diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -6241,8 +6241,8 @@
     // fold (and (masked_load) (splat_vec (x, ...))) to zext_masked_load
     auto *MLoad = dyn_cast<MaskedLoadSDNode>(N0);
     ConstantSDNode *Splat = isConstOrConstSplat(N1, true, true);
-    if (MLoad && MLoad->getExtensionType() == ISD::EXTLOAD && N0.hasOneUse() &&
-        Splat && N1.hasOneUse()) {
+    if (MLoad && MLoad->getExtensionType() == ISD::EXTLOAD && Splat &&
+        N1.hasOneUse()) {
       EVT LoadVT = MLoad->getMemoryVT();
       EVT ExtVT = VT;
       if (TLI.isLoadExtLegal(ISD::ZEXTLOAD, ExtVT, LoadVT)) {
@@ -6252,11 +6252,16 @@
         uint64_t ElementSize =
             LoadVT.getVectorElementType().getScalarSizeInBits();
         if (Splat->getAPIntValue().isMask(ElementSize)) {
-          return DAG.getMaskedLoad(
+          auto NewLoad = DAG.getMaskedLoad(
               ExtVT, SDLoc(N), MLoad->getChain(), MLoad->getBasePtr(),
               MLoad->getOffset(), MLoad->getMask(), MLoad->getPassThru(),
               LoadVT, MLoad->getMemOperand(), MLoad->getAddressingMode(),
               ISD::ZEXTLOAD, MLoad->isExpandingLoad());
+          bool LoadHasOtherUsers = !N0.hasOneUse();
+          CombineTo(N, NewLoad);
+          if (LoadHasOtherUsers)
+            CombineTo(MLoad, NewLoad.getValue(0), NewLoad.getValue(1));
+          return SDValue(N, 0);
         }
       }
     }
diff --git a/llvm/test/CodeGen/AArch64/sve-load-compare-store.ll b/llvm/test/CodeGen/AArch64/sve-load-compare-store.ll
--- a/llvm/test/CodeGen/AArch64/sve-load-compare-store.ll
+++ b/llvm/test/CodeGen/AArch64/sve-load-compare-store.ll
@@ -6,9 +6,7 @@
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    mov z1.d, z0.d
-; CHECK-NEXT:    and z1.s, z1.s, #0xffff
-; CHECK-NEXT:    cmphs p0.s, p0/z, z1.s, #0
+; CHECK-NEXT:    cmphs p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    st1b { z0.s }, p0, [x1]
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/Thumb2/mve-masked-load.ll b/llvm/test/CodeGen/Thumb2/mve-masked-load.ll
--- a/llvm/test/CodeGen/Thumb2/mve-masked-load.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-masked-load.ll
@@ -2079,7 +2079,7 @@
 ; CHECK-LE-NEXT:    vmov r0, r1, d8
 ; CHECK-LE-NEXT:    vmov r2, r3, d9
 ; CHECK-LE-NEXT:    bl foo
-; CHECK-LE-NEXT:    vmovlb.u16 q0, q4
+; CHECK-LE-NEXT:    vmov q0, q4
 ; CHECK-LE-NEXT:    vpop {d8, d9}
 ; CHECK-LE-NEXT:    pop {r7, pc}
 ;
@@ -2091,13 +2091,12 @@
 ; CHECK-BE-NEXT:    vpush {d8, d9}
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
 ; CHECK-BE-NEXT:    vpt.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vldrht.u32 q4, [r0]
-; CHECK-BE-NEXT:    vrev64.32 q0, q4
-; CHECK-BE-NEXT:    vmov r1, r0, d0
-; CHECK-BE-NEXT:    vmov r3, r2, d1
+; CHECK-BE-NEXT:    vldrht.u32 q0, [r0]
+; CHECK-BE-NEXT:    vrev64.32 q4, q0
+; CHECK-BE-NEXT:    vmov r1, r0, d8
+; CHECK-BE-NEXT:    vmov r3, r2, d9
 ; CHECK-BE-NEXT:    bl foo
-; CHECK-BE-NEXT:    vmovlb.u16 q1, q4
-; CHECK-BE-NEXT:    vrev64.32 q0, q1
+; CHECK-BE-NEXT:    vmov q0, q4
 ; CHECK-BE-NEXT:    vpop {d8, d9}
 ; CHECK-BE-NEXT:    pop {r7, pc}
 entry: