[llvm] [AggressiveInstCombine] Fold byte-reversed consecutive loads to load+bswap (PR #227517)

[Nikita Popov via llvm-commits]

nikic wrote:

>From llvm-opt-benchmark, I think the two main regressions of interest are:

 * bswap.i16 being formed, and preventing later formation of bswap.i32. Probably this transform should be able to assemble from a bswap part as well?
 * 2, 1, 0, 3 shuffles not being formed, because the 1, 0 part becomes a bswap.

https://github.com/llvm/llvm-project/pull/227517