[llvm] [CodeGen][StaticDataPartitioning]Place local-linkage global variables in hot or unlikely prefixed sections based on profile information (PR #125756)

[Snehasish Kumar via llvm-commits]

snehasish wrote:

> > Is there a reason for using .unlikely as a prefix rather than .cold? .cold seems more obvious as an opposite of .hot and we use it elsewhere in LLVM, most notably hot-cold splitting.
> 
> @petrhosek Thanks for the question. `unlikely` is used to follow the conventions of [functions](https://github.com/llvm/llvm-project/blob/69ccb1357fa6cf72063c737d06d6b29ffc465bee/llvm/lib/CodeGen/CodeGenPrepare.cpp#L598) in the codegenprepare pass. I'm open to name changes.
> 
> If identical section names (as string literals) are not de-duplicated in the relocatable object files (e.g., when compiled with `-data-sections=true -unique-section-names=false`), `cold` will give smaller section names than `unlikely` per data section where this matters.

I believe the `.cold` suffix is used for symbol names whereas the `.unlikely` suffix is for section names. It would be inconsistent to have a `.text.unlikely` section and a `.rodata.cold` section but I don't have a particularly strong opinion.

https://github.com/llvm/llvm-project/pull/125756