[llvm] Move ExpandMemCmp and MergeIcmp to the middle end (PR #77370)

[Matt Arsenault via llvm-commits]

================
@@ -25,20 +25,23 @@ define i64 @one_dimensional(ptr %a, ptr %b, i64 %N) {
 entry:
   br label %for.body
 
-for.body:                                         ; preds = %entry, %for.body
+for.body:                                         ; preds = %for.body, %entry
   %i.06 = phi i64 [ %inc, %for.body ], [ 0, %entry ]
   %sum.05 = phi i64 [ %spec.select, %for.body ], [ 0, %entry ]
   %arrayidx = getelementptr inbounds ptr, ptr %a, i64 %i.06
   %0 = load ptr, ptr %arrayidx, align 8
-  %bcmp = tail call i32 @bcmp(ptr %0, ptr %b, i64 4)
-  %tobool = icmp eq i32 %bcmp, 0
+  %1 = load i32, ptr %0, align 1
----------------
arsenm wrote:

Don't use anonymous values in tests 

https://github.com/llvm/llvm-project/pull/77370