[llvm] [LV] Autovectorization for the all-in-one histogram intrinsic (PR #91458)

[Paschalis Mpeis via llvm-commits]

================
@@ -102,6 +102,12 @@ class VPRecipeBuilder {
   VPWidenRecipe *tryToWiden(Instruction *I, ArrayRef<VPValue *> Operands,
                             VPBasicBlock *VPBB);
 
+  /// Makes Histogram count operations safe for vectorization, by emitting a
----------------
paschalis-mpeis wrote:

Maybe this can be generalized a bit more, so it doesn't refer to how AArch64 will be doing this?

Maybe something like:

```
/// Makes Histogram operations safe for vectorization, by emitting a
/// Histogram LLVM Intrinsic in place of the Load / BinOp (Add/Sub) / Store
/// operations that do the actual histogram computation.
```

https://github.com/llvm/llvm-project/pull/91458