diff --git a/openmp/libomptarget/deviceRTLs/nvptx/src/loop.cu b/openmp/libomptarget/deviceRTLs/nvptx/src/loop.cu
--- a/openmp/libomptarget/deviceRTLs/nvptx/src/loop.cu
+++ b/openmp/libomptarget/deviceRTLs/nvptx/src/loop.cu
@@ -382,10 +382,10 @@
 
   INLINE static int64_t Shuffle(unsigned active, int64_t val, int leader) {
     uint32_t lo, hi;
-    __kmpc_impl_unpack(val, lo, hi);
+    __kmpc_impl::Bits::unpack(val, lo, hi);
     hi = __kmpc_impl_shfl_sync(active, hi, leader);
     lo = __kmpc_impl_shfl_sync(active, lo, leader);
-    return __kmpc_impl_pack(lo, hi);
+    return __kmpc_impl::Bits::pack(lo, hi);
   }
 
   INLINE static uint64_t NextIter() {
diff --git a/openmp/libomptarget/deviceRTLs/nvptx/src/reduction.cu b/openmp/libomptarget/deviceRTLs/nvptx/src/reduction.cu
--- a/openmp/libomptarget/deviceRTLs/nvptx/src/reduction.cu
+++ b/openmp/libomptarget/deviceRTLs/nvptx/src/reduction.cu
@@ -29,10 +29,10 @@
 
 EXTERN int64_t __kmpc_shuffle_int64(int64_t val, int16_t delta, int16_t size) {
    uint32_t lo, hi;
-   __kmpc_impl_unpack(val, lo, hi);
+   __kmpc_impl::Bits::unpack(val, lo, hi);
    hi = __kmpc_impl_shfl_down_sync(0xFFFFFFFF, hi, delta, size);
    lo = __kmpc_impl_shfl_down_sync(0xFFFFFFFF, lo, delta, size);
-   return __kmpc_impl_pack(lo, hi);
+   return __kmpc_impl::Bits::pack(lo, hi);
 }
 
 INLINE static void gpu_regular_warp_reduce(void *reduce_data,
diff --git a/openmp/libomptarget/deviceRTLs/nvptx/src/target_api.h b/openmp/libomptarget/deviceRTLs/nvptx/src/target_api.h
new file mode 100644
--- /dev/null
+++ b/openmp/libomptarget/deviceRTLs/nvptx/src/target_api.h
@@ -0,0 +1,49 @@
+//===--- target_api.h - OpenMP GPU target abstraction interface --- c++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// CRTP style static interface for target specific functions
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _TARGET_API_H_
+#define _TARGET_API_H_
+
+#include <stdint.h>
+
+#include "option.h"
+
+namespace __kmpc_impl {
+
+// nvcc requires this to be <typename T>. Fails to compile <typename I>.
+template <typename T> class Api {
+public:
+  INLINE static uint64_t pack(uint32_t lo, uint32_t hi) {
+    return T::packImpl(lo, hi);
+  }
+  INLINE static void unpack(uint64_t val, uint32_t &lo, uint32_t &hi) {
+    T::unpackImpl(val, lo, hi);
+  }
+
+private:
+  INLINE static uint64_t packImpl(uint32_t lo, uint32_t hi);
+  INLINE static void unpackImpl(uint64_t, uint32_t &, uint32_t &) = delete;
+
+private:
+  friend T;
+  Api() = delete;
+};
+
+// Default implementations
+template <typename T>
+INLINE uint64_t Api<T>::packImpl(uint32_t lo, uint32_t hi) {
+  return (((uint64_t)hi) << 32u) | (uint64_t)lo;
+}
+
+} // namespace __kmpc_impl
+
+#endif
diff --git a/openmp/libomptarget/deviceRTLs/nvptx/src/target_impl.h b/openmp/libomptarget/deviceRTLs/nvptx/src/target_impl.h
--- a/openmp/libomptarget/deviceRTLs/nvptx/src/target_impl.h
+++ b/openmp/libomptarget/deviceRTLs/nvptx/src/target_impl.h
@@ -1,4 +1,4 @@
-//===------------ target_impl.h - NVPTX OpenMP GPU options ------- CUDA -*-===//
+//===-------- target_api.h - OpenMP GPU target abstraction ------- CUDA -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -14,20 +14,28 @@
 
 #include <stdint.h>
 
+#include "target_api.h"
 #include "option.h"
 
-INLINE void __kmpc_impl_unpack(uint64_t val, uint32_t &lo, uint32_t &hi) {
-  asm volatile("mov.b64 {%0,%1}, %2;" : "=r"(lo), "=r"(hi) : "l"(val));
-}
-
-INLINE uint64_t __kmpc_impl_pack(uint32_t lo, uint32_t hi) {
-  uint64_t val;
-  asm volatile("mov.b64 %0, {%1,%2};" : "=l"(val) : "r"(lo), "r"(hi));
-  return val;
-}
-
 typedef uint32_t __kmpc_impl_lanemask_t;
 
+namespace __kmpc_impl {
+class Bits : public Api<Bits> {
+  friend class Api<Bits>;
+
+private:
+  INLINE static uint64_t packImpl(uint32_t lo, uint32_t hi) {
+    uint64_t val;
+    asm volatile("mov.b64 %0, {%1,%2};" : "=l"(val) : "r"(lo), "r"(hi));
+    return val;
+  }
+
+  INLINE static void unpackImpl(uint64_t val, uint32_t &lo, uint32_t &hi) {
+    asm volatile("mov.b64 {%0,%1}, %2;" : "=r"(lo), "=r"(hi) : "l"(val));
+  }
+};
+} // namespace __kmpc_impl
+
 INLINE __kmpc_impl_lanemask_t __kmpc_impl_lanemask_lt() {
   __kmpc_impl_lanemask_t res;
   asm("mov.u32 %0, %%lanemask_lt;" : "=r"(res));