diff --git a/src/backend/port/atomics.c b/src/backend/port/atomics.c
index 3f53ec13300ca687fa7fc6edb1ac30fbc0b21c2d..f57e6723017560147d1348efe2161191a5c51b53 100644
--- a/src/backend/port/atomics.c
+++ b/src/backend/port/atomics.c
@@ -32,6 +32,14 @@ pg_spinlock_barrier(void)
 }
 #endif
 
+#ifdef PG_HAVE_COMPILER_BARRIER_EMULATION
+void
+pg_extern_compiler_barrier(void)
+{
+	/* do nothing */
+}
+#endif
+
 
 #ifdef PG_HAVE_ATOMIC_FLAG_SIMULATION
 
diff --git a/src/include/port/atomics/fallback.h b/src/include/port/atomics/fallback.h
index be04ed3c123f2e46b8409b8964a42e0de6c5bba4..b1a9a6a7cc743bbe291c0faec402cbbfd9367099 100644
--- a/src/include/port/atomics/fallback.h
+++ b/src/include/port/atomics/fallback.h
@@ -35,6 +35,22 @@ extern void pg_spinlock_barrier(void);
 #define pg_memory_barrier_impl pg_spinlock_barrier
 #endif
 
+#ifndef pg_compiler_barrier_impl
+/*
+ * If the compiler/arch combination does not provide compiler barriers,
+ * provide a fallback. That fallback simply consists out of a function call
+ * into a externally defined function.  That should guarantee compiler barrier
+ * semantics except for compilers that do inter translation unit/global
+ * optimization - those better provide an actual compiler barrier.
+ *
+ * Using a native compiler barrier for sure is a lot faster than this...
+ */
+#define PG_HAVE_COMPILER_BARRIER_EMULATION
+extern void pg_extern_compiler_barrier(void);
+#define pg_compiler_barrier_impl pg_extern_compiler_barrier
+#endif
+
+
 /*
  * If we have atomics implementation for this platform fall back to providing
  * the atomics API using a spinlock to protect the internal state. Possibly