[llvm] [VPlan] Materialize constant vector trip counts before final opts. (PR #142309)

[Florian Hahn via llvm-commits]

================
@@ -970,7 +970,7 @@ loop:
   %red.next = fadd double %for, %red
   %for.next = sitofp i32 %iv to double
   %iv.next = add nsw i32 %iv, 1
-  %ec = icmp eq i32 %iv.next, 1024
+  %ec = icmp eq i32 %iv.next, 1025
----------------
fhahn wrote:

Yep, for the reduction test we should probably make sure that the edge to scalar.ph remains, to make sure resume values are set up correctly.

https://github.com/llvm/llvm-project/pull/142309