diff --git a/include/coz.h b/include/coz.h
index 6f4bf1c..dc0bcc8 100644
--- a/include/coz.h
+++ b/include/coz.h
@@ -37,29 +37,55 @@ typedef struct {
   size_t backoff;  // Used to batch updates to the shared counter. Currently unused.
 } coz_counter_t;
 
-// The type of the _coz_get_counter function
-typedef coz_counter_t* (*coz_get_counter_t)(int, const char*);
-
-// Locate and invoke _coz_get_counter
 static coz_counter_t* _call_coz_get_counter(int type, const char* name) {
-  static unsigned char _initialized = 0;
-  static coz_get_counter_t fn; // The pointer to _coz_get_counter
+  // Use memcpy to avoid pedantic GCC complaint about storing function pointer in void*
+  static unsigned char initialized = dlsym == NULL;
+  static coz_counter_t* (*fn)(int, const char*);
+
+  if (!initialized) {
+    void* p = dlsym(RTLD_DEFAULT, "_coz_get_counter");
+    memcpy(&fn, &p, sizeof(p));
+    initialized = true;
+  }
 
-  if(!_initialized) {
-    if(dlsym) {
-      // Locate the _coz_get_counter method
-      void* p = dlsym(RTLD_DEFAULT, "_coz_get_counter");
+  if (fn) return fn(type, name);
+  else return 0;
+}
 
-      // Use memcpy to avoid pedantic GCC complaint about storing function pointer in void*
-      memcpy(&fn, &p, sizeof(p));
-    }
+static void _call_coz_pre_block() {
+  static unsigned char initialized = dlsym == 0;
+  static void (*fn)();
 
-    _initialized = 1;
+  if (!initialized) {
+    void* p = dlsym(RTLD_DEFAULT, "_coz_pre_block");
+    memcpy(&fn, &p, sizeof(p));
+    initialized = true;
   }
+  if (fn) return fn();
+}
 
-  // Call the function, or return null if profiler is not found
-  if(fn) return fn(type, name);
-  else return 0;
+static void _call_coz_post_block(bool skip_delays) {
+  static unsigned char initialized = dlsym == 0;
+  static void (*fn)(bool);
+
+  if (!initialized) {
+    void* p = dlsym(RTLD_DEFAULT, "_coz_post_block");
+    memcpy(&fn, &p, sizeof(p));
+    initialized = true;
+  }
+  if (fn) return fn(skip_delays);
+}
+
+static void _call_coz_wake_other() {
+  static unsigned char initialized = dlsym == 0;
+  static void (*fn)();
+
+  if (!initialized) {
+    void* p = dlsym(RTLD_DEFAULT, "_coz_wake_other");
+    memcpy(&fn, &p, sizeof(p));
+    initialized = true;
+  }
+  if (fn) return fn();
 }
 
 // Macro to initialize and increment a counter
@@ -80,12 +106,80 @@ static coz_counter_t* _call_coz_get_counter(int type, const char* name) {
 #define STR2(x) #x
 #define STR(x) STR2(x)
 
+/// Indicate progress for the counter with the given name.
+///
+/// Example:
+/// ```
+/// while (more_events()) {
+///     COZ_PROGRESS_NAMED("event_processed");
+///     process_event();
+/// }
+/// ```
 #define COZ_PROGRESS_NAMED(name) COZ_INCREMENT_COUNTER(COZ_COUNTER_TYPE_THROUGHPUT, name)
 
+/// Indicate progress for the counter named implicitly after the file and line
+/// number it is placed on.
+///
+/// Example:
+/// ```
+/// while (more_events()) {
+///     COZ_PROGRESS;
+///     process_event();
+/// }
+/// ```
 #define COZ_PROGRESS COZ_INCREMENT_COUNTER(COZ_COUNTER_TYPE_THROUGHPUT, __FILE__ ":" STR(__LINE__))
 #define COZ_BEGIN(name) COZ_INCREMENT_COUNTER(COZ_COUNTER_TYPE_BEGIN, name)
 #define COZ_END(name) COZ_INCREMENT_COUNTER(COZ_COUNTER_TYPE_END, name)
 
+/// Call before (possibly) blocking (e.g. if you're about to maybe block on a
+/// futex).
+///
+/// Example:
+/// ```
+/// lock() {
+///   int expected = 0;
+///   if (!std::atomic_compare_exchange_weak(&futex, &expected, 1)) {
+///     COZ_PRE_BLOCK();
+///     syscall(SYS_futex, &futex, FUTEX_WAIT, ...);
+///     COZ_POST_BLOCK();
+///   }
+/// }
+/// ```
+#define COZ_PRE_BLOCK() _call_coz_pre_block()
+
+/// Call after unblocking. If skip_delays is true, all delays inserted during
+/// the blocked period will be skipped.
+///
+/// Example:
+/// ```
+/// lock() {
+///   int expected = 0;
+///   if (!std::atomic_compare_exchange_weak(&futex, &expected, 1)) {
+///     COZ_PRE_BLOCK();
+///     syscall(SYS_futex, &futex, FUTEX_WAIT, ...);
+///     COZ_POST_BLOCK();
+///   }
+/// }
+/// ```
+#define COZ_POST_BLOCK(skip_delays) _call_coz_post_block(skip_delays)
+
+/// Ensure a thread has executed all the required delays before possibly
+/// unblocking another thread.
+///
+/// Example:
+/// ```
+/// // Unlocking the futex.
+/// unlock() {
+///   let have_waiters = ...;
+///   std::atomic_store(&futex, 0);
+///   if (have_waiters) {
+///     COZ_WAKE_OTHER();
+///     syscall(SYS_FUTEX, &futex, FUTEX_WAKE, ...);
+///   }
+/// }
+/// ```
+#define COZ_WAKE_OTHER() _call_coz_wake_other()
+
 #if defined(__cplusplus)
 }
 #endif
diff --git a/libcoz/libcoz.cpp b/libcoz/libcoz.cpp
index 67daff9..5ba2300 100644
--- a/libcoz/libcoz.cpp
+++ b/libcoz/libcoz.cpp
@@ -63,6 +63,18 @@ extern "C" coz_counter_t* _coz_get_counter(progress_point_type t, const char* na
   }
 }
 
+extern "C" void _coz_pre_block() {
+  profiler::get_instance().pre_block();
+}
+
+extern "C" void _coz_post_block(bool skip_delays) {
+  profiler::get_instance().post_block(skip_delays);
+}
+
+extern "C" void _coz_wake_other() {
+  profiler::get_instance().catch_up();
+}
+
 /**
  * Read a link's contents and return it as a string
  */
diff --git a/libcoz/profiler.h b/libcoz/profiler.h
index 0998a22..2ae6c97 100644
--- a/libcoz/profiler.h
+++ b/libcoz/profiler.h
@@ -160,7 +160,8 @@ class profiler {
     state->pre_block_time = _global_delay.load();
   }
 
-  /// Call after unblocking. If by_thread is true, delays will be skipped
+  /// Call after unblocking. If skip_delays is true, all delays inserted
+  /// during the blocked period will be skipped.
   void post_block(bool skip_delays) {
     thread_state* state = get_thread_state();
     if(!state)