[libcxx-commits] [libcxx] [libc++] Improve performance of std::atomic_flag on Windows (PR #163524)

[Roger Sanders via libcxx-commits]

https://github.com/RogerSanders created https://github.com/llvm/llvm-project/pull/163524

On Windows 8 and above, the WaitOnAddress, WakeByAddressSingle and WakeByAddressAll functions allow efficient implementation of the C++20 wait and notify features of std::atomic_flag. These Windows functions have never been made use of in libc++, leading to very poor performance of these features on Windows platforms, as they are implemented using a spin loop with backoff, rather than using any OS thread signalling whatsoever. This change implements the use of these OS functions where available, falling back to the original implementation on Windows versions prior to 8.

Fixes #127221

>From 63ee5dad50b8cb0e20abb8948030bd6218144b3c Mon Sep 17 00:00:00 2001
From: Roger Sanders <roger.sanders at maptek.com.au>
Date: Wed, 15 Oct 2025 19:30:41 +1100
Subject: [PATCH] [libc++] Improve performance of std::atomic_flag on Windows

On Windows 8 and above, the WaitOnAddress, WakeByAddressSingle and
WakeByAddressAll functions allow efficient implementation of the C++20
wait and notify features of std::atomic_flag. These Windows functions
have never been made use of in libc++, leading to very poor performance
of these features on Windows platforms, as they are implemented using a
spin loop with backoff, rather than using any OS thread signalling
whatsoever. This change implements the use of these OS functions where
available, falling back to the original implementation on Windows
versions prior to 8.

Fixes #127221
---
 libcxx/src/atomic.cpp | 35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

diff --git a/libcxx/src/atomic.cpp b/libcxx/src/atomic.cpp
index b214ba1fd11c0..4b9e53c264787 100644
--- a/libcxx/src/atomic.cpp
+++ b/libcxx/src/atomic.cpp
@@ -41,6 +41,10 @@
 // OpenBSD has no indirect syscalls
 #  define _LIBCPP_FUTEX(...) futex(__VA_ARGS__)
 
+#elif defined(_WIN32)
+
+#  include <windows.h>
+
 #else // <- Add other operating systems here
 
 // Baseline needs no new headers
@@ -101,6 +105,37 @@ static void __libcpp_platform_wake_by_address(__cxx_atomic_contention_t const vo
   _umtx_op(const_cast<__cxx_atomic_contention_t*>(__ptr), UMTX_OP_WAKE, __notify_one ? 1 : INT_MAX, nullptr, nullptr);
 }
 
+#elif defined(_WIN32)
+
+static void
+__libcpp_platform_wait_on_address(__cxx_atomic_contention_t const volatile* __ptr, __cxx_contention_t __val) {
+  static auto pWaitOnAddress = reinterpret_cast<BOOL(WINAPI*)(volatile void*, PVOID, SIZE_T, DWORD)>(
+      GetProcAddress(GetModuleHandleW(L"api-ms-win-core-synch-l1-2-0.dll"), "WaitOnAddress"));
+  if (pWaitOnAddress != nullptr) {
+    pWaitOnAddress(const_cast<__cxx_atomic_contention_t*>(__ptr), &__val, sizeof(__val), INFINITE);
+  } else {
+    __libcpp_thread_poll_with_backoff(
+        [=]() -> bool { return !__cxx_nonatomic_compare_equal(__cxx_atomic_load(__ptr, memory_order_relaxed), __val); },
+        __libcpp_timed_backoff_policy());
+  }
+}
+
+static void __libcpp_platform_wake_by_address(__cxx_atomic_contention_t const volatile* __ptr, bool __notify_one) {
+  if (__notify_one) {
+    static auto pWakeByAddressSingle = reinterpret_cast<void(WINAPI*)(PVOID)>(
+        GetProcAddress(GetModuleHandleW(L"api-ms-win-core-synch-l1-2-0.dll"), "WakeByAddressSingle"));
+    if (pWakeByAddressSingle != nullptr) {
+      pWakeByAddressSingle(const_cast<__cxx_atomic_contention_t*>(__ptr));
+    }
+  } else {
+    static auto pWakeByAddressAll = reinterpret_cast<void(WINAPI*)(PVOID)>(
+        GetProcAddress(GetModuleHandleW(L"api-ms-win-core-synch-l1-2-0.dll"), "WakeByAddressAll"));
+    if (pWakeByAddressAll != nullptr) {
+      pWakeByAddressAll(const_cast<__cxx_atomic_contention_t*>(__ptr));
+    }
+  }
+}
+
 #else // <- Add other operating systems here
 
 // Baseline is just a timed backoff