diff --git a/src/mapleall/maple_be/include/cg/aarch64/aarch64_args.h b/src/mapleall/maple_be/include/cg/aarch64/aarch64_args.h
index 67e6f402c70e847604218423a06a74735ec6575e..8e65824e2eee1276ab133276802c4d5c2b29d9df 100644
--- a/src/mapleall/maple_be/include/cg/aarch64/aarch64_args.h
+++ b/src/mapleall/maple_be/include/cg/aarch64/aarch64_args.h
@@ -31,6 +31,8 @@ struct ArgInfo {
   const AArch64SymbolAlloc *symLoc;
   uint8 memPairSecondRegSize;  /* struct arg requiring two regs, size of 2nd reg */
   bool doMemPairOpt;
+  bool createTwoStores;
+  bool isTwoRegParm;
 };
 
 class AArch64MoveRegArgs : public MoveRegArgs {
diff --git a/src/mapleall/maple_be/src/cg/aarch64/aarch64_args.cpp b/src/mapleall/maple_be/src/cg/aarch64/aarch64_args.cpp
index 722ac3d6750dfc17b5c8f41a89c42878e3046a9a..086b43773c59a8adbd082487aee92c969e529baf 100644
--- a/src/mapleall/maple_be/src/cg/aarch64/aarch64_args.cpp
+++ b/src/mapleall/maple_be/src/cg/aarch64/aarch64_args.cpp
@@ -89,7 +89,11 @@ ArgInfo AArch64MoveRegArgs::GetArgInfo(std::map<uint32, AArch64reg> &argsList, s
   argInfo.symSize = aarchCGFunc->GetBecommon().GetTypeSize(argInfo.mirTy->GetTypeIndex());
   argInfo.memPairSecondRegSize = 0;
   argInfo.doMemPairOpt = false;
+  argInfo.createTwoStores  = false;
+  argInfo.isTwoRegParm = false;
+
   if ((argInfo.symSize > k8ByteSize) && (argInfo.symSize <= k16ByteSize)) {
+    argInfo.isTwoRegParm = true;
     if (numFpRegs[argIndex] > kOneRegister) {
       argInfo.symSize = argInfo.stkSize = fpSize[argIndex];
     } else {
@@ -109,6 +113,7 @@ ArgInfo AArch64MoveRegArgs::GetArgInfo(std::map<uint32, AArch64reg> &argsList, s
     /* For small aggregate parameter, set to minimum of 4 bytes. */
     argInfo.symSize = argInfo.stkSize = k4ByteSize;
   } else if (numFpRegs[argIndex] > kOneRegister) {
+    argInfo.isTwoRegParm = true;
     argInfo.symSize = argInfo.stkSize = fpSize[argIndex];
   } else {
     argInfo.stkSize = (argInfo.symSize < k4ByteSize) ? k4ByteSize : argInfo.symSize;
@@ -128,6 +133,7 @@ ArgInfo AArch64MoveRegArgs::GetArgInfo(std::map<uint32, AArch64reg> &argsList, s
      */
     argInfo.symSize = kSizeOfPtr;
     argInfo.doMemPairOpt = false;
+    argInfo.createTwoStores = true;
   }
   return argInfo;
 }
@@ -254,7 +260,7 @@ void AArch64MoveRegArgs::GenerateStrInsn(ArgInfo &argInfo, AArch64reg reg2, uint
   }
   aarchCGFunc->GetCurBB()->AppendInsn(insn);
 
-  if (argInfo.doMemPairOpt) {
+  if (argInfo.createTwoStores || argInfo.doMemPairOpt) {
     /* second half of the struct passing by registers. */
     uint32 part2BitSize = argInfo.memPairSecondRegSize * kBitsPerByte;
     GenOneInsn(argInfo, *baseOpnd, part2BitSize, reg2, (stOffset + kSizeOfPtr));
@@ -302,7 +308,9 @@ void AArch64MoveRegArgs::MoveRegisterArgs() {
           static_cast<AArch64SymbolAlloc *>(aarchCGFunc->GetMemlayout()->GetSymAllocInfo(
               secondArgInfo.sym->GetStIndex()));
       /* Make sure they are in same segment if want to use stp */
-      if (IsInSameSegment(firstArgInfo, secondArgInfo)) {
+      if (((firstArgInfo.isTwoRegParm && secondArgInfo.isTwoRegParm) ||
+          (firstArgInfo.isTwoRegParm == false && secondArgInfo.isTwoRegParm == false)) &&
+          (firstArgInfo.doMemPairOpt || IsInSameSegment(firstArgInfo, secondArgInfo))) {
         GenerateStpInsn(firstArgInfo, secondArgInfo);
         it = next;
         continue;
diff --git a/src/mapleall/maple_be/src/cg/aarch64/aarch64_memlayout.cpp b/src/mapleall/maple_be/src/cg/aarch64/aarch64_memlayout.cpp
index 2da54cb6254d30b5c9d65f020dce9f14741d40ef..d8ece10f073fb3d4d701bf0b2b2788425d28a9ff 100644
--- a/src/mapleall/maple_be/src/cg/aarch64/aarch64_memlayout.cpp
+++ b/src/mapleall/maple_be/src/cg/aarch64/aarch64_memlayout.cpp
@@ -206,6 +206,10 @@ void AArch64MemLayout::LayoutFormalParams() {
         SetSizeAlignForTypeIdx(ptyIdx, size, align);
         symLoc->SetMemSegment(GetSegArgsRegPassed());
         /* the type's alignment requirement may be smaller than a registser's byte size */
+        if (ty->GetPrimType() == PTY_agg &&  be.GetTypeSize(ptyIdx) > k4ByteSize) {
+          /* struct param aligned on 8 byte boundary unless it is small enough */
+          align = kSizeOfPtr;
+        }
         segArgsRegPassed.SetSize(RoundUp(segArgsRegPassed.GetSize(), align));
         symLoc->SetOffset(segArgsRegPassed.GetSize());
         segArgsRegPassed.SetSize(segArgsRegPassed.GetSize() + size);