[llvm] LV: reuse getSmallBestKnownTC in a TC estimation (NFC) (PR #105834)

[Florian Hahn via llvm-commits]

================
@@ -1971,14 +1971,9 @@ class GeneratedRTChecks {
           // count. Assume that the outer loop executes at least twice.
           unsigned BestTripCount = 2;
 
-          // If exact trip count is known use that.
-          if (unsigned SmallTC = SE->getSmallConstantTripCount(OuterLoop))
-            BestTripCount = SmallTC;
-          else if (LoopVectorizeWithBlockFrequency) {
-            // Else use profile data if available.
-            if (auto EstimatedTC = getLoopEstimatedTripCount(OuterLoop))
-              BestTripCount = *EstimatedTC;
-          }
+          // Get the best known TC estimate.
+          if (auto EstimatedTC = getSmallBestKnownTC(*SE, OuterLoop))
----------------
fhahn wrote:

Yes, returning the max trip count here would probably severely overestimate the trip count. I *think* some of the existing callers probably should also be updated to avoid using the constant max, especially if it is very large (.e.g -1)

Would be good to have a test that shows the difference between using ` /* CanUseConstantMax=*/false` and ` /* CanUseConstantMax=*/true`

https://github.com/llvm/llvm-project/pull/105834