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
@@ -5728,6 +5728,31 @@
     if (SDValue V = combineShiftAnd1ToBitTest(N, DAG))
       return V;
 
+  // Recognize the following pattern:
+  //
+  // AndVT = (and (sign_extend NarrowVT to AndVT) #bitmask)
+  //
+  // where bitmask is a mask that clears the upper bits of AndVT. The
+  // number of bits in bitmask must be a power of two.
+  auto IsAndZeroExtMask = [](SDValue LHS, SDValue RHS) {
+    if (LHS->getOpcode() != ISD::SIGN_EXTEND)
+      return false;
+
+    auto *C = dyn_cast<ConstantSDNode>(RHS);
+    if (!C)
+      return false;
+
+    if (!C->getAPIntValue().isMask(
+            LHS.getOperand(0).getValueType().getFixedSizeInBits()))
+      return false;
+
+    return true;
+  };
+
+  // Replace (and (sign_extend ...) #bitmask) with (zero_extend ...).
+  if (IsAndZeroExtMask(N0, N1))
+    return DAG.getNode(ISD::ZERO_EXTEND, SDLoc(N), VT, N0.getOperand(0));
+
   return SDValue();
 }
 
diff --git a/llvm/test/CodeGen/ARM/ParallelDSP/multi-use-loads.ll b/llvm/test/CodeGen/ARM/ParallelDSP/multi-use-loads.ll
--- a/llvm/test/CodeGen/ARM/ParallelDSP/multi-use-loads.ll
+++ b/llvm/test/CodeGen/ARM/ParallelDSP/multi-use-loads.ll
@@ -343,8 +343,8 @@
 ;
 ; CHECK-BE-LABEL: and_user:
 ; CHECK-BE:       @ %bb.0: @ %entry
-; CHECK-BE-NEXT:    .save {r4, r5, r6, r7, lr}
-; CHECK-BE-NEXT:    push {r4, r5, r6, r7, lr}
+; CHECK-BE-NEXT:    .save {r4, r5, r6, lr}
+; CHECK-BE-NEXT:    push {r4, r5, r6, lr}
 ; CHECK-BE-NEXT:    cmp r0, #1
 ; CHECK-BE-NEXT:    blt .LBB3_4
 ; CHECK-BE-NEXT:  @ %bb.1: @ %for.body.preheader
@@ -355,24 +355,23 @@
 ; CHECK-BE-NEXT:    .p2align 2
 ; CHECK-BE-NEXT:  .LBB3_2: @ %for.body
 ; CHECK-BE-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-BE-NEXT:    ldrsh lr, [r3, #2]!
-; CHECK-BE-NEXT:    ldrsh r5, [r2, #2]!
-; CHECK-BE-NEXT:    ldrsh.w r4, [r3, #2]
-; CHECK-BE-NEXT:    ldrsh.w r7, [r2, #2]
-; CHECK-BE-NEXT:    uxth.w r6, lr
-; CHECK-BE-NEXT:    smlabb r5, r5, lr, r12
-; CHECK-BE-NEXT:    smlabb r12, r7, r4, r5
+; CHECK-BE-NEXT:    ldrh lr, [r3, #2]!
+; CHECK-BE-NEXT:    ldrsh r4, [r2, #2]!
+; CHECK-BE-NEXT:    ldrsh.w r5, [r3, #2]
+; CHECK-BE-NEXT:    ldrsh.w r6, [r2, #2]
+; CHECK-BE-NEXT:    smlabb r4, r4, lr, r12
+; CHECK-BE-NEXT:    smlabb r12, r6, r5, r4
 ; CHECK-BE-NEXT:    subs r0, #1
-; CHECK-BE-NEXT:    mul r1, r6, r1
+; CHECK-BE-NEXT:    mul r1, lr, r1
 ; CHECK-BE-NEXT:    bne .LBB3_2
 ; CHECK-BE-NEXT:  @ %bb.3: @ %for.cond.cleanup
 ; CHECK-BE-NEXT:    add.w r0, r12, r1
-; CHECK-BE-NEXT:    pop {r4, r5, r6, r7, pc}
+; CHECK-BE-NEXT:    pop {r4, r5, r6, pc}
 ; CHECK-BE-NEXT:  .LBB3_4:
 ; CHECK-BE-NEXT:    mov.w r12, #0
 ; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    add.w r0, r12, r1
-; CHECK-BE-NEXT:    pop {r4, r5, r6, r7, pc}
+; CHECK-BE-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %cmp24 = icmp sgt i32 %arg, 0
   br i1 %cmp24, label %for.body.preheader, label %for.cond.cleanup
diff --git a/llvm/test/CodeGen/ARM/and-sext-combine.ll b/llvm/test/CodeGen/ARM/and-sext-combine.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/and-sext-combine.ll
@@ -0,0 +1,29 @@
+; RUN: llc -mtriple=arm-eabi -mcpu=arm1156t2-s -mattr=+thumb2 %s -o - -O3 \
+; RUN:  -asm-verbose=0 | FileCheck %s
+
+; This tests exerts the folding of `VT = (and (sign_extend NarrowVT to
+; VT) #bitmask)` into `VT = (zero_extend NarrowVT to VT)` when
+; #bitmask value is the mask made by all ones that selects the value
+; of type NarrowVT inside the value of type VT. The folding is
+; implemented in `DAGCombiner::visitAND`.
+
+; With this the folding, the `and` of the "signed extended load" of
+; `%b` in `f_i16_i32` is rendered as a zero extended load.
+
+; CHECK-LABEL: f_i16_i32:
+; CHECK-NEXT: .fnstart
+; CHECK-NEXT: ldrh    r1, [r1]
+; CHECK-NEXT: ldrsh   r0, [r0]
+; CHECK-NEXT: smulbb  r0, r0, r1
+; CHECK-NEXT: mul     r0, r0, r1
+; CHECK-NEXT: bx      lr
+define i32 @f_i16_i32(i16* %a, i16* %b) {
+  %1 = load i16, i16* %a, align 2
+  %sext.1 = sext i16 %1 to i32
+  %2 = load i16, i16* %b, align 2
+  %sext.2 = sext i16 %2 to i32
+  %masked = and i32 %sext.2, 65535
+  %mul = mul nsw i32 %sext.2, %sext.1
+  %count.next = mul i32 %mul, %masked
+  ret i32 %count.next
+}