[Enhancement] A more customizable overlay

[Kugumin]

It would be great to have an overlay similar to Scene9 with more complex metrics, such as power usage (watts/ampere) and a frame rate monitor that analyzes the real render pipeline. It would be more accurate to call it a "screen" update rate, not a refresh rate. It would also be nice to have the ability to pin the overlay to the top center of the display and make it unclickable to avoid it overlapping other things.

Here is an example of the Scene9 mini monitor (pinned on top center by default):

[Rve27]

Regarding the power usage, this is definitely doable. We already read the battery current and voltage under the hood, so I just need to add a bit of math to calculate the watts and expose it as a new toggle in the overlay settings.

For the frame rate, our current FPS monitor actually already analyzes the real render pipeline. Instead of just reading the static display refresh rate, it hooks into Shizuku to pull the averageFPS directly from dumpsys SurfaceFlinger.

Pinning to the top center and making it unclickable are great UX ideas too. Both should be pretty straightforward to implement on the Android side (just tweaking some WindowManager flags to make the touches pass through).

I'll leave this issue open and try to get these features added in an upcoming release.

[Kugumin]

For the frame rate, our current FPS monitor actually already analyzes the real render pipeline. Instead of just reading the static display refresh rate, it hooks into Shizuku to pull the averageFPS directly from dumpsys SurfaceFlinger.

On a static image, for example, Scene9 would show 0 FPS rendering (rendering stopped), while RVSystem Monitor shows 60 FPS in the same static image. Actually I am not quite sure how does actually calculating of such metrics works and which is correct.

Sadly, there's no legit and up-to-date source code of Scene9 to find out how it calculates FPS

[Rve27]

I looked deeper into how SurfaceFlinger works and you are absolutely right. Currently, the app extracts the averageFPS variable from SurfaceFlinger's --timestats. It turns out that Android often just reports the display's configured refresh rate (e.g., 60 or 120) for averageFPS when the screen is idling or static.

To get the "Real Render FPS" like Scene9 (where it drops to 0 on a static screen), we shouldn't use averageFPS. Instead, we need to track the totalFrames (the actual number of frame page flips) and calculate the delta every second. If the screen is static, the totalFrames count won't increase, which will correctly result in 0 FPS.

I will rewrite the FPS calculation logic to use the totalFrames delta approach. This should make our metrics match Scene9's accuracy for the real render pipeline.