constexpr mutex constructor (microsoft#3824)

Co-authored-by: Stephan T. Lavavej <[email protected]>

stl/inc/mutex

@@ -31,37 +31,20 @@ _STD_BEGIN
 _EXPORT_STD class condition_variable;
 _EXPORT_STD class condition_variable_any;
 
-struct _Mtx_internal_imp_mirror {
-#ifdef _CRT_WINDOWS
-#ifdef _WIN64
-    static constexpr size_t _Critical_section_size = 16;
-#else // _WIN64
-    static constexpr size_t _Critical_section_size = 8;
-#endif // _WIN64
-#else // _CRT_WINDOWS
-#ifdef _WIN64
-    static constexpr size_t _Critical_section_size = 64;
-#else // _WIN64
-    static constexpr size_t _Critical_section_size = 36;
-#endif // _WIN64
-#endif // _CRT_WINDOWS
-
-    static constexpr size_t _Critical_section_align = alignof(void*);
-
-    int _Type;
-    _Aligned_storage_t<_Critical_section_size, _Critical_section_align> _Cs_storage;
-    long _Thread_id;
-    int _Count;
-};
-
-static_assert(sizeof(_Mtx_internal_imp_mirror) == _Mtx_internal_imp_size, "inconsistent size for mutex");
-static_assert(alignof(_Mtx_internal_imp_mirror) == _Mtx_internal_imp_alignment, "inconsistent alignment for mutex");
-
 class _Mutex_base { // base class for all mutex types
 public:
+#ifdef _DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR
     _Mutex_base(int _Flags = 0) noexcept {
         _Mtx_init_in_situ(_Mymtx(), _Flags | _Mtx_try);
     }
+#else // ^^^ _DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR / !_DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR vvv
+    constexpr _Mutex_base(int _Flags = 0) noexcept {
+        _Mtx_storage._Critical_section = {};
+        _Mtx_storage._Thread_id        = -1;
+        _Mtx_storage._Type             = _Flags | _Mtx_try;
+        _Mtx_storage._Count            = 0;
+    }
+#endif // !_DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR
 
     ~_Mutex_base() noexcept {
         _Mtx_destroy_in_situ(_Mymtx());
@@ -97,9 +80,9 @@ public:
 
 protected:
     _NODISCARD_TRY_CHANGE_STATE bool _Verify_ownership_levels() noexcept {
-        if (_Mtx_storage_mirror._Count == INT_MAX) {
+        if (_Mtx_storage._Count == INT_MAX) {
             // only occurs for recursive mutexes (N4950 [thread.mutex.recursive]/3)
-            --_Mtx_storage_mirror._Count;
+            --_Mtx_storage._Count;
             return false;
         }
 
@@ -110,23 +93,16 @@ private:
     friend condition_variable;
     friend condition_variable_any;
 
-    union {
-        _Aligned_storage_t<_Mtx_internal_imp_size, _Mtx_internal_imp_alignment> _Mtx_storage;
-        _Mtx_internal_imp_mirror _Mtx_storage_mirror;
-    };
+    _Mtx_internal_imp_t _Mtx_storage{};
 
-    _Mtx_t _Mymtx() noexcept { // get pointer to _Mtx_internal_imp_t inside _Mtx_storage
-        return reinterpret_cast<_Mtx_t>(&_Mtx_storage);
+    _Mtx_t _Mymtx() noexcept {
+        return &_Mtx_storage;
     }
 };
 
-static_assert(sizeof(_Mutex_base) == _Mtx_internal_imp_size, "inconsistent size for mutex");
-static_assert(alignof(_Mutex_base) == _Mtx_internal_imp_alignment, "inconsistent alignment for mutex");
-
 _EXPORT_STD class mutex : public _Mutex_base { // class for mutual exclusion
 public:
-    /* constexpr */ mutex() noexcept // TRANSITION, ABI
-        : _Mutex_base() {}
+    mutex() noexcept = default;
 
     mutex(const mutex&)            = delete;
     mutex& operator=(const mutex&) = delete;

stl/inc/xthreads.h

@@ -9,6 +9,7 @@
 #include <yvals_core.h>
 #if _STL_COMPILER_PREPROCESSOR
 #include <climits>
+#include <type_traits>
 #include <xtimec.h>
 
 #pragma pack(push, _CRT_PACKING)
@@ -25,40 +26,64 @@ struct _Thrd_t { // thread identifier for Win32
     _Thrd_id_t _Id;
 };
 
-// Size and alignment for _Mtx_internal_imp_t and _Cnd_internal_imp_t
+using _Smtx_t = void*;
+
+struct _Stl_critical_section {
+    void* _Unused       = nullptr; // TRANSITION, ABI: was the vptr
+    _Smtx_t _M_srw_lock = nullptr;
+};
+
+struct _Mtx_internal_imp_t {
+#ifdef _CRT_WINDOWS
+#ifdef _WIN64
+    static constexpr size_t _Critical_section_size = 16;
+#else // _WIN64
+    static constexpr size_t _Critical_section_size = 8;
+#endif // _WIN64
+#else // _CRT_WINDOWS
+#ifdef _WIN64
+    static constexpr size_t _Critical_section_size = 64;
+#else // _WIN64
+    static constexpr size_t _Critical_section_size = 36;
+#endif // _WIN64
+#endif // _CRT_WINDOWS
+
+    static constexpr size_t _Critical_section_align = alignof(void*);
+
+    int _Type{};
+    union {
+        _Stl_critical_section _Critical_section{};
+        _STD _Aligned_storage_t<_Critical_section_size, _Critical_section_align> _Cs_storage;
+    };
+    long _Thread_id{};
+    int _Count{};
+};
+
+// Size and alignment for _Cnd_internal_imp_t
 #ifdef _CRT_WINDOWS
 #ifdef _WIN64
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_size      = 32;
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_alignment = 8;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_size      = 16;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_alignment = 8;
 #else // _WIN64
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_size      = 20;
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_alignment = 4;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_size      = 8;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_alignment = 4;
 #endif // _WIN64
 #else // _CRT_WINDOWS
 #ifdef _WIN64
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_size      = 80;
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_alignment = 8;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_size      = 72;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_alignment = 8;
 #else // _WIN64
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_size      = 48;
-_INLINE_VAR constexpr size_t _Mtx_internal_imp_alignment = 4;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_size      = 40;
 _INLINE_VAR constexpr size_t _Cnd_internal_imp_alignment = 4;
 #endif // _WIN64
 #endif // _CRT_WINDOWS
 
-#ifdef _M_CEE // avoid warning LNK4248: unresolved typeref token for '_Mtx_internal_imp_t'; image may not run
-using _Mtx_t = void*;
+using _Mtx_t = _Mtx_internal_imp_t*;
+
+#ifdef _M_CEE // avoid warning LNK4248: unresolved typeref token for '_Cnd_internal_imp_t'; image may not run
 using _Cnd_t = void*;
 #else // ^^^ defined(_M_CEE) / !defined(_M_CEE) vvv
-struct _Mtx_internal_imp_t;
 struct _Cnd_internal_imp_t;
-using _Mtx_t = _Mtx_internal_imp_t*;
 using _Cnd_t = _Cnd_internal_imp_t*;
 #endif // ^^^ !defined(_M_CEE) ^^^
 
@@ -96,7 +121,6 @@ _CRTIMP2_PURE void __cdecl _Mtx_reset_owner(_Mtx_t);
 
 // shared mutex
 // these declarations must be in sync with those in sharedmutex.cpp
-using _Smtx_t = void*;
 void __cdecl _Smtx_lock_exclusive(_Smtx_t*);
 void __cdecl _Smtx_lock_shared(_Smtx_t*);
 int __cdecl _Smtx_try_lock_exclusive(_Smtx_t*);

stl/src/cond.cpp

@@ -51,7 +51,7 @@ void _Cnd_destroy(const _Cnd_t cond) { // clean up
 }
 
 int _Cnd_wait(const _Cnd_t cond, const _Mtx_t mtx) { // wait until signaled
-    const auto cs = static_cast<Concurrency::details::stl_critical_section_win7*>(_Mtx_getconcrtcs(mtx));
+    const auto cs = &mtx->_Critical_section;
     _Mtx_clear_owner(mtx);
     cond->_get_cv()->wait(cs);
     _Mtx_reset_owner(mtx);
@@ -61,7 +61,7 @@ int _Cnd_wait(const _Cnd_t cond, const _Mtx_t mtx) { // wait until signaled
 // wait until signaled or timeout
 int _Cnd_timedwait(const _Cnd_t cond, const _Mtx_t mtx, const _timespec64* const target) {
     int res       = _Thrd_success;
-    const auto cs = static_cast<Concurrency::details::stl_critical_section_win7*>(_Mtx_getconcrtcs(mtx));
+    const auto cs = &mtx->_Critical_section;
     if (target == nullptr) { // no target time specified, wait on mutex
         _Mtx_clear_owner(mtx);
         cond->_get_cv()->wait(cs);

stl/src/mutex.cpp

@@ -37,34 +37,20 @@ extern "C" [[noreturn]] _CRTIMP2_PURE void _Thrd_abort(const char* msg) { // abo
 enum class __stl_sync_api_modes_enum { normal, win7, vista, concrt };
 extern "C" _CRTIMP2 void __cdecl __set_stl_sync_api_mode(__stl_sync_api_modes_enum) {}
 
-struct _Mtx_internal_imp_t { // ConcRT mutex
-    int type;
-    typename std::_Aligned_storage<Concurrency::details::stl_critical_section_max_size,
-        Concurrency::details::stl_critical_section_max_alignment>::type cs;
-    long thread_id;
-    int count;
-    [[nodiscard]] Concurrency::details::stl_critical_section_win7* _get_cs() { // get pointer to implementation
-        return reinterpret_cast<Concurrency::details::stl_critical_section_win7*>(&cs);
-    }
-};
-
-static_assert(sizeof(_Mtx_internal_imp_t) == _Mtx_internal_imp_size, "incorrect _Mtx_internal_imp_size");
-static_assert(alignof(_Mtx_internal_imp_t) == _Mtx_internal_imp_alignment, "incorrect _Mtx_internal_imp_alignment");
-
-static_assert(
-    std::_Mtx_internal_imp_mirror::_Critical_section_size == Concurrency::details::stl_critical_section_max_size);
-static_assert(
-    std::_Mtx_internal_imp_mirror::_Critical_section_align == Concurrency::details::stl_critical_section_max_alignment);
+[[nodiscard]] static PSRWLOCK get_srw_lock(_Mtx_t mtx) {
+    return reinterpret_cast<PSRWLOCK>(&mtx->_Critical_section._M_srw_lock);
+}
 
+// TRANSITION, only used when constexpr mutex constructor is not enabled
 void _Mtx_init_in_situ(_Mtx_t mtx, int type) { // initialize mutex in situ
-    Concurrency::details::create_stl_critical_section(mtx->_get_cs());
-    mtx->thread_id = -1;
-    mtx->type      = type;
-    mtx->count     = 0;
+    Concurrency::details::create_stl_critical_section(&mtx->_Critical_section);
+    mtx->_Thread_id = -1;
+    mtx->_Type      = type;
+    mtx->_Count     = 0;
 }
 
 void _Mtx_destroy_in_situ(_Mtx_t mtx) { // destroy mutex in situ
-    _THREAD_ASSERT(mtx->count == 0, "mutex destroyed while busy");
+    _THREAD_ASSERT(mtx->_Count == 0, "mutex destroyed while busy");
     (void) mtx;
 }
 
@@ -91,27 +77,27 @@ void _Mtx_destroy(_Mtx_t mtx) { // destroy mutex
 }
 
 static int mtx_do_lock(_Mtx_t mtx, const _timespec64* target) { // lock mutex
-    if ((mtx->type & ~_Mtx_recursive) == _Mtx_plain) { // set the lock
-        if (mtx->thread_id != static_cast<long>(GetCurrentThreadId())) { // not current thread, do lock
-            mtx->_get_cs()->lock();
-            mtx->thread_id = static_cast<long>(GetCurrentThreadId());
+    if ((mtx->_Type & ~_Mtx_recursive) == _Mtx_plain) { // set the lock
+        if (mtx->_Thread_id != static_cast<long>(GetCurrentThreadId())) { // not current thread, do lock
+            AcquireSRWLockExclusive(get_srw_lock(mtx));
+            mtx->_Thread_id = static_cast<long>(GetCurrentThreadId());
         }
-        ++mtx->count;
+        ++mtx->_Count;
 
         return _Thrd_success;
     } else { // handle timed or recursive mutex
         int res = WAIT_TIMEOUT;
         if (target == nullptr) { // no target --> plain wait (i.e. infinite timeout)
-            if (mtx->thread_id != static_cast<long>(GetCurrentThreadId())) {
-                mtx->_get_cs()->lock();
+            if (mtx->_Thread_id != static_cast<long>(GetCurrentThreadId())) {
+                AcquireSRWLockExclusive(get_srw_lock(mtx));
             }
 
             res = WAIT_OBJECT_0;
 
         } else if (target->tv_sec < 0 || target->tv_sec == 0 && target->tv_nsec <= 0) {
             // target time <= 0 --> plain trylock or timed wait for time that has passed; try to lock with 0 timeout
-            if (mtx->thread_id != static_cast<long>(GetCurrentThreadId())) { // not this thread, lock it
-                if (mtx->_get_cs()->try_lock()) {
+            if (mtx->_Thread_id != static_cast<long>(GetCurrentThreadId())) { // not this thread, lock it
+                if (TryAcquireSRWLockExclusive(get_srw_lock(mtx)) != 0) {
                     res = WAIT_OBJECT_0;
                 } else {
                     res = WAIT_TIMEOUT;
@@ -125,8 +111,8 @@ static int mtx_do_lock(_Mtx_t mtx, const _timespec64* target) { // lock mutex
             _Timespec64_get_sys(&now);
             while (now.tv_sec < target->tv_sec || now.tv_sec == target->tv_sec && now.tv_nsec < target->tv_nsec) {
                 // time has not expired
-                if (mtx->thread_id == static_cast<long>(GetCurrentThreadId())
-                    || mtx->_get_cs()->try_lock()) { // stop waiting
+                if (mtx->_Thread_id == static_cast<long>(GetCurrentThreadId())
+                    || TryAcquireSRWLockExclusive(get_srw_lock(mtx)) != 0) { // stop waiting
                     res = WAIT_OBJECT_0;
                     break;
                 } else {
@@ -138,13 +124,13 @@ static int mtx_do_lock(_Mtx_t mtx, const _timespec64* target) { // lock mutex
         }
 
         if (res == WAIT_OBJECT_0 || res == WAIT_ABANDONED) {
-            if (1 < ++mtx->count) { // check count
-                if ((mtx->type & _Mtx_recursive) != _Mtx_recursive) { // not recursive, fixup count
-                    --mtx->count;
+            if (1 < ++mtx->_Count) { // check count
+                if ((mtx->_Type & _Mtx_recursive) != _Mtx_recursive) { // not recursive, fixup count
+                    --mtx->_Count;
                     res = WAIT_TIMEOUT;
                 }
             } else {
-                mtx->thread_id = static_cast<long>(GetCurrentThreadId());
+                mtx->_Thread_id = static_cast<long>(GetCurrentThreadId());
             }
         }
 
@@ -168,11 +154,14 @@ static int mtx_do_lock(_Mtx_t mtx, const _timespec64* target) { // lock mutex
 
 int _Mtx_unlock(_Mtx_t mtx) { // unlock mutex
     _THREAD_ASSERT(
-        1 <= mtx->count && mtx->thread_id == static_cast<long>(GetCurrentThreadId()), "unlock of unowned mutex");
+        1 <= mtx->_Count && mtx->_Thread_id == static_cast<long>(GetCurrentThreadId()), "unlock of unowned mutex");
+
+    if (--mtx->_Count == 0) { // leave critical section
+        mtx->_Thread_id = -1;
 
-    if (--mtx->count == 0) { // leave critical section
-        mtx->thread_id = -1;
-        mtx->_get_cs()->unlock();
+        auto srw_lock = get_srw_lock(mtx);
+        _Analysis_assume_lock_held_(*srw_lock);
+        ReleaseSRWLockExclusive(srw_lock);
     }
     return _Thrd_success; // TRANSITION, ABI: always returns _Thrd_success
 }
@@ -183,7 +172,7 @@ int _Mtx_lock(_Mtx_t mtx) { // lock mutex
 
 int _Mtx_trylock(_Mtx_t mtx) { // attempt to lock try_mutex
     _timespec64 xt;
-    _THREAD_ASSERT((mtx->type & (_Mtx_try | _Mtx_timed)) != 0, "trylock not supported by mutex");
+    _THREAD_ASSERT((mtx->_Type & (_Mtx_try | _Mtx_timed)) != 0, "trylock not supported by mutex");
     xt.tv_sec  = 0;
     xt.tv_nsec = 0;
     return mtx_do_lock(mtx, &xt);
@@ -192,27 +181,27 @@ int _Mtx_trylock(_Mtx_t mtx) { // attempt to lock try_mutex
 int _Mtx_timedlock(_Mtx_t mtx, const _timespec64* xt) { // attempt to lock timed mutex
     int res;
 
-    _THREAD_ASSERT((mtx->type & _Mtx_timed) != 0, "timedlock not supported by mutex");
+    _THREAD_ASSERT((mtx->_Type & _Mtx_timed) != 0, "timedlock not supported by mutex");
     res = mtx_do_lock(mtx, xt);
     return res == _Thrd_busy ? _Thrd_timedout : res;
 }
 
 int _Mtx_current_owns(_Mtx_t mtx) { // test if current thread owns mutex
-    return mtx->count != 0 && mtx->thread_id == static_cast<long>(GetCurrentThreadId());
+    return mtx->_Count != 0 && mtx->_Thread_id == static_cast<long>(GetCurrentThreadId());
 }
 
 void* _Mtx_getconcrtcs(_Mtx_t mtx) { // get internal cs impl
-    return mtx->_get_cs();
+    return &mtx->_Critical_section;
 }
 
 void _Mtx_clear_owner(_Mtx_t mtx) { // set owner to nobody
-    mtx->thread_id = -1;
-    --mtx->count;
+    mtx->_Thread_id = -1;
+    --mtx->_Count;
 }
 
 void _Mtx_reset_owner(_Mtx_t mtx) { // set owner to current thread
-    mtx->thread_id = static_cast<long>(GetCurrentThreadId());
-    ++mtx->count;
+    mtx->_Thread_id = static_cast<long>(GetCurrentThreadId());
+    ++mtx->_Count;
 }
 
 /*