Index: lib/Target/SystemZ/SystemZISelLowering.cpp
===================================================================
--- lib/Target/SystemZ/SystemZISelLowering.cpp
+++ lib/Target/SystemZ/SystemZISelLowering.cpp
@@ -4490,18 +4490,21 @@
     // avoid a false dependency on any previous contents of the vector
     // register.
 
-    // Use a VLREP if at least one element is a load.
-    unsigned LoadElIdx = UINT_MAX;
+    // Use a VLREP if at least one element is a load. Make sure to replicate
+    // the load with the most elements having its value.
+    std::map<const SDNode*, unsigned> UseCounts;
+    SDNode *LoadMaxUses = nullptr;
     for (unsigned I = 0; I < NumElements; ++I)
       if (Elems[I].getOpcode() == ISD::LOAD &&
           cast<LoadSDNode>(Elems[I])->isUnindexed()) {
-        LoadElIdx = I;
-        break;
+        SDNode *Ld = Elems[I].getNode();
+        UseCounts[Ld]++;
+        if (LoadMaxUses == nullptr || UseCounts[LoadMaxUses] < UseCounts[Ld])
+          LoadMaxUses = Ld;
       }
-    if (LoadElIdx != UINT_MAX) {
-      Result = DAG.getNode(SystemZISD::REPLICATE, DL, VT, Elems[LoadElIdx]);
-      Done[LoadElIdx] = true;
-      ReplicatedVal = Elems[LoadElIdx];
+    if (LoadMaxUses != nullptr) {
+      ReplicatedVal = SDValue(LoadMaxUses, 0);
+      Result = DAG.getNode(SystemZISD::REPLICATE, DL, VT, ReplicatedVal);
     } else {
       // Try to use VLVGP.
       unsigned I1 = NumElements / 2 - 1;
Index: test/CodeGen/SystemZ/vec-move-20.ll
===================================================================
--- /dev/null
+++ test/CodeGen/SystemZ/vec-move-20.ll
@@ -0,0 +1,19 @@
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 | FileCheck %s
+
+; Test a vector which is built with elements from two loads replicates the
+; load with most elements having its value.
+
+; CHECK:      vlef
+; CHECK-NOT:  vlvgf
+
+define void @update(i32* %src1, i32* %src2, <4 x i32>* %dst) {
+bb:
+  %tmp = load i32, i32* %src1
+  %tmp1 = load i32, i32* %src2
+  %tmp2 = insertelement <4 x i32> undef, i32 %tmp, i32 0
+  %tmp3 = insertelement <4 x i32> %tmp2, i32 %tmp1, i32 1
+  %tmp4 = insertelement <4 x i32> %tmp3, i32 %tmp1, i32 2
+  %tmp5 = insertelement <4 x i32> %tmp4, i32 %tmp1, i32 3
+  store <4 x i32> %tmp5, <4 x i32>* %dst
+  ret void
+}