diff --git a/src/mapleall/maple_be/src/cg/aarch64/aarch64_cgfunc.cpp b/src/mapleall/maple_be/src/cg/aarch64/aarch64_cgfunc.cpp
index fd448b6af410849a7975520c02e4e243b1efca1b..aee06045bea208c427a69ec8a48bb1987bfd47d3 100644
--- a/src/mapleall/maple_be/src/cg/aarch64/aarch64_cgfunc.cpp
+++ b/src/mapleall/maple_be/src/cg/aarch64/aarch64_cgfunc.cpp
@@ -2274,12 +2274,68 @@ void AArch64CGFunc::SelectBlkassignoff(BlkassignoffNode &bNode, Operand *src)
     IncLmbcTotalArgs();
     /* copy large agg arg to offset below */
   }
-  std::vector<Operand*> opndVec;
-  opndVec.push_back(regResult);                              /* result */
-  opndVec.push_back(PrepareMemcpyParamOpnd(offset, *dest));  /* param 0 */
-  opndVec.push_back(src);                                    /* param 1 */
-  opndVec.push_back(PrepareMemcpyParamOpnd(static_cast<uint64>(static_cast<int64>(bNode.blockSize)))); /* param 2 */
-  SelectLibCall("memcpy", opndVec, PTY_a64, PTY_a64);
+  RegOperand *param0 = PrepareMemcpyParamOpnd(offset, *dest);
+  RegOperand *param1 = static_cast<RegOperand *>(src);
+  RegOperand *param2 = PrepareMemcpyParamOpnd(bNode.blockSize);
+  if (bNode.blockSize > kParmMemcpySize) {
+    std::vector<Operand*> opndVec;
+    opndVec.push_back(regResult); /* result */
+    opndVec.push_back(param0);    /* param 0 */
+    opndVec.push_back(src);       /* param 1 */
+    opndVec.push_back(param2);    /* param 2 */
+    SelectLibCall("memcpy", opndVec, PTY_a64, PTY_a64);
+  } else {
+    int32 copyOffset = 0;
+    int32 inc = 0;
+    bool isPair;
+    for (uint32 sz = bNode.blockSize; sz > 0; ) {
+      isPair = false;
+      MOperator ldOp, stOp;
+      if (sz >= k16ByteSize) {
+        sz -= k16ByteSize;
+        inc = k16ByteSize;
+        ldOp = MOP_xldp;
+        stOp = MOP_xstp;
+        isPair = true;
+      } else if (sz >= k8ByteSize) {
+        sz -= k8ByteSize;
+        inc = k8ByteSize;
+        ldOp = MOP_xldr;
+        stOp = MOP_xstr;
+      } else if (sz >= k4ByteSize) {
+        sz -= k4ByteSize;
+        inc = k4ByteSize;
+        ldOp = MOP_wldr;
+        stOp = MOP_wstr;
+      } else if (sz >= k2ByteSize) {
+        sz -= k2ByteSize;
+        inc = k2ByteSize;
+        ldOp = MOP_wldrh;
+        stOp = MOP_wstrh;
+      } else {
+        sz -= k1ByteSize;
+        inc = k1ByteSize;
+        ldOp = MOP_wldrb;
+        stOp = MOP_wstrb;
+      }
+      AArch64reg ldBaseReg = static_cast<AArch64reg>(param1->GetRegisterNumber());
+      MemOperand &ldMem = CreateMemOpnd(ldBaseReg, copyOffset, k8ByteSize);
+
+      AArch64reg stBaseReg = static_cast<AArch64reg>(param0->GetRegisterNumber());
+      MemOperand &stMem = CreateMemOpnd(stBaseReg, copyOffset, k8ByteSize);
+      if (isPair) {
+        RegOperand &ldResult = CreateVirtualRegisterOperand(NewVReg(kRegTyInt, k8ByteSize));
+        RegOperand &ldResult2 = CreateVirtualRegisterOperand(NewVReg(kRegTyInt, k8ByteSize));
+        GetCurBB()->AppendInsn(GetCG()->BuildInstruction<AArch64Insn>(ldOp, ldResult, ldResult2, ldMem));
+        GetCurBB()->AppendInsn(GetCG()->BuildInstruction<AArch64Insn>(stOp, ldResult, ldResult2, stMem));
+      } else {
+        RegOperand &ldResult = CreateVirtualRegisterOperand(NewVReg(kRegTyInt, k8ByteSize));
+        GetCurBB()->AppendInsn(GetCG()->BuildInstruction<AArch64Insn>(ldOp, ldResult, ldMem));
+        GetCurBB()->AppendInsn(GetCG()->BuildInstruction<AArch64Insn>(stOp, ldResult, stMem));
+      }
+      copyOffset += inc;
+    }
+  }
 }
 
 void AArch64CGFunc::SelectAggIassign(IassignNode &stmt, Operand &addrOpnd) {