Question: is kanata's current CPU usage with mouse movement actions normal/optimized?

[flamingjupiter]

So, kanata is fairly light on CPU usage, at least on my system and with my configuration. Kanata's mouse movements, however, increases CPU usage quite a lot on my machine. With a lot of applications running, at idle, the CPU usage is 25% (approximate numbers). When doing light tasks e.g. note taking, CPU usage increases to 513%. When doing light tasks but using kanata's mouse movements, CPU usage spikes to 23~44%. Is this normal/expected?

System: Win 11; kanata wintercept; Ryzen 7 5700U (second last gen AMD mainstream non premium/power hungry laptop CPU); SATA SSD; 36GB RAM DDR4 3200MHz.

I guess it's relevant to say that I'm using Windows accessibility feature Filter Keys to increase key repeat rate much beyond Windows normal limits. IIRC I'm using it with a 150~200ms delay for triggering repeat of the key, and 10ms between repeats. Given that we configure mouse movement repeat manually and arbitrarily on Kanata's config file, I don't know if this can increase CPU usage.

[rszyma]

When doing light tasks but using kanata's mouse movements, CPU usage spikes to 23~44%. Is this normal/expected?

Is this percentage only kanata or whole system? If it's only kanata this doesn't seem normal to me.

Also, what interval do you use in movemouse actions? Does increasing the interval help?

I guess it's relevant to say that I'm using Windows accessibility feature Filter Keys to increase key repeat rate much beyond Windows normal limits. IIRC I'm using it with a 150~200ms delay for triggering repeat of the key, and 10ms between repeats.

Maybe try disabling the Filter Keys option and see if it helps?

[jtroo]

I took a look at kanata's CPU usage when scrolling, both with Interception and LLHOOK. Kanata itself barely uses any CPU, which is what I expected.

What I suspect is the issue is the application having to be redrawn due to the scroll event itself, and the content of the screen changing at a high rate.

[flamingjupiter]

After postponing it for a long time, I finally decided to give it a test and look and I guess @jtroo suspicion was right. Thanks again.