[libc] inline fast path of callonce

[SchrodingerZhu]

Split from #91572

[llvmbot]

@llvm/pr-subscribers-libc

Author: Schrodinger ZHU Yifan (SchrodingerZhu)

Changes

Split from #91572

---

Full diff: https://github.com/llvm/llvm-project/pull/96226.diff

2 Files Affected:

• (modified) libc/src/__support/threads/callonce.h (+27-2)
• (modified) libc/src/__support/threads/linux/callonce.cpp (+3-15)

diff --git a/libc/src/__support/threads/callonce.h b/libc/src/__support/threads/callonce.h
index b3d6813f7dda9..55199b6e30906 100644
--- a/libc/src/__support/threads/callonce.h
+++ b/libc/src/__support/threads/callonce.h
@@ -9,13 +9,38 @@
 #ifndef LLVM_LIBC_SRC___SUPPORT_THREADS_CALLONCE_H
 #define LLVM_LIBC_SRC___SUPPORT_THREADS_CALLONCE_H
 
+#include "src/__support/macros/optimization.h"
+
+#ifdef __linux__
+#include "src/__support/threads/linux/futex_utils.h"
+#else
+#error "callonce is not supported on this platform"
+#endif
+
 namespace LIBC_NAMESPACE {
 
-struct CallOnceFlag;
+#ifdef __linux__
+using CallOnceFlag = Futex;
+#endif
 using CallOnceCallback = void(void);
+namespace callonce_impl {
+static constexpr FutexWordType NOT_CALLED = 0x0;
+static constexpr FutexWordType START = 0x11;
+static constexpr FutexWordType WAITING = 0x22;
+static constexpr FutexWordType FINISH = 0x33;
+int callonce_slowpath(CallOnceFlag *flag, CallOnceCallback *callback);
+} // namespace callonce_impl
 
-int callonce(CallOnceFlag *flag, CallOnceCallback *callback);
+LIBC_INLINE int callonce(CallOnceFlag *flag, CallOnceCallback *callback) {
+  using namespace callonce_impl;
+  // Avoid cmpxchg operation if the function has already been called.
+  // The destination operand of cmpxchg may receive a write cycle without
+  // regard to the result of the comparison
+  if (LIBC_LIKELY(flag->load(cpp::MemoryOrder::RELAXED) == FINISH))
+    return 0;
 
+  return callonce_slowpath(flag, callback);
+}
 } // namespace LIBC_NAMESPACE
 
 #endif // LLVM_LIBC_SRC___SUPPORT_THREADS_CALLONCE_H
diff --git a/libc/src/__support/threads/linux/callonce.cpp b/libc/src/__support/threads/linux/callonce.cpp
index b48a514a44875..2bea2da2ae21d 100644
--- a/libc/src/__support/threads/linux/callonce.cpp
+++ b/libc/src/__support/threads/linux/callonce.cpp
@@ -7,27 +7,15 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/__support/threads/callonce.h"
-#include "src/__support/macros/optimization.h"
 #include "src/__support/threads/linux/futex_utils.h"
 
 namespace LIBC_NAMESPACE {
-
-static constexpr FutexWordType NOT_CALLED = 0x0;
-static constexpr FutexWordType START = 0x11;
-static constexpr FutexWordType WAITING = 0x22;
-static constexpr FutexWordType FINISH = 0x33;
-
-int callonce(CallOnceFlag *flag, CallOnceCallback *func) {
+namespace callonce_impl {
+int callonce_slowpath(CallOnceFlag *flag, CallOnceCallback *func) {
   auto *futex_word = reinterpret_cast<Futex *>(flag);
 
   FutexWordType not_called = NOT_CALLED;
 
-  // Avoid cmpxchg operation if the function has already been called.
-  // The destination operand of cmpxchg may receive a write cycle without
-  // regard to the result of the comparison
-  if (LIBC_LIKELY(futex_word->load(cpp::MemoryOrder::RELAXED) == FINISH))
-    return 0;
-
   // The call_once call can return only after the called function |func|
   // returns. So, we use futexes to synchronize calls with the same flag value.
   if (futex_word->compare_exchange_strong(not_called, START)) {
@@ -46,5 +34,5 @@ int callonce(CallOnceFlag *flag, CallOnceCallback *func) {
 
   return 0;
 }
-
+} // namespace callonce_impl
 } // namespace LIBC_NAMESPACE