Rethink Inlining?

[WillAyd]

ref #33886 (comment)

We use the inline keyword pretty liberally in our Cython extensions, with the thought that this probably improves performance. However, running GCC with -Winline as a flag produces a ton of warnings that we aren't actually inlining things. Some of these may even do more harm than good (just a hunch, I have no expertise with inline)

I've attached warnings from a clean build on master. Not sure all of these are necessarily even just from pandas but may be worth a look for anyone interested

warnings.log.zip

[jbrockmendel]

Maybe add to the agenda for next week's call? seems like "no strong opinion" is the modal response so far

[WillAyd]

Interestingly enough it looks like usage of the inline keyword was causing a regression in #57618