diff --git a/src/runtime/future/future.c b/src/runtime/future/future.c
index ec47ed684..1c1d1fa18 100644
--- a/src/runtime/future/future.c
+++ b/src/runtime/future/future.c
@@ -121,16 +121,16 @@ void yield(pony_actor_t* a) {
   suspend(r);
 }
 
-inline volatile bool future_fulfilled(future_t *fut) {
+inline bool future_fulfilled(future_t *fut) {
   return future_actor_get_fulfilled(fut);
 }
 
-inline volatile void *future_read_value(future_t *fut) {
+inline void *future_read_value(future_t *fut) {
   return future_actor_get_value(fut);
 }
 
-volatile void *future_get(future_t *fut, pony_actor_t* actor) {
-  volatile void *result;
+void *future_get(future_t *fut, pony_actor_t* actor) {
+  void *result;
   if (future_actor_get_value_and_fulfillment(fut, &result)) {
     return result;
   }
@@ -138,7 +138,7 @@ volatile void *future_get(future_t *fut, pony_actor_t* actor) {
   return future_read_value(fut);
 }
 
-void future_fulfil(future_t *f, volatile void *value) {
+void future_fulfil(future_t *f, void *value) {
   future_actor_set_value(f, value);
 #ifdef DEBUG_PRINT
   fprintf(stderr, "[%p]\t%p <--- fulfil\n", (void *)pthread_self(), f);
diff --git a/src/runtime/future/future.h b/src/runtime/future/future.h
index ae594841a..273b31626 100644
--- a/src/runtime/future/future.h
+++ b/src/runtime/future/future.h
@@ -18,9 +18,9 @@ typedef enum {
 // =============================================================================
 
 future_t *future_mk();
-volatile bool future_fulfilled(future_t *future);
-volatile void *future_read_value(future_t *future);
-void future_fulfil(future_t *future, volatile void *value);
+bool future_fulfilled(future_t *future);
+void *future_read_value(future_t *future);
+void future_fulfil(future_t *future, void *value);
 
 // =============================================================================
 // Actor-specific parts of the future library
@@ -29,7 +29,7 @@ void future_fulfil(future_t *future, volatile void *value);
 void future_chain(future_t *future, pony_actor_t* actor_owning_closure, struct closure *closure);
 void future_await(future_t *future, pony_actor_t* actor_awaiting_future);
 void future_block(future_t *future, pony_actor_t* actor_blocking_on_future);
-volatile void *future_get(future_t *future, pony_actor_t* actor_blocking_on_future);
+void *future_get(future_t *future, pony_actor_t* actor_blocking_on_future);
 
 // =============================================================================
 // Task-specific parts of the future library
diff --git a/src/runtime/future/future_actor.c b/src/runtime/future/future_actor.c
index 89deadc31..caca2f383 100644
--- a/src/runtime/future/future_actor.c
+++ b/src/runtime/future/future_actor.c
@@ -14,8 +14,8 @@
 #define DEBUG_PRINT 1
 
 typedef struct future_actor_fields {
-  volatile bool fulfilled;
-  volatile void *value;
+  bool fulfilled;
+  void *value;
   set_t *blocked;
   set_t *chained;
   set_t *awaiting;
@@ -189,8 +189,7 @@ void future_actor_dispatch(pony_actor_t* this, void* p, uint64_t id, int argc, p
     }
   case FUT_MSG_FULFIL:
     {
-      // Discarding volatile here because it does not matter
-      volatile void *value;
+      void *value;
       bool fulfilled;
       fulfilled = future_actor_get_value_and_fulfillment(this, &value);
 
@@ -227,24 +226,24 @@ future_actor_fields *get_fields(pony_actor_t *this) {
   return *(future_actor_fields **)this;
 }
 
-void future_actor_set_value(pony_actor_t* this, volatile void *value) {
+void future_actor_set_value(pony_actor_t* this, void *value) {
   future_actor_fields *state = get_fields(this);
   state->value = value;
   __sync_synchronize();
   state->fulfilled = true;
 }
-volatile bool future_actor_get_value_and_fulfillment(pony_actor_t* this, volatile void **value) {
+bool future_actor_get_value_and_fulfillment(pony_actor_t* this, void **value) {
   future_actor_fields *state = get_fields(this);
-  volatile bool fulfilled = state->fulfilled;
+  bool fulfilled = state->fulfilled;
   __sync_synchronize();
   *value = state->value;
   return fulfilled;
 }
-volatile bool future_actor_get_fulfilled(pony_actor_t* this) {
+bool future_actor_get_fulfilled(pony_actor_t* this) {
   future_actor_fields *state = get_fields(this);
   return state->fulfilled;
 }
-volatile void *future_actor_get_value(pony_actor_t* this) {
+void *future_actor_get_value(pony_actor_t* this) {
   future_actor_fields *state = get_fields(this);
   return state->value;
 }
diff --git a/src/runtime/future/future_actor.h b/src/runtime/future/future_actor.h
index ba99ed6db..bb71b36e0 100644
--- a/src/runtime/future/future_actor.h
+++ b/src/runtime/future/future_actor.h
@@ -16,9 +16,9 @@ enum
 pony_actor_t* future_create();
 pony_msg_t* future_actor_message_type(uint64_t id);
 void future_actor_dispatch(pony_actor_t* this, void* p, uint64_t id, int argc, pony_arg_t* argv);
-volatile bool future_actor_get_value_and_fulfillment(pony_actor_t* this, volatile void **value);
-void future_actor_set_value(pony_actor_t* this, volatile void *value);
-volatile bool future_actor_get_fulfilled(pony_actor_t* this);
-volatile void *future_actor_get_value(pony_actor_t* this);
+bool future_actor_get_value_and_fulfillment(pony_actor_t* this, void **value);
+void future_actor_set_value(pony_actor_t* this, void *value);
+bool future_actor_get_fulfilled(pony_actor_t* this);
+void *future_actor_get_value(pony_actor_t* this);
 
 #endif
diff --git a/src/runtime/future/tit_lazy.c b/src/runtime/future/tit_lazy.c
index c96e34ec9..3ea36ee40 100644
--- a/src/runtime/future/tit_lazy.c
+++ b/src/runtime/future/tit_lazy.c
@@ -31,8 +31,8 @@ static pthread_once_t stack_pool_is_initialized = PTHREAD_ONCE_INIT;
 static mpmcq_t proper_stacks_for_reuse;
 static mpmcq_t side_stacks_for_reuse;
 static __thread lazy_tit_t *current;
-static volatile unsigned int cached_side_stacks;
-static unsigned int side_stack_cache_size;
+static uint32_t cached_side_stacks;
+static uint32_t side_stack_cache_size;
 
 static void trampoline_0(lazy_tit_t *new, fun_t_0 fun);
 // static void trampoline_1(lazy_tit_t *new, fun_t_1 fun, void *a);
@@ -47,14 +47,15 @@ static inline void __end_trampoline();
 
 void mk_stack(ucontext_t *fork) {
   if (side_stack_cache_size > 0) {
-    while (fork->uc_stack.ss_sp == NULL && cached_side_stacks > 0) {
+    while (fork->uc_stack.ss_sp == NULL && __atomic_load_n(&cached_side_stacks, __ATOMIC_CONSUME) > 0) {
       fork->uc_stack.ss_sp = mpmcq_pop(&side_stacks_for_reuse);
     }
     if (fork->uc_stack.ss_sp) {
-      int cache_size;
+      uint32_t cache_size;
       do {
-        cache_size = cached_side_stacks;
-      } while (!__sync_bool_compare_and_swap(&cached_side_stacks, cache_size, cache_size - 1));
+        cache_size = __atomic_load_n(&cached_side_stacks, __ATOMIC_CONSUME); 
+      } while (!__atomic_compare_exchange_n(&cached_side_stacks, &cache_size, cache_size - 1, false, __ATOMIC_RELEASE, __ATOMIC_RELAXED));
+      // TODO: check the correctness and optimality of line above
     }
   }
 
@@ -69,10 +70,11 @@ void mk_stack(ucontext_t *fork) {
 
 void return_allocated_stack_to_pool(void *stack_pointer) {
   bool free_stack = true;
-  int current_cache_size;
+  uint32_t current_cache_size;
   do {
-    current_cache_size = cached_side_stacks;
-    if (__sync_bool_compare_and_swap(&cached_side_stacks, current_cache_size, current_cache_size + 1)) {
+    current_cache_size = __atomic_load_n(&cached_side_stacks, __ATOMIC_CONSUME); 
+    // TODO: check the correctness and optimality of line below
+    if (__atomic_compare_exchange_n(&cached_side_stacks, &current_cache_size, current_cache_size + 1, false, __ATOMIC_RELEASE, __ATOMIC_RELAXED)) {
       mpmcq_push(&side_stacks_for_reuse, stack_pointer);
       free_stack = false;
       break;