diff --git a/include/oneapi/dpl/pstl/omp/util.h b/include/oneapi/dpl/pstl/omp/util.h
index e7c4e3cbc40..b5fb399e2b5 100644
--- a/include/oneapi/dpl/pstl/omp/util.h
+++ b/include/oneapi/dpl/pstl/omp/util.h
@@ -58,33 +58,7 @@ __cancel_execution(oneapi::dpl::__internal::__omp_backend_tag)
 //------------------------------------------------------------------------
 
 template <typename _ExecutionPolicy, typename _Tp>
-class __buffer_impl
-{
-    std::allocator<_Tp> __allocator_;
-    _Tp* __ptr_;
-    const std::size_t __buf_size_;
-    __buffer_impl(const __buffer_impl&) = delete;
-    void
-    operator=(const __buffer_impl&) = delete;
-
-  public:
-    __buffer_impl(_ExecutionPolicy /*__exec*/, std::size_t __n)
-        : __allocator_(), __ptr_(__allocator_.allocate(__n)), __buf_size_(__n)
-    {
-    }
-
-    operator bool() const { return __ptr_ != nullptr; }
-
-    _Tp*
-    get() const
-    {
-        return __ptr_;
-    }
-    ~__buffer_impl() { __allocator_.deallocate(__ptr_, __buf_size_); }
-};
-
-template <typename _ExecutionPolicy, typename _Tp>
-using __buffer = __buffer_impl<::std::decay_t<_ExecutionPolicy>, _Tp>;
+using __buffer = oneapi::dpl::__utils::__buffer_impl<std::decay_t<_ExecutionPolicy>, _Tp, std::allocator>;
 
 // Preliminary size of each chunk: requires further discussion
 constexpr std::size_t __default_chunk_size = 2048;
diff --git a/include/oneapi/dpl/pstl/parallel_backend_serial.h b/include/oneapi/dpl/pstl/parallel_backend_serial.h
index edd6652d359..6acd4b617f9 100644
--- a/include/oneapi/dpl/pstl/parallel_backend_serial.h
+++ b/include/oneapi/dpl/pstl/parallel_backend_serial.h
@@ -25,6 +25,8 @@
 #include <utility>
 #include <type_traits>
 
+#include "parallel_backend_utils.h"
+
 namespace oneapi
 {
 namespace dpl
@@ -33,32 +35,7 @@ namespace __serial_backend
 {
 
 template <typename _ExecutionPolicy, typename _Tp>
-class __buffer_impl
-{
-    ::std::allocator<_Tp> __allocator_;
-    _Tp* __ptr_;
-    const ::std::size_t __buf_size_;
-    __buffer_impl(const __buffer_impl&) = delete;
-    void
-    operator=(const __buffer_impl&) = delete;
-
-  public:
-    __buffer_impl(_ExecutionPolicy /*__exec*/, ::std::size_t __n)
-        : __allocator_(), __ptr_(__allocator_.allocate(__n)), __buf_size_(__n)
-    {
-    }
-
-    operator bool() const { return __ptr_ != nullptr; }
-    _Tp*
-    get() const
-    {
-        return __ptr_;
-    }
-    ~__buffer_impl() { __allocator_.deallocate(__ptr_, __buf_size_); }
-};
-
-template <typename _ExecutionPolicy, typename _Tp>
-using __buffer = __buffer_impl<::std::decay_t<_ExecutionPolicy>, _Tp>;
+using __buffer = oneapi::dpl::__utils::__buffer_impl<std::decay_t<_ExecutionPolicy>, _Tp, std::allocator>;
 
 inline void
 __cancel_execution(oneapi::dpl::__internal::__serial_backend_tag)
diff --git a/include/oneapi/dpl/pstl/parallel_backend_tbb.h b/include/oneapi/dpl/pstl/parallel_backend_tbb.h
index 2ddfa61007a..e7b1fa2d646 100644
--- a/include/oneapi/dpl/pstl/parallel_backend_tbb.h
+++ b/include/oneapi/dpl/pstl/parallel_backend_tbb.h
@@ -55,35 +55,7 @@ not an initialize array, because initialization/destruction
 would make the span be at least O(N). */
 // tbb::allocator can improve performance in some cases.
 template <typename _ExecutionPolicy, typename _Tp>
-class __buffer_impl
-{
-    tbb::tbb_allocator<_Tp> _M_allocator;
-    _Tp* _M_ptr;
-    const ::std::size_t _M_buf_size;
-    __buffer_impl(const __buffer_impl&) = delete;
-    void
-    operator=(const __buffer_impl&) = delete;
-
-  public:
-    //! Try to obtain buffer of given size to store objects of _Tp type
-    __buffer_impl(_ExecutionPolicy /*__exec*/, const ::std::size_t __n)
-        : _M_allocator(), _M_ptr(_M_allocator.allocate(__n)), _M_buf_size(__n)
-    {
-    }
-    //! True if buffer was successfully obtained, zero otherwise.
-    operator bool() const { return _M_ptr != nullptr; }
-    //! Return pointer to buffer, or nullptr if buffer could not be obtained.
-    _Tp*
-    get() const
-    {
-        return _M_ptr;
-    }
-    //! Destroy buffer
-    ~__buffer_impl() { _M_allocator.deallocate(_M_ptr, _M_buf_size); }
-};
-
-template <typename _ExecutionPolicy, typename _Tp>
-using __buffer = __buffer_impl<::std::decay_t<_ExecutionPolicy>, _Tp>;
+using __buffer = oneapi::dpl::__utils::__buffer_impl<std::decay_t<_ExecutionPolicy>, _Tp, tbb::tbb_allocator>;
 
 // Wrapper for tbb::task
 inline void
diff --git a/include/oneapi/dpl/pstl/parallel_backend_utils.h b/include/oneapi/dpl/pstl/parallel_backend_utils.h
index c539d0f7e4e..502c5b92bdd 100644
--- a/include/oneapi/dpl/pstl/parallel_backend_utils.h
+++ b/include/oneapi/dpl/pstl/parallel_backend_utils.h
@@ -25,10 +25,42 @@ namespace oneapi
 {
 namespace dpl
 {
-
 namespace __utils
 {
 
+//------------------------------------------------------------------------
+// raw buffer (with specified _TAllocator)
+//------------------------------------------------------------------------
+
+template <typename _ExecutionPolicy, typename _Tp, template <typename _T> typename _TAllocator>
+class __buffer_impl
+{
+    _TAllocator<_Tp> _M_allocator;
+    _Tp* _M_ptr = nullptr;
+    const ::std::size_t _M_buf_size = 0;
+
+    __buffer_impl(const __buffer_impl&) = delete;
+    void
+    operator=(const __buffer_impl&) = delete;
+
+  public:
+    //! Try to obtain buffer of given size to store objects of _Tp type
+    __buffer_impl(_ExecutionPolicy /*__exec*/, const ::std::size_t __n)
+        : _M_allocator(), _M_ptr(_M_allocator.allocate(__n)), _M_buf_size(__n)
+    {
+    }
+    //! True if buffer was successfully obtained, zero otherwise.
+    operator bool() const { return _M_ptr != nullptr; }
+    //! Return pointer to buffer, or nullptr if buffer could not be obtained.
+    _Tp*
+    get() const
+    {
+        return _M_ptr;
+    }
+    //! Destroy buffer
+    ~__buffer_impl() { _M_allocator.deallocate(_M_ptr, _M_buf_size); }
+};
+
 //! Destroy sequence [xs,xe)
 struct __serial_destroy
 {