[llvm] [VPlan] Use VPInstruction for uniform binops. (PR #141429)

[via llvm-commits]

================
@@ -11,18 +11,19 @@ loop:
   %arrayidx = getelementptr inbounds i64, ptr %ascast, i64 %next
   store i64 %next, ptr %arrayidx, align 4
 
-; check that we find the two interleaved blocks with ascast, gep and store:
+; check that we find the loop-invariant ascast followed by two interleaved
+; blocks with gep and store:
+; CHECK: [[AS1:%.*]] = addrspacecast ptr addrspace(1) %in to ptr
----------------
ayalz wrote:

LICM rather than sinking scalar operands.

https://github.com/llvm/llvm-project/pull/141429