diff --git a/common.py b/common.py
index 7d6685b..a245d8f 100644
--- a/common.py
+++ b/common.py
@@ -1541,6 +1541,8 @@ https://github.com/cirosantilli/linux-kernel-module-cheat#gem5-debug-build
                     not my_path_properties['extra_objs_disable_baremetal_bootloader']
                 ):
                     extra_objs.extend(extra_objs_baremetal_bootloader)
+                if self.env['mode'] == 'userland':
+                    cc_flags_after.extend(['-pthread', LF])
             if self.need_rebuild([in_path] + extra_objs + extra_deps, out_path):
                 cc_flags.extend(my_path_properties['cc_flags'])
                 cc_flags_after.extend(my_path_properties['cc_flags_after'])
diff --git a/path_properties.py b/path_properties.py
index 9efb79e..d9ddb3c 100644
--- a/path_properties.py
+++ b/path_properties.py
@@ -507,7 +507,6 @@ path_properties_tuples = (
                         'proc_events.c': {'requires_sudo': True},
                         'sched_getaffinity.c': {'requires_syscall_getcpu': True},
                         'sched_getaffinity_threads.c': {
-                            'cc_flags_after': ['-pthread', LF],
                             'more_than_1s': True,
                             'requires_syscall_getcpu': True,
                         },
diff --git a/userland/cpp/atomic.cpp b/userland/cpp/atomic.cpp
index 64a041f..34e9aab 100644
--- a/userland/cpp/atomic.cpp
+++ b/userland/cpp/atomic.cpp
@@ -16,12 +16,31 @@
 #if __cplusplus >= 201103L
 std::atomic_ulong my_atomic_ulong(0);
 unsigned long my_non_atomic_ulong = 0;
+#if defined(__x86_64__)
+unsigned long my_arch_atomic_ulong = 0;
+unsigned long my_arch_non_atomic_ulong = 0;
+#endif
 size_t niters;
 
 void threadMain() {
     for (size_t i = 0; i < niters; ++i) {
         my_atomic_ulong++;
         my_non_atomic_ulong++;
+#if defined(__x86_64__)
+        __asm__ __volatile__ (
+            "incq %0;"
+            : "+m" (my_arch_non_atomic_ulong)
+            :
+            :
+        );
+        __asm__ __volatile__ (
+            "lock;"
+            "incq %0;"
+            : "+m" (my_arch_atomic_ulong)
+            :
+            :
+        );
+#endif
     }
 }
 #endif
@@ -37,7 +56,7 @@ int main(int argc, char **argv) {
     if (argc > 2) {
         niters = std::stoull(argv[2], NULL, 0);
     } else {
-        niters = 1000;
+        niters = 10000;
     }
     std::vector<std::thread> threads(nthreads);
     for (size_t i = 0; i < nthreads; ++i)
@@ -45,8 +64,12 @@ int main(int argc, char **argv) {
     for (size_t i = 0; i < nthreads; ++i)
         threads[i].join();
     assert(my_atomic_ulong.load() == nthreads * niters);
-    // Same as above through `operator T`.
-    assert(my_atomic_ulong == nthreads * niters);
-    std::cout << my_non_atomic_ulong << std::endl;
+    // We can also use the atomics direclty through `operator T` conversion.
+    assert(my_atomic_ulong == my_atomic_ulong.load());
+    std::cout << "my_non_atomic_ulong " << my_non_atomic_ulong << std::endl;
+#if defined(__x86_64__)
+    assert(my_arch_atomic_ulong == nthreads * niters);
+    std::cout << "my_arch_non_atomic_ulong " << my_arch_non_atomic_ulong << std::endl;
+#endif
 #endif
 }