[llvm] [InstCombine] fold nested bitreverse and bswap (PR #202236)

[Nikita Popov via llvm-commits]

https://github.com/nikic requested changes to this pull request.

This should live in VectorCombine with proper cost modelling. The transform only makes sense if the target actually has a bytewise vector bitreverse.

https://github.com/llvm/llvm-project/pull/202236