Feature request: Pinned to core

[raddy]

Why not have the backend init allow for fixing the backend to a specific core? Is the thought that this is too architecture specific?

[RafaGago]

It would have been much better for the logger to just avoid starting a logging thread by itself.

Providing a function that runs the logging loop (now this code is on void thread() on "src/mal_log/backend.hpp") which returns sometimes ("maybe a timeout") would be better IMHO:

-It would make easy to run the logger inside a worker thread that is used for other purposes too.
-Many logger instances could run from the same thread.
-The user would be able to manage the thread as he requires, e.g. tying it to a core.

The change is not very hard to implement, it's just to tweak (initialization on first call) and expose the "thread()" function to the outside with a better name, e.g. "backend_run_some()".

As there is code depending on the current behavior, a flag on the init function or configuration to signal that the thread is managed externally would be required.

I can provide guidance to implement this and merge the changes back to master.

[funkyshoc]

FYI you can use sched_getaffinity/sched_setaffinity before/after the m_fe.init_backend so the thread that is spawned inherits your desired core. Easy enough solution that I implemented rather than lean on a configuration option within the code.