[llvm] [LoopUnroll] Add `prepare-for-lto` pass option (PR #192153)

[Aiden Grossman via llvm-commits]

boomanaiden154 wrote:

> FYI: The LTO workloads I have access to are GPU workloads that do not use LLVM's LTO pipeline. So I'd be hesitant to extrapolate to other use cases just based on the data I'll provide.

I don't think the exact workload matters too much given this isn't proposing to change a default. But having some numbers wouldn't hurt to help motivate this.

Mostly out of my own curiosity, where are you planning on utilizing this pass option?

https://github.com/llvm/llvm-project/pull/192153