Best way to reduce flight controller computational load

[bnsgeyer (Bill Geyer)]

With the recent post regarding high computational loading causing aircraft twitch, I looked at the loading of my CUAV-V5 for my heli. It is flying Copter 4.1.5. It uses EKF3 for two IMUs and one harmonic notch attenuating 3 harmonics. the computational load runs around 65 to 78%. I looked at reducing the number of EKF instances and switch from EKF3 to EKF2. Only running one EKF3 instance reduced the load to around 50-55%. Running 2 instances of EKF2 reduces the load to 45-50%. So my question is whether I should only run 1 instance of EKF3 or 2 instance of EKF2?

[dotanb7 (Dotan)]

Hey bill, can you refer me to the original post you mentioned?
I am having other issues with what I believe is computational load on the MCU and I’d like to read anything that relates to this in Ardupilot.
Even some keywords to use in the search would be helpful.

(I searched, and still am searching.)

[bnsgeyer (Bill Geyer)]

@dotanb7 here is the link to the post I was referring to.

Internal errors 0x100000 I:921 flow_of_ctr Twitching motion in loiter during forward flight

This is a follow up post on: The internal error in the title appeared along with the symptoms described in the initial post. It has come to my attention that the same internal error appeared on the new rover release as well. I got the suggestion to create a new post to bring the attention to this internal error in particular. To read about the behavior in flight, pleas read the linked post. This is the initial log file where the issue first appeared (the first flight of 4.3.0) https://drive…

[xfacta (Shawn)]

Hi Bill,
I experienced twitches on Copter with CUAV V5 Nano and the CPU load did appear to be an issue, but in reality it was fixed properly by updating to 4.3.3 (and later now) since Leonard fixed some loop/timing issues.
This is a QuadX8 running DSHOT600.

I need to get that copter out again and fly some more, but I think the load issue went away too. I’ve not had trouble running EKF3 and all other default params, along with ESC/RPM driven harmonic notch filter.
These days I even leave all these in place, that I used to reduce to minimise load in some OCD fashion:
INS_LOG_BAT_MASK,1
INS_LOG_BAT_OPT,4
LOG_BITMASK,180222
and the little V5 Nano has to work hard to pay it’s way

EDIT: Fixed firmware version numbers
I should also have said that the load was reduced slightly by careful tailoring of params, but the twitching issue did slightly remain - so it wasnt just CPU load that was a propblem, but a symptom.
And then the real fix was Leonards updates.

[dotanb7 (Dotan)]

Well I’m deviating a bit from the original topic, our current issue is with logging and serial communication (at least this is what i think).
So maybe the computing load is not affecting flight - but other things that are slightly less critical.
See here

The parameters you mentioned should help a bit with load on the CPU?
Can you think of other parameters that might affect the CPU load?