[llvm] [MachineLICM] Allow hoisting loads from invariant address (PR #70796)

[David Sherwood via llvm-commits]

https://github.com/david-arm commented:

Seems like a useful improvement to have @igogo-x86! I'd expect the IR version of LICM to hoist out most cases, but things like ExpandMemCmp, CodeGenPrepare or other pre-isel passes may introduce loads in loops that we do want to hoist out where possible.

https://github.com/llvm/llvm-project/pull/70796