[llvm] Extending LoopVersioningLICM to handle cases where loopbound is invariant (PR #192902)

[Pawan Nirpal via llvm-commits]

pawan-nirpal-031 wrote:

@nema-ashutosh @mk-srivastava Is there any interest in enablement of this pass by default? would you be open to collaborate on perhaps adding a vectorization aware cost model to this pass to minimize rt checks induced perf regressions? I have a few heuristics for AArch64 that prove to be useful in limiting such regressions. 

https://github.com/llvm/llvm-project/pull/192902