diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -734,8 +734,11 @@
     NextI = next_nodbg(NextI, E);
 
   unsigned Opc = I->getOpcode();
+  unsigned MergeMIOpc = MergeMI->getOpcode();
   bool IsScaled = !TII->hasUnscaledLdStOffset(Opc);
-  int OffsetStride = IsScaled ? 1 : TII->getMemScale(*I);
+  bool IsMergedMIScaled = !TII->hasUnscaledLdStOffset(MergeMIOpc);
+  int OffsetStride = IsScaled ? TII->getMemScale(*I) : 1;
+  int MergeMIOffsetStride = IsMergedMIScaled ? TII->getMemScale(*MergeMI) : 1;
 
   bool MergeForward = Flags.getMergeForward();
   // Insert our new paired instruction after whichever of the paired
@@ -749,15 +752,20 @@
 
   // Which register is Rt and which is Rt2 depends on the offset order.
   MachineInstr *RtMI;
-  if (AArch64InstrInfo::getLdStOffsetOp(*I).getImm() ==
-      AArch64InstrInfo::getLdStOffsetOp(*MergeMI).getImm() + OffsetStride)
+  int64_t ScaledIAddress =
+      AArch64InstrInfo::getLdStOffsetOp(*I).getImm() * OffsetStride;
+  int64_t ScaledMergeMIAddress =
+      AArch64InstrInfo::getLdStOffsetOp(*MergeMI).getImm() *
+      MergeMIOffsetStride;
+  if (ScaledIAddress == ScaledMergeMIAddress + OffsetStride)
     RtMI = &*MergeMI;
   else
     RtMI = &*I;
 
   int OffsetImm = AArch64InstrInfo::getLdStOffsetOp(*RtMI).getImm();
+  bool IsScaledRtMI = !TII->hasUnscaledLdStOffset(RtMI->getOpcode());
   // Change the scaled offset from small to large type.
-  if (IsScaled) {
+  if (IsScaledRtMI) {
     assert(((OffsetImm & 1) == 0) && "Unexpected offset to merge");
     OffsetImm /= 2;
   }
diff --git a/llvm/test/CodeGen/AArch64/str-narrow-zero-merge.ll b/llvm/test/CodeGen/AArch64/str-narrow-zero-merge.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/str-narrow-zero-merge.ll
@@ -0,0 +1,52 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-linux-gnu | FileCheck %s
+
+%struct.d = type { i16, i16, i16 }
+
+define void @merge_unscaled_str_with_scaled() {
+; CHECK-LABEL: merge_unscaled_str_with_scaled:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    sub sp, sp, #16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    str xzr, [sp, #64]
+; CHECK-NEXT:    add sp, sp, #16
+; CHECK-NEXT:    ret
+entry:
+  %ptr.1 = alloca i32, align 4
+  %ptr.2 = alloca %struct.d, align 2
+  store i32 0, ptr %ptr.1, align 4
+  %0 = getelementptr inbounds i8, ptr %ptr.2, i64 2
+  store i32 0, ptr %0, align 2
+  ret void
+}
+
+define void @merge_scaled_str_with_scaled_str(){
+; CHECK-LABEL: merge_scaled_str_with_scaled_str:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub sp, sp, #16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    str xzr, [sp, #8]
+; CHECK-NEXT:    add sp, sp, #16
+; CHECK-NEXT:    ret
+  %ptr.1 = alloca i32, align 4
+  %ptr.2 = alloca i32, align 4
+  store i32 0, ptr %ptr.1, align 4
+  store i32 0, ptr %ptr.2, align 4
+  ret void
+}
+
+define void @merge_scaled_str_with_scaled_str_2() {
+; CHECK-LABEL: merge_scaled_str_with_scaled_str_2:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    sub sp, sp, #16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    str wzr, [sp, #12]
+; CHECK-NEXT:    add sp, sp, #16
+; CHECK-NEXT:    ret
+entry:
+  %g = alloca i16, align 2
+  %h = alloca i16, align 2
+  store i16 0, ptr %g, align 2
+  store i16 0, ptr %h, align 2
+  ret void
+}