[llvm] [LV] Cache call vectorization decisions (PR #66521)

[David Sherwood via llvm-commits]

================
@@ -1450,6 +1482,7 @@ class LoopVectorizationCostModel {
     if (VF.isScalar() || Uniforms.contains(VF))
       return;
     setCostBasedWideningDecision(VF);
+    setVectorizedCallDecision(VF);
----------------
david-arm wrote:

This is a comment for line 1479, since github reviews rather helpfully don't let you comment on the actual line you care about. :face_palm

It looks like the comment above collectUniformsAndScalars is out of date. I think it was already out of date before this change, but now it's even worse! Could you add something to the comment about collecting decisions regarding call vectorisation?

https://github.com/llvm/llvm-project/pull/66521