Merge pull request #40078 from Algunenano/threadlocal_perf

Optimize CurrentMemoryTracker alloc and free

cmake/warnings.cmake

@@ -23,6 +23,7 @@ if (COMPILER_CLANG)
     no_warning(zero-length-array)
     no_warning(c++98-compat-pedantic)
     no_warning(c++98-compat)
+    no_warning(c++20-compat) # Use constinit in C++20 without warnings
     no_warning(conversion)
     no_warning(ctad-maybe-unsupported) # clang 9+, linux-only
     no_warning(disabled-macro-expansion)

src/Common/CurrentThread.cpp

@@ -43,13 +43,6 @@ ProfileEvents::Counters & CurrentThread::getProfileEvents()
     return current_thread ? current_thread->performance_counters : ProfileEvents::global_counters;
 }
 
-MemoryTracker * CurrentThread::getMemoryTracker()
-{
-    if (unlikely(!current_thread))
-        return nullptr;
-    return &current_thread->memory_tracker;
-}
-
 void CurrentThread::updateProgressIn(const Progress & value)
 {
     if (unlikely(!current_thread))

src/Common/CurrentThread.h

@@ -54,7 +54,12 @@ public:
     static void updatePerformanceCounters();
 
     static ProfileEvents::Counters & getProfileEvents();
-    static MemoryTracker * getMemoryTracker();
+    inline ALWAYS_INLINE static MemoryTracker * getMemoryTracker()
+    {
+        if (unlikely(!current_thread))
+            return nullptr;
+        return &current_thread->memory_tracker;
+    }
 
     /// Update read and write rows (bytes) statistics (used in system.query_thread_log)
     static void updateProgressIn(const Progress & value);

src/Common/ThreadStatus.cpp

@@ -24,9 +24,7 @@ namespace ErrorCodes
     extern const int LOGICAL_ERROR;
 }
 
-
-thread_local ThreadStatus * current_thread = nullptr;
-thread_local ThreadStatus * main_thread = nullptr;
+thread_local ThreadStatus constinit * current_thread = nullptr;
 
 #if !defined(SANITIZER)
 namespace

src/Common/ThreadStatus.h

@@ -102,8 +102,16 @@ public:
 
 using ThreadGroupStatusPtr = std::shared_ptr<ThreadGroupStatus>;
 
-
-extern thread_local ThreadStatus * current_thread;
+/**
+ * We use **constinit** here to tell the compiler the current_thread variable is initialized.
+ * If we didn't help the compiler, then it would most likely add a check before every use of the variable to initialize it if needed.
+ * Instead it will trust that we are doing the right thing (and we do initialize it to nullptr) and emit more optimal code.
+ * This is noticeable in functions like CurrentMemoryTracker::free and CurrentMemoryTracker::allocImpl
+ * See also:
+ * - https://en.cppreference.com/w/cpp/language/constinit
+ * - https://github.com/ClickHouse/ClickHouse/pull/40078
+ */
+extern thread_local constinit ThreadStatus * current_thread;
 
 /** Encapsulates all per-thread info (ProfileEvents, MemoryTracker, query_id, query context, etc.).
   * The object must be created in thread function and destroyed in the same thread before the exit.