Increase RemoteDebuggerPeerTCP poll to 6.9ms to reduce CPU usage

[tavurth]

Closes #57256

And profile report:

[Faless]

Please use a descriptive commit name, e.g.:

Increase RemoteDebuggerPeerTCP poll interval to 10ms

Fix high CPU usage on MacOS.

Did you try with 1ms instead? Does it make much difference?

[tavurth]

@Faless I will check that now, and I will add a more descriptive name & force push

[tavurth]

1ms seems to result in an increase of around 6% CPU with sleep_usec set to 40,000.

I will try @ 4.1ms to fit with the 144hz refresh rate

[tavurth]

@Faless 4.1ms seems to be a good compromise as it leads to only 1% increase in CPU while updating at 144hz.

Thoughts?

[Faless]

@Faless 4.1ms seems to be a good compromise as it leads to only 1% increase in CPU while updating at 144hz.

Thoughts?

@tavurth that sounds good. Thanks!

[tavurth]

Updated commit message and force pushed

[Calinou]

4.1 ms doesn't match a 144 Hz refresh rate, it matches (roughly) a 240 Hz refresh rate. Nonetheless, if it works well, you can keep this value 🙂

[tavurth]

6.9ms seems to give a slight cpu reduction (on average dipping to 9%) but not a huge amount.

I will push that version so we can eak the maximum out of it as well as for correctness.

[tavurth]

Seems ready

[akien-mga]

Nice :)

[akien-mga]

Thanks!

[akien-mga]

Let's give this value a try, and if some users find it still too high, or too low for real-time updates, we could also consider making it configurable.

[tavurth]

@akien-mga configurable would be very nice!

I'm not sure if this needs to be cherry picked back to 3.x because as far as Calinou stated the system was rewritten for 4.x

[Faless]

Yeah, we don't need the cherry pick, the new debugger code (which uses threads) was not ported to 3.x