[libcxx-commits] [libcxx] [libc++] Optimize vector::erase (PR #213916)

[Nikolas Klauser via libcxx-commits]

================
@@ -431,6 +431,44 @@ _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 void __uninitialized_allocat
   }
 }
 
+// __uninitialized_allocator_relocate_forward relocates the objects in [__first, __last) into __result
+// element-by-element from __first to __last. Relocation means that the objects in [__first, __last) are placed into
+// __result as-if by move-construct and destroy, except that the move constructor and destructor may never be called if
+// they are known to be equivalent to a memcpy.
+//
+// Preconditions:  At __result there is no object and [__first, __last) contains objects
+// Postconditions: __result contains the objects from [__first, __last) and
+//                 [__first, __last) doesn't contain any objects
+//
+// The strong exception guarantee is provided.
+template <class _Alloc, class _ContiguousIterator>
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 void __uninitialized_allocator_relocate_forward(
+    _Alloc& __alloc, _ContiguousIterator __first, _ContiguousIterator __last, _ContiguousIterator __result) {
+  static_assert(__libcpp_is_contiguous_iterator<_ContiguousIterator>::value, "");
+  using _ValueType = typename iterator_traits<_ContiguousIterator>::value_type;
+  static_assert(
+      __is_nothrow_relocatable_v<_ValueType> && __allocator_has_trivial_move_construct_v<_Alloc, _ValueType> &&
----------------
philnik777 wrote:

That is true, but I think I'd rather extend the handling once we can actually test it (e.g. by using it in `insert`). Once we have that we can weaken the `static_assert`. In `erase` we can only use it in the trivially relocatable case anyways for now.

https://github.com/llvm/llvm-project/pull/213916