[llvm] [X86] Improve llvm.get.active.lane.mask codegen (PR #225832)

[via llvm-commits]

Andarwinux wrote:

> There may also need some TTI changes to make loop vectorizer to use llvm.get.active.lane.mask for tail folding (-mllvm -tail-folding-policy=prefer-fold-tail) loops?

https://github.com/llvm/llvm-project/blob/2aa63fe0458bb5265756d05fa123df3d1b3a1dd6/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h#L465-L468

I think could use DataAndControlFlow for AVX512 and keep the default DataWithoutLaneMask like aarch64 did.

Since tail folding is not enabled by default on x86, there is no practical impact for now.

https://github.com/llvm/llvm-project/pull/225832