[andai]

What's the CPU usage? Like 2%?

I had a manual fs polling thing a while back. It was ugly (low time budget, didn't wanna mess with the native watchers), just scanned the whole thing once per second. It averaged out to like 0.3% CPU.

Not elegant, but acceptable for my purposes! (Small-ish directory, and "ping me within a second or two" was realtime enough for this use case.)

[kimixa]

If this stops the core being able to drop to a lower power state it can be whole multiples of power use on some devices.

Wake ups are death for mobile form factors, even if not really doing much work.

[andai]

This is a pretty good argument against the way we do operating systems now, right?

[kimixa]

Why? Most modern OSs are "tickless" - where there's no regular scheduling tick and it can sleep pretty much indefinitely if there's no work.

[booi]

i mean, technically this is once per millisecond, so this would happen 1000x more. In your case due to the kernel overhead you would likely not even be able to do it (300% CPU?).

Either way this does seem like a very large overhead due to the fact that there's just no other way to do it without a deeper kernel integration which might be outside the scope of what sqlite is trying to do.

[nine_k]

If the fs tree scanned once per second had 1000 files, it would be once per millisecond for a file.