[llvm] [AArch64] Improve code generation for experimental.cttz.elts (PR #91505)

[David Sherwood via llvm-commits]

================
@@ -156,53 +137,178 @@ define i64 @vscale_4096_poison(<vscale x 16 x i8> %a) #1 {
   ret i64 %res
 }
 
-; NO VSCALE RANGE
+; MATCH WITH BRKB + CNTP
----------------
david-arm wrote:

nit: This isn't quite true as it might match with BRKB + INCP, although I understand what you mean. Perhaps you could rewrite this as:

`; EFFICIENT LOWERING USING BRKB`

?

https://github.com/llvm/llvm-project/pull/91505