[llvm] [LoopVectorize] Don't scalarize predicated instruction with optsize (PR #129265)

[David Sherwood via llvm-commits]

================
@@ -257,80 +151,25 @@ define void @example2(i32 %n, i32 %x) optsize {
 }
 
 ; Loop has no primary induction as its integer IV has step -1 starting at
-; unknown N, but can still be vectorized.
+; unknown N, but can still be vectorized. But that would require tail folding,
----------------
david-arm wrote:

I think this should be something like:

```
; Loop has no primary induction as its integer IV has step -1 starting at
; unknown N, but can still be vectorized. However, this will require tail
; folding and since the memory operations need scalarising it will
; significantly increase code size. Due to the optsize attribute we decide
; not to vectorize.
```

https://github.com/llvm/llvm-project/pull/129265