[llvm] Reapply "[LV] Use ExtractLane(LastActiveLane, V) live outs when tail-folding. (#149042)" (PR #168738)

[Luke Lau via llvm-commits]

================
@@ -1095,8 +1095,19 @@ class LLVM_ABI_FOR_TEST VPInstruction : public VPRecipeWithIRFlags,
     AnyOf,
     // Calculates the first active lane index of the vector predicate operands.
     // It produces the lane index across all unrolled iterations. Unrolling will
-    // add all copies of its original operand as additional operands.
+    // add all copies of its original operand as additional operands. If any of
+    // the vector predicate operands are all zeroes then this returns poison.
----------------
lukel97 wrote:

Yup, split off into #169298. After thinking about this for a bit I didn't bother adding the FirstActiveLaneZeroNotPoison opcode back, since I'm not sure if the extra complexity is worth it if we can likely optimize it away on RISC-V with e.g. tail folding whenever that's supported with early exit loops.

https://github.com/llvm/llvm-project/pull/168738