never had to do with a smartphone before now but it is a normal shape of activity in a Linux embedded system.
Deactivate and observe?
2 Likes
Putting the daemons to sleep is gratifying, indeed.
2 Likes
Never!
1 Like
These messages in the system log are NOT completely gone but mitigate that rarely appear again not solved but worked around, AFAIK. The problem is known from 4.4.0.68, at least - as you can read here.
[ +0.000108] OOM killer enabled.
[ +0.000001] Restarting tasks ... done.
[ +0.013891] PM: PM: suspend exit 2023-07-03 18:52:53.726781389 UTC
[ +0.000002] PM: suspend exit
[ +0.048486] ## mmc1: mmc_gpio_set_uim2_en: gpio=101 value=1
[ +0.057462] PM: PM: suspend entry 2023-07-03 18:52:53.832776345 UTC
[ +0.000005] PM: suspend entry (deep)
[ +0.000003] PM: Syncing filesystems ... done.
[ +0.003319] Freezing user space processes ...
[ +0.011845] PM: Wakeup pending, aborting suspend
[ +0.000066] Freezing of tasks aborted after 0.011 seconds
These below, continues to pollute the syslog but I am investigating about them as well.
[ +1.003158] binder: 2784:2784 transaction failed 29189/-22, size 32-0 line 3096
[ +1.001224] binder: 2784:2784 transaction failed 29189/-22, size 32-0 line 3096
It is time that SFOS starts to honoring the OS postfix (aka became a serious Operative System - well, there are also few kernel OOPs to take care). Unfortunately my new smartphone arrived this afternoon and I have to move to another bay… 
UPDATE
The transaction failed messages are from ofono (pid: 2784). If killed, it re-spawns and start again to make such a show. Instead, activating the Android Support, it calms down.
Like already said, this thread is about the 10iii and not the 10ii. My cpus are all sleeping fine and i normaly don’t use bluetooth…and still i got the battery drain.
I suggest to change the title including the Xperia 10 II because, after all the two smartphones are enough similar to go with very similar settings.
This some information that are good for both the smartphone.
Please, install the System Monitor and check the system log with dmesg -Hw. Much probably you will see that your CPUs are not sleeping well but constantly awaken by contest switch (processes that switch the CPUs around).
As soon as, I will have the chance to buy a cheap Xperia 10 III, I will equipped my personal lab with that model as well. If you like to fund raise that buying, the link is here.
If you have read the thread, you would know that there are no sleeping issues on the 10iii. I just checked the dmesg log and there is nothing suspicious. Everything works as expected.
If your 10ii suffers from battery drain(the 6% per hour you wrote someehere) there must be something wrong wth your phone. I never heard of such a huge drain on the 10ii.
Since the 10ii and the 10iii don’t share the same Sony binaries and AOSP base they are not comparable.
If you still think that your drain is a generel problem and not only an issue on your phone/ your SFOS installation please start your own thread about the drainage on the 10ii and stop polluting this thread with nonsense.
Ok, thanks for the information.
While using the phone a 6%/h does not seem a problem for me. It is a 16h of working, even with 12% is about 8h of working. I have no experience about Android smartphone that can last much longer than this.
This is a very good point about hardware support and brings me to the idea that having a Xperia 10 III in my personal lab is a necessity in order to support it. Thanks again.
1 Like
At first I was going to suggest including the Xperia 10 (and other Z80-based (1) designs) as well, since the those phones look almost identical, but then I had an epiphany.
We really should put everything in a single thread. Not only would this relieve us of updating thread titles, crossposting and being off topic would cease being possible at all. Unity, Mir (2) and understanding (3) will follow.
After all we’re here because of Sillfish Opera System and as you know Forum means city square in Latvian (4).
(1) That chip was designed by deflectors from the intel community in a story simliar to that of the 6502 A.D., but different
(2) …and, surely, the Snap package format and its related forum, Snapchat
(3) Please, just remember to sudo apt dist-upgrade regularly
(4) The citizens of ancient Vilnius who gathered in the square were not constrained by “titles” (5) set by anyone. We should follow their example and be equally free
(5) Titles were in fact bestowed upon the Caliph by God
Edit to add: Please disregard, my account was taken over by CheatGDP. Hitting Ctrl-C repeatedly saved you from more of the same.
4 Likes
@miau answered in a very consistent way:
AFAIK, this is the most significative difference between these two system apart 5G support and a different arrangement of the CPUs sets. At, least from a system admin point of view.
BTW, does the binaries package contain also the Linux kernel?
Differences include, but are not limited to:
- different SoC (Qualcomm SDM665 Snapdragon 665 11 nm vs. Qualcomm SM6350 Snapdragon 690 5G 8 nm, i.e. different CPU, different GPU, different modem, different technological process
- different AOSP binaries, different kernel version
- different battery (and probably the entire BMS)
- different other hardware components (display panel, camera module) and possibly also internal components like power circuitry, etc.
- software differences on the SFOS side, for example individually implemented VoLTE support for each of those models separately.
To recap, these two devices are actually almost completely different, with not much in common but the look. Any (or all) of the above can be responsible for different power consumption. These two devices are NOT comparable when it comes to power consumption. This thread is about the 10 III. If you want to talk about 10 II’s power consumption, please find (or start if it doesn’t exist) a separate thread about that model’s power drain.
2 Likes
Thanks @wetab73, I got your point.
However, from the PoV of a system administration those differences can bring different contributes and requires some adaptation in the shell scripting (eg. CPU arrangement 4:4 vs 6:2 or 2:6) but it almost the same old story:
- we need power saving templates (here).
When applying a pm template will be as easy as push a button or select an option and bush a button, then we can improve the templates. Improving both in generalization (policies) and in customization (smartphone models).
If this would not be possible, it means that the SFOS abstraction layer from the underlying hardware is not good enough (from the PoV of a system administrator).
Anyway, I am not interested in continue to write in this thread because I collected enough information from here.
What we need is Jolla working on this. But there isn’t even a notice, nothing, they just don’t care.
5 Likes