[llvm] [SimpleLoopUnswitch] Record loops from unswitching non-trivial conditions (PR #141121)

[Björn Pettersson via llvm-commits]

bjope wrote:

For the record, this did cause benchmark regressions for our downstream target. Worst case I've seen was a cycle increase by a factor 1.3 for a benchmark. I don't have much more to say about it right now (haven't had time to look at it closer to understand the exact scenario). Just wanted to mention this, if it turns out more people have problems.

https://github.com/llvm/llvm-project/pull/141121