[llvm] [AArch64][InstCombine] Eliminate redundant barrier intrinsics (PR #112023)

[David Sherwood via llvm-commits]

david-arm wrote:

> @david-arm This change mostly parallels optimizations done for fences in `InstCombinerImpl::visitFenceInst` and at least there it was beneficial. It's hard for me to come up with a situation where `dmb; some instrs; dmb` would be faster than `some instrs; dmb` except some really weird hw quirk not sure we could account for in IR anyway (and again, fences would be affected in the same way).

No that's fair enough and thanks for the explanation!

https://github.com/llvm/llvm-project/pull/112023