[llvm] Adding the support for vectorization of loops with load based tripcount (PR #209390)

[Ramkumar Ramachandra via llvm-commits]

================
@@ -0,0 +1,35 @@
+; RUN: opt < %s -passes=loop-vectorize -enable-vectorize-loads-as-bound -force-vector-width=4 -force-vector-interleave=1 -S | FileCheck %s
+;
+; Negative test: the exit condition is an fcmp (not icmp). 
+; SCEV cannot compute the trip count of a float-bounded loop, so the
+; pass enters the dynamic-bound path.  It then rejects because the
+; branch condition is an FCmpInst, not an ICmpInst.
+;
+;   void foo(float *A, float *B, float *Limit) {
+;     for (float x = 0.0f; x < *Limit; x += 1.0f)
+;       A[(int)x] = B[(int)x];
+;   }
+
+define void @foo(ptr %A, ptr %B, ptr %Limit) {
----------------
artagnon wrote:

```suggestion
define void @foo(ptr %a, ptr %b, ptr %limit) {
```

best to use lower-case names as FileCheck uses case-insensitive patterns?

https://github.com/llvm/llvm-project/pull/209390