diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -20298,11 +20298,13 @@
   // Zihintntl
   case RISCV::BI__builtin_riscv_ntl_load: {
     llvm::Type *ResTy = ConvertType(E->getType());
-    ConstantInt *Mode = cast<ConstantInt>(Ops[1]);
+    unsigned DomainVal = 5; // Default __RISCV_NTLH_ALL
+    if (Ops.size() == 2)
+      DomainVal = cast<ConstantInt>(Ops[1])->getZExtValue();
 
     llvm::MDNode *RISCVDomainNode = llvm::MDNode::get(
         getLLVMContext(),
-        llvm::ConstantAsMetadata::get(Builder.getInt32(Mode->getZExtValue())));
+        llvm::ConstantAsMetadata::get(Builder.getInt32(DomainVal)));
     llvm::MDNode *NontemporalNode = llvm::MDNode::get(
         getLLVMContext(), llvm::ConstantAsMetadata::get(Builder.getInt32(1)));
 
@@ -20324,11 +20326,13 @@
     return Load;
   }
   case RISCV::BI__builtin_riscv_ntl_store: {
-    ConstantInt *Mode = cast<ConstantInt>(Ops[2]);
+    unsigned DomainVal = 5; // Default __RISCV_NTLH_ALL
+    if (Ops.size() == 3)
+      DomainVal = cast<ConstantInt>(Ops[2])->getZExtValue();
 
     llvm::MDNode *RISCVDomainNode = llvm::MDNode::get(
         getLLVMContext(),
-        llvm::ConstantAsMetadata::get(Builder.getInt32(Mode->getZExtValue())));
+        llvm::ConstantAsMetadata::get(Builder.getInt32(DomainVal)));
     llvm::MDNode *NontemporalNode = llvm::MDNode::get(
         getLLVMContext(), llvm::ConstantAsMetadata::get(Builder.getInt32(1)));
 
diff --git a/clang/lib/Headers/riscv_ntlh.h b/clang/lib/Headers/riscv_ntlh.h
--- a/clang/lib/Headers/riscv_ntlh.h
+++ b/clang/lib/Headers/riscv_ntlh.h
@@ -21,8 +21,6 @@
   __RISCV_NTLH_ALL
 };
 
-#define __riscv_ntl_load(PTR, DOMAIN) __builtin_riscv_ntl_load((PTR), (DOMAIN))
-#define __riscv_ntl_store(PTR, VAL, DOMAIN)                                    \
-  __builtin_riscv_ntl_store((PTR), (VAL), (DOMAIN))
-
-#endif
\ No newline at end of file
+#define __riscv_ntl_load __builtin_riscv_ntl_load
+#define __riscv_ntl_store __builtin_riscv_ntl_store
+#endif
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -5287,12 +5287,16 @@
     bool IsStore = BuiltinID == RISCV::BI__builtin_riscv_ntl_store;
     unsigned NumArgs = IsStore ? 3 : 2;
 
-    if (checkArgCount(*this, TheCall, NumArgs))
+    if (checkArgCountAtLeast(*this, TheCall, NumArgs - 1))
+      return true;
+
+    if (checkArgCountAtMost(*this, TheCall, NumArgs))
       return true;
 
     // Domain value should be compile-time constant.
     // 2 <= domain <= 5
-    if (SemaBuiltinConstantArgRange(TheCall, NumArgs - 1, 2, 5))
+    if (TheCall->getNumArgs() == NumArgs &&
+        SemaBuiltinConstantArgRange(TheCall, NumArgs - 1, 2, 5))
       return true;
 
     Expr *PointerArg = TheCall->getArg(0);
diff --git a/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c b/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
--- a/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
+++ b/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
@@ -100,6 +100,24 @@
   *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL);   // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
   *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL);   // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
 
+  uc = __riscv_ntl_load(&sc);   // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
+  sc = __riscv_ntl_load(&uc);   // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
+  us = __riscv_ntl_load(&ss);   // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
+  ss = __riscv_ntl_load(&us);   // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
+  ui = __riscv_ntl_load(&si);   // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
+  si = __riscv_ntl_load(&ui);   // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
+  ull = __riscv_ntl_load(&sll); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  sll = __riscv_ntl_load(&ull); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  h1 = __riscv_ntl_load(&h2);   // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
+  f1 = __riscv_ntl_load(&f2);   // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
+  d1 = __riscv_ntl_load(&d2);   // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  v4si1 = __riscv_ntl_load(&v4si2);   // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
+  v8ss1 = __riscv_ntl_load(&v8ss2);   // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
+  v16sc1 = __riscv_ntl_load(&v16sc2);   // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
+  *scvi1 = __riscv_ntl_load(scvi2);   // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  *scvs1 = __riscv_ntl_load(scvs2);   // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  *scvc1 = __riscv_ntl_load(scvc2);   // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+
   __riscv_ntl_store(&uc, 1, __RISCV_NTLH_INNERMOST_PRIVATE);    // CHECK: store i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !5
   __riscv_ntl_store(&sc, 1, __RISCV_NTLH_INNERMOST_PRIVATE);    // CHECK: store i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !5
   __riscv_ntl_store(&us, 1, __RISCV_NTLH_INNERMOST_PRIVATE);    // CHECK: store i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
@@ -172,6 +190,23 @@
   __riscv_ntl_store(scvs2, *scvs1, __RISCV_NTLH_ALL);  // CHECK: store <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
   __riscv_ntl_store(scvc2, *scvc1, __RISCV_NTLH_ALL);  // CHECK: store <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
 
+  __riscv_ntl_store(&uc, 1);    // CHECK: store i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&sc, 1);    // CHECK: store i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&us, 1);    // CHECK: store i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&ss, 1);    // CHECK: store i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&ui, 1);    // CHECK: store i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&si, 1);    // CHECK: store i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&ull, 1);   // CHECK: store i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&sll, 1);   // CHECK: store i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&h1, 1.0);  // CHECK: store half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&f1, 1.0);  // CHECK: store float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&d1, 1.0);  // CHECK: store double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&v4si1, v4si2);  // CHECK: store <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&v8ss1, v8ss2);  // CHECK: store <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(&v16sc1, v16sc2);  // CHECK: store <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(scvi2, *scvi1);  // CHECK: store <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(scvs2, *scvs1);  // CHECK: store <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+  __riscv_ntl_store(scvc2, *scvc1);  // CHECK: store <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
 }
 // clang-format on
 
@@ -179,4 +214,4 @@
 // CHECK: !5 = !{i32 2}
 // CHECK: !6 = !{i32 3}
 // CHECK: !7 = !{i32 4}
-// CHECK: !8 = !{i32 5}
\ No newline at end of file
+// CHECK: !8 = !{i32 5}