Multi threaded parallel interpolation

[HongLouyemeng]

can inverse_distance_to_grid use ThreadPoolExecutor to accelerate it？

[dopplershift]

With standard Python, the Global Interpreter Lock (GIL) will keep multiple threads of code from running concurrently within the Python interpreter, so this won't help you see any reduction in runtime. You'd need to use something like multiprocessing to get that through the use of multiple interpreters.

Starting with Python 3.13, there are now (experimental) builds of the Python interpreter available using "free-threading", which removes the GIL and permits multiple threads to run concurrently inside a single interpreter instance. Numpy is shipping wheels that support this build of Python, so in theory inverse_distance_to_grid() may work with this; this shouldn't be changing any global shared state, so I don't see any particular roadblocks on the MetPy side. That said, this is all experimental so I wouldn't be shocked if there were some problems.

If you give it a try, we'd love to hear how it works (though can't offer any help in getting it working at this time).

[HongLouyemeng]

thanks for your answer😄