Fix PR22634 - std::allocator doesn't respect over-aligned types.

This patch fixes std::allocator, and more specifically, all users
of __libcpp_allocate and __libcpp_deallocate, to support over-aligned
types.

__libcpp_allocate/deallocate now take an alignment parameter, and when
the specified alignment is greater than that supported by malloc/new,
the aligned version of operator new is called (assuming it's available).

When aligned new isn't available, the old behavior has been kept, and the
alignment parameter is ignored.

This patch depends on recent changes to __builtin_operator_new/delete which
allow them to be used to call any regular new/delete operator. By using
__builtin_operator_new/delete when possible, the new/delete erasure optimization
is maintained.

git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@328180 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/include/__config b/include/__config
index f96d743..d150ede 100644
--- a/include/__config
+++ b/include/__config
@@ -955,6 +955,7 @@
 #  endif
 #endif // defined(__APPLE__)
 
+
 #if defined(__APPLE__) || defined(__FreeBSD__)
 #define _LIBCPP_HAS_DEFAULTRUNELOCALE
 #endif
diff --git a/include/__sso_allocator b/include/__sso_allocator
index 8147e75..4002736 100644
--- a/include/__sso_allocator
+++ b/include/__sso_allocator
@@ -55,14 +55,14 @@
             __allocated_ = true;
             return (pointer)&buf_;
         }
-        return static_cast<pointer>(_VSTD::__allocate(__n * sizeof(_Tp)));
+        return static_cast<pointer>(_VSTD::__libcpp_allocate(__n * sizeof(_Tp), __alignof(_Tp)));
     }
     _LIBCPP_INLINE_VISIBILITY void deallocate(pointer __p, size_type)
     {
         if (__p == (pointer)&buf_)
             __allocated_ = false;
         else
-            _VSTD::__libcpp_deallocate(__p);
+            _VSTD::__libcpp_deallocate(__p, __alignof(_Tp));
     }
     _LIBCPP_INLINE_VISIBILITY size_type max_size() const throw() {return size_type(~0) / sizeof(_Tp);}
 
diff --git a/include/memory b/include/memory
index ea52ba2..021bd13 100644
--- a/include/memory
+++ b/include/memory
@@ -1796,10 +1796,10 @@
         if (__n > max_size())
             __throw_length_error("allocator<T>::allocate(size_t n)"
                                  " 'n' exceeds maximum supported size");
-        return static_cast<pointer>(_VSTD::__allocate(__n * sizeof(_Tp)));
+        return static_cast<pointer>(_VSTD::__libcpp_allocate(__n * sizeof(_Tp), __alignof(_Tp)));
         }
     _LIBCPP_INLINE_VISIBILITY void deallocate(pointer __p, size_type) _NOEXCEPT
-        {_VSTD::__libcpp_deallocate((void*)__p);}
+        {_VSTD::__libcpp_deallocate((void*)__p, __alignof(_Tp));}
     _LIBCPP_INLINE_VISIBILITY size_type max_size() const _NOEXCEPT
         {return size_type(~0) / sizeof(_Tp);}
 #if !defined(_LIBCPP_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCPP_HAS_NO_VARIADICS)
@@ -1897,10 +1897,10 @@
         if (__n > max_size())
             __throw_length_error("allocator<const T>::allocate(size_t n)"
                                  " 'n' exceeds maximum supported size");
-        return static_cast<pointer>(_VSTD::__allocate(__n * sizeof(_Tp)));
+        return static_cast<pointer>(_VSTD::__libcpp_allocate(__n * sizeof(_Tp), __alignof(_Tp)));
     }
     _LIBCPP_INLINE_VISIBILITY void deallocate(pointer __p, size_type) _NOEXCEPT
-        {_VSTD::__libcpp_deallocate((void*) const_cast<_Tp *>(__p));}
+        {_VSTD::__libcpp_deallocate((void*) const_cast<_Tp *>(__p), __alignof(_Tp));}
     _LIBCPP_INLINE_VISIBILITY size_type max_size() const _NOEXCEPT
         {return size_type(~0) / sizeof(_Tp);}
 #if !defined(_LIBCPP_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCPP_HAS_NO_VARIADICS)
@@ -2016,12 +2016,7 @@
     while (__n > 0)
     {
 #if !defined(_LIBCPP_HAS_NO_ALIGNED_ALLOCATION)
-#if defined(__STDCPP_DEFAULT_NEW_ALIGNMENT__)
-        if (std::alignment_of<_Tp>::value > __STDCPP_DEFAULT_NEW_ALIGNMENT__)
-#else
-        if (std::alignment_of<_Tp>::value >
-            std::alignment_of<std::max_align_t>::value)
-#endif
+    if (__is_overaligned_for_new(__alignof(_Tp)))
         {
             std::align_val_t __al =
                 std::align_val_t(std::alignment_of<_Tp>::value);
@@ -2032,12 +2027,7 @@
                 __n * sizeof(_Tp), nothrow));
         }
 #else
-#if defined(__STDCPP_DEFAULT_NEW_ALIGNMENT__)
-        if (std::alignment_of<_Tp>::value > __STDCPP_DEFAULT_NEW_ALIGNMENT__)
-#else
-        if (std::alignment_of<_Tp>::value >
-            std::alignment_of<std::max_align_t>::value)
-#endif
+    if (__is_overaligned_for_new(__alignof(_Tp)))
         {
             // Since aligned operator new is unavailable, return an empty
             // buffer rather than one with invalid alignment.
@@ -2061,20 +2051,7 @@
 inline _LIBCPP_INLINE_VISIBILITY
 void return_temporary_buffer(_Tp* __p) _NOEXCEPT
 {
-#if !defined(_LIBCPP_HAS_NO_ALIGNED_ALLOCATION)
-#if defined(__STDCPP_DEFAULT_NEW_ALIGNMENT__)
-    if (std::alignment_of<_Tp>::value > __STDCPP_DEFAULT_NEW_ALIGNMENT__)
-#else
-    if (std::alignment_of<_Tp>::value >
-        std::alignment_of<std::max_align_t>::value)
-#endif
-    {
-        std::align_val_t __al = std::align_val_t(std::alignment_of<_Tp>::value);
-        ::operator delete(__p, __al);
-        return;
-    }
-#endif
-    ::operator delete(__p);
+  _VSTD::__libcpp_deallocate((void*)__p, __alignof(_Tp));
 }
 
 #if _LIBCPP_STD_VER <= 14 || defined(_LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR)
diff --git a/include/new b/include/new
index 27c248c..e42ffad 100644
--- a/include/new
+++ b/include/new
@@ -89,6 +89,7 @@
 
 #include <__config>
 #include <exception>
+#include <type_traits>
 #include <cstddef>
 #ifdef _LIBCPP_NO_EXCEPTIONS
 #include <cstdlib>
@@ -113,6 +114,14 @@
 # define _LIBCPP_HAS_NO_ALIGNED_ALLOCATION
 #endif
 
+
+#if !__has_builtin(__builtin_operator_new) || \
+   __has_builtin(__builtin_operator_new) < 201802L || \
+   defined(_LIBCPP_HAS_NO_ALIGNED_ALLOCATION) || \
+   !defined(__cpp_aligned_new) || __cpp_aligned_new < 201606
+#define _LIBCPP_HAS_NO_BUILTIN_ALIGNED_OPERATOR_NEW_DELETE
+#endif
+
 namespace std  // purposefully not using versioning namespace
 {
 
@@ -223,7 +232,27 @@
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-inline _LIBCPP_INLINE_VISIBILITY void *__allocate(size_t __size) {
+_LIBCPP_CONSTEXPR inline _LIBCPP_INLINE_VISIBILITY bool __is_overaligned_for_new(size_t __align) _NOEXCEPT {
+#ifdef __STDCPP_DEFAULT_NEW_ALIGNMENT__
+  return __align > __STDCPP_DEFAULT_NEW_ALIGNMENT__;
+#else
+  return __align > alignment_of<max_align_t>::value;
+#endif
+}
+
+inline _LIBCPP_INLINE_VISIBILITY void *__libcpp_allocate(size_t __size, size_t __align) {
+#ifndef _LIBCPP_HAS_NO_ALIGNED_ALLOCATION
+  if (__is_overaligned_for_new(__align)) {
+    const align_val_t __align_val = static_cast<align_val_t>(__align);
+# ifdef _LIBCPP_HAS_NO_BUILTIN_ALIGNED_OPERATOR_NEW_DELETE
+    return ::operator new(__size, __align_val);
+# else
+    return __builtin_operator_new(__size, __align_val);
+# endif
+  }
+#else
+  ((void)__align);
+#endif
 #ifdef _LIBCPP_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE
   return ::operator new(__size);
 #else
@@ -231,11 +260,23 @@
 #endif
 }
 
-inline _LIBCPP_INLINE_VISIBILITY void __libcpp_deallocate(void *__ptr) {
-#ifdef _LIBCPP_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE
-  ::operator delete(__ptr);
+inline _LIBCPP_INLINE_VISIBILITY void __libcpp_deallocate(void* __ptr, size_t __align) {
+#ifndef _LIBCPP_HAS_NO_ALIGNED_ALLOCATION
+  if (__is_overaligned_for_new(__align)) {
+    const align_val_t __align_val = static_cast<align_val_t>(__align);
+# ifdef _LIBCPP_HAS_NO_BUILTIN_ALIGNED_OPERATOR_NEW_DELETE
+    return ::operator delete(__ptr, __align_val);
+# else
+    return __builtin_operator_delete(__ptr, __align_val);
+# endif
+  }
 #else
-  __builtin_operator_delete(__ptr);
+  ((void)__align);
+#endif
+#ifdef _LIBCPP_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE
+  return ::operator delete(__ptr);
+#else
+  return __builtin_operator_delete(__ptr);
 #endif
 }
 
diff --git a/include/valarray b/include/valarray
index b495ccf..8d3892a 100644
--- a/include/valarray
+++ b/include/valarray
@@ -2738,7 +2738,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<result_type*>(_VSTD::__allocate(__n * sizeof(result_type)));
+                static_cast<result_type*>(
+                    _VSTD::__libcpp_allocate(__n * sizeof(result_type), __alignof(result_type)));
         for (size_t __i = 0; __i != __n; ++__r.__end_, ++__i)
             ::new (__r.__end_) result_type(__expr_[__i]);
     }
@@ -2755,7 +2756,8 @@
 {
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+            _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2789,7 +2791,8 @@
 {
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+            _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2814,7 +2817,8 @@
 {
     if (__v.size())
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__v.size() * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+            _VSTD::__libcpp_allocate(__v.size() * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2851,7 +2855,8 @@
     size_t __n = __il.size();
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+_VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2879,7 +2884,8 @@
     size_t __n = __sa.__size_;
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+          _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2905,7 +2911,8 @@
     size_t __n = __ga.__1d_.size();
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2934,7 +2941,8 @@
     size_t __n = __ma.__1d_.size();
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2963,7 +2971,8 @@
     size_t __n = __ia.__1d_.size();
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
@@ -2999,7 +3008,8 @@
     if (size() != __n)
     {
         __clear();
-        __begin_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         __end_ = __begin_ + __n;
         _VSTD::uninitialized_copy(__f, __l, __begin_);
     } else {
@@ -3254,7 +3264,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
             ::new (__r.__end_) value_type(+*__p);
     }
@@ -3271,7 +3282,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
             ::new (__r.__end_) value_type(-*__p);
     }
@@ -3288,7 +3300,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
             ::new (__r.__end_) value_type(~*__p);
     }
@@ -3305,7 +3318,7 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<bool*>(_VSTD::__allocate(__n * sizeof(bool)));
+                static_cast<bool*>(_VSTD::__libcpp_allocate(__n * sizeof(bool), __alignof(bool)));
         for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
             ::new (__r.__end_) bool(!*__p);
     }
@@ -3625,7 +3638,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         const value_type* __sb;
         value_type* __tb;
         value_type* __te;
@@ -3663,7 +3677,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         __i %= static_cast<int>(__n);
         const value_type* __m = __i >= 0 ? __begin_ + __i : __end_ + __i;
         for (const value_type* __s = __m; __s != __end_; ++__r.__end_, ++__s)
@@ -3684,7 +3699,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
             ::new (__r.__end_) value_type(__f(*__p));
     }
@@ -3701,7 +3717,8 @@
     {
         __r.__begin_ =
             __r.__end_ =
-                static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+                static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
         for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
             ::new (__r.__end_) value_type(__f(*__p));
     }
@@ -3716,7 +3733,7 @@
     {
         while (__end_ != __begin_)
             (--__end_)->~value_type();
-        _VSTD::__libcpp_deallocate(__begin_);
+        _VSTD::__libcpp_deallocate(__begin_, __alignof(value_type));
         __begin_ = __end_ = nullptr;
     }
 }
@@ -3728,7 +3745,8 @@
     __clear();
     if (__n)
     {
-        __begin_ = __end_ = static_cast<value_type*>(_VSTD::__allocate(__n * sizeof(value_type)));
+        __begin_ = __end_ = static_cast<value_type*>(
+           _VSTD::__libcpp_allocate(__n * sizeof(value_type), __alignof(value_type)));
 #ifndef _LIBCPP_NO_EXCEPTIONS
         try
         {
diff --git a/src/experimental/memory_resource.cpp b/src/experimental/memory_resource.cpp
index c4dc1ca..a6eca37 100644
--- a/src/experimental/memory_resource.cpp
+++ b/src/experimental/memory_resource.cpp
@@ -31,10 +31,10 @@
 
 protected:
     virtual void* do_allocate(size_t __size, size_t __align)
-        { return __allocate(__size); }
+        { return _VSTD::__libcpp_allocate(__size, __align); /* FIXME */}
 
-    virtual void do_deallocate(void * __p, size_t, size_t)
-        { _VSTD::__libcpp_deallocate(__p); }
+    virtual void do_deallocate(void * __p, size_t, size_t __align)
+        { _VSTD::__libcpp_deallocate(__p, __align); /* FIXME */ }
 
     virtual bool do_is_equal(memory_resource const & __other) const _NOEXCEPT
         { return &__other == this; }
diff --git a/test/std/utilities/memory/default.allocator/allocator.members/allocate.pass.cpp b/test/std/utilities/memory/default.allocator/allocator.members/allocate.pass.cpp
index f2cf9f2..930da0b 100644
--- a/test/std/utilities/memory/default.allocator/allocator.members/allocate.pass.cpp
+++ b/test/std/utilities/memory/default.allocator/allocator.members/allocate.pass.cpp
@@ -15,39 +15,93 @@
 #include <memory>
 #include <cassert>
 
+#include "test_macros.h"
 #include "count_new.hpp"
 
-int A_constructed = 0;
 
-struct A
-{
-    int data;
-    A() {++A_constructed;}
-    A(const A&) {++A_constructed;}
-    ~A() {--A_constructed;}
+#ifdef TEST_HAS_NO_ALIGNED_ALLOCATION
+static const bool UsingAlignedNew = false;
+#else
+static const bool UsingAlignedNew = true;
+#endif
+
+#ifdef __STDCPP_DEFAULT_NEW_ALIGNMENT__
+static const size_t MaxAligned = __STDCPP_DEFAULT_NEW_ALIGNMENT__;
+#else
+static const size_t MaxAligned = std::alignment_of<std::max_align_t>::value;
+#endif
+
+static const size_t OverAligned = MaxAligned * 2;
+
+
+template <size_t Align>
+struct TEST_ALIGNAS(Align) AlignedType {
+  char data;
+  static int constructed;
+  AlignedType() { ++constructed; }
+  AlignedType(AlignedType const&) { ++constructed; }
+  ~AlignedType() { --constructed; }
 };
+template <size_t Align>
+int AlignedType<Align>::constructed = 0;
 
-int main()
-{
-    globalMemCounter.reset();
-    std::allocator<A> a;
+
+template <size_t Align>
+void test_aligned() {
+  typedef AlignedType<Align> T;
+  T::constructed = 0;
+  globalMemCounter.reset();
+  std::allocator<T> a;
+  const bool IsOverAlignedType = Align > MaxAligned;
+  const bool ExpectAligned = IsOverAlignedType && UsingAlignedNew;
+  {
     assert(globalMemCounter.checkOutstandingNewEq(0));
-    assert(A_constructed == 0);
+    assert(T::constructed == 0);
     globalMemCounter.last_new_size = 0;
-    A* volatile ap = a.allocate(3);
+    globalMemCounter.last_new_align = 0;
+    T* volatile ap = a.allocate(3);
     assert(globalMemCounter.checkOutstandingNewEq(1));
-    assert(globalMemCounter.checkLastNewSizeEq(3 * sizeof(int)));
-    assert(A_constructed == 0);
+    assert(globalMemCounter.checkNewCalledEq(1));
+    assert(globalMemCounter.checkAlignedNewCalledEq(ExpectAligned));
+    assert(globalMemCounter.checkLastNewSizeEq(3 * sizeof(T)));
+    assert(globalMemCounter.checkLastNewAlignEq(ExpectAligned ? Align : 0));
+    assert(T::constructed == 0);
+    globalMemCounter.last_delete_align = 0;
     a.deallocate(ap, 3);
     assert(globalMemCounter.checkOutstandingNewEq(0));
-    assert(A_constructed == 0);
-
+    assert(globalMemCounter.checkDeleteCalledEq(1));
+    assert(globalMemCounter.checkAlignedDeleteCalledEq(ExpectAligned));
+    assert(globalMemCounter.checkLastDeleteAlignEq(ExpectAligned ? Align : 0));
+    assert(T::constructed == 0);
+  }
+  globalMemCounter.reset();
+  {
     globalMemCounter.last_new_size = 0;
-    A* volatile ap2 = a.allocate(3, (const void*)5);
+    globalMemCounter.last_new_align = 0;
+    T* volatile ap2 = a.allocate(11, (const void*)5);
     assert(globalMemCounter.checkOutstandingNewEq(1));
-    assert(globalMemCounter.checkLastNewSizeEq(3 * sizeof(int)));
-    assert(A_constructed == 0);
-    a.deallocate(ap2, 3);
+    assert(globalMemCounter.checkNewCalledEq(1));
+    assert(globalMemCounter.checkAlignedNewCalledEq(ExpectAligned));
+    assert(globalMemCounter.checkLastNewSizeEq(11 * sizeof(T)));
+    assert(globalMemCounter.checkLastNewAlignEq(ExpectAligned ? Align : 0));
+    assert(T::constructed == 0);
+    globalMemCounter.last_delete_align = 0;
+    a.deallocate(ap2, 11);
     assert(globalMemCounter.checkOutstandingNewEq(0));
-    assert(A_constructed == 0);
+    assert(globalMemCounter.checkDeleteCalledEq(1));
+    assert(globalMemCounter.checkAlignedDeleteCalledEq(ExpectAligned));
+    assert(globalMemCounter.checkLastDeleteAlignEq(ExpectAligned ? Align : 0));
+    assert(T::constructed == 0);
+  }
+}
+
+int main() {
+    test_aligned<1>();
+    test_aligned<2>();
+    test_aligned<4>();
+    test_aligned<8>();
+    test_aligned<16>();
+    test_aligned<MaxAligned>();
+    test_aligned<OverAligned>();
+    test_aligned<OverAligned * 2>();
 }
diff --git a/test/support/count_new.hpp b/test/support/count_new.hpp
index c001c03..e3111e7 100644
--- a/test/support/count_new.hpp
+++ b/test/support/count_new.hpp
@@ -59,12 +59,20 @@
     int outstanding_new;
     int new_called;
     int delete_called;
+    int aligned_new_called;
+    int aligned_delete_called;
     std::size_t last_new_size;
+    std::size_t last_new_align;
+    std::size_t last_delete_align;
 
     int outstanding_array_new;
     int new_array_called;
     int delete_array_called;
+    int aligned_new_array_called;
+    int aligned_delete_array_called;
     std::size_t last_new_array_size;
+    std::size_t last_new_array_align;
+    std::size_t last_delete_array_align;
 
 public:
     void newCalled(std::size_t s)
@@ -82,6 +90,12 @@
         last_new_size = s;
     }
 
+    void alignedNewCalled(std::size_t s, std::size_t a) {
+      newCalled(s);
+      ++aligned_new_called;
+      last_new_align = a;
+    }
+
     void deleteCalled(void * p)
     {
         assert(p);
@@ -89,6 +103,12 @@
         ++delete_called;
     }
 
+    void alignedDeleteCalled(void *p, std::size_t a) {
+      deleteCalled(p);
+      ++aligned_delete_called;
+      last_delete_align = a;
+    }
+
     void newArrayCalled(std::size_t s)
     {
         assert(disable_allocations == false);
@@ -104,6 +124,12 @@
         last_new_array_size = s;
     }
 
+    void alignedNewArrayCalled(std::size_t s, std::size_t a) {
+      newArrayCalled(s);
+      ++aligned_new_array_called;
+      last_new_array_align = a;
+    }
+
     void deleteArrayCalled(void * p)
     {
         assert(p);
@@ -111,6 +137,12 @@
         ++delete_array_called;
     }
 
+    void alignedDeleteArrayCalled(void * p, std::size_t a) {
+      deleteArrayCalled(p);
+      ++aligned_delete_array_called;
+      last_delete_array_align = a;
+    }
+
     void disableAllocations()
     {
         disable_allocations = true;
@@ -121,7 +153,6 @@
         disable_allocations = false;
     }
 
-
     void reset()
     {
         disable_allocations = false;
@@ -130,12 +161,18 @@
         outstanding_new = 0;
         new_called = 0;
         delete_called = 0;
+        aligned_new_called = 0;
+        aligned_delete_called = 0;
         last_new_size = 0;
+        last_new_align = 0;
 
         outstanding_array_new = 0;
         new_array_called = 0;
         delete_array_called = 0;
+        aligned_new_array_called = 0;
+        aligned_delete_array_called = 0;
         last_new_array_size = 0;
+        last_new_array_align = 0;
     }
 
 public:
@@ -174,6 +211,31 @@
         return disable_checking || n != delete_called;
     }
 
+    bool checkAlignedNewCalledEq(int n) const
+    {
+        return disable_checking || n == aligned_new_called;
+    }
+
+    bool checkAlignedNewCalledNotEq(int n) const
+    {
+        return disable_checking || n != aligned_new_called;
+    }
+
+    bool checkAlignedNewCalledGreaterThan(int n) const
+    {
+        return disable_checking || aligned_new_called > n;
+    }
+
+    bool checkAlignedDeleteCalledEq(int n) const
+    {
+        return disable_checking || n == aligned_delete_called;
+    }
+
+    bool checkAlignedDeleteCalledNotEq(int n) const
+    {
+        return disable_checking || n != aligned_delete_called;
+    }
+
     bool checkLastNewSizeEq(std::size_t n) const
     {
         return disable_checking || n == last_new_size;
@@ -184,6 +246,26 @@
         return disable_checking || n != last_new_size;
     }
 
+    bool checkLastNewAlignEq(std::size_t n) const
+    {
+        return disable_checking || n == last_new_align;
+    }
+
+    bool checkLastNewAlignNotEq(std::size_t n) const
+    {
+        return disable_checking || n != last_new_align;
+    }
+
+    bool checkLastDeleteAlignEq(std::size_t n) const
+    {
+        return disable_checking || n == last_delete_align;
+    }
+
+    bool checkLastDeleteAlignNotEq(std::size_t n) const
+    {
+        return disable_checking || n != last_delete_align;
+    }
+
     bool checkOutstandingArrayNewEq(int n) const
     {
         return disable_checking || n == outstanding_array_new;
@@ -214,6 +296,31 @@
         return disable_checking || n != delete_array_called;
     }
 
+    bool checkAlignedNewArrayCalledEq(int n) const
+    {
+        return disable_checking || n == aligned_new_array_called;
+    }
+
+    bool checkAlignedNewArrayCalledNotEq(int n) const
+    {
+        return disable_checking || n != aligned_new_array_called;
+    }
+
+    bool checkAlignedNewArrayCalledGreaterThan(int n) const
+    {
+        return disable_checking || aligned_new_array_called > n;
+    }
+
+    bool checkAlignedDeleteArrayCalledEq(int n) const
+    {
+        return disable_checking || n == aligned_delete_array_called;
+    }
+
+    bool checkAlignedDeleteArrayCalledNotEq(int n) const
+    {
+        return disable_checking || n != aligned_delete_array_called;
+    }
+
     bool checkLastNewArraySizeEq(std::size_t n) const
     {
         return disable_checking || n == last_new_array_size;
@@ -223,6 +330,16 @@
     {
         return disable_checking || n != last_new_array_size;
     }
+
+    bool checkLastNewArrayAlignEq(std::size_t n) const
+    {
+        return disable_checking || n == last_new_array_align;
+    }
+
+    bool checkLastNewArrayAlignNotEq(std::size_t n) const
+    {
+        return disable_checking || n != last_new_array_align;
+    }
 };
 
 #ifdef DISABLE_NEW_COUNT
@@ -254,22 +371,65 @@
     std::free(p);
 }
 
-
 void* operator new[](std::size_t s) TEST_THROW_SPEC(std::bad_alloc)
 {
     getGlobalMemCounter()->newArrayCalled(s);
     return operator new(s);
 }
 
-
 void operator delete[](void* p) TEST_NOEXCEPT
 {
     getGlobalMemCounter()->deleteArrayCalled(p);
     operator delete(p);
 }
 
-#endif // DISABLE_NEW_COUNT
+#ifndef TEST_HAS_NO_ALIGNED_ALLOCATION
+#if defined(_LIBCPP_MSVCRT_LIKE) || \
+  (!defined(_LIBCPP_VERSION) && defined(_WIN32))
+#define USE_ALIGNED_ALLOC
+#endif
 
+void* operator new(std::size_t s, std::align_val_t av) TEST_THROW_SPEC(std::bad_alloc) {
+  const std::size_t a = static_cast<std::size_t>(av);
+  getGlobalMemCounter()->alignedNewCalled(s, a);
+  void *ret;
+#ifdef USE_ALIGNED_ALLOC
+  ret = _aligned_malloc(s, a);
+#else
+  posix_memalign(&ret, a, s);
+#endif
+  if (ret == nullptr)
+    detail::throw_bad_alloc_helper();
+  return ret;
+}
+
+void operator delete(void *p, std::align_val_t av) TEST_NOEXCEPT {
+  const std::size_t a = static_cast<std::size_t>(av);
+  getGlobalMemCounter()->alignedDeleteCalled(p, a);
+  if (p) {
+#ifdef USE_ALIGNED_ALLOC
+    ::_aligned_free(p);
+#else
+    ::free(p);
+#endif
+  }
+}
+
+void* operator new[](std::size_t s, std::align_val_t av) TEST_THROW_SPEC(std::bad_alloc) {
+  const std::size_t a = static_cast<std::size_t>(av);
+  getGlobalMemCounter()->alignedNewArrayCalled(s, a);
+  return operator new(s, av);
+}
+
+void operator delete[](void *p, std::align_val_t av) TEST_NOEXCEPT {
+  const std::size_t a = static_cast<std::size_t>(av);
+  getGlobalMemCounter()->alignedDeleteArrayCalled(p, a);
+  return operator delete(p, av);
+}
+
+#endif // TEST_HAS_NO_ALIGNED_ALLOCATION
+
+#endif // DISABLE_NEW_COUNT
 
 struct DisableAllocationGuard {
     explicit DisableAllocationGuard(bool disable = true) : m_disabled(disable)
@@ -295,7 +455,6 @@
     DisableAllocationGuard& operator=(DisableAllocationGuard const&);
 };
 
-
 struct RequireAllocationGuard {
     explicit RequireAllocationGuard(std::size_t RequireAtLeast = 1)
             : m_req_alloc(RequireAtLeast),
diff --git a/test/support/test_macros.h b/test/support/test_macros.h
index 257875a..06800bd 100644
--- a/test/support/test_macros.h
+++ b/test/support/test_macros.h
@@ -157,6 +157,11 @@
 #define TEST_NORETURN [[noreturn]]
 #endif
 
+#if !defined(__cpp_aligned_new) || __cpp_aligned_new < 201606L || \
+    defined(_LIBCPP_HAS_NO_ALIGNED_ALLOCATION)
+#define TEST_HAS_NO_ALIGNED_ALLOCATION
+#endif
+
 #if defined(_LIBCPP_SAFE_STATIC)
 #define TEST_SAFE_STATIC _LIBCPP_SAFE_STATIC
 #else
@@ -228,6 +233,7 @@
 }
 #endif
 
+
 #if defined(__GNUC__)
 #pragma GCC diagnostic pop
 #endif