[libc++][PSTL][NFC] Rename to `pstl/` to `__pstl/`

[philnik]

Update CMakeLists.txt

[philnik]

Rebased

[philnik]

Rebased

[philnik]

Rebased

[philnik]

Try to fix CI

libcxx/include/CMakeLists.txt

set(files
   memory
   memory_resource
   mutex
   new
   numbers
   numeric
   optional
   ostream
-  pstl/internal/algorithm_fwd.h
-  pstl/internal/algorithm_impl.h
-  pstl/internal/execution_defs.h
-  pstl/internal/execution_impl.h
-  pstl/internal/glue_algorithm_defs.h
-  pstl/internal/glue_algorithm_impl.h
-  pstl/internal/glue_execution_defs.h
-  pstl/internal/glue_memory_defs.h
-  pstl/internal/glue_memory_impl.h
-  pstl/internal/glue_numeric_defs.h
-  pstl/internal/glue_numeric_impl.h
-  pstl/internal/memory_impl.h
-  pstl/internal/numeric_fwd.h
-  pstl/internal/numeric_impl.h
-  pstl/internal/omp/parallel_for.h
-  pstl/internal/omp/parallel_for_each.h
-  pstl/internal/omp/parallel_invoke.h
-  pstl/internal/omp/parallel_merge.h
-  pstl/internal/omp/parallel_scan.h
-  pstl/internal/omp/parallel_stable_partial_sort.h
-  pstl/internal/omp/parallel_stable_sort.h
-  pstl/internal/omp/parallel_transform_reduce.h
-  pstl/internal/omp/parallel_transform_scan.h
-  pstl/internal/omp/util.h
-  pstl/internal/parallel_backend.h
-  pstl/internal/parallel_backend_omp.h
-  pstl/internal/parallel_backend_serial.h
-  pstl/internal/parallel_backend_tbb.h
-  pstl/internal/parallel_backend_utils.h
-  pstl/internal/parallel_impl.h
-  pstl/internal/pstl_config.h
-  pstl/internal/unseq_backend_simd.h
-  pstl/internal/utils.h
+  __pstl/internal/algorithm_fwd.h
+  __pstl/internal/algorithm_impl.h
+  __pstl/internal/execution_defs.h
+  __pstl/internal/execution_impl.h
+  __pstl/internal/glue_algorithm_defs.h
+  __pstl/internal/glue_algorithm_impl.h
+  __pstl/internal/glue_execution_defs.h
+  __pstl/internal/glue_memory_defs.h
+  __pstl/internal/glue_memory_impl.h
+  __pstl/internal/glue_numeric_defs.h
+  __pstl/internal/glue_numeric_impl.h
+  __pstl/internal/memory_impl.h
+  __pstl/internal/numeric_fwd.h
+  __pstl/internal/numeric_impl.h
+  __pstl/internal/omp/parallel_for.h
+  __pstl/internal/omp/parallel_for_each.h
+  __pstl/internal/omp/parallel_invoke.h
+  __pstl/internal/omp/parallel_merge.h
+  __pstl/internal/omp/parallel_scan.h
+  __pstl/internal/omp/parallel_stable_partial_sort.h
+  __pstl/internal/omp/parallel_stable_sort.h
+  __pstl/internal/omp/parallel_transform_reduce.h
+  __pstl/internal/omp/parallel_transform_scan.h
+  __pstl/internal/omp/util.h
+  __pstl/internal/parallel_backend.h
+  __pstl/internal/parallel_backend_omp.h
+  __pstl/internal/parallel_backend_serial.h
+  __pstl/internal/parallel_backend_tbb.h
+  __pstl/internal/parallel_backend_utils.h
+  __pstl/internal/parallel_impl.h
+  __pstl/internal/pstl_config.h
+  __pstl/internal/unseq_backend_simd.h
+  __pstl/internal/utils.h
   queue
   random
   ranges
   ratio
   regex
   scoped_allocator
   semaphore
   set

libcxx/include/__pstl/internal/algorithm_fwd.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_ALGORITHM_FWD_H
 #define _PSTL_ALGORITHM_FWD_H
 
 #include <iterator>
 #include <type_traits>
 #include <utility>
 
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 //------------------------------------------------------------------------
 // any_of
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Pred>
 bool __brick_any_of(const _ForwardIterator,
                     const _ForwardIterator,
                     _Pred,
                     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Pred>
 bool __brick_any_of(const _RandomAccessIterator,
                     const _RandomAccessIterator,
                     _Pred,
                     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Pred>
 bool __pattern_any_of(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Pred) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Pred>
 bool __pattern_any_of(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Pred);
 
 //------------------------------------------------------------------------
 // walk1 (pseudo)
 //
 // walk1 evaluates f(x) for each dereferenced value x drawn from [first,last)
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Function>
 void __brick_walk1(_ForwardIterator,
                    _ForwardIterator,
                    _Function,
                    /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Function>
 void __brick_walk1(_RandomAccessIterator,
                    _RandomAccessIterator,
                    _Function,
                    /*vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Function>
 void __pattern_walk1(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Function) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Function>
 void __pattern_walk1(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Function);
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Brick>
 void __pattern_walk_brick(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Brick) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Brick>
 void __pattern_walk_brick(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Brick);
 
 //------------------------------------------------------------------------
 // walk1_n
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Size, class _Function>
 _ForwardIterator __brick_walk1_n(
     _ForwardIterator,
     _Size,
     _Function,
     /*_IsVectorTag=*/std::false_type);
 
 template <class _RandomAccessIterator, class _DifferenceType, class _Function>
 _RandomAccessIterator __brick_walk1_n(
     _RandomAccessIterator,
     _DifferenceType,
     _Function,
     /*vectorTag=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Function>
 _ForwardIterator __pattern_walk1_n(_Tag, _ExecutionPolicy&&, _ForwardIterator, _Size, _Function) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Function>
 _RandomAccessIterator
 __pattern_walk1_n(__parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _Size, _Function);
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Brick>
 _ForwardIterator __pattern_walk_brick_n(_Tag, _ExecutionPolicy&&, _ForwardIterator, _Size, _Brick) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Brick>
 _RandomAccessIterator
 __pattern_walk_brick_n(__parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _Size, _Brick);
 
 //------------------------------------------------------------------------
 // walk2 (pseudo)
 //
 // walk2 evaluates f(x,y) for deferenced values (x,y) drawn from [first1,last1) and [first2,...)
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Function>
 _ForwardIterator2 __brick_walk2(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _Function,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Function>
 _RandomAccessIterator2 __brick_walk2(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _Function,
     /*vector=*/std::true_type) noexcept;
 
 template <class _ForwardIterator1, class _Size, class _ForwardIterator2, class _Function>
 _ForwardIterator2 __brick_walk2_n(
     _ForwardIterator1,
     _Size,
     _ForwardIterator2,
     _Function,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _Size, class _RandomAccessIterator2, class _Function>
 _RandomAccessIterator2 __brick_walk2_n(
     _RandomAccessIterator1,
     _Size,
     _RandomAccessIterator2,
     _Function,
     /*vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Function>
 _ForwardIterator2
 __pattern_walk2(_Tag, _ExecutionPolicy&&, _ForwardIterator1, _ForwardIterator1, _ForwardIterator2, _Function) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Function>
 _RandomAccessIterator2 __pattern_walk2(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _Function);
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _Size,
           class _ForwardIterator2,
           class _Function>
 _ForwardIterator2
 __pattern_walk2_n(_Tag, _ExecutionPolicy&&, _ForwardIterator1, _Size, _ForwardIterator2, _Function) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _Size,
           class _RandomAccessIterator2,
           class _Function>
 _RandomAccessIterator2 __pattern_walk2_n(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator1, _Size, _RandomAccessIterator2, _Function);
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Brick>
 _ForwardIterator2 __pattern_walk2_brick(
     _Tag, _ExecutionPolicy&&, _ForwardIterator1, _ForwardIterator1, _ForwardIterator2, _Brick) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Brick>
 _RandomAccessIterator2 __pattern_walk2_brick(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _Brick);
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _Size,
           class _ForwardIterator2,
           class _Brick>
 _ForwardIterator2
 __pattern_walk2_brick_n(_Tag, _ExecutionPolicy&&, _ForwardIterator1, _Size, _ForwardIterator2, _Brick) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _Size,
           class _RandomAccessIterator2,
           class _Brick>
 _RandomAccessIterator2 __pattern_walk2_brick_n(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator1, _Size, _RandomAccessIterator2, _Brick);
 
 //------------------------------------------------------------------------
 // walk3 (pseudo)
 //
 // walk3 evaluates f(x,y,z) for (x,y,z) drawn from [first1,last1), [first2,...), [first3,...)
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator3, class _Function>
 _ForwardIterator3 __brick_walk3(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator3,
     _Function,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Function>
 _RandomAccessIterator3 __brick_walk3(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator3,
     _Function,
     /*vector=*/std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator3,
           class _Function>
 _ForwardIterator3 __pattern_walk3(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator3,
     _Function) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _Function>
 _RandomAccessIterator3 __pattern_walk3(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator3,
     _Function);
 
 //------------------------------------------------------------------------
 // equal
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __brick_equal(_ForwardIterator1,
                    _ForwardIterator1,
                    _ForwardIterator2,
                    _BinaryPredicate,
                    /* is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 bool __brick_equal(_RandomAccessIterator1,
                    _RandomAccessIterator1,
                    _RandomAccessIterator2,
                    _BinaryPredicate,
                    /* is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __pattern_equal(
     _Tag, _ExecutionPolicy&&, _ForwardIterator1, _ForwardIterator1, _ForwardIterator2, _BinaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 bool __pattern_equal(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _BinaryPredicate);
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __brick_equal(_ForwardIterator1,
                    _ForwardIterator1,
                    _ForwardIterator2,
                    _ForwardIterator2,
                    _BinaryPredicate,
                    /* is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 bool __brick_equal(_RandomAccessIterator1,
                    _RandomAccessIterator1,
                    _RandomAccessIterator2,
                    _RandomAccessIterator2,
                    _BinaryPredicate,
                    /* is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __pattern_equal(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 bool __pattern_equal(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate);
 
 //------------------------------------------------------------------------
 // find_if
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Predicate>
 _ForwardIterator __brick_find_if(
     _ForwardIterator,
     _ForwardIterator,
     _Predicate,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Predicate>
 _RandomAccessIterator __brick_find_if(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Predicate,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 _ForwardIterator __pattern_find_if(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Predicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Predicate>
 _RandomAccessIterator __pattern_find_if(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Predicate);
 
 //------------------------------------------------------------------------
 // find_end
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __brick_find_end(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_find_end(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __pattern_find_end(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator1 __pattern_find_end(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // find_first_of
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __brick_find_first_of(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_find_first_of(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __pattern_find_first_of(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator1 __pattern_find_first_of(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // search
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __brick_search(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_search(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate,
     /*vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __pattern_search(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _BinaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator1 __pattern_search(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _BinaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // search_n
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
 _ForwardIterator __brick_search_n(
     _ForwardIterator,
     _ForwardIterator,
     _Size,
     const _Tp&,
     _BinaryPredicate,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Size, class _Tp, class _BinaryPredicate>
 _RandomAccessIterator __brick_search_n(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Size,
     const _Tp&,
     _BinaryPredicate,
     /*vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
 _ForwardIterator __pattern_search_n(
     _Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Size, const _Tp&, _BinaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _Size,
           class _Tp,
           class _BinaryPredicate>
 _RandomAccessIterator __pattern_search_n(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Size,
     const _Tp&,
     _BinaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // copy_n
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Size, class _OutputIterator>
 _OutputIterator __brick_copy_n(_ForwardIterator,
                                _Size,
                                _OutputIterator,
                                /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Size, class _OutputIterator>
 _OutputIterator __brick_copy_n(_RandomAccessIterator,
                                _Size,
                                _OutputIterator,
                                /*vector=*/std::true_type) noexcept;
 
 //------------------------------------------------------------------------
 // copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator __brick_copy(_ForwardIterator,
                              _ForwardIterator,
                              _OutputIterator,
                              /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __brick_copy(_RandomAccessIterator,
                              _RandomAccessIterator,
                              _OutputIterator,
                              /*vector=*/std::true_type) noexcept;
 
 //------------------------------------------------------------------------
 // move
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator __brick_move(_ForwardIterator,
                              _ForwardIterator,
                              _OutputIterator,
                              /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __brick_move(_RandomAccessIterator,
                              _RandomAccessIterator,
                              _OutputIterator,
                              /*vector=*/std::true_type) noexcept;
 
 //------------------------------------------------------------------------
 // swap_ranges
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator __brick_swap_ranges(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __brick_swap_ranges(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     /*vector=*/std::true_type) noexcept;
 
 //------------------------------------------------------------------------
 // copy_if
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator, class _UnaryPredicate>
 _OutputIterator __brick_copy_if(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     _UnaryPredicate,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator, class _UnaryPredicate>
 _OutputIterator __brick_copy_if(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _UnaryPredicate,
     /*vector=*/std::true_type) noexcept;
 
 template <class _DifferenceType, class _ForwardIterator, class _UnaryPredicate>
 std::pair<_DifferenceType, _DifferenceType> __brick_calc_mask_1(
     _ForwardIterator,
     _ForwardIterator,
     bool* __restrict,
     _UnaryPredicate,
     /*vector=*/std::false_type) noexcept;
 template <class _DifferenceType, class _RandomAccessIterator, class _UnaryPredicate>
 std::pair<_DifferenceType, _DifferenceType> __brick_calc_mask_1(
     _RandomAccessIterator,
     _RandomAccessIterator,
     bool* __restrict,
     _UnaryPredicate,
     /*vector=*/std::true_type) noexcept;
 
 template <class _ForwardIterator, class _OutputIterator>
 void __brick_copy_by_mask(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     bool*,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator>
 void __brick_copy_by_mask(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     bool* __restrict,
     /*vector=*/std::true_type) noexcept;
 
 template <class _ForwardIterator, class _OutputIterator1, class _OutputIterator2>
 void __brick_partition_by_mask(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator1,
     _OutputIterator2,
     bool*,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator1, class _OutputIterator2>
 void __brick_partition_by_mask(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator1,
     _OutputIterator2,
     bool*,
     /*vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator, class _UnaryPredicate>
 _OutputIterator __pattern_copy_if(
     _Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _OutputIterator, _UnaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryPredicate>
 _OutputIterator __pattern_copy_if(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _UnaryPredicate);
 
 //------------------------------------------------------------------------
 // count
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _Predicate>
 typename std::iterator_traits<_RandomAccessIterator>::difference_type __brick_count(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Predicate,
     /* is_vector = */ std::true_type) noexcept;
 
 template <class _ForwardIterator, class _Predicate>
 typename std::iterator_traits<_ForwardIterator>::difference_type __brick_count(
     _ForwardIterator,
     _ForwardIterator,
     _Predicate,
     /* is_vector = */ std::false_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 typename std::iterator_traits<_ForwardIterator>::difference_type
 __pattern_count(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Predicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Predicate>
 typename std::iterator_traits<_RandomAccessIterator>::difference_type __pattern_count(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Predicate);
 
 //------------------------------------------------------------------------
 // unique
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator __brick_unique(
     _ForwardIterator,
     _ForwardIterator,
     _BinaryPredicate,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __brick_unique(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _BinaryPredicate,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator
 __pattern_unique(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _BinaryPredicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __pattern_unique(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _BinaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // unique_copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class OutputIterator, class _BinaryPredicate>
 OutputIterator __brick_unique_copy(
     _ForwardIterator,
     _ForwardIterator,
     OutputIterator,
     _BinaryPredicate,
     /*vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator, class _BinaryPredicate>
 _OutputIterator __brick_unique_copy(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _BinaryPredicate,
     /*vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator, class _BinaryPredicate>
 _OutputIterator __pattern_unique_copy(
     _Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _OutputIterator, _BinaryPredicate) noexcept;
 
 template <class _ExecutionPolicy, class _DifferenceType, class _RandomAccessIterator, class _BinaryPredicate>
 _DifferenceType __brick_calc_mask_2(
     _RandomAccessIterator,
     _RandomAccessIterator,
     bool* __restrict,
     _BinaryPredicate,
     /*vector=*/std::false_type) noexcept;
 
 template <class _DifferenceType, class _RandomAccessIterator, class _BinaryPredicate>
 _DifferenceType __brick_calc_mask_2(
     _RandomAccessIterator,
     _RandomAccessIterator,
     bool* __restrict,
     _BinaryPredicate,
     /*vector=*/std::true_type) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _BinaryPredicate>
 _OutputIterator __pattern_unique_copy(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _BinaryPredicate);
 
 //------------------------------------------------------------------------
 // reverse
 //------------------------------------------------------------------------
 
 template <class _BidirectionalIterator>
 void __brick_reverse(_BidirectionalIterator,
                      _BidirectionalIterator,
                      /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator>
 void __brick_reverse(_RandomAccessIterator,
                      _RandomAccessIterator,
                      /*__is_vector=*/std::true_type) noexcept;
 
 template <class _BidirectionalIterator>
 void __brick_reverse(_BidirectionalIterator,
                      _BidirectionalIterator,
                      _BidirectionalIterator,
                      /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator>
 void __brick_reverse(_RandomAccessIterator,
                      _RandomAccessIterator,
                      _RandomAccessIterator,
                      /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator>
 void __pattern_reverse(_Tag, _ExecutionPolicy&&, _BidirectionalIterator, _BidirectionalIterator) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator>
 void __pattern_reverse(__parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator);
 
 //------------------------------------------------------------------------
 // reverse_copy
 //------------------------------------------------------------------------
 
 template <class _BidirectionalIterator, class _OutputIterator>
 _OutputIterator __brick_reverse_copy(
     _BidirectionalIterator,
     _BidirectionalIterator,
     _OutputIterator,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __brick_reverse_copy(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator, class _OutputIterator>
 _OutputIterator __pattern_reverse_copy(
     _Tag, _ExecutionPolicy&&, _BidirectionalIterator, _BidirectionalIterator, _OutputIterator) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __pattern_reverse_copy(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _OutputIterator);
 
 //------------------------------------------------------------------------
 // rotate
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator>
 _ForwardIterator __brick_rotate(
     _ForwardIterator,
     _ForwardIterator,
     _ForwardIterator,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator>
 _RandomAccessIterator __brick_rotate(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator>
 _ForwardIterator
 __pattern_rotate(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _ForwardIterator) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator>
 _RandomAccessIterator __pattern_rotate(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _RandomAccessIterator);
 
 //------------------------------------------------------------------------
 // rotate_copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator __brick_rotate_copy(
     _ForwardIterator,
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __brick_rotate_copy(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator>
 _OutputIterator __pattern_rotate_copy(
     _Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _ForwardIterator, _OutputIterator) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _OutputIterator>
 _OutputIterator __pattern_rotate_copy(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator);
 
 //------------------------------------------------------------------------
 // is_partitioned
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _UnaryPredicate>
 bool __brick_is_partitioned(_ForwardIterator,
                             _ForwardIterator,
                             _UnaryPredicate,
                             /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 bool __brick_is_partitioned(_RandomAccessIterator,
                             _RandomAccessIterator,
                             _UnaryPredicate,
                             /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 bool __pattern_is_partitioned(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _UnaryPredicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 bool __pattern_is_partitioned(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _UnaryPredicate);
 
 //------------------------------------------------------------------------
 // partition
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator __brick_partition(
     _ForwardIterator,
     _ForwardIterator,
     _UnaryPredicate,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __brick_partition(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _UnaryPredicate,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator
 __pattern_partition(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _UnaryPredicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __pattern_partition(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _UnaryPredicate);
 
 //------------------------------------------------------------------------
 // stable_partition
 //------------------------------------------------------------------------
 
 template <class _BidirectionalIterator, class _UnaryPredicate>
 _BidirectionalIterator __brick_stable_partition(
     _BidirectionalIterator,
     _BidirectionalIterator,
     _UnaryPredicate,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __brick_stable_partition(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _UnaryPredicate,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator, class _UnaryPredicate>
 _BidirectionalIterator __pattern_stable_partition(
     _Tag, _ExecutionPolicy&&, _BidirectionalIterator, _BidirectionalIterator, _UnaryPredicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __pattern_stable_partition(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _UnaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // partition_copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator1, class _OutputIterator2, class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2> __brick_partition_copy(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator1,
     _OutputIterator2,
     _UnaryPredicate,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator1, class _OutputIterator2, class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2> __brick_partition_copy(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator1,
     _OutputIterator2,
     _UnaryPredicate,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator,
           class _OutputIterator1,
           class _OutputIterator2,
           class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2> __pattern_partition_copy(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator1,
     _OutputIterator2,
     _UnaryPredicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator1,
           class _OutputIterator2,
           class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2> __pattern_partition_copy(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator1,
     _OutputIterator2,
     _UnaryPredicate);
 
 //------------------------------------------------------------------------
 // sort
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare, class _IsMoveConstructible>
 void __pattern_sort(
     _Tag, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare, _IsMoveConstructible) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_sort(__parallel_tag<_IsVector>,
                     _ExecutionPolicy&&,
                     _RandomAccessIterator,
                     _RandomAccessIterator,
                     _Compare,
                     /*is_move_constructible=*/std::true_type);
 
 //------------------------------------------------------------------------
 // stable_sort
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_stable_sort(_Tag, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_stable_sort(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare);
 
 //------------------------------------------------------------------------
 // partial_sort
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_partial_sort(
     _Tag, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _RandomAccessIterator, _Compare) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_partial_sort(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // partial_sort_copy
 //------------------------------------------------------------------------
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 _RandomAccessIterator2 __pattern_partial_sort_copy(
     _Tag,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 _RandomAccessIterator2 __pattern_partial_sort_copy(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Compare);
 
 //------------------------------------------------------------------------
 // adjacent_find
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __brick_adjacent_find(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _BinaryPredicate,
     /* IsVector = */ std::true_type,
     bool) noexcept;
 
 template <class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator __brick_adjacent_find(
     _ForwardIterator,
     _ForwardIterator,
     _BinaryPredicate,
     /* IsVector = */ std::false_type,
     bool) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator
 __pattern_adjacent_find(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _BinaryPredicate, bool) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __pattern_adjacent_find(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _BinaryPredicate,
     bool);
 
 //------------------------------------------------------------------------
 // nth_element
 //------------------------------------------------------------------------
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_nth_element(
     _Tag, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _RandomAccessIterator, _Compare) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_nth_element(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare) noexcept;
 
 //------------------------------------------------------------------------
 // fill, fill_n
 //------------------------------------------------------------------------
 template <class _RandomAccessIterator, class _Tp>
 void __brick_fill(_RandomAccessIterator,
                   _RandomAccessIterator,
                   const _Tp&,
                   /* __is_vector = */ std::true_type) noexcept;
 
 template <class _ForwardIterator, class _Tp>
 void __brick_fill(_ForwardIterator,
                   _ForwardIterator,
                   const _Tp&,
                   /* __is_vector = */ std::false_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 void __pattern_fill(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, const _Tp&) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Tp>
 _RandomAccessIterator
 __pattern_fill(__parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, const _Tp&);
 
 template <class _RandomAccessIterator, class _Size, class _Tp>
 _RandomAccessIterator
 __brick_fill_n(_RandomAccessIterator,
                _Size,
                const _Tp&,
                /* __is_vector = */ std::true_type) noexcept;
 
 template <class _OutputIterator, class _Size, class _Tp>
 _OutputIterator
 __brick_fill_n(_OutputIterator,
                _Size,
                const _Tp&,
                /* __is_vector = */ std::false_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _OutputIterator, class _Size, class _Tp>
 _OutputIterator __pattern_fill_n(_Tag, _ExecutionPolicy&&, _OutputIterator, _Size, const _Tp&) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Tp>
 _RandomAccessIterator
 __pattern_fill_n(__parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _Size, const _Tp&);
 
 //------------------------------------------------------------------------
 // generate, generate_n
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _Generator>
 void __brick_generate(_RandomAccessIterator,
                       _RandomAccessIterator,
                       _Generator,
                       /* is_vector = */ std::true_type) noexcept;
 
 template <class _ForwardIterator, class _Generator>
 void __brick_generate(_ForwardIterator,
                       _ForwardIterator,
                       _Generator,
                       /* is_vector = */ std::false_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Generator>
 void __pattern_generate(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Generator) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Generator>
 _RandomAccessIterator __pattern_generate(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Generator);
 
 template <class _RandomAccessIterator, class Size, class _Generator>
 _RandomAccessIterator __brick_generate_n(
     _RandomAccessIterator,
     Size,
     _Generator,
     /* is_vector = */ std::true_type) noexcept;
 
 template <class OutputIterator, class Size, class _Generator>
 OutputIterator __brick_generate_n(
     OutputIterator,
     Size,
     _Generator,
     /* is_vector = */ std::false_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class OutputIterator, class Size, class _Generator>
 OutputIterator __pattern_generate_n(_Tag, _ExecutionPolicy&&, OutputIterator, Size, _Generator) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class Size, class _Generator>
 _RandomAccessIterator
 __pattern_generate_n(__parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, Size, _Generator);
 
 //------------------------------------------------------------------------
 // remove
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator __brick_remove_if(
     _ForwardIterator,
     _ForwardIterator,
     _UnaryPredicate,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __brick_remove_if(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _UnaryPredicate,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator
 __pattern_remove_if(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _UnaryPredicate) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __pattern_remove_if(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _UnaryPredicate) noexcept;
 
 //------------------------------------------------------------------------
 // merge
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_merge(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_merge(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_merge(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_merge(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // inplace_merge
 //------------------------------------------------------------------------
 
 template <class _BidirectionalIterator, class _Compare>
 void __brick_inplace_merge(
     _BidirectionalIterator,
     _BidirectionalIterator,
     _BidirectionalIterator,
     _Compare,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Compare>
 void __brick_inplace_merge(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator, class _Compare>
 void __pattern_inplace_merge(
     _Tag,
     _ExecutionPolicy&&,
     _BidirectionalIterator,
     _BidirectionalIterator,
     _BidirectionalIterator,
     _Compare) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_inplace_merge(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // includes
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 bool __pattern_includes(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 bool __pattern_includes(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Compare);
 
 //------------------------------------------------------------------------
 // set_union
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_union(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_union(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_union(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_union(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // set_intersection
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_intersection(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_intersection(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_intersection(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_intersection(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // set_difference
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_difference(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_difference(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_difference(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_difference(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // set_symmetric_difference
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_symmetric_difference(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_symmetric_difference(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_symmetric_difference(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _OutputIterator,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_symmetric_difference(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _OutputIterator,
     _Compare);
 
 //------------------------------------------------------------------------
 // is_heap_until
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __brick_is_heap_until(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __brick_is_heap_until(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator
 __pattern_is_heap_until(_Tag, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare) noexcept;
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __pattern_is_heap_until(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare) noexcept;
 
 //------------------------------------------------------------------------
 // min_element
 //------------------------------------------------------------------------
 
 template <typename _ForwardIterator, typename _Compare>
 _ForwardIterator __brick_min_element(
     _ForwardIterator,
     _ForwardIterator,
     _Compare,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <typename _RandomAccessIterator, typename _Compare>
 _RandomAccessIterator __brick_min_element(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <typename _Tag, typename _ExecutionPolicy, typename _ForwardIterator, typename _Compare>
 _ForwardIterator __pattern_min_element(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Compare) noexcept;
 
 template <typename _IsVector, typename _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare>
 _RandomAccessIterator __pattern_min_element(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare);
 
 //------------------------------------------------------------------------
 // minmax_element
 //------------------------------------------------------------------------
 
 template <typename _ForwardIterator, typename _Compare>
 std::pair<_ForwardIterator, _ForwardIterator> __brick_minmax_element(
     _ForwardIterator,
     _ForwardIterator,
     _Compare,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <typename _RandomAccessIterator, typename _Compare>
 std::pair<_RandomAccessIterator, _RandomAccessIterator> __brick_minmax_element(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Compare,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <typename _Tag, typename _ExecutionPolicy, typename _ForwardIterator, typename _Compare>
 std::pair<_ForwardIterator, _ForwardIterator>
 __pattern_minmax_element(_Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Compare) noexcept;
 
 template <typename _IsVector, typename _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare>
 std::pair<_RandomAccessIterator, _RandomAccessIterator> __pattern_minmax_element(
     __parallel_tag<_IsVector>, _ExecutionPolicy&&, _RandomAccessIterator, _RandomAccessIterator, _Compare);
 
 //------------------------------------------------------------------------
 // mismatch
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 std::pair<_ForwardIterator1, _ForwardIterator2> __brick_mismatch(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _Predicate,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Predicate>
 std::pair<_RandomAccessIterator1, _RandomAccessIterator2> __brick_mismatch(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Predicate,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 std::pair<_ForwardIterator1, _ForwardIterator2> __pattern_mismatch(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _Predicate) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Predicate>
 std::pair<_RandomAccessIterator1, _RandomAccessIterator2> __pattern_mismatch(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Predicate) noexcept;
 
 //------------------------------------------------------------------------
 // lexicographical_compare
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 bool __brick_lexicographical_compare(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _Compare,
     /* __is_vector = */ std::false_type) noexcept;
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Compare>
 bool __brick_lexicographical_compare(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Compare,
     /* __is_vector = */ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 bool __pattern_lexicographical_compare(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _ForwardIterator2,
     _Compare) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 bool __pattern_lexicographical_compare(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _RandomAccessIterator2,
     _Compare) noexcept;
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_ALGORITHM_FWD_H */

libcxx/include/__pstl/internal/algorithm_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_ALGORITHM_IMPL_H
 #define _PSTL_ALGORITHM_IMPL_H
 
 #include <algorithm>
 #include <functional>
 #include <iterator>
 #include <type_traits>
 #include <utility>
 
 #include "execution_impl.h"
 #include "memory_impl.h"
 #include "parallel_backend.h"
 #include "parallel_backend_utils.h"
 #include "parallel_impl.h"
 #include "pstl_config.h"
 #include "unseq_backend_simd.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 //------------------------------------------------------------------------
 // any_of
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Pred>
 bool __brick_any_of(const _ForwardIterator __first,
                     const _ForwardIterator __last,
                     _Pred __pred,
                     /*__is_vector=*/std::false_type) noexcept {
   return std::any_of(__first, __last, __pred);
 };
 
 template <class _RandomAccessIterator, class _Pred>
 bool __brick_any_of(const _RandomAccessIterator __first,
                     const _RandomAccessIterator __last,
                     _Pred __pred,
                     /*__is_vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_or(__first, __last - __first, __pred);
 };
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Pred>
 bool __pattern_any_of(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Pred __pred) noexcept {
   return __internal::__brick_any_of(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Pred>
 bool __pattern_any_of(__parallel_tag<_IsVector> __tag,
                       _ExecutionPolicy&& __exec,
                       _RandomAccessIterator __first,
                       _RandomAccessIterator __last,
                       _Pred __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __internal::__parallel_or(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__pred](_RandomAccessIterator __i, _RandomAccessIterator __j) {
           return __internal::__brick_any_of(__i, __j, __pred, _IsVector{});
         });
   });
 }
 
 // [alg.foreach]
 // for_each_n with no policy
 
 template <class _ForwardIterator, class _Size, class _Function>
 _ForwardIterator __for_each_n_it_serial(_ForwardIterator __first, _Size __n, _Function __f) {
   for (; __n > 0; ++__first, --__n)
     __f(__first);
   return __first;
 }
 
 //------------------------------------------------------------------------
 // walk1 (pseudo)
 //
 // walk1 evaluates f(x) for each dereferenced value x drawn from [first,last)
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _Function>
 void __brick_walk1(
     _ForwardIterator __first, _ForwardIterator __last, _Function __f, /*vector=*/std::false_type) noexcept {
   std::for_each(__first, __last, __f);
 }
 
 template <class _RandomAccessIterator, class _Function>
 void __brick_walk1(_RandomAccessIterator __first,
                    _RandomAccessIterator __last,
                    _Function __f,
                    /*vector=*/std::true_type) noexcept {
   __unseq_backend::__simd_walk_1(__first, __last - __first, __f);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Function>
 void __pattern_walk1(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Function __f) noexcept {
   __internal::__brick_walk1(__first, __last, __f, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Function>
 void __pattern_walk1(__parallel_tag<_IsVector> __tag,
                      _ExecutionPolicy&& __exec,
                      _RandomAccessIterator __first,
                      _RandomAccessIterator __last,
                      _Function __f) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__f](_RandomAccessIterator __i, _RandomAccessIterator __j) {
           __internal::__brick_walk1(__i, __j, __f, _IsVector{});
         });
   });
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Brick>
 void __pattern_walk_brick(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Brick __brick) noexcept {
   __brick(__first, __last);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Brick>
 void __pattern_walk_brick(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Brick __brick) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__brick](_RandomAccessIterator __i, _RandomAccessIterator __j) { __brick(__i, __j); });
   });
 }
 
 //------------------------------------------------------------------------
 // walk1_n
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _Size, class _Function>
 _ForwardIterator __brick_walk1_n(_ForwardIterator __first, _Size __n, _Function __f, /*_IsVectorTag=*/std::false_type) {
   return __internal::__for_each_n_it_serial(__first, __n, [&__f](_ForwardIterator __it) {
     __f(*__it);
   }); // calling serial version
 }
 
 template <class _RandomAccessIterator, class _DifferenceType, class _Function>
 _RandomAccessIterator
 __brick_walk1_n(_RandomAccessIterator __first,
                 _DifferenceType __n,
                 _Function __f,
                 /*vectorTag=*/std::true_type) noexcept {
   return __unseq_backend::__simd_walk_1(__first, __n, __f);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Function>
 _ForwardIterator
 __pattern_walk1_n(_Tag, _ExecutionPolicy&&, _ForwardIterator __first, _Size __n, _Function __f) noexcept {
   return __internal::__brick_walk1_n(__first, __n, __f, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Function>
 _RandomAccessIterator __pattern_walk1_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _Size __n,
     _Function __f) {
   __internal::__pattern_walk1(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __first + __n, __f);
 
   return __first + __n;
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Brick>
 _ForwardIterator
 __pattern_walk_brick_n(_Tag, _ExecutionPolicy&&, _ForwardIterator __first, _Size __n, _Brick __brick) noexcept {
   return __brick(__first, __n);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Brick>
 _RandomAccessIterator __pattern_walk_brick_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _Size __n,
     _Brick __brick) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __first + __n,
         [__brick](_RandomAccessIterator __i, _RandomAccessIterator __j) { __brick(__i, __j - __i); });
     return __first + __n;
   });
 }
 
 //------------------------------------------------------------------------
 // walk2 (pseudo)
 //
 // walk2 evaluates f(x,y) for deferenced values (x,y) drawn from [first1,last1) and [first2,...)
 //------------------------------------------------------------------------
 template <class _ForwardIterator1, class _ForwardIterator2, class _Function>
 _ForwardIterator2
 __brick_walk2(_ForwardIterator1 __first1,
               _ForwardIterator1 __last1,
               _ForwardIterator2 __first2,
               _Function __f,
               /*vector=*/std::false_type) noexcept {
   for (; __first1 != __last1; ++__first1, ++__first2)
     __f(*__first1, *__first2);
   return __first2;
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Function>
 _RandomAccessIterator2
 __brick_walk2(_RandomAccessIterator1 __first1,
               _RandomAccessIterator1 __last1,
               _RandomAccessIterator2 __first2,
               _Function __f,
               /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_walk_2(__first1, __last1 - __first1, __first2, __f);
 }
 
 template <class _ForwardIterator1, class _Size, class _ForwardIterator2, class _Function>
 _ForwardIterator2 __brick_walk2_n(
     _ForwardIterator1 __first1,
     _Size __n,
     _ForwardIterator2 __first2,
     _Function __f,
     /*vector=*/std::false_type) noexcept {
   for (; __n > 0; --__n, ++__first1, ++__first2)
     __f(*__first1, *__first2);
   return __first2;
 }
 
 template <class _RandomAccessIterator1, class _Size, class _RandomAccessIterator2, class _Function>
 _RandomAccessIterator2 __brick_walk2_n(
     _RandomAccessIterator1 __first1,
     _Size __n,
     _RandomAccessIterator2 __first2,
     _Function __f,
     /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_walk_2(__first1, __n, __first2, __f);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Function>
 _ForwardIterator2 __pattern_walk2(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Function __f) noexcept {
   return __internal::__brick_walk2(__first1, __last1, __first2, __f, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Function>
 _RandomAccessIterator2 __pattern_walk2(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _Function __f) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         [__f, __first1, __first2](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           __internal::__brick_walk2(__i, __j, __first2 + (__i - __first1), __f, _IsVector{});
         });
     return __first2 + (__last1 - __first1);
   });
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _Size,
           class _ForwardIterator2,
           class _Function>
 _ForwardIterator2 __pattern_walk2_n(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _Size __n,
     _ForwardIterator2 __first2,
     _Function __f) noexcept {
   return __internal::__brick_walk2_n(__first1, __n, __first2, __f, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _Size,
           class _RandomAccessIterator2,
           class _Function>
 _RandomAccessIterator2 __pattern_walk2_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _Size __n,
     _RandomAccessIterator2 __first2,
     _Function __f) {
   return __internal::__pattern_walk2(
       __tag, std::forward<_ExecutionPolicy>(__exec), __first1, __first1 + __n, __first2, __f);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Brick>
 _ForwardIterator2 __pattern_walk2_brick(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Brick __brick) noexcept {
   return __brick(__first1, __last1, __first2);
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Brick>
 _RandomAccessIterator2 __pattern_walk2_brick(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _Brick __brick) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         [__first1, __first2, __brick](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           __brick(__i, __j, __first2 + (__i - __first1));
         });
     return __first2 + (__last1 - __first1);
   });
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _Size,
           class _ForwardIterator2,
           class _Brick>
 _ForwardIterator2 __pattern_walk2_brick_n(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _Size __n,
     _ForwardIterator2 __first2,
     _Brick __brick) noexcept {
   return __brick(__first1, __n, __first2);
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _Size,
           class _RandomAccessIterator2,
           class _Brick>
 _RandomAccessIterator2 __pattern_walk2_brick_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _Size __n,
     _RandomAccessIterator2 __first2,
     _Brick __brick) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __first1 + __n,
         [__first1, __first2, __brick](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           __brick(__i, __j - __i, __first2 + (__i - __first1));
         });
     return __first2 + __n;
   });
 }
 
 //------------------------------------------------------------------------
 // walk3 (pseudo)
 //
 // walk3 evaluates f(x,y,z) for (x,y,z) drawn from [first1,last1), [first2,...), [first3,...)
 //------------------------------------------------------------------------
 template <class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator3, class _Function>
 _ForwardIterator3 __brick_walk3(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator3 __first3,
     _Function __f,
     /*vector=*/std::false_type) noexcept {
   for (; __first1 != __last1; ++__first1, ++__first2, ++__first3)
     __f(*__first1, *__first2, *__first3);
   return __first3;
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Function>
 _RandomAccessIterator3 __brick_walk3(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator3 __first3,
     _Function __f,
     /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_walk_3(__first1, __last1 - __first1, __first2, __first3, __f);
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator3,
           class _Function>
 _ForwardIterator3 __pattern_walk3(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator3 __first3,
     _Function __f) noexcept {
   return __internal::__brick_walk3(__first1, __last1, __first2, __first3, __f, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _Function>
 _RandomAccessIterator3 __pattern_walk3(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator3 __first3,
     _Function __f) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         [__f, __first1, __first2, __first3](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           __internal::__brick_walk3(
               __i, __j, __first2 + (__i - __first1), __first3 + (__i - __first1), __f, _IsVector{});
         });
     return __first3 + (__last1 - __first1);
   });
 }
 
 //------------------------------------------------------------------------
 // equal
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __brick_equal(_ForwardIterator1 __first1,
                    _ForwardIterator1 __last1,
                    _ForwardIterator2 __first2,
                    _ForwardIterator2 __last2,
                    _BinaryPredicate __p,
                    /* IsVector = */ std::false_type) noexcept {
   return std::equal(__first1, __last1, __first2, __last2, __p);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 bool __brick_equal(_RandomAccessIterator1 __first1,
                    _RandomAccessIterator1 __last1,
                    _RandomAccessIterator2 __first2,
                    _RandomAccessIterator2 __last2,
                    _BinaryPredicate __p,
                    /* is_vector = */ std::true_type) noexcept {
   if (__last1 - __first1 != __last2 - __first2)
     return false;
 
   return __unseq_backend::__simd_first(__first1, __last1 - __first1, __first2, std::not_fn(__p)).first == __last1;
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __pattern_equal(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _BinaryPredicate __p) noexcept {
   return __internal::__brick_equal(__first1, __last1, __first2, __last2, __p, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 bool __pattern_equal(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _BinaryPredicate __p) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__last1 - __first1 != __last2 - __first2)
     return false;
 
   return __internal::__except_handler([&]() {
     return !__internal::__parallel_or(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         [__first1, __first2, __p](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           return !__internal::__brick_equal(
               __i, __j, __first2 + (__i - __first1), __first2 + (__j - __first1), __p, _IsVector{});
         });
   });
 }
 
 //------------------------------------------------------------------------
 // equal version for sequences with equal length
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __brick_equal(_ForwardIterator1 __first1,
                    _ForwardIterator1 __last1,
                    _ForwardIterator2 __first2,
                    _BinaryPredicate __p,
                    /* IsVector = */ std::false_type) noexcept {
   return std::equal(__first1, __last1, __first2, __p);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 bool __brick_equal(_RandomAccessIterator1 __first1,
                    _RandomAccessIterator1 __last1,
                    _RandomAccessIterator2 __first2,
                    _BinaryPredicate __p,
                    /* is_vector = */ std::true_type) noexcept {
   return __unseq_backend::__simd_first(__first1, __last1 - __first1, __first2, std::not_fn(__p)).first == __last1;
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 bool __pattern_equal(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _BinaryPredicate __p) noexcept {
   return __internal::__brick_equal(__first1, __last1, __first2, __p, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 bool __pattern_equal(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _BinaryPredicate __p) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return !__internal::__parallel_or(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         [__first1, __first2, __p](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           return !__internal::__brick_equal(__i, __j, __first2 + (__i - __first1), __p, _IsVector{});
         });
   });
 }
 
 //------------------------------------------------------------------------
 // find_if
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _Predicate>
 _ForwardIterator
 __brick_find_if(_ForwardIterator __first,
                 _ForwardIterator __last,
                 _Predicate __pred,
                 /*is_vector=*/std::false_type) noexcept {
   return std::find_if(__first, __last, __pred);
 }
 
 template <class _RandomAccessIterator, class _Predicate>
 _RandomAccessIterator
 __brick_find_if(_RandomAccessIterator __first,
                 _RandomAccessIterator __last,
                 _Predicate __pred,
                 /*is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _SizeType;
   return __unseq_backend::__simd_first(
       __first, _SizeType(0), __last - __first, [&__pred](_RandomAccessIterator __it, _SizeType __i) {
         return __pred(__it[__i]);
       });
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 _ForwardIterator __pattern_find_if(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) noexcept {
   return __internal::__brick_find_if(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Predicate>
 _RandomAccessIterator __pattern_find_if(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Predicate __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __internal::__parallel_find(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__pred](_RandomAccessIterator __i, _RandomAccessIterator __j) {
           return __internal::__brick_find_if(__i, __j, __pred, _IsVector{});
         },
         std::less<typename std::iterator_traits<_RandomAccessIterator>::difference_type>(),
         /*is_first=*/true);
   });
 }
 
 //------------------------------------------------------------------------
 // find_end
 //------------------------------------------------------------------------
 
 // find the first occurrence of the subsequence [s_first, s_last)
 //   or the  last occurrence of the subsequence in the range [first, last)
 // b_first determines what occurrence we want to find (first or last)
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate, class _IsVector>
 _RandomAccessIterator1 __find_subrange(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator1 __global_last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred,
     bool __b_first,
     _IsVector __is_vector) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator2>::value_type _ValueType;
   auto __n2 = __s_last - __s_first;
   if (__n2 < 1) {
     return __b_first ? __first : __last;
   }
 
   auto __n1 = __global_last - __first;
   if (__n1 < __n2) {
     return __last;
   }
 
   auto __cur = __last;
   while (__first != __last && (__global_last - __first >= __n2)) {
     // find position of *s_first in [first, last) (it can be start of subsequence)
     __first = __internal::__brick_find_if(
         __first, __last, __equal_value_by_pred<_ValueType, _BinaryPredicate>(*__s_first, __pred), __is_vector);
 
     // if position that was found previously is the start of subsequence
     // then we can exit the loop (b_first == true) or keep the position
     // (b_first == false)
     if (__first != __last && (__global_last - __first >= __n2) &&
         __internal::__brick_equal(__s_first + 1, __s_last, __first + 1, __pred, __is_vector)) {
       if (__b_first) {
         return __first;
       } else {
         __cur = __first;
       }
     } else if (__first == __last) {
       break;
     } else {
     }
 
     // in case of b_first == false we try to find new start position
     // for the next subsequence
     ++__first;
   }
   return __cur;
 }
 
 template <class _RandomAccessIterator, class _Size, class _Tp, class _BinaryPredicate, class _IsVector>
 _RandomAccessIterator __find_subrange(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _RandomAccessIterator __global_last,
     _Size __count,
     const _Tp& __value,
     _BinaryPredicate __pred,
     _IsVector __is_vector) noexcept {
   if (static_cast<_Size>(__global_last - __first) < __count || __count < 1) {
     return __last; // According to the standard last shall be returned when count < 1
   }
 
   auto __unary_pred = __equal_value_by_pred<_Tp, _BinaryPredicate>(__value, __pred);
   while (__first != __last && (static_cast<_Size>(__global_last - __first) >= __count)) {
     __first = __internal::__brick_find_if(__first, __last, __unary_pred, __is_vector);
 
     // check that all of elements in [first+1, first+count) equal to value
     if (__first != __last && (static_cast<_Size>(__global_last - __first) >= __count) &&
         !__internal::__brick_any_of(__first + 1, __first + __count, std::not_fn(__unary_pred), __is_vector)) {
       return __first;
     } else if (__first == __last) {
       break;
     } else {
       ++__first;
     }
   }
   return __last;
 }
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __brick_find_end(
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred,
     /*__is_vector=*/std::false_type) noexcept {
   return std::find_end(__first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_find_end(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred,
     /*__is_vector=*/std::true_type) noexcept {
   return __find_subrange(__first, __last, __last, __s_first, __s_last, __pred, false, std::true_type());
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __pattern_find_end(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred) noexcept {
   return __internal::__brick_find_end(__first, __last, __s_first, __s_last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator1 __pattern_find_end(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__last - __first == __s_last - __s_first) {
     const bool __res =
         __internal::__pattern_equal(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __pred);
     return __res ? __first : __last;
   } else {
     return __internal::__except_handler([&]() {
       return __internal::__parallel_find(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __first,
           __last,
           [__last, __s_first, __s_last, __pred](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
             return __internal::__find_subrange(__i, __j, __last, __s_first, __s_last, __pred, false, _IsVector{});
           },
           std::greater<typename std::iterator_traits<_RandomAccessIterator1>::difference_type>(),
           /*is_first=*/false);
     });
   }
 }
 
 //------------------------------------------------------------------------
 // find_first_of
 //------------------------------------------------------------------------
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __brick_find_first_of(
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred,
     /*__is_vector=*/std::false_type) noexcept {
   return std::find_first_of(__first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_find_first_of(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred,
     /*__is_vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_find_first_of(__first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __pattern_find_first_of(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred) noexcept {
   return __internal::__brick_find_first_of(__first, __last, __s_first, __s_last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator1 __pattern_find_first_of(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __internal::__parallel_find(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__s_first, __s_last, __pred](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           return __internal::__brick_find_first_of(__i, __j, __s_first, __s_last, __pred, _IsVector{});
         },
         std::less<typename std::iterator_traits<_RandomAccessIterator1>::difference_type>(),
         /*is_first=*/true);
   });
 }
 
 //------------------------------------------------------------------------
 // search
 //------------------------------------------------------------------------
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_search(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred,
     /*vector=*/std::false_type) noexcept {
   return std::search(__first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator1 __brick_search(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred,
     /*vector=*/std::true_type) noexcept {
   return __internal::__find_subrange(__first, __last, __last, __s_first, __s_last, __pred, true, std::true_type());
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1 __pattern_search(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred) noexcept {
   return __internal::__brick_search(__first, __last, __s_first, __s_last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator1 __pattern_search(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __s_first,
     _RandomAccessIterator2 __s_last,
     _BinaryPredicate __pred) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__last - __first == __s_last - __s_first) {
     const bool __res =
         __internal::__pattern_equal(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __pred);
     return __res ? __first : __last;
   } else {
     return __internal::__except_handler([&]() {
       return __internal::__parallel_find(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __first,
           __last,
           [__last, __s_first, __s_last, __pred](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
             return __internal::__find_subrange(__i, __j, __last, __s_first, __s_last, __pred, true, _IsVector{});
           },
           std::less<typename std::iterator_traits<_RandomAccessIterator1>::difference_type>(),
           /*is_first=*/true);
     });
   }
 }
 
 //------------------------------------------------------------------------
 // search_n
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
 _ForwardIterator __brick_search_n(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Size __count,
     const _Tp& __value,
     _BinaryPredicate __pred,
     /*vector=*/std::false_type) noexcept {
   return std::search_n(__first, __last, __count, __value, __pred);
 }
 
 template <class _RandomAccessIterator, class _Size, class _Tp, class _BinaryPredicate>
 _RandomAccessIterator __brick_search_n(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Size __count,
     const _Tp& __value,
     _BinaryPredicate __pred,
     /*vector=*/std::true_type) noexcept {
   return __internal::__find_subrange(__first, __last, __last, __count, __value, __pred, std::true_type());
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
 _ForwardIterator __pattern_search_n(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Size __count,
     const _Tp& __value,
     _BinaryPredicate __pred) noexcept {
   return __internal::__brick_search_n(__first, __last, __count, __value, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _Size,
           class _Tp,
           class _BinaryPredicate>
 _RandomAccessIterator __pattern_search_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Size __count,
     const _Tp& __value,
     _BinaryPredicate __pred) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (static_cast<_Size>(__last - __first) == __count) {
     const bool __result = !__internal::__pattern_any_of(
         __tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, [&__value, &__pred](const _Tp& __val) {
           return !__pred(__val, __value);
         });
     return __result ? __first : __last;
   } else {
     return __internal::__except_handler([&__exec, __first, __last, __count, &__value, __pred]() {
       return __internal::__parallel_find(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __first,
           __last,
           [__last, __count, &__value, __pred](_RandomAccessIterator __i, _RandomAccessIterator __j) {
             return __internal::__find_subrange(__i, __j, __last, __count, __value, __pred, _IsVector{});
           },
           std::less<typename std::iterator_traits<_RandomAccessIterator>::difference_type>(),
           /*is_first=*/true);
     });
   }
 }
 
 //------------------------------------------------------------------------
 // copy_n
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Size, class _OutputIterator>
 _OutputIterator
 __brick_copy_n(_ForwardIterator __first, _Size __n, _OutputIterator __result, /*vector=*/std::false_type) noexcept {
   return std::copy_n(__first, __n, __result);
 }
 
 template <class _RandomAccessIterator1, class _Size, class _RandomAccessIterator2>
 _RandomAccessIterator2
 __brick_copy_n(_RandomAccessIterator1 __first,
                _Size __n,
                _RandomAccessIterator2 __result,
                /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_assign(
       __first, __n, __result, [](_RandomAccessIterator1 __first, _RandomAccessIterator2 __result) {
         *__result = *__first;
       });
 }
 
 //------------------------------------------------------------------------
 // copy
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator
 __brick_copy(_ForwardIterator __first,
              _ForwardIterator __last,
              _OutputIterator __result,
              /*vector=*/std::false_type) noexcept {
   return std::copy(__first, __last, __result);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2
 __brick_copy(_RandomAccessIterator1 __first,
              _RandomAccessIterator1 __last,
              _RandomAccessIterator2 __result,
              /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_assign(
       __first, __last - __first, __result, [](_RandomAccessIterator1 __first, _RandomAccessIterator2 __result) {
         *__result = *__first;
       });
 }
 
 //------------------------------------------------------------------------
 // move
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator
 __brick_move(_ForwardIterator __first,
              _ForwardIterator __last,
              _OutputIterator __result,
              /*vector=*/std::false_type) noexcept {
   return std::move(__first, __last, __result);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2
 __brick_move(_RandomAccessIterator1 __first,
              _RandomAccessIterator1 __last,
              _RandomAccessIterator2 __result,
              /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_assign(
       __first, __last - __first, __result, [](_RandomAccessIterator1 __first, _RandomAccessIterator2 __result) {
         *__result = std::move(*__first);
       });
 }
 
 struct __brick_move_destroy {
   template <typename _RandomAccessIterator1, typename _RandomAccessIterator2>
   _RandomAccessIterator2
   operator()(_RandomAccessIterator1 __first,
              _RandomAccessIterator1 __last,
              _RandomAccessIterator2 __result,
              /*vec*/ std::true_type) const {
     using _IteratorValueType = typename std::iterator_traits<_RandomAccessIterator1>::value_type;
 
     return __unseq_backend::__simd_assign(
         __first, __last - __first, __result, [](_RandomAccessIterator1 __first, _RandomAccessIterator2 __result) {
           *__result = std::move(*__first);
           (*__first).~_IteratorValueType();
         });
   }
 
   template <typename _RandomAccessIterator1, typename _RandomAccessIterator2>
   _RandomAccessIterator2
   operator()(_RandomAccessIterator1 __first,
              _RandomAccessIterator1 __last,
              _RandomAccessIterator2 __result,
              /*vec*/ std::false_type) const {
     using _IteratorValueType = typename std::iterator_traits<_RandomAccessIterator1>::value_type;
 
     for (; __first != __last; ++__first, ++__result) {
       *__result = std::move(*__first);
       (*__first).~_IteratorValueType();
     }
     return __result;
   }
 };
 
 //------------------------------------------------------------------------
 // swap_ranges
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator __brick_swap_ranges(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     /*vector=*/std::false_type) noexcept {
   return std::swap_ranges(__first, __last, __result);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2 __brick_swap_ranges(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     /*vector=*/std::true_type) noexcept {
   using std::iter_swap;
   return __unseq_backend::__simd_assign(
       __first, __last - __first, __result, iter_swap<_RandomAccessIterator1, _RandomAccessIterator2>);
 }
 
 //------------------------------------------------------------------------
 // copy_if
 //------------------------------------------------------------------------
 template <class _ForwardIterator, class _OutputIterator, class _UnaryPredicate>
 _OutputIterator __brick_copy_if(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     _UnaryPredicate __pred,
     /*vector=*/std::false_type) noexcept {
   return std::copy_if(__first, __last, __result, __pred);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _UnaryPredicate>
 _RandomAccessIterator2 __brick_copy_if(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     _UnaryPredicate __pred,
     /*vector=*/std::true_type) noexcept {
 #if defined(_PSTL_MONOTONIC_PRESENT)
   return __unseq_backend::__simd_copy_if(__first, __last - __first, __result, __pred);
 #else
   return std::copy_if(__first, __last, __result, __pred);
 #endif
 }
 
 // TODO: Try to use transform_reduce for combining __brick_copy_if_phase1 on IsVector.
 template <class _DifferenceType, class _ForwardIterator, class _UnaryPredicate>
 std::pair<_DifferenceType, _DifferenceType> __brick_calc_mask_1(
     _ForwardIterator __first,
     _ForwardIterator __last,
     bool* __restrict __mask,
     _UnaryPredicate __pred,
     /*vector=*/std::false_type) noexcept {
   auto __count_true = _DifferenceType(0);
   auto __size       = __last - __first;
 
   static_assert(__are_random_access_iterators<_ForwardIterator>::value,
                 "Pattern-brick error. Should be a random access iterator.");
 
   for (; __first != __last; ++__first, ++__mask) {
     *__mask = __pred(*__first);
     if (*__mask) {
       ++__count_true;
     }
   }
   return std::make_pair(__count_true, __size - __count_true);
 }
 
 template <class _DifferenceType, class _RandomAccessIterator, class _UnaryPredicate>
 std::pair<_DifferenceType, _DifferenceType> __brick_calc_mask_1(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     bool* __mask,
     _UnaryPredicate __pred,
     /*vector=*/std::true_type) noexcept {
   auto __result = __unseq_backend::__simd_calc_mask_1(__first, __last - __first, __mask, __pred);
   return std::make_pair(__result, (__last - __first) - __result);
 }
 
 template <class _ForwardIterator, class _OutputIterator, class _Assigner>
 void __brick_copy_by_mask(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     bool* __mask,
     _Assigner __assigner,
     /*vector=*/std::false_type) noexcept {
   for (; __first != __last; ++__first, ++__mask) {
     if (*__mask) {
       __assigner(__first, __result);
       ++__result;
     }
   }
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Assigner>
 void __brick_copy_by_mask(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     bool* __restrict __mask,
     _Assigner __assigner,
     /*vector=*/std::true_type) noexcept {
 #if defined(_PSTL_MONOTONIC_PRESENT)
   __unseq_backend::__simd_copy_by_mask(__first, __last - __first, __result, __mask, __assigner);
 #else
   __internal::__brick_copy_by_mask(__first, __last, __result, __mask, __assigner, std::false_type());
 #endif
 }
 
 template <class _ForwardIterator, class _OutputIterator1, class _OutputIterator2>
 void __brick_partition_by_mask(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator1 __out_true,
     _OutputIterator2 __out_false,
     bool* __mask,
     /*vector=*/std::false_type) noexcept {
   for (; __first != __last; ++__first, ++__mask) {
     if (*__mask) {
       *__out_true = *__first;
       ++__out_true;
     } else {
       *__out_false = *__first;
       ++__out_false;
     }
   }
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3>
 void __brick_partition_by_mask(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __out_true,
     _RandomAccessIterator3 __out_false,
     bool* __mask,
     /*vector=*/std::true_type) noexcept {
 #if defined(_PSTL_MONOTONIC_PRESENT)
   __unseq_backend::__simd_partition_by_mask(__first, __last - __first, __out_true, __out_false, __mask);
 #else
   __internal::__brick_partition_by_mask(__first, __last, __out_true, __out_false, __mask, std::false_type());
 #endif
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator, class _UnaryPredicate>
 _OutputIterator __pattern_copy_if(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     _UnaryPredicate __pred) noexcept {
   return __internal::__brick_copy_if(__first, __last, __result, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _UnaryPredicate>
 _RandomAccessIterator2 __pattern_copy_if(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     _UnaryPredicate __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType;
   const _DifferenceType __n = __last - __first;
   if (_DifferenceType(1) < __n) {
     __par_backend::__buffer<bool> __mask_buf(__n);
     return __internal::__except_handler([&__exec, __n, __first, __result, __pred, &__mask_buf]() {
       bool* __mask = __mask_buf.get();
       _DifferenceType __m{};
       __par_backend::__parallel_strict_scan(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __n,
           _DifferenceType(0),
           [=](_DifferenceType __i, _DifferenceType __len) { // Reduce
             return __internal::__brick_calc_mask_1<_DifferenceType>(
                        __first + __i, __first + (__i + __len), __mask + __i, __pred, _IsVector{})
                 .first;
           },
           std::plus<_DifferenceType>(),                                                // Combine
           [=](_DifferenceType __i, _DifferenceType __len, _DifferenceType __initial) { // Scan
             __internal::__brick_copy_by_mask(
                 __first + __i,
                 __first + (__i + __len),
                 __result + __initial,
                 __mask + __i,
                 [](_RandomAccessIterator1 __x, _RandomAccessIterator2 __z) { *__z = *__x; },
                 _IsVector{});
           },
           [&__m](_DifferenceType __total) { __m = __total; });
       return __result + __m;
     });
   }
   // trivial sequence - use serial algorithm
   return __internal::__brick_copy_if(__first, __last, __result, __pred, _IsVector{});
 }
 
 //------------------------------------------------------------------------
 // count
 //------------------------------------------------------------------------
 template <class _RandomAccessIterator, class _Predicate>
 typename std::iterator_traits<_RandomAccessIterator>::difference_type
 __brick_count(_RandomAccessIterator __first,
               _RandomAccessIterator __last,
               _Predicate __pred,
               /* is_vector = */ std::true_type) noexcept {
   return __unseq_backend::__simd_count(__first, __last - __first, __pred);
 }
 
 template <class _ForwardIterator, class _Predicate>
 typename std::iterator_traits<_ForwardIterator>::difference_type
 __brick_count(_ForwardIterator __first,
               _ForwardIterator __last,
               _Predicate __pred,
               /* is_vector = */ std::false_type) noexcept {
   return std::count_if(__first, __last, __pred);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 typename std::iterator_traits<_ForwardIterator>::difference_type __pattern_count(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) noexcept {
   return __internal::__brick_count(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Predicate>
 typename std::iterator_traits<_RandomAccessIterator>::difference_type __pattern_count(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Predicate __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _SizeType;
   return __internal::__except_handler([&]() {
     return __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         _SizeType(0),
         [__pred](_RandomAccessIterator __begin, _RandomAccessIterator __end, _SizeType __value) -> _SizeType {
           return __value + __internal::__brick_count(__begin, __end, __pred, _IsVector{});
         },
         std::plus<_SizeType>());
   });
 }
 
 //------------------------------------------------------------------------
 // unique
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator
 __brick_unique(_RandomAccessIterator __first,
                _RandomAccessIterator __last,
                _BinaryPredicate __pred,
                /*is_vector=*/std::false_type) noexcept {
   return std::unique(__first, __last, __pred);
 }
 
 template <class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator
 __brick_unique(_RandomAccessIterator __first,
                _RandomAccessIterator __last,
                _BinaryPredicate __pred,
                /*is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::unique(__first, __last, __pred);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator __pattern_unique(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred) noexcept {
   return __internal::__brick_unique(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 // That function is shared between two algorithms - remove_if (__pattern_remove_if) and unique (pattern unique). But a
 // mask calculation is different. So, a caller passes _CalcMask brick into remove_elements.
 template <class _IsVector, class _ExecutionPolicy, class _ForwardIterator, class _CalcMask>
 _ForwardIterator __remove_elements(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _CalcMask __calc_mask) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_ForwardIterator>::difference_type _DifferenceType;
   typedef typename std::iterator_traits<_ForwardIterator>::value_type _Tp;
   _DifferenceType __n = __last - __first;
   __par_backend::__buffer<bool> __mask_buf(__n);
   // 1. find a first iterator that should be removed
   return __internal::__except_handler([&]() {
     bool* __mask          = __mask_buf.get();
     _DifferenceType __min = __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         _DifferenceType(0),
         __n,
         __n,
         [__first, __mask, &__calc_mask](
             _DifferenceType __i, _DifferenceType __j, _DifferenceType __local_min) -> _DifferenceType {
           // Create mask
           __calc_mask(__mask + __i, __mask + __j, __first + __i);
 
           // if minimum was found in a previous range we shouldn't do anymore
           if (__local_min < __i) {
             return __local_min;
           }
           // find first iterator that should be removed
           bool* __result = __internal::__brick_find_if(
               __mask + __i, __mask + __j, [](bool __val) { return !__val; }, _IsVector{});
           if (__result - __mask == __j) {
             return __local_min;
           }
           return std::min(__local_min, _DifferenceType(__result - __mask));
         },
         [](_DifferenceType __local_min1, _DifferenceType __local_min2) -> _DifferenceType {
           return std::min(__local_min1, __local_min2);
         });
 
     // No elements to remove - exit
     if (__min == __n) {
       return __last;
     }
     __n -= __min;
     __first += __min;
 
     __par_backend::__buffer<_Tp> __buf(__n);
     _Tp* __result = __buf.get();
     __mask += __min;
     _DifferenceType __m{};
     // 2. Elements that doesn't satisfy pred are moved to result
     __par_backend::__parallel_strict_scan(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __n,
         _DifferenceType(0),
         [__mask](_DifferenceType __i, _DifferenceType __len) {
           return __internal::__brick_count(
               __mask + __i, __mask + __i + __len, [](bool __val) { return __val; }, _IsVector{});
         },
         std::plus<_DifferenceType>(),
         [=](_DifferenceType __i, _DifferenceType __len, _DifferenceType __initial) {
           __internal::__brick_copy_by_mask(
               __first + __i,
               __first + __i + __len,
               __result + __initial,
               __mask + __i,
               [](_ForwardIterator __x, _Tp* __z) {
                 __internal::__invoke_if_else(
                     std::is_trivial<_Tp>(),
                     [&]() { *__z = std::move(*__x); },
                     [&]() { ::new (std::addressof(*__z)) _Tp(std::move(*__x)); });
               },
               _IsVector{});
         },
         [&__m](_DifferenceType __total) { __m = __total; });
 
     // 3. Elements from result are moved to [first, last)
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __result,
         __result + __m,
         [__result, __first](_Tp* __i, _Tp* __j) {
           __invoke_if_else(
               std::is_trivial<_Tp>(),
               [&]() { __brick_move(__i, __j, __first + (__i - __result), _IsVector{}); },
               [&]() { __brick_move_destroy()(__i, __j, __first + (__i - __result), _IsVector{}); });
         });
     return __first + __m;
   });
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __pattern_unique(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _BinaryPredicate __pred) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::reference _ReferenceType;
 
   if (__first == __last) {
     return __last;
   }
   if (__first + 1 == __last || __first + 2 == __last) {
     // Trivial sequence - use serial algorithm
     return __internal::__brick_unique(__first, __last, __pred, _IsVector{});
   }
   return __internal::__remove_elements(
       __tag,
       std::forward<_ExecutionPolicy>(__exec),
       ++__first,
       __last,
       [&__pred](bool* __b, bool* __e, _RandomAccessIterator __it) {
         __internal::__brick_walk3(
             __b,
             __e,
             __it - 1,
             __it,
             [&__pred](bool& __x, _ReferenceType __y, _ReferenceType __z) { __x = !__pred(__y, __z); },
             _IsVector{});
       });
 }
 
 //------------------------------------------------------------------------
 // unique_copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class OutputIterator, class _BinaryPredicate>
 OutputIterator __brick_unique_copy(
     _ForwardIterator __first,
     _ForwardIterator __last,
     OutputIterator __result,
     _BinaryPredicate __pred,
     /*vector=*/std::false_type) noexcept {
   return std::unique_copy(__first, __last, __result, __pred);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _BinaryPredicate>
 _RandomAccessIterator2 __brick_unique_copy(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     _BinaryPredicate __pred,
     /*vector=*/std::true_type) noexcept {
 #if defined(_PSTL_MONOTONIC_PRESENT)
   return __unseq_backend::__simd_unique_copy(__first, __last - __first, __result, __pred);
 #else
   return std::unique_copy(__first, __last, __result, __pred);
 #endif
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator, class _BinaryPredicate>
 _OutputIterator __pattern_unique_copy(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     _BinaryPredicate __pred) noexcept {
   return __internal::__brick_unique_copy(__first, __last, __result, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _DifferenceType, class _RandomAccessIterator, class _BinaryPredicate>
 _DifferenceType __brick_calc_mask_2(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     bool* __restrict __mask,
     _BinaryPredicate __pred,
     /*vector=*/std::false_type) noexcept {
   _DifferenceType __count = 0;
   for (; __first != __last; ++__first, ++__mask) {
     *__mask = !__pred(*__first, *(__first - 1));
     __count += *__mask;
   }
   return __count;
 }
 
 template <class _DifferenceType, class _RandomAccessIterator, class _BinaryPredicate>
 _DifferenceType __brick_calc_mask_2(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     bool* __restrict __mask,
     _BinaryPredicate __pred,
     /*vector=*/std::true_type) noexcept {
   return __unseq_backend::__simd_calc_mask_2(__first, __last - __first, __mask, __pred);
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryPredicate>
 _RandomAccessIterator2 __pattern_unique_copy(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     _BinaryPredicate __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType;
   const _DifferenceType __n = __last - __first;
   if (_DifferenceType(2) < __n) {
     __par_backend::__buffer<bool> __mask_buf(__n);
     if (_DifferenceType(2) < __n) {
       return __internal::__except_handler([&__exec, __n, __first, __result, __pred, &__mask_buf]() {
         bool* __mask = __mask_buf.get();
         _DifferenceType __m{};
         __par_backend::__parallel_strict_scan(
             __backend_tag{},
             std::forward<_ExecutionPolicy>(__exec),
             __n,
             _DifferenceType(0),
             [=](_DifferenceType __i, _DifferenceType __len) -> _DifferenceType { // Reduce
               _DifferenceType __extra = 0;
               if (__i == 0) {
                 // Special boundary case
                 __mask[__i] = true;
                 if (--__len == 0)
                   return 1;
                 ++__i;
                 ++__extra;
               }
               return __internal::__brick_calc_mask_2<_DifferenceType>(
                          __first + __i, __first + (__i + __len), __mask + __i, __pred, _IsVector{}) +
                      __extra;
             },
             std::plus<_DifferenceType>(),                                                // Combine
             [=](_DifferenceType __i, _DifferenceType __len, _DifferenceType __initial) { // Scan
               // Phase 2 is same as for __pattern_copy_if
               __internal::__brick_copy_by_mask(
                   __first + __i,
                   __first + (__i + __len),
                   __result + __initial,
                   __mask + __i,
                   [](_RandomAccessIterator1 __x, _RandomAccessIterator2 __z) { *__z = *__x; },
                   _IsVector{});
             },
             [&__m](_DifferenceType __total) { __m = __total; });
         return __result + __m;
       });
     }
   }
   // trivial sequence - use serial algorithm
   return __internal::__brick_unique_copy(__first, __last, __result, __pred, _IsVector{});
 }
 
 //------------------------------------------------------------------------
 // reverse
 //------------------------------------------------------------------------
 template <class _BidirectionalIterator>
 void __brick_reverse(
     _BidirectionalIterator __first, _BidirectionalIterator __last, /*__is_vector=*/std::false_type) noexcept {
   std::reverse(__first, __last);
 }
 
 template <class _RandomAccessIterator>
 void __brick_reverse(
     _RandomAccessIterator __first, _RandomAccessIterator __last, /*__is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::reference _ReferenceType;
 
   const auto __n = (__last - __first) / 2;
   __unseq_backend::__simd_walk_2(
       __first, __n, std::reverse_iterator<_RandomAccessIterator>(__last), [](_ReferenceType __x, _ReferenceType __y) {
         using std::swap;
         swap(__x, __y);
       });
 }
 
 // this brick is called in parallel version, so we can use iterator arithmetic
 template <class _BidirectionalIterator>
 void __brick_reverse(_BidirectionalIterator __first,
                      _BidirectionalIterator __last,
                      _BidirectionalIterator __d_last,
                      /*is_vector=*/std::false_type) noexcept {
   for (--__d_last; __first != __last; ++__first, --__d_last) {
     using std::iter_swap;
     iter_swap(__first, __d_last);
   }
 }
 
 // this brick is called in parallel version, so we can use iterator arithmetic
 template <class _RandomAccessIterator>
 void __brick_reverse(_RandomAccessIterator __first,
                      _RandomAccessIterator __last,
                      _RandomAccessIterator __d_last,
                      /*is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::reference _ReferenceType;
 
   __unseq_backend::__simd_walk_2(
       __first,
       __last - __first,
       std::reverse_iterator<_RandomAccessIterator>(__d_last),
       [](_ReferenceType __x, _ReferenceType __y) {
         using std::swap;
         swap(__x, __y);
       });
 }
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator>
 void __pattern_reverse(
     _Tag, _ExecutionPolicy&&, _BidirectionalIterator __first, _BidirectionalIterator __last) noexcept {
   __internal::__brick_reverse(__first, __last, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator>
 void __pattern_reverse(__parallel_tag<_IsVector> __tag,
                        _ExecutionPolicy&& __exec,
                        _RandomAccessIterator __first,
                        _RandomAccessIterator __last) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __par_backend::__parallel_for(
       __backend_tag{},
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __first + (__last - __first) / 2,
       [__first, __last](_RandomAccessIterator __inner_first, _RandomAccessIterator __inner_last) {
         __internal::__brick_reverse(__inner_first, __inner_last, __last - (__inner_first - __first), _IsVector{});
       });
 }
 
 //------------------------------------------------------------------------
 // reverse_copy
 //------------------------------------------------------------------------
 
 template <class _BidirectionalIterator, class _OutputIterator>
 _OutputIterator __brick_reverse_copy(
     _BidirectionalIterator __first,
     _BidirectionalIterator __last,
     _OutputIterator __d_first,
     /*is_vector=*/std::false_type) noexcept {
   return std::reverse_copy(__first, __last, __d_first);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2 __brick_reverse_copy(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __d_first,
     /*is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator1>::reference _ReferenceType1;
   typedef typename std::iterator_traits<_RandomAccessIterator2>::reference _ReferenceType2;
 
   return __unseq_backend::__simd_walk_2(
       std::reverse_iterator<_RandomAccessIterator1>(__last),
       __last - __first,
       __d_first,
       [](_ReferenceType1 __x, _ReferenceType2 __y) { __y = __x; });
 }
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator, class _OutputIterator>
 _OutputIterator __pattern_reverse_copy(
     _Tag,
     _ExecutionPolicy&&,
     _BidirectionalIterator __first,
     _BidirectionalIterator __last,
     _OutputIterator __d_first) noexcept {
   return __internal::__brick_reverse_copy(__first, __last, __d_first, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2 __pattern_reverse_copy(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __d_first) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   auto __len = __last - __first;
   __par_backend::__parallel_for(
       __backend_tag{},
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       [__first, __len, __d_first](_RandomAccessIterator1 __inner_first, _RandomAccessIterator1 __inner_last) {
         __internal::__brick_reverse_copy(
             __inner_first, __inner_last, __d_first + (__len - (__inner_last - __first)), _IsVector{});
       });
   return __d_first + __len;
 }
 
 //------------------------------------------------------------------------
 // rotate
 //------------------------------------------------------------------------
 template <class _ForwardIterator>
 _ForwardIterator
 __brick_rotate(_ForwardIterator __first,
                _ForwardIterator __middle,
                _ForwardIterator __last,
                /*is_vector=*/std::false_type) noexcept {
 #if defined(_PSTL_CPP11_STD_ROTATE_BROKEN)
   std::rotate(__first, __middle, __last);
   return std::next(__first, std::distance(__middle, __last));
 #else
   return std::rotate(__first, __middle, __last);
 #endif
 }
 
 template <class _RandomAccessIterator>
 _RandomAccessIterator
 __brick_rotate(_RandomAccessIterator __first,
                _RandomAccessIterator __middle,
                _RandomAccessIterator __last,
                /*is_vector=*/std::true_type) noexcept {
   auto __n                          = __last - __first;
   auto __m                          = __middle - __first;
   const _RandomAccessIterator __ret = __first + (__last - __middle);
 
   bool __is_left = (__m <= __n / 2);
   if (!__is_left)
     __m = __n - __m;
 
   while (__n > 1 && __m > 0) {
     using std::iter_swap;
     const auto __m_2 = __m * 2;
     if (__is_left) {
       for (; __last - __first >= __m_2; __first += __m) {
         __unseq_backend::__simd_assign(
             __first, __m, __first + __m, iter_swap<_RandomAccessIterator, _RandomAccessIterator>);
       }
     } else {
       for (; __last - __first >= __m_2; __last -= __m) {
         __unseq_backend::__simd_assign(
             __last - __m, __m, __last - __m_2, iter_swap<_RandomAccessIterator, _RandomAccessIterator>);
       }
     }
     __is_left = !__is_left;
     __m       = __n % __m;
     __n       = __last - __first;
   }
 
   return __ret;
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator>
 _ForwardIterator __pattern_rotate(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last) noexcept {
   return __internal::__brick_rotate(__first, __middle, __last, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator>
 _RandomAccessIterator __pattern_rotate(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __middle,
     _RandomAccessIterator __last) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _Tp;
   auto __n = __last - __first;
   auto __m = __middle - __first;
   if (__m <= __n / 2) {
     __par_backend::__buffer<_Tp> __buf(__n - __m);
     return __internal::__except_handler([&__exec, __n, __m, __first, __middle, __last, &__buf]() {
       _Tp* __result = __buf.get();
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __middle,
           __last,
           [__middle, __result](_RandomAccessIterator __b, _RandomAccessIterator __e) {
             __internal::__brick_uninitialized_move(__b, __e, __result + (__b - __middle), _IsVector{});
           });
 
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __first,
           __middle,
           [__last, __middle](_RandomAccessIterator __b, _RandomAccessIterator __e) {
             __internal::__brick_move(__b, __e, __b + (__last - __middle), _IsVector{});
           });
 
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __result,
           __result + (__n - __m),
           [__first, __result](_Tp* __b, _Tp* __e) {
             __brick_move_destroy()(__b, __e, __first + (__b - __result), _IsVector{});
           });
 
       return __first + (__last - __middle);
     });
   } else {
     __par_backend::__buffer<_Tp> __buf(__m);
     return __internal::__except_handler([&__exec, __n, __m, __first, __middle, __last, &__buf]() {
       _Tp* __result = __buf.get();
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __first,
           __middle,
           [__first, __result](_RandomAccessIterator __b, _RandomAccessIterator __e) {
             __internal::__brick_uninitialized_move(__b, __e, __result + (__b - __first), _IsVector{});
           });
 
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __middle,
           __last,
           [__first, __middle](_RandomAccessIterator __b, _RandomAccessIterator __e) {
             __internal::__brick_move(__b, __e, __first + (__b - __middle), _IsVector{});
           });
 
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __result,
           __result + __m,
           [__n, __m, __first, __result](_Tp* __b, _Tp* __e) {
             __brick_move_destroy()(__b, __e, __first + ((__n - __m) + (__b - __result)), _IsVector{});
           });
 
       return __first + (__last - __middle);
     });
   }
 }
 
 //------------------------------------------------------------------------
 // rotate_copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator>
 _OutputIterator __brick_rotate_copy(
     _ForwardIterator __first,
     _ForwardIterator __middle,
     _ForwardIterator __last,
     _OutputIterator __result,
     /*__is_vector=*/std::false_type) noexcept {
   return std::rotate_copy(__first, __middle, __last, __result);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2 __brick_rotate_copy(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __middle,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result,
     /*__is_vector=*/std::true_type) noexcept {
   _RandomAccessIterator2 __res = __internal::__brick_copy(__middle, __last, __result, std::true_type());
   return __internal::__brick_copy(__first, __middle, __res, std::true_type());
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator>
 _OutputIterator __pattern_rotate_copy(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __middle,
     _ForwardIterator __last,
     _OutputIterator __result) noexcept {
   return __internal::__brick_rotate_copy(__first, __middle, __last, __result, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator1, class _RandomAccessIterator2>
 _RandomAccessIterator2 __pattern_rotate_copy(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __middle,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __result) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __par_backend::__parallel_for(
       __backend_tag{},
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       [__first, __last, __middle, __result](_RandomAccessIterator1 __b, _RandomAccessIterator1 __e) {
         if (__b > __middle) {
           __internal::__brick_copy(__b, __e, __result + (__b - __middle), _IsVector{});
         } else {
           _RandomAccessIterator2 __new_result = __result + ((__last - __middle) + (__b - __first));
           if (__e < __middle) {
             __internal::__brick_copy(__b, __e, __new_result, _IsVector{});
           } else {
             __internal::__brick_copy(__b, __middle, __new_result, _IsVector{});
             __internal::__brick_copy(__middle, __e, __result, _IsVector{});
           }
         }
       });
   return __result + (__last - __first);
 }
 
 //------------------------------------------------------------------------
 // is_partitioned
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _UnaryPredicate>
 bool __brick_is_partitioned(_ForwardIterator __first,
                             _ForwardIterator __last,
                             _UnaryPredicate __pred,
                             /*is_vector=*/std::false_type) noexcept {
   return std::is_partitioned(__first, __last, __pred);
 }
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 bool __brick_is_partitioned(_RandomAccessIterator __first,
                             _RandomAccessIterator __last,
                             _UnaryPredicate __pred,
                             /*is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _SizeType;
   if (__first == __last) {
     return true;
   } else {
     _RandomAccessIterator __result = __unseq_backend::__simd_first(
         __first, _SizeType(0), __last - __first, [&__pred](_RandomAccessIterator __it, _SizeType __i) {
           return !__pred(__it[__i]);
         });
     if (__result == __last) {
       return true;
     } else {
       ++__result;
       return !__unseq_backend::__simd_or(__result, __last - __result, __pred);
     }
   }
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 bool __pattern_is_partitioned(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred) noexcept {
   return __internal::__brick_is_partitioned(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 bool __pattern_is_partitioned(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred) {
   if (__first == __last) {
     return true;
   } else {
     using __backend_tag = typename decltype(__tag)::__backend_tag;
 
     return __internal::__except_handler([&]() {
       // State of current range:
       // broken     - current range is not partitioned by pred
       // all_true   - all elements in current range satisfy pred
       // all_false  - all elements in current range don't satisfy pred
       // true_false - elements satisfy pred are placed before elements that don't satisfy pred
       enum _ReduceType { __not_init = -1, __broken, __all_true, __all_false, __true_false };
       _ReduceType __init = __not_init;
 
       // Array with states that we'll have when state from the left branch is merged with state from the right branch.
       // State is calculated by formula: new_state = table[left_state * 4 + right_state]
       _ReduceType __table[] = {
           __broken,
           __broken,
           __broken,
           __broken,
           __broken,
           __all_true,
           __true_false,
           __true_false,
           __broken,
           __broken,
           __all_false,
           __broken,
           __broken,
           __broken,
           __true_false,
           __broken};
 
       __init = __par_backend::__parallel_reduce(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __first,
           __last,
           __init,
           [&__pred,
            &__table](_RandomAccessIterator __i, _RandomAccessIterator __j, _ReduceType __value) -> _ReduceType {
             if (__value == __broken) {
               return __broken;
             }
             _ReduceType __res = __not_init;
             // if first element satisfy pred
             if (__pred(*__i)) {
               // find first element that don't satisfy pred
               _RandomAccessIterator __x = __internal::__brick_find_if(__i + 1, __j, std::not_fn(__pred), _IsVector{});
               if (__x != __j) {
                 // find first element after "x" that satisfy pred
                 _RandomAccessIterator __y = __internal::__brick_find_if(__x + 1, __j, __pred, _IsVector{});
                 // if it was found then range isn't partitioned by pred
                 if (__y != __j) {
                   return __broken;
                 } else {
                   __res = __true_false;
                 }
               } else {
                 __res = __all_true;
               }
             } else { // if first element doesn't satisfy pred
               // then we should find the first element that satisfy pred.
               // If we found it then range isn't partitioned by pred
               if (__internal::__brick_find_if(__i + 1, __j, __pred, _IsVector{}) != __j) {
                 return __broken;
               } else {
                 __res = __all_false;
               }
             }
             // if we have value from left range then we should calculate the result
             return (__value == -1) ? __res : __table[__value * 4 + __res];
           },
 
           [&__table](_ReduceType __val1, _ReduceType __val2) -> _ReduceType {
             if (__val1 == __broken || __val2 == __broken) {
               return __broken;
             }
             // calculate the result for new big range
             return __table[__val1 * 4 + __val2];
           });
       return __init != __broken;
     });
   }
 }
 
 //------------------------------------------------------------------------
 // partition
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator __brick_partition(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _UnaryPredicate __pred,
     /*is_vector=*/std::false_type) noexcept {
   return std::partition(__first, __last, __pred);
 }
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __brick_partition(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred,
     /*is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::partition(__first, __last, __pred);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator __pattern_partition(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred) noexcept {
   return __internal::__brick_partition(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __pattern_partition(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   // partitioned range: elements before pivot satisfy pred (true part),
   //                    elements after pivot don't satisfy pred (false part)
   struct _PartitionRange {
     _RandomAccessIterator __begin;
     _RandomAccessIterator __pivot;
     _RandomAccessIterator __end;
   };
 
   return __internal::__except_handler([&]() {
     _PartitionRange __init{__last, __last, __last};
 
     // lambda for merging two partitioned ranges to one partitioned range
     auto __reductor = [&__exec](_PartitionRange __val1, _PartitionRange __val2) -> _PartitionRange {
       auto __size1     = __val1.__end - __val1.__pivot;
       auto __size2     = __val2.__pivot - __val2.__begin;
       auto __new_begin = __val2.__begin - (__val1.__end - __val1.__begin);
 
       // if all elements in left range satisfy pred then we can move new pivot to pivot of right range
       if (__val1.__end == __val1.__pivot) {
         return {__new_begin, __val2.__pivot, __val2.__end};
       }
       // if true part of right range greater than false part of left range
       // then we should swap the false part of left range and last part of true part of right range
       else if (__size2 > __size1) {
         __par_backend::__parallel_for(
             __backend_tag{},
             std::forward<_ExecutionPolicy>(__exec),
             __val1.__pivot,
             __val1.__pivot + __size1,
             [__val1, __val2, __size1](_RandomAccessIterator __i, _RandomAccessIterator __j) {
               __internal::__brick_swap_ranges(
                   __i, __j, (__val2.__pivot - __size1) + (__i - __val1.__pivot), _IsVector{});
             });
         return {__new_begin, __val2.__pivot - __size1, __val2.__end};
       }
       // else we should swap the first part of false part of left range and true part of right range
       else {
         __par_backend::__parallel_for(
             __backend_tag{},
             std::forward<_ExecutionPolicy>(__exec),
             __val1.__pivot,
             __val1.__pivot + __size2,
             [__val1, __val2](_RandomAccessIterator __i, _RandomAccessIterator __j) {
               __internal::__brick_swap_ranges(__i, __j, __val2.__begin + (__i - __val1.__pivot), _IsVector{});
             });
         return {__new_begin, __val1.__pivot + __size2, __val2.__end};
       }
     };
 
     _PartitionRange __result = __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         __init,
         [__pred,
          __reductor](_RandomAccessIterator __i, _RandomAccessIterator __j, _PartitionRange __value) -> _PartitionRange {
           // 1. serial partition
           _RandomAccessIterator __pivot = __internal::__brick_partition(__i, __j, __pred, _IsVector{});
 
           // 2. merging of two ranges (left and right respectively)
           return __reductor(__value, {__i, __pivot, __j});
         },
         __reductor);
     return __result.__pivot;
   });
 }
 
 //------------------------------------------------------------------------
 // stable_partition
 //------------------------------------------------------------------------
 
 template <class _BidirectionalIterator, class _UnaryPredicate>
 _BidirectionalIterator __brick_stable_partition(
     _BidirectionalIterator __first,
     _BidirectionalIterator __last,
     _UnaryPredicate __pred,
     /*__is_vector=*/std::false_type) noexcept {
   return std::stable_partition(__first, __last, __pred);
 }
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __brick_stable_partition(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred,
     /*__is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::stable_partition(__first, __last, __pred);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator, class _UnaryPredicate>
 _BidirectionalIterator __pattern_stable_partition(
     _Tag,
     _ExecutionPolicy&&,
     _BidirectionalIterator __first,
     _BidirectionalIterator __last,
     _UnaryPredicate __pred) noexcept {
   return __internal::__brick_stable_partition(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __pattern_stable_partition(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   // partitioned range: elements before pivot satisfy pred (true part),
   //                    elements after pivot don't satisfy pred (false part)
   struct _PartitionRange {
     _RandomAccessIterator __begin;
     _RandomAccessIterator __pivot;
     _RandomAccessIterator __end;
   };
 
   return __internal::__except_handler([&]() {
     _PartitionRange __init{__last, __last, __last};
 
     // lambda for merging two partitioned ranges to one partitioned range
     auto __reductor = [](_PartitionRange __val1, _PartitionRange __val2) -> _PartitionRange {
       auto __size1     = __val1.__end - __val1.__pivot;
       auto __new_begin = __val2.__begin - (__val1.__end - __val1.__begin);
 
       // if all elements in left range satisfy pred then we can move new pivot to pivot of right range
       if (__val1.__end == __val1.__pivot) {
         return {__new_begin, __val2.__pivot, __val2.__end};
       }
       // if true part of right range greater than false part of left range
       // then we should swap the false part of left range and last part of true part of right range
       else {
         __internal::__brick_rotate(__val1.__pivot, __val2.__begin, __val2.__pivot, _IsVector{});
         return {__new_begin, __val2.__pivot - __size1, __val2.__end};
       }
     };
 
     _PartitionRange __result = __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         __init,
         [&__pred,
          __reductor](_RandomAccessIterator __i, _RandomAccessIterator __j, _PartitionRange __value) -> _PartitionRange {
           // 1. serial stable_partition
           _RandomAccessIterator __pivot = __internal::__brick_stable_partition(__i, __j, __pred, _IsVector{});
 
           // 2. merging of two ranges (left and right respectively)
           return __reductor(__value, {__i, __pivot, __j});
         },
         __reductor);
     return __result.__pivot;
   });
 }
 
 //------------------------------------------------------------------------
 // partition_copy
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator1, class _OutputIterator2, class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2> __brick_partition_copy(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator1 __out_true,
     _OutputIterator2 __out_false,
     _UnaryPredicate __pred,
     /*is_vector=*/std::false_type) noexcept {
   return std::partition_copy(__first, __last, __out_true, __out_false, __pred);
 }
 
 template <class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _UnaryPredicate>
 std::pair<_RandomAccessIterator2, _RandomAccessIterator3> __brick_partition_copy(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __out_true,
     _RandomAccessIterator3 __out_false,
     _UnaryPredicate __pred,
     /*is_vector=*/std::true_type) noexcept {
 #if defined(_PSTL_MONOTONIC_PRESENT)
   return __unseq_backend::__simd_partition_copy(__first, __last - __first, __out_true, __out_false, __pred);
 #else
   return std::partition_copy(__first, __last, __out_true, __out_false, __pred);
 #endif
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator,
           class _OutputIterator1,
           class _OutputIterator2,
           class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2> __pattern_partition_copy(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator1 __out_true,
     _OutputIterator2 __out_false,
     _UnaryPredicate __pred) noexcept {
   return __internal::__brick_partition_copy(
       __first, __last, __out_true, __out_false, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _UnaryPredicate>
 std::pair<_RandomAccessIterator2, _RandomAccessIterator3> __pattern_partition_copy(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __out_true,
     _RandomAccessIterator3 __out_false,
     _UnaryPredicate __pred) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType;
   typedef std::pair<_DifferenceType, _DifferenceType> _ReturnType;
   const _DifferenceType __n = __last - __first;
   if (_DifferenceType(1) < __n) {
     __par_backend::__buffer<bool> __mask_buf(__n);
     return __internal::__except_handler([&__exec, __n, __first, __out_true, __out_false, __pred, &__mask_buf]() {
       bool* __mask = __mask_buf.get();
       _ReturnType __m{};
       __par_backend::__parallel_strict_scan(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __n,
           std::make_pair(_DifferenceType(0), _DifferenceType(0)),
           [=](_DifferenceType __i, _DifferenceType __len) { // Reduce
             return __internal::__brick_calc_mask_1<_DifferenceType>(
                 __first + __i, __first + (__i + __len), __mask + __i, __pred, _IsVector{});
           },
           [](const _ReturnType& __x, const _ReturnType& __y) -> _ReturnType {
             return std::make_pair(__x.first + __y.first, __x.second + __y.second);
           },                                                                       // Combine
           [=](_DifferenceType __i, _DifferenceType __len, _ReturnType __initial) { // Scan
             __internal::__brick_partition_by_mask(
                 __first + __i,
                 __first + (__i + __len),
                 __out_true + __initial.first,
                 __out_false + __initial.second,
                 __mask + __i,
                 _IsVector{});
           },
           [&__m](_ReturnType __total) { __m = __total; });
       return std::make_pair(__out_true + __m.first, __out_false + __m.second);
     });
   }
   // trivial sequence - use serial algorithm
   return __internal::__brick_partition_copy(__first, __last, __out_true, __out_false, __pred, _IsVector{});
 }
 
 //------------------------------------------------------------------------
 // sort
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare, class _IsMoveConstructible>
 void __pattern_sort(_Tag,
                     _ExecutionPolicy&&,
                     _RandomAccessIterator __first,
                     _RandomAccessIterator __last,
                     _Compare __comp,
                     _IsMoveConstructible) noexcept {
   std::sort(__first, __last, __comp);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_sort(__parallel_tag<_IsVector> __tag,
                     _ExecutionPolicy&& __exec,
                     _RandomAccessIterator __first,
                     _RandomAccessIterator __last,
                     _Compare __comp,
                     /*is_move_constructible=*/std::true_type) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __internal::__except_handler([&]() {
     __par_backend::__parallel_stable_sort(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         __comp,
         [](_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
           std::sort(__first, __last, __comp);
         });
   });
 }
 
 //------------------------------------------------------------------------
 // stable_sort
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_stable_sort(
     _Tag, _ExecutionPolicy&&, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) noexcept {
   std::stable_sort(__first, __last, __comp);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_stable_sort(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __internal::__except_handler([&]() {
     __par_backend::__parallel_stable_sort(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         __comp,
         [](_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
           std::stable_sort(__first, __last, __comp);
         });
   });
 }
 
 //------------------------------------------------------------------------
 // partial_sort
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_partial_sort(
     _Tag,
     _ExecutionPolicy&&,
     _RandomAccessIterator __first,
     _RandomAccessIterator __middle,
     _RandomAccessIterator __last,
     _Compare __comp) noexcept {
   std::partial_sort(__first, __middle, __last, __comp);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_partial_sort(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __middle,
     _RandomAccessIterator __last,
     _Compare __comp) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   const auto __n = __middle - __first;
   if (__n == 0)
     return;
 
   __internal::__except_handler([&]() {
     __par_backend::__parallel_stable_sort(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         __comp,
         [__n](_RandomAccessIterator __begin, _RandomAccessIterator __end, _Compare __comp) {
           if (__n < __end - __begin)
             std::partial_sort(__begin, __begin + __n, __end, __comp);
           else
             std::sort(__begin, __end, __comp);
         },
         __n);
   });
 }
 
 //------------------------------------------------------------------------
 // partial_sort_copy
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __pattern_partial_sort_copy(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _RandomAccessIterator __d_first,
     _RandomAccessIterator __d_last,
     _Compare __comp) noexcept {
   return std::partial_sort_copy(__first, __last, __d_first, __d_last, __comp);
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 _RandomAccessIterator2 __pattern_partial_sort_copy(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __d_first,
     _RandomAccessIterator2 __d_last,
     _Compare __comp) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__last == __first || __d_last == __d_first) {
     return __d_first;
   }
   auto __n1 = __last - __first;
   auto __n2 = __d_last - __d_first;
   return __internal::__except_handler([&]() {
     if (__n2 >= __n1) {
       __par_backend::__parallel_stable_sort(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __d_first,
           __d_first + __n1,
           __comp,
           [__first, __d_first](_RandomAccessIterator2 __i, _RandomAccessIterator2 __j, _Compare __comp) {
             _RandomAccessIterator1 __i1 = __first + (__i - __d_first);
             _RandomAccessIterator1 __j1 = __first + (__j - __d_first);
 
         // 1. Copy elements from input to output
 #if !defined(_PSTL_ICC_18_OMP_SIMD_BROKEN)
             __internal::__brick_copy(__i1, __j1, __i, _IsVector{});
 #else
             std::copy(__i1, __j1, __i);
 #endif
             // 2. Sort elements in output sequence
             std::sort(__i, __j, __comp);
           },
           __n1);
       return __d_first + __n1;
     } else {
       typedef typename std::iterator_traits<_RandomAccessIterator1>::value_type _T1;
       typedef typename std::iterator_traits<_RandomAccessIterator2>::value_type _T2;
       __par_backend::__buffer<_T1> __buf(__n1);
       _T1* __r = __buf.get();
 
       __par_backend::__parallel_stable_sort(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __r,
           __r + __n1,
           __comp,
           [__n2, __first, __r](_T1* __i, _T1* __j, _Compare __comp) {
             _RandomAccessIterator1 __it = __first + (__i - __r);
 
             // 1. Copy elements from input to raw memory
             for (_T1* __k = __i; __k != __j; ++__k, ++__it) {
               ::new (__k) _T2(*__it);
             }
 
             // 2. Sort elements in temporary __buffer
             if (__n2 < __j - __i)
               std::partial_sort(__i, __i + __n2, __j, __comp);
             else
               std::sort(__i, __j, __comp);
           },
           __n2);
 
       // 3. Move elements from temporary __buffer to output
       __par_backend::__parallel_for(
           __backend_tag{},
           std::forward<_ExecutionPolicy>(__exec),
           __r,
           __r + __n2,
           [__r, __d_first](_T1* __i, _T1* __j) {
             __brick_move_destroy()(__i, __j, __d_first + (__i - __r), _IsVector{});
           });
       __par_backend::__parallel_for(
           __backend_tag{}, std::forward<_ExecutionPolicy>(__exec), __r + __n2, __r + __n1, [](_T1* __i, _T1* __j) {
             __brick_destroy(__i, __j, _IsVector{});
           });
 
       return __d_first + __n2;
     }
   });
 }
 
 //------------------------------------------------------------------------
 // adjacent_find
 //------------------------------------------------------------------------
 template <class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __brick_adjacent_find(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _BinaryPredicate __pred,
     /* IsVector = */ std::true_type,
     bool __or_semantic) noexcept {
   return __unseq_backend::__simd_adjacent_find(__first, __last, __pred, __or_semantic);
 }
 
 template <class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator __brick_adjacent_find(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _BinaryPredicate __pred,
     /* IsVector = */ std::false_type,
     bool) noexcept {
   return std::adjacent_find(__first, __last, __pred);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 _ForwardIterator __pattern_adjacent_find(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _BinaryPredicate __pred,
     bool __or_semantic) noexcept {
   return __internal::__brick_adjacent_find(__first, __last, __pred, typename _Tag::__is_vector{}, __or_semantic);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _BinaryPredicate>
 _RandomAccessIterator __pattern_adjacent_find(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _BinaryPredicate __pred,
     bool __or_semantic) {
   if (__last - __first < 2)
     return __last;
 
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         __last,
         [__last, __pred, __or_semantic](
             _RandomAccessIterator __begin, _RandomAccessIterator __end, _RandomAccessIterator __value)
             -> _RandomAccessIterator {
           // TODO: investigate performance benefits from the use of shared variable for the result,
           // checking (compare_and_swap idiom) its __value at __first.
           if (__or_semantic && __value < __last) { // found
             __par_backend::__cancel_execution();
             return __value;
           }
 
           if (__value > __begin) {
             // modify __end to check the predicate on the boundary __values;
             // TODO: to use a custom range with boundaries overlapping
             // TODO: investigate what if we remove "if" below and run algorithm on range [__first, __last-1)
             // then check the pair [__last-1, __last)
             if (__end != __last)
               ++__end;
 
             // correct the global result iterator if the "brick" returns a local "__last"
             const _RandomAccessIterator __res =
                 __internal::__brick_adjacent_find(__begin, __end, __pred, _IsVector{}, __or_semantic);
             if (__res < __end)
               __value = __res;
           }
           return __value;
         },
         [](_RandomAccessIterator __x, _RandomAccessIterator __y) -> _RandomAccessIterator {
           return __x < __y ? __x : __y;
         } // reduce a __value
     );
   });
 }
 
 //------------------------------------------------------------------------
 // nth_element
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_nth_element(
     _Tag,
     _ExecutionPolicy&&,
     _RandomAccessIterator __first,
     _RandomAccessIterator __nth,
     _RandomAccessIterator __last,
     _Compare __comp) noexcept {
   std::nth_element(__first, __nth, __last, __comp);
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_nth_element(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __nth,
     _RandomAccessIterator __last,
     _Compare __comp) noexcept {
   if (__first == __last || __nth == __last) {
     return;
   }
 
   using std::iter_swap;
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _Tp;
   _RandomAccessIterator __x;
   do {
     __x = __internal::__pattern_partition(
         __tag, std::forward<_ExecutionPolicy>(__exec), __first + 1, __last, [&__comp, __first](const _Tp& __x) {
           return __comp(__x, *__first);
         });
     --__x;
     if (__x != __first) {
       iter_swap(__first, __x);
     }
     // if x > nth then our new range for partition is [first, x)
     if (__x - __nth > 0) {
       __last = __x;
     }
     // if x < nth then our new range for partition is [x, last)
     else if (__x - __nth < 0) {
       // if *x == *nth then we can start new partition with x+1
       if (!__comp(*__nth, *__x) && !__comp(*__x, *__nth)) {
         ++__x;
       } else {
         iter_swap(__nth, __x);
       }
       __first = __x;
     }
   } while (__x != __nth);
 }
 
 //------------------------------------------------------------------------
 // fill, fill_n
 //------------------------------------------------------------------------
 template <class _RandomAccessIterator, class _Tp>
 void __brick_fill(_RandomAccessIterator __first,
                   _RandomAccessIterator __last,
                   const _Tp& __value,
                   /* __is_vector = */ std::true_type) noexcept {
   __unseq_backend::__simd_fill_n(__first, __last - __first, __value);
 }
 
 template <class _ForwardIterator, class _Tp>
 void __brick_fill(_ForwardIterator __first,
                   _ForwardIterator __last,
                   const _Tp& __value,
                   /* __is_vector = */ std::false_type) noexcept {
   std::fill(__first, __last, __value);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 void __pattern_fill(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) noexcept {
   __internal::__brick_fill(__first, __last, __value, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Tp>
 _RandomAccessIterator __pattern_fill(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     const _Tp& __value) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&__exec, __first, __last, &__value]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [&__value](_RandomAccessIterator __begin, _RandomAccessIterator __end) {
           __internal::__brick_fill(__begin, __end, __value, _IsVector{});
         });
     return __last;
   });
 }
 
 template <class _RandomAccessIterator, class _Size, class _Tp>
 _RandomAccessIterator
 __brick_fill_n(_RandomAccessIterator __first,
                _Size __count,
                const _Tp& __value,
                /* __is_vector = */ std::true_type) noexcept {
   return __unseq_backend::__simd_fill_n(__first, __count, __value);
 }
 
 template <class _OutputIterator, class _Size, class _Tp>
 _OutputIterator __brick_fill_n(
     _OutputIterator __first, _Size __count, const _Tp& __value, /* __is_vector = */ std::false_type) noexcept {
   return std::fill_n(__first, __count, __value);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _OutputIterator, class _Size, class _Tp>
 _OutputIterator
 __pattern_fill_n(_Tag, _ExecutionPolicy&&, _OutputIterator __first, _Size __count, const _Tp& __value) noexcept {
   return __internal::__brick_fill_n(__first, __count, __value, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Tp>
 _RandomAccessIterator __pattern_fill_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _Size __count,
     const _Tp& __value) {
   return __internal::__pattern_fill(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __first + __count, __value);
 }
 
 //------------------------------------------------------------------------
 // generate, generate_n
 //------------------------------------------------------------------------
 template <class _RandomAccessIterator, class _Generator>
 void __brick_generate(_RandomAccessIterator __first,
                       _RandomAccessIterator __last,
                       _Generator __g,
                       /* is_vector = */ std::true_type) noexcept {
   __unseq_backend::__simd_generate_n(__first, __last - __first, __g);
 }
 
 template <class _ForwardIterator, class _Generator>
 void __brick_generate(_ForwardIterator __first,
                       _ForwardIterator __last,
                       _Generator __g,
                       /* is_vector = */ std::false_type) noexcept {
   std::generate(__first, __last, __g);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _Generator>
 void __pattern_generate(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Generator __g) noexcept {
   __internal::__brick_generate(__first, __last, __g, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Generator>
 _RandomAccessIterator __pattern_generate(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Generator __g) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_for(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__g](_RandomAccessIterator __begin, _RandomAccessIterator __end) {
           __internal::__brick_generate(__begin, __end, __g, _IsVector{});
         });
     return __last;
   });
 }
 
 template <class _RandomAccessIterator, class Size, class _Generator>
 _RandomAccessIterator __brick_generate_n(
     _RandomAccessIterator __first,
     Size __count,
     _Generator __g,
     /* is_vector = */ std::true_type) noexcept {
   return __unseq_backend::__simd_generate_n(__first, __count, __g);
 }
 
 template <class OutputIterator, class Size, class _Generator>
 OutputIterator
 __brick_generate_n(OutputIterator __first, Size __count, _Generator __g, /* is_vector = */ std::false_type) noexcept {
   return std::generate_n(__first, __count, __g);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _OutputIterator, class _Size, class _Generator>
 _OutputIterator
 __pattern_generate_n(_Tag, _ExecutionPolicy&&, _OutputIterator __first, _Size __count, _Generator __g) noexcept {
   return __internal::__brick_generate_n(__first, __count, __g, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Size, class _Generator>
 _RandomAccessIterator __pattern_generate_n(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _Size __count,
     _Generator __g) {
   static_assert(__are_random_access_iterators<_RandomAccessIterator>::value,
                 "Pattern-brick error. Should be a random access iterator.");
   return __internal::__pattern_generate(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __first + __count, __g);
 }
 
 //------------------------------------------------------------------------
 // remove
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator __brick_remove_if(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _UnaryPredicate __pred,
     /* __is_vector = */ std::false_type) noexcept {
   return std::remove_if(__first, __last, __pred);
 }
 
 template <class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __brick_remove_if(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred,
     /* __is_vector = */ std::true_type) noexcept {
 #if defined(_PSTL_MONOTONIC_PRESENT)
   return __unseq_backend::__simd_remove_if(__first, __last - __first, __pred);
 #else
   return std::remove_if(__first, __last, __pred);
 #endif
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 _ForwardIterator __pattern_remove_if(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred) noexcept {
   return __internal::__brick_remove_if(__first, __last, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryPredicate>
 _RandomAccessIterator __pattern_remove_if(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _UnaryPredicate __pred) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::reference _ReferenceType;
 
   if (__first == __last || __first + 1 == __last) {
     // Trivial sequence - use serial algorithm
     return __internal::__brick_remove_if(__first, __last, __pred, _IsVector{});
   }
 
   return __internal::__remove_elements(
       __tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       [&__pred](bool* __b, bool* __e, _RandomAccessIterator __it) {
         __internal::__brick_walk2(
             __b, __e, __it, [&__pred](bool& __x, _ReferenceType __y) { __x = !__pred(__y); }, _IsVector{});
       });
 }
 
 //------------------------------------------------------------------------
 // merge
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_merge(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __d_first,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   return std::merge(__first1, __last1, __first2, __last2, __d_first, __comp);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Compare>
 _RandomAccessIterator3 __brick_merge(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __d_first,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::merge(__first1, __last1, __first2, __last2, __d_first, __comp);
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_merge(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __d_first,
     _Compare __comp) noexcept {
   return __internal::__brick_merge(
       __first1, __last1, __first2, __last2, __d_first, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _Compare>
 _RandomAccessIterator3 __pattern_merge(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __d_first,
     _Compare __comp) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   __par_backend::__parallel_merge(
       __backend_tag{},
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __last2,
       __d_first,
       __comp,
       [](_RandomAccessIterator1 __f1,
          _RandomAccessIterator1 __l1,
          _RandomAccessIterator2 __f2,
          _RandomAccessIterator2 __l2,
          _RandomAccessIterator3 __f3,
          _Compare __comp) { return __internal::__brick_merge(__f1, __l1, __f2, __l2, __f3, __comp, _IsVector{}); });
   return __d_first + (__last1 - __first1) + (__last2 - __first2);
 }
 
 //------------------------------------------------------------------------
 // inplace_merge
 //------------------------------------------------------------------------
 template <class _BidirectionalIterator, class _Compare>
 void __brick_inplace_merge(
     _BidirectionalIterator __first,
     _BidirectionalIterator __middle,
     _BidirectionalIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   std::inplace_merge(__first, __middle, __last, __comp);
 }
 
 template <class _RandomAccessIterator, class _Compare>
 void __brick_inplace_merge(
     _RandomAccessIterator __first,
     _RandomAccessIterator __middle,
     _RandomAccessIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial")
   std::inplace_merge(__first, __middle, __last, __comp);
 }
 
 template <class _Tag, class _ExecutionPolicy, class _BidirectionalIterator, class _Compare>
 void __pattern_inplace_merge(
     _Tag,
     _ExecutionPolicy&&,
     _BidirectionalIterator __first,
     _BidirectionalIterator __middle,
     _BidirectionalIterator __last,
     _Compare __comp) noexcept {
   __internal::__brick_inplace_merge(__first, __middle, __last, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 void __pattern_inplace_merge(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __middle,
     _RandomAccessIterator __last,
     _Compare __comp) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__first == __last || __first == __middle || __middle == __last) {
     return;
   }
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _Tp;
   auto __n = __last - __first;
   __par_backend::__buffer<_Tp> __buf(__n);
   _Tp* __r = __buf.get();
   __internal::__except_handler([&]() {
     auto __move_values = [](_RandomAccessIterator __x, _Tp* __z) {
       __internal::__invoke_if_else(
           std::is_trivial<_Tp>(),
           [&]() { *__z = std::move(*__x); },
           [&]() { ::new (std::addressof(*__z)) _Tp(std::move(*__x)); });
     };
 
     auto __move_sequences = [](_RandomAccessIterator __first1, _RandomAccessIterator __last1, _Tp* __first2) {
       return __internal::__brick_uninitialized_move(__first1, __last1, __first2, _IsVector());
     };
 
     __par_backend::__parallel_merge(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __middle,
         __middle,
         __last,
         __r,
         __comp,
         [__n, __move_values, __move_sequences](
             _RandomAccessIterator __f1,
             _RandomAccessIterator __l1,
             _RandomAccessIterator __f2,
             _RandomAccessIterator __l2,
             _Tp* __f3,
             _Compare __comp) {
           (__utils::__serial_move_merge(__n))(
               __f1, __l1, __f2, __l2, __f3, __comp, __move_values, __move_values, __move_sequences, __move_sequences);
           return __f3 + (__l1 - __f1) + (__l2 - __f2);
         });
     __par_backend::__parallel_for(
         __backend_tag{}, std::forward<_ExecutionPolicy>(__exec), __r, __r + __n, [__r, __first](_Tp* __i, _Tp* __j) {
           __brick_move_destroy()(__i, __j, __first + (__i - __r), _IsVector{});
         });
   });
 }
 
 //------------------------------------------------------------------------
 // includes
 //------------------------------------------------------------------------
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 bool __pattern_includes(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Compare __comp) noexcept {
   return std::includes(__first1, __last1, __first2, __last2, __comp);
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 bool __pattern_includes(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _Compare __comp) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__first2 >= __last2)
     return true;
 
   if (__first1 >= __last1 || __comp(*__first2, *__first1) || __comp(*(__last1 - 1), *(__last2 - 1)))
     return false;
 
   __first1 = std::lower_bound(__first1, __last1, *__first2, __comp);
   if (__first1 == __last1)
     return false;
 
   if (__last2 - __first2 == 1)
     return !__comp(*__first1, *__first2) && !__comp(*__first2, *__first1);
 
   return __internal::__except_handler([&]() {
     return !__internal::__parallel_or(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first2,
         __last2,
         [__first1, __last1, __first2, __last2, &__comp](_RandomAccessIterator2 __i, _RandomAccessIterator2 __j) {
           _PSTL_ASSERT(__j > __i);
           //_PSTL_ASSERT(__j - __i > 1);
 
           // 1. moving boundaries to "consume" subsequence of equal elements
           auto __is_equal = [&__comp](_RandomAccessIterator2 __a, _RandomAccessIterator2 __b) -> bool {
             return !__comp(*__a, *__b) && !__comp(*__b, *__a);
           };
 
           // 1.1 left bound, case "aaa[aaaxyz...]" - searching "x"
           if (__i > __first2 && __is_equal(__i, __i - 1)) {
             // whole subrange continues to content equal elements - return "no op"
             if (__is_equal(__i, __j - 1))
               return false;
 
             __i = std::upper_bound(__i, __last2, *__i, __comp);
           }
 
           // 1.2 right bound, case "[...aaa]aaaxyz" - searching "x"
           if (__j < __last2 && __is_equal(__j - 1, __j))
             __j = std::upper_bound(__j, __last2, *__j, __comp);
 
           // 2. testing is __a subsequence of the second range included into the first range
           auto __b = std::lower_bound(__first1, __last1, *__i, __comp);
 
           _PSTL_ASSERT(!__comp(*(__last1 - 1), *__b));
           _PSTL_ASSERT(!__comp(*(__j - 1), *__i));
           return !std::includes(__b, __last1, __i, __j, __comp);
         });
   });
 }
 
 constexpr auto __set_algo_cut_off = 1000;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare,
           class _SizeFunction,
           class _SetOP>
 _OutputIterator __parallel_set_op(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     _SizeFunction __size_func,
     _SetOP __set_op) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_ForwardIterator1>::difference_type _DifferenceType;
   typedef typename std::iterator_traits<_OutputIterator>::value_type _Tp;
 
   struct _SetRange {
     _DifferenceType __pos, __len, __buf_pos;
     bool empty() const { return __len == 0; }
   };
 
   const _DifferenceType __n1 = __last1 - __first1;
   const _DifferenceType __n2 = __last2 - __first2;
 
   __par_backend::__buffer<_Tp> __buf(__size_func(__n1, __n2));
 
   return __internal::__except_handler(
       [&__exec, __n1, __first1, __last1, __first2, __last2, __result, __comp, __size_func, __set_op, &__buf]() {
         auto __buffer = __buf.get();
         _DifferenceType __m{};
         auto __scan = [=](_DifferenceType, _DifferenceType, const _SetRange& __s) { // Scan
           if (!__s.empty())
             __brick_move_destroy()(
                 __buffer + __s.__buf_pos, __buffer + (__s.__buf_pos + __s.__len), __result + __s.__pos, _IsVector{});
         };
         __par_backend::__parallel_strict_scan(
             __backend_tag{},
             std::forward<_ExecutionPolicy>(__exec),
             __n1,
             _SetRange{0, 0, 0},                               //-1, 0},
             [=](_DifferenceType __i, _DifferenceType __len) { // Reduce
               //[__b; __e) - a subrange of the first sequence, to reduce
               _ForwardIterator1 __b = __first1 + __i, __e = __first1 + (__i + __len);
 
               // try searching for the first element which not equal to *__b
               if (__b != __first1)
                 __b = std::upper_bound(__b, __last1, *__b, __comp);
 
               // try searching for the first element which not equal to *__e
               if (__e != __last1)
                 __e = std::upper_bound(__e, __last1, *__e, __comp);
 
               // check is [__b; __e) empty
               if (__e - __b < 1) {
                 _ForwardIterator2 __bb = __last2;
                 if (__b != __last1)
                   __bb = std::lower_bound(__first2, __last2, *__b, __comp);
 
                 const _DifferenceType __buf_pos = __size_func((__b - __first1), (__bb - __first2));
                 return _SetRange{0, 0, __buf_pos};
               }
 
               // try searching for "corresponding" subrange [__bb; __ee) in the second sequence
               _ForwardIterator2 __bb = __first2;
               if (__b != __first1)
                 __bb = std::lower_bound(__first2, __last2, *__b, __comp);
 
               _ForwardIterator2 __ee = __last2;
               if (__e != __last1)
                 __ee = std::lower_bound(__bb, __last2, *__e, __comp);
 
               const _DifferenceType __buf_pos = __size_func((__b - __first1), (__bb - __first2));
               auto __buffer_b                 = __buffer + __buf_pos;
               auto __res                      = __set_op(__b, __e, __bb, __ee, __buffer_b, __comp);
 
               return _SetRange{0, __res - __buffer_b, __buf_pos};
             },
             [](const _SetRange& __a, const _SetRange& __b) { // Combine
               if (__b.__buf_pos > __a.__buf_pos || ((__b.__buf_pos == __a.__buf_pos) && !__b.empty()))
                 return _SetRange{__a.__pos + __a.__len + __b.__pos, __b.__len, __b.__buf_pos};
               return _SetRange{__b.__pos + __b.__len + __a.__pos, __a.__len, __a.__buf_pos};
             },
             __scan,                                     // Scan
             [&__m, &__scan](const _SetRange& __total) { // Apex
               // final scan
               __scan(0, 0, __total);
               __m = __total.__pos + __total.__len;
             });
         return __result + __m;
       });
 }
 
 // a shared parallel pattern for '__pattern_set_union' and '__pattern_set_symmetric_difference'
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare,
           class _SetUnionOp>
 _OutputIterator __parallel_set_union_op(
     _Tag __tag,
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     _SetUnionOp __set_union_op) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_ForwardIterator1>::difference_type _DifferenceType;
 
   const auto __n1 = __last1 - __first1;
   const auto __n2 = __last2 - __first2;
 
   auto copy_range1 = [](_ForwardIterator1 __begin, _ForwardIterator1 __end, _OutputIterator __res) {
     return __internal::__brick_copy(__begin, __end, __res, typename _Tag::__is_vector{});
   };
   auto copy_range2 = [](_ForwardIterator2 __begin, _ForwardIterator2 __end, _OutputIterator __res) {
     return __internal::__brick_copy(__begin, __end, __res, typename _Tag::__is_vector{});
   };
 
   // {1} {}: parallel copying just first sequence
   if (__n2 == 0)
     return __internal::__pattern_walk2_brick(
         __tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __result, copy_range1);
 
   // {} {2}: parallel copying justmake  second sequence
   if (__n1 == 0)
     return __internal::__pattern_walk2_brick(
         __tag, std::forward<_ExecutionPolicy>(__exec), __first2, __last2, __result, copy_range2);
 
   // testing  whether the sequences are intersected
   _ForwardIterator1 __left_bound_seq_1 = std::lower_bound(__first1, __last1, *__first2, __comp);
 
   if (__left_bound_seq_1 == __last1) {
     //{1} < {2}: seq2 is wholly greater than seq1, so, do parallel copying seq1 and seq2
     __par_backend::__parallel_invoke(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         [=] {
           __internal::__pattern_walk2_brick(
               __tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __result, copy_range1);
         },
         [=] {
           __internal::__pattern_walk2_brick(
               __tag, std::forward<_ExecutionPolicy>(__exec), __first2, __last2, __result + __n1, copy_range2);
         });
     return __result + __n1 + __n2;
   }
 
   // testing  whether the sequences are intersected
   _ForwardIterator2 __left_bound_seq_2 = std::lower_bound(__first2, __last2, *__first1, __comp);
 
   if (__left_bound_seq_2 == __last2) {
     //{2} < {1}: seq2 is wholly greater than seq1, so, do parallel copying seq1 and seq2
     __par_backend::__parallel_invoke(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         [=] {
           __internal::__pattern_walk2_brick(
               __tag, std::forward<_ExecutionPolicy>(__exec), __first2, __last2, __result, copy_range2);
         },
         [=] {
           __internal::__pattern_walk2_brick(
               __tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __result + __n2, copy_range1);
         });
     return __result + __n1 + __n2;
   }
 
   const auto __m1 = __left_bound_seq_1 - __first1;
   if (__m1 > __set_algo_cut_off) {
     auto __res_or = __result;
     __result += __m1; // we know proper offset due to [first1; left_bound_seq_1) < [first2; last2)
     __par_backend::__parallel_invoke(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         // do parallel copying of [first1; left_bound_seq_1)
         [=] {
           __internal::__pattern_walk2_brick(
               __tag, std::forward<_ExecutionPolicy>(__exec), __first1, __left_bound_seq_1, __res_or, copy_range1);
         },
         [=, &__result] {
           __result = __internal::__parallel_set_op(
               __tag,
               std::forward<_ExecutionPolicy>(__exec),
               __left_bound_seq_1,
               __last1,
               __first2,
               __last2,
               __result,
               __comp,
               [](_DifferenceType __n, _DifferenceType __m) { return __n + __m; },
               __set_union_op);
         });
     return __result;
   }
 
   const auto __m2 = __left_bound_seq_2 - __first2;
   _PSTL_ASSERT(__m1 == 0 || __m2 == 0);
   if (__m2 > __set_algo_cut_off) {
     auto __res_or = __result;
     __result += __m2; // we know proper offset due to [first2; left_bound_seq_2) < [first1; last1)
     __par_backend::__parallel_invoke(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         // do parallel copying of [first2; left_bound_seq_2)
         [=] {
           __internal::__pattern_walk2_brick(
               __tag, std::forward<_ExecutionPolicy>(__exec), __first2, __left_bound_seq_2, __res_or, copy_range2);
         },
         [=, &__result] {
           __result = __internal::__parallel_set_op(
               __tag,
               std::forward<_ExecutionPolicy>(__exec),
               __first1,
               __last1,
               __left_bound_seq_2,
               __last2,
               __result,
               __comp,
               [](_DifferenceType __n, _DifferenceType __m) { return __n + __m; },
               __set_union_op);
         });
     return __result;
   }
 
   return __internal::__parallel_set_op(
       __tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __last2,
       __result,
       __comp,
       [](_DifferenceType __n, _DifferenceType __m) { return __n + __m; },
       __set_union_op);
 }
 
 //------------------------------------------------------------------------
 // set_union
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_union(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     /*__is_vector=*/std::false_type) noexcept {
   return std::set_union(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <typename _IsVector>
 struct __BrickCopyConstruct {
   template <typename _ForwardIterator, typename _OutputIterator>
   _OutputIterator operator()(_ForwardIterator __first, _ForwardIterator __last, _OutputIterator __result) {
     return __brick_uninitialized_copy(__first, __last, __result, _IsVector());
   }
 };
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_union(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     /*__is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::set_union(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_union(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp) noexcept {
   return __internal::__brick_set_union(
       __first1, __last1, __first2, __last2, __result, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_union(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp) {
   const auto __n1 = __last1 - __first1;
   const auto __n2 = __last2 - __first2;
 
   // use serial algorithm
   if (__n1 + __n2 <= __set_algo_cut_off)
     return std::set_union(__first1, __last1, __first2, __last2, __result, __comp);
 
   typedef typename std::iterator_traits<_OutputIterator>::value_type _Tp;
   return __parallel_set_union_op(
       __tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __last2,
       __result,
       __comp,
       [](_RandomAccessIterator1 __first1,
          _RandomAccessIterator1 __last1,
          _RandomAccessIterator2 __first2,
          _RandomAccessIterator2 __last2,
          _Tp* __result,
          _Compare __comp) {
         return __pstl::__utils::__set_union_construct(
             __first1, __last1, __first2, __last2, __result, __comp, __BrickCopyConstruct<_IsVector>());
       });
 }
 
 //------------------------------------------------------------------------
 // set_intersection
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_intersection(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     /*__is_vector=*/std::false_type) noexcept {
   return std::set_intersection(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Compare>
 _RandomAccessIterator3 __brick_set_intersection(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __result,
     _Compare __comp,
     /*__is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::set_intersection(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_intersection(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp) noexcept {
   return __internal::__brick_set_intersection(
       __first1, __last1, __first2, __last2, __result, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _Compare>
 _RandomAccessIterator3 __pattern_set_intersection(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __result,
     _Compare __comp) {
   typedef typename std::iterator_traits<_RandomAccessIterator3>::value_type _Tp;
   typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType;
 
   const auto __n1 = __last1 - __first1;
   const auto __n2 = __last2 - __first2;
 
   // intersection is empty
   if (__n1 == 0 || __n2 == 0)
     return __result;
 
   // testing  whether the sequences are intersected
   _RandomAccessIterator1 __left_bound_seq_1 = std::lower_bound(__first1, __last1, *__first2, __comp);
   //{1} < {2}: seq 2 is wholly greater than seq 1, so, the intersection is empty
   if (__left_bound_seq_1 == __last1)
     return __result;
 
   // testing  whether the sequences are intersected
   _RandomAccessIterator2 __left_bound_seq_2 = std::lower_bound(__first2, __last2, *__first1, __comp);
   //{2} < {1}: seq 1 is wholly greater than seq 2, so, the intersection is empty
   if (__left_bound_seq_2 == __last2)
     return __result;
 
   const auto __m1 = __last1 - __left_bound_seq_1 + __n2;
   if (__m1 > __set_algo_cut_off) {
     // we know proper offset due to [first1; left_bound_seq_1) < [first2; last2)
     return __internal::__parallel_set_op(
         __tag,
         std::forward<_ExecutionPolicy>(__exec),
         __left_bound_seq_1,
         __last1,
         __first2,
         __last2,
         __result,
         __comp,
         [](_DifferenceType __n, _DifferenceType __m) { return std::min(__n, __m); },
         [](_RandomAccessIterator1 __first1,
            _RandomAccessIterator1 __last1,
            _RandomAccessIterator2 __first2,
            _RandomAccessIterator2 __last2,
            _Tp* __result,
            _Compare __comp) {
           return __pstl::__utils::__set_intersection_construct(__first1, __last1, __first2, __last2, __result, __comp);
         });
   }
 
   const auto __m2 = __last2 - __left_bound_seq_2 + __n1;
   if (__m2 > __set_algo_cut_off) {
     // we know proper offset due to [first2; left_bound_seq_2) < [first1; last1)
     __result = __internal::__parallel_set_op(
         __tag,
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         __left_bound_seq_2,
         __last2,
         __result,
         __comp,
         [](_DifferenceType __n, _DifferenceType __m) { return std::min(__n, __m); },
         [](_RandomAccessIterator1 __first1,
            _RandomAccessIterator1 __last1,
            _RandomAccessIterator2 __first2,
            _RandomAccessIterator2 __last2,
            _Tp* __result,
            _Compare __comp) {
           return __pstl::__utils::__set_intersection_construct(__first2, __last2, __first1, __last1, __result, __comp);
         });
     return __result;
   }
 
   // [left_bound_seq_1; last1) and [left_bound_seq_2; last2) - use serial algorithm
   return std::set_intersection(__left_bound_seq_1, __last1, __left_bound_seq_2, __last2, __result, __comp);
 }
 
 //------------------------------------------------------------------------
 // set_difference
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_difference(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     /*__is_vector=*/std::false_type) noexcept {
   return std::set_difference(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Compare>
 _RandomAccessIterator3 __brick_set_difference(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __result,
     _Compare __comp,
     /*__is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::set_difference(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_difference(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp) noexcept {
   return __internal::__brick_set_difference(
       __first1, __last1, __first2, __last2, __result, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _Compare>
 _RandomAccessIterator3 __pattern_set_difference(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __result,
     _Compare __comp) {
   typedef typename std::iterator_traits<_RandomAccessIterator3>::value_type _Tp;
   typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType;
 
   const auto __n1 = __last1 - __first1;
   const auto __n2 = __last2 - __first2;
 
   // {} \ {2}: the difference is empty
   if (__n1 == 0)
     return __result;
 
   // {1} \ {}: parallel copying just first sequence
   if (__n2 == 0)
     return __internal::__pattern_walk2_brick(
         __tag,
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         __result,
         [](_RandomAccessIterator1 __begin, _RandomAccessIterator1 __end, _RandomAccessIterator3 __res) {
           return __internal::__brick_copy(__begin, __end, __res, _IsVector{});
         });
 
   // testing  whether the sequences are intersected
   _RandomAccessIterator1 __left_bound_seq_1 = std::lower_bound(__first1, __last1, *__first2, __comp);
   //{1} < {2}: seq 2 is wholly greater than seq 1, so, parallel copying just first sequence
   if (__left_bound_seq_1 == __last1)
     return __internal::__pattern_walk2_brick(
         __tag,
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         __result,
         [](_RandomAccessIterator1 __begin, _RandomAccessIterator1 __end, _RandomAccessIterator3 __res) {
           return __internal::__brick_copy(__begin, __end, __res, _IsVector{});
         });
 
   // testing  whether the sequences are intersected
   _RandomAccessIterator2 __left_bound_seq_2 = std::lower_bound(__first2, __last2, *__first1, __comp);
   //{2} < {1}: seq 1 is wholly greater than seq 2, so, parallel copying just first sequence
   if (__left_bound_seq_2 == __last2)
     return __internal::__pattern_walk2_brick(
         __tag,
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         __result,
         [](_RandomAccessIterator1 __begin, _RandomAccessIterator1 __end, _RandomAccessIterator3 __res) {
           return __internal::__brick_copy(__begin, __end, __res, _IsVector{});
         });
 
   if (__n1 + __n2 > __set_algo_cut_off)
     return __parallel_set_op(
         __tag,
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         __first2,
         __last2,
         __result,
         __comp,
         [](_DifferenceType __n, _DifferenceType) { return __n; },
         [](_RandomAccessIterator1 __first1,
            _RandomAccessIterator1 __last1,
            _RandomAccessIterator2 __first2,
            _RandomAccessIterator2 __last2,
            _Tp* __result,
            _Compare __comp) {
           return __pstl::__utils::__set_difference_construct(
               __first1, __last1, __first2, __last2, __result, __comp, __BrickCopyConstruct<_IsVector>());
         });
 
   // use serial algorithm
   return std::set_difference(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 //------------------------------------------------------------------------
 // set_symmetric_difference
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _OutputIterator, class _Compare>
 _OutputIterator __brick_set_symmetric_difference(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp,
     /*__is_vector=*/std::false_type) noexcept {
   return std::set_symmetric_difference(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Compare>
 _RandomAccessIterator3 __brick_set_symmetric_difference(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __result,
     _Compare __comp,
     /*__is_vector=*/std::true_type) noexcept {
   _PSTL_PRAGMA_MESSAGE("Vectorized algorithm unimplemented, redirected to serial");
   return std::set_symmetric_difference(__first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _OutputIterator,
           class _Compare>
 _OutputIterator __pattern_set_symmetric_difference(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _OutputIterator __result,
     _Compare __comp) noexcept {
   return __internal::__brick_set_symmetric_difference(
       __first1, __last1, __first2, __last2, __result, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _RandomAccessIterator3,
           class _Compare>
 _RandomAccessIterator3 __pattern_set_symmetric_difference(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _RandomAccessIterator3 __result,
     _Compare __comp) {
   const auto __n1 = __last1 - __first1;
   const auto __n2 = __last2 - __first2;
 
   // use serial algorithm
   if (__n1 + __n2 <= __set_algo_cut_off)
     return std::set_symmetric_difference(__first1, __last1, __first2, __last2, __result, __comp);
 
   typedef typename std::iterator_traits<_RandomAccessIterator3>::value_type _Tp;
   return __internal::__parallel_set_union_op(
       __tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __last2,
       __result,
       __comp,
       [](_RandomAccessIterator1 __first1,
          _RandomAccessIterator1 __last1,
          _RandomAccessIterator2 __first2,
          _RandomAccessIterator2 __last2,
          _Tp* __result,
          _Compare __comp) {
         return __pstl::__utils::__set_symmetric_difference_construct(
             __first1, __last1, __first2, __last2, __result, __comp, __BrickCopyConstruct<_IsVector>());
       });
 }
 
 //------------------------------------------------------------------------
 // is_heap_until
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __brick_is_heap_until(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   return std::is_heap_until(__first, __last, __comp);
 }
 
 template <class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __brick_is_heap_until(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
   if (__last - __first < 2)
     return __last;
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _SizeType;
   return __unseq_backend::__simd_first(
       __first, _SizeType(0), __last - __first, [&__comp](_RandomAccessIterator __it, _SizeType __i) {
         return __comp(__it[(__i - 1) / 2], __it[__i]);
       });
 }
 
 template <class _Tag, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __pattern_is_heap_until(
     _Tag, _ExecutionPolicy&&, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) noexcept {
   return __internal::__brick_is_heap_until(__first, __last, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _RandomAccessIterator, class _DifferenceType, class _Compare>
 _RandomAccessIterator __is_heap_until_local(
     _RandomAccessIterator __first,
     _DifferenceType __begin,
     _DifferenceType __end,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   _DifferenceType __i = __begin;
   for (; __i < __end; ++__i) {
     if (__comp(__first[(__i - 1) / 2], __first[__i])) {
       break;
     }
   }
   return __first + __i;
 }
 
 template <class _RandomAccessIterator, class _DifferenceType, class _Compare>
 _RandomAccessIterator __is_heap_until_local(
     _RandomAccessIterator __first,
     _DifferenceType __begin,
     _DifferenceType __end,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
   return __unseq_backend::__simd_first(
       __first, __begin, __end, [&__comp](_RandomAccessIterator __it, _DifferenceType __i) {
         return __comp(__it[(__i - 1) / 2], __it[__i]);
       });
 }
 
 template <class _IsVector, class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 _RandomAccessIterator __pattern_is_heap_until(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__last - __first < 2)
     return __last;
 
   return __internal::__except_handler([&]() {
     return __parallel_find(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__first, __comp](_RandomAccessIterator __i, _RandomAccessIterator __j) {
           return __internal::__is_heap_until_local(__first, __i - __first, __j - __first, __comp, _IsVector{});
         },
         std::less<typename std::iterator_traits<_RandomAccessIterator>::difference_type>(),
         /*is_first=*/true);
   });
 }
 
 //------------------------------------------------------------------------
 // min_element
 //------------------------------------------------------------------------
 
 template <typename _ForwardIterator, typename _Compare>
 _ForwardIterator __brick_min_element(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   return std::min_element(__first, __last, __comp);
 }
 
 template <typename _RandomAccessIterator, typename _Compare>
 _RandomAccessIterator __brick_min_element(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
 #if defined(_PSTL_UDR_PRESENT)
   return __unseq_backend::__simd_min_element(__first, __last - __first, __comp);
 #else
   return std::min_element(__first, __last, __comp);
 #endif
 }
 
 template <typename _Tag, typename _ExecutionPolicy, typename _ForwardIterator, typename _Compare>
 _ForwardIterator __pattern_min_element(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) noexcept {
   return __internal::__brick_min_element(__first, __last, __comp, typename _Tag::__is_vector{});
 }
 
 template <typename _IsVector, typename _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare>
 _RandomAccessIterator __pattern_min_element(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp) {
   if (__first == __last)
     return __last;
 
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first + 1,
         __last,
         __first,
         [=](_RandomAccessIterator __begin, _RandomAccessIterator __end, _RandomAccessIterator __init)
             -> _RandomAccessIterator {
           const _RandomAccessIterator subresult = __internal::__brick_min_element(__begin, __end, __comp, _IsVector{});
           return __internal::__cmp_iterators_by_values(__init, subresult, __comp);
         },
         [=](_RandomAccessIterator __it1, _RandomAccessIterator __it2) -> _RandomAccessIterator {
           return __internal::__cmp_iterators_by_values(__it1, __it2, __comp);
         });
   });
 }
 
 //------------------------------------------------------------------------
 // minmax_element
 //------------------------------------------------------------------------
 
 template <typename _ForwardIterator, typename _Compare>
 std::pair<_ForwardIterator, _ForwardIterator> __brick_minmax_element(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   return std::minmax_element(__first, __last, __comp);
 }
 
 template <typename _RandomAccessIterator, typename _Compare>
 std::pair<_RandomAccessIterator, _RandomAccessIterator> __brick_minmax_element(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
 #if defined(_PSTL_UDR_PRESENT)
   return __unseq_backend::__simd_minmax_element(__first, __last - __first, __comp);
 #else
   return std::minmax_element(__first, __last, __comp);
 #endif
 }
 
 template <typename _Tag, typename _ExecutionPolicy, typename _ForwardIterator, typename _Compare>
 std::pair<_ForwardIterator, _ForwardIterator> __pattern_minmax_element(
     _Tag, _ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) noexcept {
   return __internal::__brick_minmax_element(__first, __last, __comp, typename _Tag::__is_vector{});
 }
 
 template <typename _IsVector, typename _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare>
 std::pair<_RandomAccessIterator, _RandomAccessIterator> __pattern_minmax_element(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Compare __comp) {
   if (__first == __last)
     return std::make_pair(__first, __first);
 
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     typedef std::pair<_RandomAccessIterator, _RandomAccessIterator> _Result;
 
     return __par_backend::__parallel_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first + 1,
         __last,
         std::make_pair(__first, __first),
         [=](_RandomAccessIterator __begin, _RandomAccessIterator __end, _Result __init) -> _Result {
           const _Result __subresult = __internal::__brick_minmax_element(__begin, __end, __comp, _IsVector{});
           return std::make_pair(
               __internal::__cmp_iterators_by_values(__subresult.first, __init.first, __comp),
               __internal::__cmp_iterators_by_values(__init.second, __subresult.second, std::not_fn(__comp)));
         },
         [=](_Result __p1, _Result __p2) -> _Result {
           return std::make_pair(__internal::__cmp_iterators_by_values(__p1.first, __p2.first, __comp),
                                 __internal::__cmp_iterators_by_values(__p2.second, __p1.second, std::not_fn(__comp)));
         });
   });
 }
 
 //------------------------------------------------------------------------
 // mismatch
 //------------------------------------------------------------------------
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 std::pair<_ForwardIterator1, _ForwardIterator2> __mismatch_serial(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _BinaryPredicate __pred) {
 #if defined(_PSTL_CPP14_2RANGE_MISMATCH_EQUAL_PRESENT)
   return std::mismatch(__first1, __last1, __first2, __last2, __pred);
 #else
   for (; __first1 != __last1 && __first2 != __last2 && __pred(*__first1, *__first2); ++__first1, ++__first2) {
   }
   return std::make_pair(__first1, __first2);
 #endif
 }
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 std::pair<_ForwardIterator1, _ForwardIterator2> __brick_mismatch(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Predicate __pred,
     /* __is_vector = */ std::false_type) noexcept {
   return __mismatch_serial(__first1, __last1, __first2, __last2, __pred);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Predicate>
 std::pair<_RandomAccessIterator1, _RandomAccessIterator2> __brick_mismatch(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _Predicate __pred,
     /* __is_vector = */ std::true_type) noexcept {
   auto __n = std::min(__last1 - __first1, __last2 - __first2);
   return __unseq_backend::__simd_first(__first1, __n, __first2, std::not_fn(__pred));
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 std::pair<_ForwardIterator1, _ForwardIterator2> __pattern_mismatch(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Predicate __pred) noexcept {
   return __internal::__brick_mismatch(__first1, __last1, __first2, __last2, __pred, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Predicate>
 std::pair<_RandomAccessIterator1, _RandomAccessIterator2> __pattern_mismatch(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _Predicate __pred) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     auto __n      = std::min(__last1 - __first1, __last2 - __first2);
     auto __result = __internal::__parallel_find(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __first1 + __n,
         [__first1, __first2, __pred](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           return __internal::__brick_mismatch(
                      __i, __j, __first2 + (__i - __first1), __first2 + (__j - __first1), __pred, _IsVector{})
               .first;
         },
         std::less<typename std::iterator_traits<_RandomAccessIterator1>::difference_type>(),
         /*is_first=*/true);
     return std::make_pair(__result, __first2 + (__result - __first1));
   });
 }
 
 //------------------------------------------------------------------------
 // lexicographical_compare
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 bool __brick_lexicographical_compare(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Compare __comp,
     /* __is_vector = */ std::false_type) noexcept {
   return std::lexicographical_compare(__first1, __last1, __first2, __last2, __comp);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _Compare>
 bool __brick_lexicographical_compare(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _Compare __comp,
     /* __is_vector = */ std::true_type) noexcept {
   if (__first2 == __last2) {        // if second sequence is empty
     return false;
   } else if (__first1 == __last1) { // if first sequence is empty
     return true;
   } else {
     typedef typename std::iterator_traits<_RandomAccessIterator1>::reference ref_type1;
     typedef typename std::iterator_traits<_RandomAccessIterator2>::reference ref_type2;
     --__last1;
     --__last2;
     auto __n = std::min(__last1 - __first1, __last2 - __first2);
     std::pair<_RandomAccessIterator1, _RandomAccessIterator2> __result = __unseq_backend::__simd_first(
         __first1, __n, __first2, [__comp](const ref_type1 __x, const ref_type2 __y) mutable {
           return __comp(__x, __y) || __comp(__y, __x);
         });
 
     if (__result.first == __last1 && __result.second != __last2) { // if first sequence shorter than second
       return !__comp(*__result.second, *__result.first);
     } else { // if second sequence shorter than first or both have the same number of elements
       return __comp(*__result.first, *__result.second);
     }
   }
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 bool __pattern_lexicographical_compare(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Compare __comp) noexcept {
   return __internal::__brick_lexicographical_compare(
       __first1, __last1, __first2, __last2, __comp, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Compare>
 bool __pattern_lexicographical_compare(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _RandomAccessIterator2 __last2,
     _Compare __comp) noexcept {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   if (__first2 == __last2) {        // if second sequence is empty
     return false;
   } else if (__first1 == __last1) { // if first sequence is empty
     return true;
   } else {
     typedef typename std::iterator_traits<_RandomAccessIterator1>::reference _RefType1;
     typedef typename std::iterator_traits<_RandomAccessIterator2>::reference _RefType2;
     --__last1;
     --__last2;
     auto __n      = std::min(__last1 - __first1, __last2 - __first2);
     auto __result = __internal::__parallel_find(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __first1 + __n,
         [__first1, __first2, &__comp](_RandomAccessIterator1 __i, _RandomAccessIterator1 __j) {
           return __internal::__brick_mismatch(
                      __i,
                      __j,
                      __first2 + (__i - __first1),
                      __first2 + (__j - __first1),
                      [&__comp](const _RefType1 __x, const _RefType2 __y) {
                        return !__comp(__x, __y) && !__comp(__y, __x);
                      },
                      _IsVector{})
               .first;
         },
         std::less<typename std::iterator_traits<_RandomAccessIterator1>::difference_type>(),
         /*is_first=*/true);
 
     if (__result == __last1 && __first2 + (__result - __first1) != __last2) { // if first sequence shorter than second
       return !__comp(*(__first2 + (__result - __first1)), *__result);
     } else { // if second sequence shorter than first or both have the same number of elements
       return __comp(*__result, *(__first2 + (__result - __first1)));
     }
   }
 }
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_ALGORITHM_IMPL_H */

libcxx/include/__pstl/internal/execution_defs.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_EXECUTION_POLICY_DEFS_H
 #define _PSTL_EXECUTION_POLICY_DEFS_H
 
 #include <type_traits>
 
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace execution {
 inline namespace v1 {
 
 // 2.4, Sequential execution policy
 class sequenced_policy {};
 
 // 2.5, Parallel execution policy
 class parallel_policy {};
 
 // 2.6, Parallel+Vector execution policy
 class parallel_unsequenced_policy {};
 
 class unsequenced_policy {};
 
 // 2.8, Execution policy objects
 constexpr sequenced_policy seq{};
 constexpr parallel_policy par{};
 constexpr parallel_unsequenced_policy par_unseq{};
 constexpr unsequenced_policy unseq{};
 
 // 2.3, Execution policy type trait
 template <class T>
 struct is_execution_policy : std::false_type {};
 
 template <>
 struct is_execution_policy<__pstl::execution::sequenced_policy> : std::true_type {};
 template <>
 struct is_execution_policy<__pstl::execution::parallel_policy> : std::true_type {};
 template <>
 struct is_execution_policy<__pstl::execution::parallel_unsequenced_policy> : std::true_type {};
 template <>
 struct is_execution_policy<__pstl::execution::unsequenced_policy> : std::true_type {};
 
 #if defined(_PSTL_CPP14_VARIABLE_TEMPLATES_PRESENT)
 template <class T>
 constexpr bool is_execution_policy_v = __pstl::execution::is_execution_policy<T>::value;
 #endif
 
 } // namespace v1
 } // namespace execution
 
 namespace __internal {
 template <class ExecPolicy, class T>
 using __enable_if_execution_policy =
     typename std::enable_if<__pstl::execution::is_execution_policy<typename std::decay<ExecPolicy>::type>::value,
                             T>::type;
 
 template <class _IsVector>
 struct __serial_tag;
 template <class _IsVector>
 struct __parallel_tag;
 
 } // namespace __internal
 
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_EXECUTION_POLICY_DEFS_H */

libcxx/include/__pstl/internal/execution_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_EXECUTION_IMPL_H
 #define _PSTL_EXECUTION_IMPL_H
 
 #include <iterator>
 #include <type_traits>
 
 #include "execution_defs.h"
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 template <typename _IteratorTag, typename... _IteratorTypes>
 using __are_iterators_of = std::conjunction<
     std::is_base_of<_IteratorTag, typename std::iterator_traits<std::decay_t<_IteratorTypes>>::iterator_category>...>;
 
 template <typename... _IteratorTypes>
 using __are_random_access_iterators = __are_iterators_of<std::random_access_iterator_tag, _IteratorTypes...>;
 
 struct __serial_backend_tag {};
 struct __tbb_backend_tag {};
 struct __openmp_backend_tag {};
 
 #if defined(_PSTL_PAR_BACKEND_TBB)
 using __par_backend_tag = __tbb_backend_tag;
 #elif defined(_PSTL_PAR_BACKEND_OPENMP)
 using __par_backend_tag = __openmp_backend_tag;
 #elif defined(_PSTL_PAR_BACKEND_SERIAL)
 using __par_backend_tag = __serial_backend_tag;
 #else
 #  error "A parallel backend must be specified";
 #endif
 
 template <class _IsVector>
 struct __serial_tag {
   using __is_vector = _IsVector;
 };
 
 template <class _IsVector>
 struct __parallel_tag {
   using __is_vector = _IsVector;
   // backend tag can be change depending on
   // TBB availability in the environment
   using __backend_tag = __par_backend_tag;
 };
 
 template <class _IsVector, class... _IteratorTypes>
 using __tag_type =
     typename std::conditional<__internal::__are_random_access_iterators<_IteratorTypes...>::value,
                               __parallel_tag<_IsVector>,
                               __serial_tag<_IsVector>>::type;
 
 template <class... _IteratorTypes>
 __serial_tag</*_IsVector = */ std::false_type>
 __select_backend(__pstl::execution::sequenced_policy, _IteratorTypes&&...) {
   return {};
 }
 
 template <class... _IteratorTypes>
 __serial_tag<__internal::__are_random_access_iterators<_IteratorTypes...>>
 __select_backend(__pstl::execution::unsequenced_policy, _IteratorTypes&&...) {
   return {};
 }
 
 template <class... _IteratorTypes>
 __tag_type</*_IsVector = */ std::false_type, _IteratorTypes...>
 __select_backend(__pstl::execution::parallel_policy, _IteratorTypes&&...) {
   return {};
 }
 
 template <class... _IteratorTypes>
 __tag_type<__internal::__are_random_access_iterators<_IteratorTypes...>, _IteratorTypes...>
 __select_backend(__pstl::execution::parallel_unsequenced_policy, _IteratorTypes&&...) {
   return {};
 }
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_EXECUTION_IMPL_H */

libcxx/include/__pstl/internal/glue_algorithm_defs.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_ALGORITHM_DEFS_H
 #define _PSTL_GLUE_ALGORITHM_DEFS_H
 
 #include <functional>
 #include <iterator>
 
 #include "execution_defs.h"
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace std {
 
 // [alg.any_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 any_of(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred);
 
 // [alg.all_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 all_of(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred);
 
 // [alg.none_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 none_of(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred);
 
 // [alg.foreach]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Function>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 for_each(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Function __f);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Function>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 for_each_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n, _Function __f);
 
 // [alg.find]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 find_if(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 find_if_not(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 find(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value);
 
 // [alg.find.end]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 find_end(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first,
          _ForwardIterator1 __last,
          _ForwardIterator2 __s_first,
          _ForwardIterator2 __s_last,
          _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 find_end(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first,
          _ForwardIterator1 __last,
          _ForwardIterator2 __s_first,
          _ForwardIterator2 __s_last);
 
 // [alg.find_first_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1> find_first_of(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 find_first_of(_ExecutionPolicy&& __exec,
               _ForwardIterator1 __first,
               _ForwardIterator1 __last,
               _ForwardIterator2 __s_first,
               _ForwardIterator2 __s_last);
 
 // [alg.adjacent_find]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 adjacent_find(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 adjacent_find(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred);
 
 // [alg.count]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy,
                                                  typename iterator_traits<_ForwardIterator>::difference_type>
 count(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy,
                                                  typename iterator_traits<_ForwardIterator>::difference_type>
 count_if(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred);
 
 // [alg.search]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 search(_ExecutionPolicy&& __exec,
        _ForwardIterator1 __first,
        _ForwardIterator1 __last,
        _ForwardIterator2 __s_first,
        _ForwardIterator2 __s_last,
        _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 search(_ExecutionPolicy&& __exec,
        _ForwardIterator1 __first,
        _ForwardIterator1 __last,
        _ForwardIterator2 __s_first,
        _ForwardIterator2 __s_last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 search_n(_ExecutionPolicy&& __exec,
          _ForwardIterator __first,
          _ForwardIterator __last,
          _Size __count,
          const _Tp& __value,
          _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> search_n(
     _ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Size __count, const _Tp& __value);
 
 // [alg.copy]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 copy(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __result);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _Size, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 copy_n(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _Size __n, _ForwardIterator2 __result);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 copy_if(_ExecutionPolicy&& __exec,
         _ForwardIterator1 __first,
         _ForwardIterator1 __last,
         _ForwardIterator2 result,
         _Predicate __pred);
 
 // [alg.swap]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> swap_ranges(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2);
 
 // [alg.transform]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _UnaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 transform(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first,
           _ForwardIterator1 __last,
           _ForwardIterator2 __result,
           _UnaryOperation __op);
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 transform(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first1,
           _ForwardIterator1 __last1,
           _ForwardIterator2 __first2,
           _ForwardIterator __result,
           _BinaryOperation __op);
 
 // [alg.replace]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 replace_if(_ExecutionPolicy&& __exec,
            _ForwardIterator __first,
            _ForwardIterator __last,
            _UnaryPredicate __pred,
            const _Tp& __new_value);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 replace(_ExecutionPolicy&& __exec,
         _ForwardIterator __first,
         _ForwardIterator __last,
         const _Tp& __old_value,
         const _Tp& __new_value);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _UnaryPredicate, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> replace_copy_if(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _UnaryPredicate __pred,
     const _Tp& __new_value);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> replace_copy(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     const _Tp& __old_value,
     const _Tp& __new_value);
 
 // [alg.fill]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 fill(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 fill_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __count, const _Tp& __value);
 
 // [alg.generate]
 template <class _ExecutionPolicy, class _ForwardIterator, class _Generator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 generate(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Generator __g);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Generator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 generate_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size count, _Generator __g);
 
 // [alg.remove]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> remove_copy_if(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Predicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 remove_copy(_ExecutionPolicy&& __exec,
             _ForwardIterator1 __first,
             _ForwardIterator1 __last,
             _ForwardIterator2 __result,
             const _Tp& __value);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 remove_if(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 remove(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value);
 
 // [alg.unique]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 unique(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 unique(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 unique_copy(_ExecutionPolicy&& __exec,
             _ForwardIterator1 __first,
             _ForwardIterator1 __last,
             _ForwardIterator2 __result,
             _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 unique_copy(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __result);
 
 // [alg.reverse]
 
 template <class _ExecutionPolicy, class _BidirectionalIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 reverse(_ExecutionPolicy&& __exec, _BidirectionalIterator __first, _BidirectionalIterator __last);
 
 template <class _ExecutionPolicy, class _BidirectionalIterator, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 reverse_copy(_ExecutionPolicy&& __exec,
              _BidirectionalIterator __first,
              _BidirectionalIterator __last,
              _ForwardIterator __d_first);
 
 // [alg.rotate]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 rotate(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 rotate_copy(_ExecutionPolicy&& __exec,
             _ForwardIterator1 __first,
             _ForwardIterator1 __middle,
             _ForwardIterator1 __last,
             _ForwardIterator2 __result);
 
 // [alg.partitions]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_partitioned(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 partition(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _BidirectionalIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _BidirectionalIterator> stable_partition(
     _ExecutionPolicy&& __exec, _BidirectionalIterator __first, _BidirectionalIterator __last, _UnaryPredicate __pred);
 
 template <class _ExecutionPolicy,
           class _ForwardIterator,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 partition_copy(_ExecutionPolicy&& __exec,
                _ForwardIterator __first,
                _ForwardIterator __last,
                _ForwardIterator1 __out_true,
                _ForwardIterator2 __out_false,
                _UnaryPredicate __pred);
 
 // [alg.sort]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last);
 
 // [stable.sort]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 stable_sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 stable_sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last);
 
 // [mismatch]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2,
          _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _BinaryPredicate __pred);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2);
 
 // [alg.equal]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _BinaryPredicate __p);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _BinaryPredicate __p);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2);
 
 // [alg.move]
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 move(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __d_first);
 
 // [partial.sort]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 partial_sort(_ExecutionPolicy&& __exec,
              _RandomAccessIterator __first,
              _RandomAccessIterator __middle,
              _RandomAccessIterator __last,
              _Compare __comp);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 partial_sort(_ExecutionPolicy&& __exec,
              _RandomAccessIterator __first,
              _RandomAccessIterator __middle,
              _RandomAccessIterator __last);
 
 // [partial.sort.copy]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator> partial_sort_copy(
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _RandomAccessIterator __d_first,
     _RandomAccessIterator __d_last,
     _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator> partial_sort_copy(
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _RandomAccessIterator __d_first,
     _RandomAccessIterator __d_last);
 
 // [is.sorted]
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 is_sorted_until(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 is_sorted_until(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_sorted(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_sorted(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 // [alg.nth.element]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 nth_element(_ExecutionPolicy&& __exec,
             _RandomAccessIterator __first,
             _RandomAccessIterator __nth,
             _RandomAccessIterator __last,
             _Compare __comp);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 nth_element(_ExecutionPolicy&& __exec,
             _RandomAccessIterator __first,
             _RandomAccessIterator __nth,
             _RandomAccessIterator __last);
 
 // [alg.merge]
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 merge(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _ForwardIterator __d_first,
       _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 merge(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _ForwardIterator __d_first);
 
 template <class _ExecutionPolicy, class _BidirectionalIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 inplace_merge(_ExecutionPolicy&& __exec,
               _BidirectionalIterator __first,
               _BidirectionalIterator __middle,
               _BidirectionalIterator __last,
               _Compare __comp);
 
 template <class _ExecutionPolicy, class _BidirectionalIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 inplace_merge(_ExecutionPolicy&& __exec,
               _BidirectionalIterator __first,
               _BidirectionalIterator __middle,
               _BidirectionalIterator __last);
 
 // [includes]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 includes(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2,
          _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 includes(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2);
 
 // [set.union]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 set_union(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first1,
           _ForwardIterator1 __last1,
           _ForwardIterator2 __first2,
           _ForwardIterator2 __last2,
           _ForwardIterator __result,
           _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 set_union(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first1,
           _ForwardIterator1 __last1,
           _ForwardIterator2 __first2,
           _ForwardIterator2 __last2,
           _ForwardIterator __result);
 
 // [set.intersection]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_intersection(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result,
     _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_intersection(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result);
 
 // [set.difference]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result,
     _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result);
 
 // [set.symmetric.difference]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_symmetric_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator result,
     _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_symmetric_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result);
 
 // [is.heap]
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator>
 is_heap_until(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator>
 is_heap_until(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_heap(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_heap(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last);
 
 // [alg.min.max]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 min_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 min_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 max_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 max_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator, _ForwardIterator>>
 minmax_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator, _ForwardIterator>>
 minmax_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 // [alg.lex.comparison]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool> lexicographical_compare(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Compare __comp);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool> lexicographical_compare(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2);
 
 } // namespace std
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_GLUE_ALGORITHM_DEFS_H */

libcxx/include/__pstl/internal/glue_algorithm_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_ALGORITHM_IMPL_H
 #define _PSTL_GLUE_ALGORITHM_IMPL_H
 
 #include <functional>
 
 #include "pstl_config.h"
 
 #include "algorithm_fwd.h"
 #include "execution_defs.h"
 #include "numeric_fwd.h" /* count and count_if use __pattern_transform_reduce */
 #include "utils.h"
 
 #include "execution_impl.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace std {
 
 // [alg.any_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 any_of(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_any_of(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 // [alg.all_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Pred>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 all_of(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Pred __pred) {
   return !std::any_of(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::not_fn(__pred));
 }
 
 // [alg.none_of]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 none_of(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   return !std::any_of(std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 // [alg.foreach]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Function>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 for_each(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Function __f) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_walk1(__dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __f);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Function>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 for_each_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n, _Function __f) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_walk1_n(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __n, __f);
 }
 
 // [alg.find]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 find_if(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_find_if(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 find_if_not(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   return std::find_if(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::not_fn(__pred));
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 find(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   return std::find_if(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __pstl::__internal::__equal_value<_Tp>(__value));
 }
 
 // [alg.find.end]
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 find_end(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first,
          _ForwardIterator1 __last,
          _ForwardIterator2 __s_first,
          _ForwardIterator2 __s_last,
          _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __s_first);
 
   return __pstl::__internal::__pattern_find_end(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 find_end(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first,
          _ForwardIterator1 __last,
          _ForwardIterator2 __s_first,
          _ForwardIterator2 __s_last) {
   return std::find_end(std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __s_last, std::equal_to<>());
 }
 
 // [alg.find_first_of]
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1> find_first_of(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __s_first,
     _ForwardIterator2 __s_last,
     _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __s_first);
 
   return __pstl::__internal::__pattern_find_first_of(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 find_first_of(_ExecutionPolicy&& __exec,
               _ForwardIterator1 __first,
               _ForwardIterator1 __last,
               _ForwardIterator2 __s_first,
               _ForwardIterator2 __s_last) {
   return std::find_first_of(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __s_last, std::equal_to<>());
 }
 
 // [alg.adjacent_find]
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 adjacent_find(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   return __pstl::__internal::__pattern_adjacent_find(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       std::equal_to<_ValueType>(),
       /*first_semantic*/ false);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 adjacent_find(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_adjacent_find(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred, /*first_semantic*/ false);
 }
 
 // [alg.count]
 
 // Implementation note: count and count_if call the pattern directly instead of calling std::transform_reduce
 // so that we do not have to include <numeric>.
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy,
                                                  typename iterator_traits<_ForwardIterator>::difference_type>
 count(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   return __pstl::__internal::__pattern_count(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, [&__value](const _ValueType& __x) {
         return __value == __x;
       });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy,
                                                  typename iterator_traits<_ForwardIterator>::difference_type>
 count_if(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_count(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 // [alg.search]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 search(_ExecutionPolicy&& __exec,
        _ForwardIterator1 __first,
        _ForwardIterator1 __last,
        _ForwardIterator2 __s_first,
        _ForwardIterator2 __s_last,
        _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __s_first);
 
   return __pstl::__internal::__pattern_search(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __s_last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator1>
 search(_ExecutionPolicy&& __exec,
        _ForwardIterator1 __first,
        _ForwardIterator1 __last,
        _ForwardIterator2 __s_first,
        _ForwardIterator2 __s_last) {
   return std::search(std::forward<_ExecutionPolicy>(__exec), __first, __last, __s_first, __s_last, std::equal_to<>());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 search_n(_ExecutionPolicy&& __exec,
          _ForwardIterator __first,
          _ForwardIterator __last,
          _Size __count,
          const _Tp& __value,
          _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_search_n(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __count, __value, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> search_n(
     _ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Size __count, const _Tp& __value) {
   return std::search_n(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __count,
       __value,
       std::equal_to<typename iterator_traits<_ForwardIterator>::value_type>());
 }
 
 // [alg.copy]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 copy(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __result) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__pattern_walk2_brick(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       [](_ForwardIterator1 __begin, _ForwardIterator1 __end, _ForwardIterator2 __res) {
         return __pstl::__internal::__brick_copy(__begin, __end, __res, __is_vector{});
       });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _Size, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 copy_n(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _Size __n, _ForwardIterator2 __result) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__pattern_walk2_brick_n(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __n,
       __result,
       [](_ForwardIterator1 __begin, _Size __sz, _ForwardIterator2 __res) {
         return __pstl::__internal::__brick_copy_n(__begin, __sz, __res, __is_vector{});
       });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 copy_if(_ExecutionPolicy&& __exec,
         _ForwardIterator1 __first,
         _ForwardIterator1 __last,
         _ForwardIterator2 __result,
         _Predicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_copy_if(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __result, __pred);
 }
 
 // [alg.swap]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> swap_ranges(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2) {
   typedef typename iterator_traits<_ForwardIterator1>::reference _ReferenceType1;
   typedef typename iterator_traits<_ForwardIterator2>::reference _ReferenceType2;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   return __pstl::__internal::__pattern_walk2(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       [](_ReferenceType1 __x, _ReferenceType2 __y) {
         using std::swap;
         swap(__x, __y);
       });
 }
 
 // [alg.transform]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _UnaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 transform(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first,
           _ForwardIterator1 __last,
           _ForwardIterator2 __result,
           _UnaryOperation __op) {
   typedef typename iterator_traits<_ForwardIterator1>::reference _InputType;
   typedef typename iterator_traits<_ForwardIterator2>::reference _OutputType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_walk2(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       [__op](_InputType __x, _OutputType __y) mutable { __y = __op(__x); });
 }
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 transform(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first1,
           _ForwardIterator1 __last1,
           _ForwardIterator2 __first2,
           _ForwardIterator __result,
           _BinaryOperation __op) {
   typedef typename iterator_traits<_ForwardIterator1>::reference _Input1Type;
   typedef typename iterator_traits<_ForwardIterator2>::reference _Input2Type;
   typedef typename iterator_traits<_ForwardIterator>::reference _OutputType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2, __result);
 
   return __pstl::__internal::__pattern_walk3(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __result,
       [__op](_Input1Type x, _Input2Type y, _OutputType z) mutable { z = __op(x, y); });
 }
 
 // [alg.replace]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 replace_if(_ExecutionPolicy&& __exec,
            _ForwardIterator __first,
            _ForwardIterator __last,
            _UnaryPredicate __pred,
            const _Tp& __new_value) {
   typedef typename iterator_traits<_ForwardIterator>::reference _ElementType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_walk1(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       [&__pred, &__new_value](_ElementType __elem) {
         if (__pred(__elem)) {
           __elem = __new_value;
         }
       });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 replace(_ExecutionPolicy&& __exec,
         _ForwardIterator __first,
         _ForwardIterator __last,
         const _Tp& __old_value,
         const _Tp& __new_value) {
   std::replace_if(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __pstl::__internal::__equal_value<_Tp>(__old_value),
       __new_value);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _UnaryPredicate, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> replace_copy_if(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _UnaryPredicate __pred,
     const _Tp& __new_value) {
   typedef typename iterator_traits<_ForwardIterator1>::reference _InputType;
   typedef typename iterator_traits<_ForwardIterator2>::reference _OutputType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_walk2(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       [__pred, &__new_value](_InputType __x, _OutputType __y) mutable { __y = __pred(__x) ? __new_value : __x; });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> replace_copy(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     const _Tp& __old_value,
     const _Tp& __new_value) {
   return std::replace_copy_if(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __pstl::__internal::__equal_value<_Tp>(__old_value),
       __new_value);
 }
 
 // [alg.fill]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 fill(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_fill(__dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __value);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 fill_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __count, const _Tp& __value) {
   if (__count <= 0)
     return __first;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_fill_n(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __count, __value);
 }
 
 // [alg.generate]
 template <class _ExecutionPolicy, class _ForwardIterator, class _Generator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 generate(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Generator __g) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_generate(__dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __g);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Generator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 generate_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __count, _Generator __g) {
   if (__count <= 0)
     return __first;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_generate_n(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __count, __g);
 }
 
 // [alg.remove]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Predicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> remove_copy_if(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Predicate __pred) {
   return std::copy_if(std::forward<_ExecutionPolicy>(__exec), __first, __last, __result, std::not_fn(__pred));
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 remove_copy(_ExecutionPolicy&& __exec,
             _ForwardIterator1 __first,
             _ForwardIterator1 __last,
             _ForwardIterator2 __result,
             const _Tp& __value) {
   return std::copy_if(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __pstl::__internal::__not_equal_value<_Tp>(__value));
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 remove_if(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_remove_if(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 remove(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   return std::remove_if(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __pstl::__internal::__equal_value<_Tp>(__value));
 }
 
 // [alg.unique]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 unique(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_unique(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 unique(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   return std::unique(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::equal_to<>());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 unique_copy(_ExecutionPolicy&& __exec,
             _ForwardIterator1 __first,
             _ForwardIterator1 __last,
             _ForwardIterator2 __result,
             _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_unique_copy(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __result, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> unique_copy(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __result) {
   return std::unique_copy(__exec, __first, __last, __result, std::equal_to<>());
 }
 
 // [alg.reverse]
 
 template <class _ExecutionPolicy, class _BidirectionalIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 reverse(_ExecutionPolicy&& __exec, _BidirectionalIterator __first, _BidirectionalIterator __last) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_reverse(__dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last);
 }
 
 template <class _ExecutionPolicy, class _BidirectionalIterator, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 reverse_copy(_ExecutionPolicy&& __exec,
              _BidirectionalIterator __first,
              _BidirectionalIterator __last,
              _ForwardIterator __d_first) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __d_first);
 
   return __pstl::__internal::__pattern_reverse_copy(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __d_first);
 }
 
 // [alg.rotate]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 rotate(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_rotate(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __middle, __last);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 rotate_copy(_ExecutionPolicy&& __exec,
             _ForwardIterator1 __first,
             _ForwardIterator1 __middle,
             _ForwardIterator1 __last,
             _ForwardIterator2 __result) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_rotate_copy(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __middle, __last, __result);
 }
 
 // [alg.partitions]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_partitioned(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_is_partitioned(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 partition(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _UnaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_partition(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 template <class _ExecutionPolicy, class _BidirectionalIterator, class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _BidirectionalIterator> stable_partition(
     _ExecutionPolicy&& __exec, _BidirectionalIterator __first, _BidirectionalIterator __last, _UnaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_stable_partition(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __pred);
 }
 
 template <class _ExecutionPolicy,
           class _ForwardIterator,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _UnaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 partition_copy(_ExecutionPolicy&& __exec,
                _ForwardIterator __first,
                _ForwardIterator __last,
                _ForwardIterator1 __out_true,
                _ForwardIterator2 __out_false,
                _UnaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __out_true, __out_false);
 
   return __pstl::__internal::__pattern_partition_copy(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __out_true, __out_false, __pred);
 }
 
 // [alg.sort]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   typedef typename iterator_traits<_RandomAccessIterator>::value_type _InputType;
   return __pstl::__internal::__pattern_sort(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __comp,
       typename std::is_move_constructible<_InputType>::type());
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last) {
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _InputType;
   std::sort(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_InputType>());
 }
 
 // [stable.sort]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 stable_sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_stable_sort(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 stable_sort(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last) {
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _InputType;
   std::stable_sort(__exec, __first, __last, std::less<_InputType>());
 }
 
 // [mismatch]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2,
          _BinaryPredicate __pred) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   return __pstl::__internal::__pattern_mismatch(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _BinaryPredicate __pred) {
   return std::mismatch(
       __exec, __first1, __last1, __first2, std::next(__first2, std::distance(__first1, __last1)), __pred);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2) {
   return std::mismatch(std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, std::equal_to<>());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator1, _ForwardIterator2>>
 mismatch(_ExecutionPolicy&& __exec, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2) {
   // TODO: to get rid of "distance"
   return std::mismatch(
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       std::next(__first2, std::distance(__first1, __last1)));
 }
 
 // [alg.equal]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _BinaryPredicate __p) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   return __pstl::__internal::__pattern_equal(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __p);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2) {
   return std::equal(std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, std::equal_to<>());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _BinaryPredicate __p) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   return __pstl::__internal::__pattern_equal(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __p);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 equal(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2) {
   return equal(std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, std::equal_to<>());
 }
 
 // [alg.move]
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2>
 move(_ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __d_first) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __d_first);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__pattern_walk2_brick(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __d_first,
       [](_ForwardIterator1 __begin, _ForwardIterator1 __end, _ForwardIterator2 __res) {
         return __pstl::__internal::__brick_move(__begin, __end, __res, __is_vector{});
       });
 }
 
 // [partial.sort]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 partial_sort(_ExecutionPolicy&& __exec,
              _RandomAccessIterator __first,
              _RandomAccessIterator __middle,
              _RandomAccessIterator __last,
              _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_partial_sort(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __middle, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 partial_sort(_ExecutionPolicy&& __exec,
              _RandomAccessIterator __first,
              _RandomAccessIterator __middle,
              _RandomAccessIterator __last) {
   typedef typename iterator_traits<_RandomAccessIterator>::value_type _InputType;
   std::partial_sort(__exec, __first, __middle, __last, std::less<_InputType>());
 }
 
 // [partial.sort.copy]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator> partial_sort_copy(
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _RandomAccessIterator __d_first,
     _RandomAccessIterator __d_last,
     _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __d_first);
 
   return __pstl::__internal::__pattern_partial_sort_copy(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __d_first, __d_last, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator> partial_sort_copy(
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _RandomAccessIterator __d_first,
     _RandomAccessIterator __d_last) {
   return std::partial_sort_copy(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __d_first, __d_last, std::less<>());
 }
 
 // [is.sorted]
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 is_sorted_until(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) {
   auto __dispatch_tag          = __pstl::__internal::__select_backend(__exec, __first);
   const _ForwardIterator __res = __pstl::__internal::__pattern_adjacent_find(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __pstl::__internal::__reorder_pred<_Compare>(__comp),
       /*first_semantic*/ false);
   return __res == __last ? __last : std::next(__res);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 is_sorted_until(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename std::iterator_traits<_ForwardIterator>::value_type _InputType;
   return is_sorted_until(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_InputType>());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_sorted(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_adjacent_find(
              __dispatch_tag,
              std::forward<_ExecutionPolicy>(__exec),
              __first,
              __last,
              __pstl::__internal::__reorder_pred<_Compare>(__comp),
              /*or_semantic*/ true) == __last;
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_sorted(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename std::iterator_traits<_ForwardIterator>::value_type _InputType;
   return std::is_sorted(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_InputType>());
 }
 
 // [alg.merge]
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 merge(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _ForwardIterator __d_first,
       _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2, __d_first);
 
   return __pstl::__internal::__pattern_merge(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __d_first, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 merge(_ExecutionPolicy&& __exec,
       _ForwardIterator1 __first1,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _ForwardIterator __d_first) {
   return std::merge(
       std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __d_first, std::less<>());
 }
 
 template <class _ExecutionPolicy, class _BidirectionalIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 inplace_merge(_ExecutionPolicy&& __exec,
               _BidirectionalIterator __first,
               _BidirectionalIterator __middle,
               _BidirectionalIterator __last,
               _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_inplace_merge(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __middle, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _BidirectionalIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 inplace_merge(_ExecutionPolicy&& __exec,
               _BidirectionalIterator __first,
               _BidirectionalIterator __middle,
               _BidirectionalIterator __last) {
   typedef typename std::iterator_traits<_BidirectionalIterator>::value_type _InputType;
   std::inplace_merge(__exec, __first, __middle, __last, std::less<_InputType>());
 }
 
 // [includes]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 includes(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2,
          _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   return __pstl::__internal::__pattern_includes(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 includes(_ExecutionPolicy&& __exec,
          _ForwardIterator1 __first1,
          _ForwardIterator1 __last1,
          _ForwardIterator2 __first2,
          _ForwardIterator2 __last2) {
   return std::includes(std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, std::less<>());
 }
 
 // [set.union]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 set_union(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first1,
           _ForwardIterator1 __last1,
           _ForwardIterator2 __first2,
           _ForwardIterator2 __last2,
           _ForwardIterator __result,
           _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2, __result);
 
   return __pstl::__internal::__pattern_set_union(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 set_union(_ExecutionPolicy&& __exec,
           _ForwardIterator1 __first1,
           _ForwardIterator1 __last1,
           _ForwardIterator2 __first2,
           _ForwardIterator2 __last2,
           _ForwardIterator __result) {
   return std::set_union(
       std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, std::less<>());
 }
 
 // [set.intersection]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_intersection(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result,
     _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2, __result);
 
   return __pstl::__internal::__pattern_set_intersection(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_intersection(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result) {
   return std::set_intersection(
       std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, std::less<>());
 }
 
 // [set.difference]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result,
     _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2, __result);
 
   return __pstl::__internal::__pattern_set_difference(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result) {
   return std::set_difference(
       std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, std::less<>());
 }
 
 // [set.symmetric.difference]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _ForwardIterator,
           class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_symmetric_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result,
     _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2, __result);
 
   return __pstl::__internal::__pattern_set_symmetric_difference(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> set_symmetric_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _ForwardIterator __result) {
   return std::set_symmetric_difference(
       std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __result, std::less<>());
 }
 
 // [is.heap]
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator>
 is_heap_until(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__pattern_is_heap_until(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _RandomAccessIterator>
 is_heap_until(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last) {
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _InputType;
   return std::is_heap_until(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_InputType>());
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_heap(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
   return std::is_heap_until(std::forward<_ExecutionPolicy>(__exec), __first, __last, __comp) == __last;
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool>
 is_heap(_ExecutionPolicy&& __exec, _RandomAccessIterator __first, _RandomAccessIterator __last) {
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _InputType;
   return std::is_heap(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_InputType>());
 }
 
 // [alg.min.max]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 min_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_min_element(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 min_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename std::iterator_traits<_ForwardIterator>::value_type _InputType;
   return std::min_element(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_InputType>());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 max_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) {
   return min_element(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __pstl::__internal::__reorder_pred<_Compare>(__comp));
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 max_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename std::iterator_traits<_ForwardIterator>::value_type _InputType;
   return std::min_element(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __pstl::__internal::__reorder_pred<std::less<_InputType>>(std::less<_InputType>()));
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator, _ForwardIterator>>
 minmax_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_minmax_element(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, std::pair<_ForwardIterator, _ForwardIterator>>
 minmax_element(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   return std::minmax_element(std::forward<_ExecutionPolicy>(__exec), __first, __last, std::less<_ValueType>());
 }
 
 // [alg.nth.element]
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 nth_element(_ExecutionPolicy&& __exec,
             _RandomAccessIterator __first,
             _RandomAccessIterator __nth,
             _RandomAccessIterator __last,
             _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__pattern_nth_element(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __nth, __last, __comp);
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 nth_element(_ExecutionPolicy&& __exec,
             _RandomAccessIterator __first,
             _RandomAccessIterator __nth,
             _RandomAccessIterator __last) {
   typedef typename iterator_traits<_RandomAccessIterator>::value_type _InputType;
   std::nth_element(std::forward<_ExecutionPolicy>(__exec), __first, __nth, __last, std::less<_InputType>());
 }
 
 // [alg.lex.comparison]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Compare>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool> lexicographical_compare(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2,
     _Compare __comp) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   return __pstl::__internal::__pattern_lexicographical_compare(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, __comp);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, bool> lexicographical_compare(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _ForwardIterator2 __last2) {
   return std::lexicographical_compare(
       std::forward<_ExecutionPolicy>(__exec), __first1, __last1, __first2, __last2, std::less<>());
 }
 
 } // namespace std
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_GLUE_ALGORITHM_IMPL_H */

libcxx/include/__pstl/internal/glue_execution_defs.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_EXECUTION_DEFS_H
 #define _PSTL_GLUE_EXECUTION_DEFS_H
 
 #include <type_traits>
 
 #include "execution_defs.h"
 #include "pstl_config.h"
 
 namespace std {
 // Type trait
 using __pstl::execution::is_execution_policy;
 #if defined(_PSTL_CPP14_VARIABLE_TEMPLATES_PRESENT)
 #  if defined(__INTEL_COMPILER)
 template <class T>
 constexpr bool is_execution_policy_v = is_execution_policy<T>::value;
 #  else
 using __pstl::execution::is_execution_policy_v;
 #  endif
 #endif
 
 namespace execution {
 // Standard C++ policy classes
 using __pstl::execution::parallel_policy;
 using __pstl::execution::parallel_unsequenced_policy;
 using __pstl::execution::sequenced_policy;
 
 // Standard predefined policy instances
 using __pstl::execution::par;
 using __pstl::execution::par_unseq;
 using __pstl::execution::seq;
 
 // Implementation-defined names
 // Unsequenced policy is not yet standard, but for consistency
 // we include it into namespace std::execution as well
 using __pstl::execution::unseq;
 using __pstl::execution::unsequenced_policy;
 } // namespace execution
 } // namespace std
 
 #include "algorithm_impl.h"
 #include "numeric_impl.h"
 #include "parallel_backend.h"
 
 #endif /* _PSTL_GLUE_EXECUTION_DEFS_H */

libcxx/include/__pstl/internal/glue_memory_defs.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_MEMORY_DEFS_H
 #define _PSTL_GLUE_MEMORY_DEFS_H
 
 #include "execution_defs.h"
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace std {
 
 // [uninitialized.copy]
 
 template <class _ExecutionPolicy, class _InputIterator, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_copy(_ExecutionPolicy&& __exec, _InputIterator __first, _InputIterator __last, _ForwardIterator __result);
 
 template <class _ExecutionPolicy, class _InputIterator, class _Size, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_copy_n(_ExecutionPolicy&& __exec, _InputIterator __first, _Size __n, _ForwardIterator __result);
 
 // [uninitialized.move]
 
 template <class _ExecutionPolicy, class _InputIterator, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_move(_ExecutionPolicy&& __exec, _InputIterator __first, _InputIterator __last, _ForwardIterator __result);
 
 template <class _ExecutionPolicy, class _InputIterator, class _Size, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_move_n(_ExecutionPolicy&& __exec, _InputIterator __first, _Size __n, _ForwardIterator __result);
 
 // [uninitialized.fill]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 uninitialized_fill(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_fill_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n, const _Tp& __value);
 
 // [specialized.destroy]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 destroy(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 destroy_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n);
 
 // [uninitialized.construct.default]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 uninitialized_default_construct(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_default_construct_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n);
 
 // [uninitialized.construct.value]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 uninitialized_value_construct(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_value_construct_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n);
 
 } //  namespace std
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_GLUE_MEMORY_DEFS_H */

libcxx/include/__pstl/internal/glue_memory_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_MEMORY_IMPL_H
 #define _PSTL_GLUE_MEMORY_IMPL_H
 
 #include "pstl_config.h"
 
 #include "algorithm_fwd.h"
 #include "execution_defs.h"
 #include "utils.h"
 
 #include "execution_impl.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace std {
 
 // [uninitialized.copy]
 
 template <class _ExecutionPolicy, class _InputIterator, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> uninitialized_copy(
     _ExecutionPolicy&& __exec, _InputIterator __first, _InputIterator __last, _ForwardIterator __result) {
   typedef typename iterator_traits<_InputIterator>::value_type _ValueType1;
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType2;
   typedef typename iterator_traits<_InputIterator>::reference _ReferenceType1;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType2;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__invoke_if_else(
       std::integral_constant < bool,
       std::is_trivial<_ValueType1>::value&& std::is_trivial<_ValueType2>::value > (),
       [&]() {
         return __pstl::__internal::__pattern_walk2_brick(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __last,
             __result,
             [](_InputIterator __begin, _InputIterator __end, _ForwardIterator __res) {
               return __pstl::__internal::__brick_copy(__begin, __end, __res, __is_vector{});
             });
       },
       [&]() {
         return __pstl::__internal::__pattern_walk2(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __last,
             __result,
             [](_ReferenceType1 __val1, _ReferenceType2 __val2) { ::new (std::addressof(__val2)) _ValueType2(__val1); });
       });
 }
 
 template <class _ExecutionPolicy, class _InputIterator, class _Size, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_copy_n(_ExecutionPolicy&& __exec, _InputIterator __first, _Size __n, _ForwardIterator __result) {
   typedef typename iterator_traits<_InputIterator>::value_type _ValueType1;
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType2;
   typedef typename iterator_traits<_InputIterator>::reference _ReferenceType1;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType2;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__invoke_if_else(
       std::integral_constant < bool,
       std::is_trivial<_ValueType1>::value&& std::is_trivial<_ValueType2>::value > (),
       [&]() {
         return __pstl::__internal::__pattern_walk2_brick_n(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __n,
             __result,
             [](_InputIterator __begin, _Size __sz, _ForwardIterator __res) {
               return __pstl::__internal::__brick_copy_n(__begin, __sz, __res, __is_vector{});
             });
       },
       [&]() {
         return __pstl::__internal::__pattern_walk2_n(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __n,
             __result,
             [](_ReferenceType1 __val1, _ReferenceType2 __val2) { ::new (std::addressof(__val2)) _ValueType2(__val1); });
       });
 }
 
 // [uninitialized.move]
 
 template <class _ExecutionPolicy, class _InputIterator, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator> uninitialized_move(
     _ExecutionPolicy&& __exec, _InputIterator __first, _InputIterator __last, _ForwardIterator __result) {
   typedef typename iterator_traits<_InputIterator>::value_type _ValueType1;
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType2;
   typedef typename iterator_traits<_InputIterator>::reference _ReferenceType1;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType2;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__invoke_if_else(
       std::integral_constant < bool,
       std::is_trivial<_ValueType1>::value&& std::is_trivial<_ValueType2>::value > (),
       [&]() {
         return __pstl::__internal::__pattern_walk2_brick(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __last,
             __result,
             [](_InputIterator __begin, _InputIterator __end, _ForwardIterator __res) {
               return __pstl::__internal::__brick_copy(__begin, __end, __res, __is_vector{});
             });
       },
       [&]() {
         return __pstl::__internal::__pattern_walk2(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __last,
             __result,
             [](_ReferenceType1 __val1, _ReferenceType2 __val2) {
               ::new (std::addressof(__val2)) _ValueType2(std::move(__val1));
             });
       });
 }
 
 template <class _ExecutionPolicy, class _InputIterator, class _Size, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_move_n(_ExecutionPolicy&& __exec, _InputIterator __first, _Size __n, _ForwardIterator __result) {
   typedef typename iterator_traits<_InputIterator>::value_type _ValueType1;
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType2;
   typedef typename iterator_traits<_InputIterator>::reference _ReferenceType1;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType2;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__invoke_if_else(
       std::integral_constant < bool,
       std::is_trivial<_ValueType1>::value&& std::is_trivial<_ValueType2>::value > (),
       [&]() {
         return __pstl::__internal::__pattern_walk2_brick_n(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __n,
             __result,
             [](_InputIterator __begin, _Size __sz, _ForwardIterator __res) {
               return __pstl::__internal::__brick_copy_n(__begin, __sz, __res, __is_vector{});
             });
       },
       [&]() {
         return __pstl::__internal::__pattern_walk2_n(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __n,
             __result,
             [](_ReferenceType1 __val1, _ReferenceType2 __val2) {
               ::new (std::addressof(__val2)) _ValueType2(std::move(__val1));
             });
       });
 }
 
 // [uninitialized.fill]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 uninitialized_fill(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   __pstl::__internal::__invoke_if_else(
       std::is_arithmetic<_ValueType>(),
       [&]() {
         __pstl::__internal::__pattern_walk_brick(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __last,
             [&__value](_ForwardIterator __begin, _ForwardIterator __end) {
               __pstl::__internal::__brick_fill(__begin, __end, _ValueType(__value), __is_vector{});
             });
       },
       [&]() {
         __pstl::__internal::__pattern_walk1(
             __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, [&__value](_ReferenceType __val) {
               ::new (std::addressof(__val)) _ValueType(__value);
             });
       });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_fill_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n, const _Tp& __value) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__invoke_if_else(
       std::is_arithmetic<_ValueType>(),
       [&]() {
         return __pstl::__internal::__pattern_walk_brick_n(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __n,
             [&__value](_ForwardIterator __begin, _Size __count) {
               return __pstl::__internal::__brick_fill_n(__begin, __count, _ValueType(__value), __is_vector{});
             });
       },
       [&]() {
         return __pstl::__internal::__pattern_walk1_n(
             __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __n, [&__value](_ReferenceType __val) {
               ::new (std::addressof(__val)) _ValueType(__value);
             });
       });
 }
 
 // [specialized.destroy]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 destroy(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__invoke_if_not(std::is_trivially_destructible<_ValueType>(), [&]() {
     __pstl::__internal::__pattern_walk1(
         __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, [](_ReferenceType __val) {
           __val.~_ValueType();
         });
   });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 destroy_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__invoke_if_else(
       std::is_trivially_destructible<_ValueType>(),
       [&]() { return std::next(__first, __n); },
       [&]() {
         return __pstl::__internal::__pattern_walk1_n(
             __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __n, [](_ReferenceType __val) {
               __val.~_ValueType();
             });
       });
 }
 
 // [uninitialized.construct.default]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 uninitialized_default_construct(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   __pstl::__internal::__invoke_if_not(std::is_trivial<_ValueType>(), [&]() {
     __pstl::__internal::__pattern_walk1(
         __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, [](_ReferenceType __val) {
           ::new (std::addressof(__val)) _ValueType;
         });
   });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_default_construct_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   return __pstl::__internal::__invoke_if_else(
       std::is_trivial<_ValueType>(),
       [&]() { return std::next(__first, __n); },
       [&]() {
         return __pstl::__internal::__pattern_walk1_n(
             __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __n, [](_ReferenceType __val) {
               ::new (std::addressof(__val)) _ValueType;
             });
       });
 }
 
 // [uninitialized.construct.value]
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, void>
 uninitialized_value_construct(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   __pstl::__internal::__invoke_if_else(
       std::is_trivial<_ValueType>(),
       [&]() {
         __pstl::__internal::__pattern_walk_brick(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __last,
             [](_ForwardIterator __begin, _ForwardIterator __end) {
               __pstl::__internal::__brick_fill(__begin, __end, _ValueType(), __is_vector{});
             });
       },
       [&]() {
         __pstl::__internal::__pattern_walk1(
             __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, [](_ReferenceType __val) {
               ::new (std::addressof(__val)) _ValueType();
             });
       });
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Size>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator>
 uninitialized_value_construct_n(_ExecutionPolicy&& __exec, _ForwardIterator __first, _Size __n) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   typedef typename iterator_traits<_ForwardIterator>::reference _ReferenceType;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
 
   using __is_vector = typename decltype(__dispatch_tag)::__is_vector;
 
   return __pstl::__internal::__invoke_if_else(
       std::is_trivial<_ValueType>(),
       [&]() {
         return __pstl::__internal::__pattern_walk_brick_n(
             __dispatch_tag,
             std::forward<_ExecutionPolicy>(__exec),
             __first,
             __n,
             [](_ForwardIterator __begin, _Size __count) {
               return __pstl::__internal::__brick_fill_n(__begin, __count, _ValueType(), __is_vector{});
             });
       },
       [&]() {
         return __pstl::__internal::__pattern_walk1_n(
             __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __n, [](_ReferenceType __val) {
               ::new (std::addressof(__val)) _ValueType();
             });
       });
 }
 
 } // namespace std
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_GLUE_MEMORY_IMPL_H */

libcxx/include/__pstl/internal/glue_numeric_defs.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_NUMERIC_DEFS_H
 #define _PSTL_GLUE_NUMERIC_DEFS_H
 
 #include <iterator>
 
 #include "execution_defs.h"
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace std {
 // [reduce]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp>
 reduce(_ExecutionPolicy&& __exec,
        _ForwardIterator __first,
        _ForwardIterator __last,
        _Tp __init,
        _BinaryOperation __binary_op);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp>
 reduce(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Tp __init);
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy,
                                                  typename iterator_traits<_ForwardIterator>::value_type>
 reduce(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp> transform_reduce(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init);
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp> transform_reduce(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init,
     _BinaryOperation1 __binary_op1,
     _BinaryOperation2 __binary_op2);
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp, class _BinaryOperation, class _UnaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp> transform_reduce(
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op);
 
 // [exclusive.scan]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> exclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Tp __init);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> exclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Tp __init,
     _BinaryOperation __binary_op);
 
 // [inclusive.scan]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> inclusive_scan(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __result);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op,
     _Tp __init);
 
 // [transform.exclusive.scan]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _Tp,
           class _BinaryOperation,
           class _UnaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> transform_exclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op);
 
 // [transform.inclusive.scan]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _BinaryOperation,
           class _UnaryOperation,
           class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> transform_inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op,
     _Tp __init);
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _UnaryOperation,
           class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> transform_inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op);
 
 // [adjacent.difference]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> adjacent_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __d_first,
     _BinaryOperation op);
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> adjacent_difference(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __d_first);
 
 } // namespace std
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_GLUE_NUMERIC_DEFS_H */

libcxx/include/__pstl/internal/glue_numeric_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_GLUE_NUMERIC_IMPL_H
 #define _PSTL_GLUE_NUMERIC_IMPL_H
 
 #include <functional>
 
 #include "pstl_config.h"
 
 #include "execution_impl.h"
 #include "numeric_fwd.h"
 #include "utils.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace std {
 
 // [reduce]
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp>
 reduce(_ExecutionPolicy&& __exec,
        _ForwardIterator __first,
        _ForwardIterator __last,
        _Tp __init,
        _BinaryOperation __binary_op) {
   return transform_reduce(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __init, __binary_op, __pstl::__internal::__no_op());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp>
 reduce(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last, _Tp __init) {
   return transform_reduce(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __init, std::plus<_Tp>(), __pstl::__internal::__no_op());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy,
                                                  typename iterator_traits<_ForwardIterator>::value_type>
 reduce(_ExecutionPolicy&& __exec, _ForwardIterator __first, _ForwardIterator __last) {
   typedef typename iterator_traits<_ForwardIterator>::value_type _ValueType;
   return transform_reduce(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       _ValueType{},
       std::plus<_ValueType>(),
       __pstl::__internal::__no_op());
 }
 
 // [transform.reduce]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp> transform_reduce(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
 
   typedef typename iterator_traits<_ForwardIterator1>::value_type _InputType;
   return __pstl::__internal::__pattern_transform_reduce(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __init,
       std::plus<_InputType>(),
       std::multiplies<_InputType>());
 }
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp> transform_reduce(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init,
     _BinaryOperation1 __binary_op1,
     _BinaryOperation2 __binary_op2) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first1, __first2);
   return __pstl::__internal::__pattern_transform_reduce(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first1,
       __last1,
       __first2,
       __init,
       __binary_op1,
       __binary_op2);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Tp, class _BinaryOperation, class _UnaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _Tp> transform_reduce(
     _ExecutionPolicy&& __exec,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first);
   return __pstl::__internal::__pattern_transform_reduce(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __init, __binary_op, __unary_op);
 }
 
 // [exclusive.scan]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> exclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Tp __init) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using namespace __pstl;
   return __internal::__pattern_transform_scan(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __pstl::__internal::__no_op(),
       __init,
       std::plus<_Tp>(),
       /*inclusive=*/std::false_type());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> exclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Tp __init,
     _BinaryOperation __binary_op) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   using namespace __pstl;
   return __internal::__pattern_transform_scan(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __pstl::__internal::__no_op(),
       __init,
       __binary_op,
       /*inclusive=*/std::false_type());
 }
 
 // [inclusive.scan]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> inclusive_scan(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __result) {
   typedef typename iterator_traits<_ForwardIterator1>::value_type _InputType;
   return transform_inclusive_scan(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       std::plus<_InputType>(),
       __pstl::__internal::__no_op());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op) {
   return transform_inclusive_scan(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __result, __binary_op, __pstl::__internal::__no_op());
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _Tp, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op,
     _Tp __init) {
   return transform_inclusive_scan(
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __binary_op,
       __pstl::__internal::__no_op(),
       __init);
 }
 
 // [transform.exclusive.scan]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _Tp,
           class _BinaryOperation,
           class _UnaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> transform_exclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_transform_scan(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __unary_op,
       __init,
       __binary_op,
       /*inclusive=*/std::false_type());
 }
 
 // [transform.inclusive.scan]
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _BinaryOperation,
           class _UnaryOperation,
           class _Tp>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> transform_inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op,
     _Tp __init) {
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __result);
 
   return __pstl::__internal::__pattern_transform_scan(
       __dispatch_tag,
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last,
       __result,
       __unary_op,
       __init,
       __binary_op,
       /*inclusive=*/std::true_type());
 }
 
 template <class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _UnaryOperation,
           class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> transform_inclusive_scan(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __result,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op) {
   if (__first != __last) {
     auto __tmp = __unary_op(*__first);
     *__result  = __tmp;
     return transform_inclusive_scan(
         std::forward<_ExecutionPolicy>(__exec), ++__first, __last, ++__result, __binary_op, __unary_op, __tmp);
   } else {
     return __result;
   }
 }
 
 // [adjacent.difference]
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryOperation>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> adjacent_difference(
     _ExecutionPolicy&& __exec,
     _ForwardIterator1 __first,
     _ForwardIterator1 __last,
     _ForwardIterator2 __d_first,
     _BinaryOperation __op) {
   if (__first == __last)
     return __d_first;
 
   auto __dispatch_tag = __pstl::__internal::__select_backend(__exec, __first, __d_first);
 
   return __pstl::__internal::__pattern_adjacent_difference(
       __dispatch_tag, std::forward<_ExecutionPolicy>(__exec), __first, __last, __d_first, __op);
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator1, class _ForwardIterator2>
 __pstl::__internal::__enable_if_execution_policy<_ExecutionPolicy, _ForwardIterator2> adjacent_difference(
     _ExecutionPolicy&& __exec, _ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __d_first) {
   typedef typename iterator_traits<_ForwardIterator1>::value_type _ValueType;
   return adjacent_difference(
       std::forward<_ExecutionPolicy>(__exec), __first, __last, __d_first, std::minus<_ValueType>());
 }
 
 } // namespace std
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_GLUE_NUMERIC_IMPL_H_ */

libcxx/include/__pstl/internal/memory_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_MEMORY_IMPL_H
 #define _PSTL_MEMORY_IMPL_H
 
 #include <iterator>
 
 #include "pstl_config.h"
 #include "unseq_backend_simd.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 //------------------------------------------------------------------------
 // uninitialized_move
 //------------------------------------------------------------------------
 
 template <typename _ForwardIterator, typename _OutputIterator>
 _OutputIterator __brick_uninitialized_move(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     /*vector=*/std::false_type) noexcept {
   using _ValueType = typename std::iterator_traits<_OutputIterator>::value_type;
   for (; __first != __last; ++__first, ++__result) {
     ::new (std::addressof(*__result)) _ValueType(std::move(*__first));
   }
   return __result;
 }
 
 template <typename _RandomAccessIterator, typename _OutputIterator>
 _OutputIterator __brick_uninitialized_move(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     /*vector=*/std::true_type) noexcept {
   using __ValueType     = typename std::iterator_traits<_OutputIterator>::value_type;
   using _ReferenceType1 = typename std::iterator_traits<_RandomAccessIterator>::reference;
   using _ReferenceType2 = typename std::iterator_traits<_OutputIterator>::reference;
 
   return __unseq_backend::__simd_walk_2(
       __first, __last - __first, __result, [](_ReferenceType1 __x, _ReferenceType2 __y) {
         ::new (std::addressof(__y)) __ValueType(std::move(__x));
       });
 }
 
 template <typename _Iterator>
 void __brick_destroy(_Iterator __first, _Iterator __last, /*vector*/ std::false_type) noexcept {
   using _ValueType = typename std::iterator_traits<_Iterator>::value_type;
 
   for (; __first != __last; ++__first)
     __first->~_ValueType();
 }
 
 template <typename _RandomAccessIterator>
 void __brick_destroy(_RandomAccessIterator __first, _RandomAccessIterator __last, /*vector*/ std::true_type) noexcept {
   using _ValueType     = typename std::iterator_traits<_RandomAccessIterator>::value_type;
   using _ReferenceType = typename std::iterator_traits<_RandomAccessIterator>::reference;
 
   __unseq_backend::__simd_walk_1(__first, __last - __first, [](_ReferenceType __x) { __x.~_ValueType(); });
 }
 
 //------------------------------------------------------------------------
 // uninitialized copy
 //------------------------------------------------------------------------
 
 template <typename _ForwardIterator, typename _OutputIterator>
 _OutputIterator __brick_uninitialized_copy(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     /*vector=*/std::false_type) noexcept {
   using _ValueType = typename std::iterator_traits<_OutputIterator>::value_type;
   for (; __first != __last; ++__first, ++__result) {
     ::new (std::addressof(*__result)) _ValueType(*__first);
   }
   return __result;
 }
 
 template <typename _RandomAccessIterator, typename _OutputIterator>
 _OutputIterator __brick_uninitialized_copy(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     /*vector=*/std::true_type) noexcept {
   using __ValueType     = typename std::iterator_traits<_OutputIterator>::value_type;
   using _ReferenceType1 = typename std::iterator_traits<_RandomAccessIterator>::reference;
   using _ReferenceType2 = typename std::iterator_traits<_OutputIterator>::reference;
 
   return __unseq_backend::__simd_walk_2(
       __first, __last - __first, __result, [](_ReferenceType1 __x, _ReferenceType2 __y) {
         ::new (std::addressof(__y)) __ValueType(__x);
       });
 }
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_MEMORY_IMPL_H */

libcxx/include/__pstl/internal/numeric_fwd.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_NUMERIC_FWD_H
 #define _PSTL_NUMERIC_FWD_H
 
 #include <type_traits>
 #include <utility>
 
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 //------------------------------------------------------------------------
 // transform_reduce (version with two binary functions, according to draft N4659)
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 _Tp __brick_transform_reduce(
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _Tp,
     _BinaryOperation1,
     _BinaryOperation2,
     /*__is_vector=*/std::true_type) noexcept;
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Tp, class _BinaryOperation1, class _BinaryOperation2>
 _Tp __brick_transform_reduce(
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _Tp,
     _BinaryOperation1,
     _BinaryOperation2,
     /*__is_vector=*/std::false_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 _Tp __pattern_transform_reduce(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1,
     _ForwardIterator1,
     _ForwardIterator2,
     _Tp,
     _BinaryOperation1,
     _BinaryOperation2) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 _Tp __pattern_transform_reduce(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator1,
     _RandomAccessIterator1,
     _RandomAccessIterator2,
     _Tp,
     _BinaryOperation1,
     _BinaryOperation2);
 
 //------------------------------------------------------------------------
 // transform_reduce (version with unary and binary functions)
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _Tp, class _UnaryOperation, class _BinaryOperation>
 _Tp __brick_transform_reduce(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Tp,
     _BinaryOperation,
     _UnaryOperation,
     /*is_vector=*/std::true_type) noexcept;
 
 template <class _ForwardIterator, class _Tp, class _BinaryOperation, class _UnaryOperation>
 _Tp __brick_transform_reduce(
     _ForwardIterator,
     _ForwardIterator,
     _Tp,
     _BinaryOperation,
     _UnaryOperation,
     /*is_vector=*/std::false_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator,
           class _Tp,
           class _BinaryOperation,
           class _UnaryOperation>
 _Tp __pattern_transform_reduce(
     _Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _Tp, _BinaryOperation, _UnaryOperation) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _Tp,
           class _BinaryOperation,
           class _UnaryOperation>
 _Tp __pattern_transform_reduce(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _Tp,
     _BinaryOperation,
     _UnaryOperation);
 
 //------------------------------------------------------------------------
 // transform_exclusive_scan
 //
 // walk3 evaluates f(x,y,z) for (x,y,z) drawn from [first1,last1), [first2,...), [first3,...)
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator, class _UnaryOperation, class _Tp, class _BinaryOperation>
 std::pair<_OutputIterator, _Tp> __brick_transform_scan(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     _UnaryOperation,
     _Tp,
     _BinaryOperation,
     /*Inclusive*/ std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator, class _UnaryOperation, class _Tp, class _BinaryOperation>
 std::pair<_OutputIterator, _Tp> __brick_transform_scan(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _UnaryOperation,
     _Tp,
     _BinaryOperation,
     /*Inclusive*/ std::true_type) noexcept;
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 _OutputIterator __pattern_transform_scan(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     _UnaryOperation,
     _Tp,
     _BinaryOperation,
     _Inclusive) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 typename std::enable_if<!std::is_floating_point<_Tp>::value, _OutputIterator>::type __pattern_transform_scan(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _UnaryOperation,
     _Tp,
     _BinaryOperation,
     _Inclusive);
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 typename std::enable_if<std::is_floating_point<_Tp>::value, _OutputIterator>::type __pattern_transform_scan(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _UnaryOperation,
     _Tp,
     _BinaryOperation,
     _Inclusive);
 
 //------------------------------------------------------------------------
 // adjacent_difference
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator, class _BinaryOperation>
 _OutputIterator __brick_adjacent_difference(
     _ForwardIterator,
     _ForwardIterator,
     _OutputIterator,
     _BinaryOperation,
     /*is_vector*/ std::false_type) noexcept;
 
 template <class _RandomAccessIterator, class _OutputIterator, class _BinaryOperation>
 _OutputIterator __brick_adjacent_difference(
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _BinaryOperation,
     /*is_vector*/ std::true_type) noexcept;
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator, class _BinaryOperation>
 _OutputIterator __pattern_adjacent_difference(
     _Tag, _ExecutionPolicy&&, _ForwardIterator, _ForwardIterator, _OutputIterator, _BinaryOperation) noexcept;
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _BinaryOperation>
 _OutputIterator __pattern_adjacent_difference(
     __parallel_tag<_IsVector>,
     _ExecutionPolicy&&,
     _RandomAccessIterator,
     _RandomAccessIterator,
     _OutputIterator,
     _BinaryOperation);
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_NUMERIC_FWD_H */

libcxx/include/__pstl/internal/numeric_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_NUMERIC_IMPL_H
 #define _PSTL_NUMERIC_IMPL_H
 
 #include <iterator>
 #include <type_traits>
 #include <numeric>
 
 #include "parallel_backend.h"
 #include "pstl_config.h"
 #include "execution_impl.h"
 #include "unseq_backend_simd.h"
 #include "algorithm_fwd.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 //------------------------------------------------------------------------
 // transform_reduce (version with two binary functions, according to draft N4659)
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _Tp, class _BinaryOperation1, class _BinaryOperation2>
 _Tp __brick_transform_reduce(
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init,
     _BinaryOperation1 __binary_op1,
     _BinaryOperation2 __binary_op2,
     /*is_vector=*/std::false_type) noexcept {
   return std::inner_product(__first1, __last1, __first2, __init, __binary_op1, __binary_op2);
 }
 
 template <class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 _Tp __brick_transform_reduce(
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _Tp __init,
     _BinaryOperation1 __binary_op1,
     _BinaryOperation2 __binary_op2,
     /*is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType;
   return __unseq_backend::__simd_transform_reduce(
       __last1 - __first1, __init, __binary_op1, [=, &__binary_op2](_DifferenceType __i) {
         return __binary_op2(__first1[__i], __first2[__i]);
       });
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator1,
           class _ForwardIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 _Tp __pattern_transform_reduce(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator1 __first1,
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init,
     _BinaryOperation1 __binary_op1,
     _BinaryOperation2 __binary_op2) noexcept {
   return __brick_transform_reduce(
       __first1, __last1, __first2, __init, __binary_op1, __binary_op2, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _Tp,
           class _BinaryOperation1,
           class _BinaryOperation2>
 _Tp __pattern_transform_reduce(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first1,
     _RandomAccessIterator1 __last1,
     _RandomAccessIterator2 __first2,
     _Tp __init,
     _BinaryOperation1 __binary_op1,
     _BinaryOperation2 __binary_op2) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __par_backend::__parallel_transform_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first1,
         __last1,
         [__first1, __first2, __binary_op2](_RandomAccessIterator1 __i) mutable {
           return __binary_op2(*__i, *(__first2 + (__i - __first1)));
         },
         __init,
         __binary_op1, // Combine
         [__first1, __first2, __binary_op1, __binary_op2](
             _RandomAccessIterator1 __i, _RandomAccessIterator1 __j, _Tp __init) -> _Tp {
           return __internal::__brick_transform_reduce(
               __i, __j, __first2 + (__i - __first1), __init, __binary_op1, __binary_op2, _IsVector{});
         });
   });
 }
 
 //------------------------------------------------------------------------
 // transform_reduce (version with unary and binary functions)
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _Tp, class _BinaryOperation, class _UnaryOperation>
 _Tp __brick_transform_reduce(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op,
     /*is_vector=*/std::false_type) noexcept {
   return std::transform_reduce(__first, __last, __init, __binary_op, __unary_op);
 }
 
 template <class _RandomAccessIterator, class _Tp, class _UnaryOperation, class _BinaryOperation>
 _Tp __brick_transform_reduce(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op,
     /*is_vector=*/std::true_type) noexcept {
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _DifferenceType;
   return __unseq_backend::__simd_transform_reduce(
       __last - __first, __init, __binary_op, [=, &__unary_op](_DifferenceType __i) {
         return __unary_op(__first[__i]);
       });
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator,
           class _Tp,
           class _BinaryOperation,
           class _UnaryOperation>
 _Tp __pattern_transform_reduce(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op) noexcept {
   return __internal::__brick_transform_reduce(
       __first, __last, __init, __binary_op, __unary_op, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _Tp,
           class _BinaryOperation,
           class _UnaryOperation>
 _Tp __pattern_transform_reduce(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _Tp __init,
     _BinaryOperation __binary_op,
     _UnaryOperation __unary_op) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   return __internal::__except_handler([&]() {
     return __par_backend::__parallel_transform_reduce(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __first,
         __last,
         [__unary_op](_RandomAccessIterator __i) mutable { return __unary_op(*__i); },
         __init,
         __binary_op,
         [__unary_op, __binary_op](_RandomAccessIterator __i, _RandomAccessIterator __j, _Tp __init) {
           return __internal::__brick_transform_reduce(__i, __j, __init, __binary_op, __unary_op, _IsVector{});
         });
   });
 }
 
 //------------------------------------------------------------------------
 // transform_exclusive_scan
 //
 // walk3 evaluates f(x,y,z) for (x,y,z) drawn from [first1,last1), [first2,...), [first3,...)
 //------------------------------------------------------------------------
 
 // Exclusive form
 template <class _ForwardIterator, class _OutputIterator, class _UnaryOperation, class _Tp, class _BinaryOperation>
 std::pair<_OutputIterator, _Tp> __brick_transform_scan(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     /*Inclusive*/ std::false_type,
     /*is_vector=*/std::false_type) noexcept {
   for (; __first != __last; ++__first, ++__result) {
     *__result = __init;
     _PSTL_PRAGMA_FORCEINLINE
     __init = __binary_op(__init, __unary_op(*__first));
   }
   return std::make_pair(__result, __init);
 }
 
 // Inclusive form
 template <class _RandomAccessIterator, class _OutputIterator, class _UnaryOperation, class _Tp, class _BinaryOperation>
 std::pair<_OutputIterator, _Tp> __brick_transform_scan(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     /*Inclusive*/ std::true_type,
     /*is_vector=*/std::false_type) noexcept {
   for (; __first != __last; ++__first, ++__result) {
     _PSTL_PRAGMA_FORCEINLINE
     __init    = __binary_op(__init, __unary_op(*__first));
     *__result = __init;
   }
   return std::make_pair(__result, __init);
 }
 
 // type is arithmetic and binary operation is a user defined operation.
 template <typename _Tp, typename _BinaryOperation>
 using is_arithmetic_udop =
     std::integral_constant<bool,
                            std::is_arithmetic<_Tp>::value && !std::is_same<_BinaryOperation, std::plus<_Tp>>::value>;
 
 // [restriction] - T shall be DefaultConstructible.
 // [violation] - default ctor of T shall set the identity value for binary_op.
 template <class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 typename std::enable_if<!is_arithmetic_udop<_Tp, _BinaryOperation>::value, std::pair<_OutputIterator, _Tp>>::type
 __brick_transform_scan(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     _Inclusive,
     /*is_vector=*/std::true_type) noexcept {
 #if defined(_PSTL_UDS_PRESENT)
   return __unseq_backend::__simd_scan(
       __first, __last - __first, __result, __unary_op, __init, __binary_op, _Inclusive());
 #else
   // We need to call serial brick here to call function for inclusive and exclusive scan that depends on _Inclusive()
   // value
   return __internal::__brick_transform_scan(
       __first,
       __last,
       __result,
       __unary_op,
       __init,
       __binary_op,
       _Inclusive(),
       /*is_vector=*/std::false_type());
 #endif
 }
 
 template <class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 typename std::enable_if<is_arithmetic_udop<_Tp, _BinaryOperation>::value, std::pair<_OutputIterator, _Tp>>::type
 __brick_transform_scan(
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     _Inclusive,
     /*is_vector=*/std::true_type) noexcept {
   return __internal::__brick_transform_scan(
       __first,
       __last,
       __result,
       __unary_op,
       __init,
       __binary_op,
       _Inclusive(),
       /*is_vector=*/std::false_type());
 }
 
 template <class _Tag,
           class _ExecutionPolicy,
           class _ForwardIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 _OutputIterator __pattern_transform_scan(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     _Inclusive) noexcept {
   return __internal::__brick_transform_scan(
              __first, __last, __result, __unary_op, __init, __binary_op, _Inclusive(), typename _Tag::__is_vector{})
       .first;
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 typename std::enable_if<!std::is_floating_point<_Tp>::value, _OutputIterator>::type __pattern_transform_scan(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     _Inclusive) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _DifferenceType;
 
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_transform_scan(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __last - __first,
         [__first, __unary_op](_DifferenceType __i) mutable { return __unary_op(__first[__i]); },
         __init,
         __binary_op,
         [__first, __unary_op, __binary_op](_DifferenceType __i, _DifferenceType __j, _Tp __init) {
           // Execute serial __brick_transform_reduce, due to the explicit SIMD vectorization (reduction) requires a
           // commutative operation for the guarantee of correct scan.
           return __internal::__brick_transform_reduce(
               __first + __i,
               __first + __j,
               __init,
               __binary_op,
               __unary_op,
               /*__is_vector*/ std::false_type());
         },
         [__first, __unary_op, __binary_op, __result](_DifferenceType __i, _DifferenceType __j, _Tp __init) {
           return __internal::__brick_transform_scan(
                      __first + __i,
                      __first + __j,
                      __result + __i,
                      __unary_op,
                      __init,
                      __binary_op,
                      _Inclusive(),
                      _IsVector{})
               .second;
         });
     return __result + (__last - __first);
   });
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator,
           class _OutputIterator,
           class _UnaryOperation,
           class _Tp,
           class _BinaryOperation,
           class _Inclusive>
 typename std::enable_if<std::is_floating_point<_Tp>::value, _OutputIterator>::type __pattern_transform_scan(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator __first,
     _RandomAccessIterator __last,
     _OutputIterator __result,
     _UnaryOperation __unary_op,
     _Tp __init,
     _BinaryOperation __binary_op,
     _Inclusive) {
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _DifferenceType;
   _DifferenceType __n = __last - __first;
 
   if (__n <= 0) {
     return __result;
   }
   return __internal::__except_handler([&]() {
     __par_backend::__parallel_strict_scan(
         __backend_tag{},
         std::forward<_ExecutionPolicy>(__exec),
         __n,
         __init,
         [__first, __unary_op, __binary_op, __result](_DifferenceType __i, _DifferenceType __len) {
           return __internal::__brick_transform_scan(
                      __first + __i,
                      __first + (__i + __len),
                      __result + __i,
                      __unary_op,
                      _Tp{},
                      __binary_op,
                      _Inclusive(),
                      _IsVector{})
               .second;
         },
         __binary_op,
         [__result, &__binary_op](_DifferenceType __i, _DifferenceType __len, _Tp __initial) {
           return *(std::transform(__result + __i,
                                   __result + __i + __len,
                                   __result + __i,
                                   [&__initial, &__binary_op](const _Tp& __x) {
                                     _PSTL_PRAGMA_FORCEINLINE
                                     return __binary_op(__initial, __x);
                                   }) -
                    1);
         },
         [](_Tp) {});
     return __result + (__last - __first);
   });
 }
 
 //------------------------------------------------------------------------
 // adjacent_difference
 //------------------------------------------------------------------------
 
 template <class _ForwardIterator, class _OutputIterator, class _BinaryOperation>
 _OutputIterator __brick_adjacent_difference(
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __d_first,
     _BinaryOperation __op,
     /*is_vector*/ std::false_type) noexcept {
   return std::adjacent_difference(__first, __last, __d_first, __op);
 }
 
 template <class _RandomAccessIterator1, class _RandomAccessIterator2, class BinaryOperation>
 _RandomAccessIterator2 __brick_adjacent_difference(
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __d_first,
     BinaryOperation __op,
     /*is_vector=*/std::true_type) noexcept {
   _PSTL_ASSERT(__first != __last);
 
   typedef typename std::iterator_traits<_RandomAccessIterator1>::reference _ReferenceType1;
   typedef typename std::iterator_traits<_RandomAccessIterator2>::reference _ReferenceType2;
 
   auto __n   = __last - __first;
   *__d_first = *__first;
   return __unseq_backend::__simd_walk_3(
       __first + 1,
       __n - 1,
       __first,
       __d_first + 1,
       [&__op](_ReferenceType1 __x, _ReferenceType1 __y, _ReferenceType2 __z) { __z = __op(__x, __y); });
 }
 
 template <class _Tag, class _ExecutionPolicy, class _ForwardIterator, class _OutputIterator, class _BinaryOperation>
 _OutputIterator __pattern_adjacent_difference(
     _Tag,
     _ExecutionPolicy&&,
     _ForwardIterator __first,
     _ForwardIterator __last,
     _OutputIterator __d_first,
     _BinaryOperation __op) noexcept {
   return __internal::__brick_adjacent_difference(__first, __last, __d_first, __op, typename _Tag::__is_vector{});
 }
 
 template <class _IsVector,
           class _ExecutionPolicy,
           class _RandomAccessIterator1,
           class _RandomAccessIterator2,
           class _BinaryOperation>
 _RandomAccessIterator2 __pattern_adjacent_difference(
     __parallel_tag<_IsVector> __tag,
     _ExecutionPolicy&& __exec,
     _RandomAccessIterator1 __first,
     _RandomAccessIterator1 __last,
     _RandomAccessIterator2 __d_first,
     _BinaryOperation __op) {
   _PSTL_ASSERT(__first != __last);
   typedef typename std::iterator_traits<_RandomAccessIterator1>::reference _ReferenceType1;
   typedef typename std::iterator_traits<_RandomAccessIterator2>::reference _ReferenceType2;
 
   using __backend_tag = typename decltype(__tag)::__backend_tag;
 
   *__d_first = *__first;
   __par_backend::__parallel_for(
       __backend_tag{},
       std::forward<_ExecutionPolicy>(__exec),
       __first,
       __last - 1,
       [&__op, __d_first, __first](_RandomAccessIterator1 __b, _RandomAccessIterator1 __e) {
         _RandomAccessIterator2 __d_b = __d_first + (__b - __first);
         __internal::__brick_walk3(
             __b,
             __e,
             __b + 1,
             __d_b + 1,
             [&__op](_ReferenceType1 __x, _ReferenceType1 __y, _ReferenceType2 __z) { __z = __op(__y, __x); },
             _IsVector{});
       });
   return __d_first + (__last - __first);
 }
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_NUMERIC_IMPL_H */

libcxx/include/__pstl/internal/omp/parallel_for.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_FOR_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_FOR_H
 
 #include <cstddef>
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <class _Index, class _Fp>
 void
 __parallel_for_body(_Index __first, _Index __last, _Fp __f)
 {
     // initial partition of the iteration space into chunks
     auto __policy = __omp_backend::__chunk_partitioner(__first, __last);
 
     // To avoid over-subscription we use taskloop for the nested parallelism
     _PSTL_PRAGMA(omp taskloop untied mergeable)
     for (std::size_t __chunk = 0; __chunk < __policy.__n_chunks; ++__chunk)
     {
         __pstl::__omp_backend::__process_chunk(__policy, __first, __chunk, __f);
     }
 }
 
 //------------------------------------------------------------------------
 // Notation:
 // Evaluation of brick f[i,j) for each subrange [i,j) of [first, last)
 //------------------------------------------------------------------------
 
 template <class _ExecutionPolicy, class _Index, class _Fp>
 void
 __parallel_for(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last, _Fp __f)
 {
     if (omp_in_parallel())
     {
         // we don't create a nested parallel region in an existing parallel
         // region: just create tasks
         __pstl::__omp_backend::__parallel_for_body(__first, __last, __f);
     }
     else
     {
         // in any case (nested or non-nested) one parallel region is created and
         // only one thread creates a set of tasks
         _PSTL_PRAGMA(omp parallel)
         _PSTL_PRAGMA(omp single nowait) { __pstl::__omp_backend::__parallel_for_body(__first, __last, __f); }
     }
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_FOR_H

libcxx/include/__pstl/internal/omp/parallel_for_each.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_FOR_EACH_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_FOR_EACH_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <class _ForwardIterator, class _Fp>
 void
 __parallel_for_each_body(_ForwardIterator __first, _ForwardIterator __last, _Fp __f)
 {
     using DifferenceType = typename std::iterator_traits<_ForwardIterator>::difference_type;
     // TODO: Think of an approach to remove the std::distance call
     auto __size = std::distance(__first, __last);
 
     _PSTL_PRAGMA(omp taskloop untied mergeable)
     for (DifferenceType __index = 0; __index < __size; ++__index)
     {
         // TODO: Think of an approach to remove the increment here each time.
         auto __iter = std::next(__first, __index);
         __f(*__iter);
     }
 }
 
 template <class _ExecutionPolicy, class _ForwardIterator, class _Fp>
 void
 __parallel_for_each(_ExecutionPolicy&&, _ForwardIterator __first, _ForwardIterator __last, _Fp __f)
 {
     if (omp_in_parallel())
     {
         // we don't create a nested parallel region in an existing parallel
         // region: just create tasks
         __pstl::__omp_backend::__parallel_for_each_body(__first, __last, __f);
     }
     else
     {
         // in any case (nested or non-nested) one parallel region is created and
         // only one thread creates a set of tasks
         _PSTL_PRAGMA(omp parallel)
         _PSTL_PRAGMA(omp single nowait) { __pstl::__omp_backend::__parallel_for_each_body(__first, __last, __f); }
     }
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_FOR_EACH_H

libcxx/include/__pstl/internal/omp/parallel_invoke.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_INVOKE_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_INVOKE_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <typename _F1, typename _F2>
 void
 __parallel_invoke_body(_F1&& __f1, _F2&& __f2)
 {
     _PSTL_PRAGMA(omp taskgroup)
     {
         _PSTL_PRAGMA(omp task untied mergeable) { std::forward<_F1>(__f1)(); }
         _PSTL_PRAGMA(omp task untied mergeable) { std::forward<_F2>(__f2)(); }
     }
 }
 
 template <class _ExecutionPolicy, typename _F1, typename _F2>
 void
 __parallel_invoke(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&&, _F1&& __f1, _F2&& __f2)
 {
     if (omp_in_parallel())
     {
         __pstl::__omp_backend::__parallel_invoke_body(std::forward<_F1>(__f1), std::forward<_F2>(__f2));
     }
     else
     {
         _PSTL_PRAGMA(omp parallel)
         _PSTL_PRAGMA(omp single nowait)
         __pstl::__omp_backend::__parallel_invoke_body(std::forward<_F1>(__f1), std::forward<_F2>(__f2));
     }
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_INVOKE_H

libcxx/include/__pstl/internal/omp/parallel_merge.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_MERGE_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_MERGE_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename _RandomAccessIterator3,
           typename _Compare, typename _LeafMerge>
 void
 __parallel_merge_body(std::size_t __size_x, std::size_t __size_y, _RandomAccessIterator1 __xs,
                       _RandomAccessIterator1 __xe, _RandomAccessIterator2 __ys, _RandomAccessIterator2 __ye,
                       _RandomAccessIterator3 __zs, _Compare __comp, _LeafMerge __leaf_merge)
 {
 
     if (__size_x + __size_y <= __omp_backend::__default_chunk_size)
     {
         __leaf_merge(__xs, __xe, __ys, __ye, __zs, __comp);
         return;
     }
 
     _RandomAccessIterator1 __xm;
     _RandomAccessIterator2 __ym;
 
     if (__size_x < __size_y)
     {
         __ym = __ys + (__size_y / 2);
         __xm = std::upper_bound(__xs, __xe, *__ym, __comp);
     }
     else
     {
         __xm = __xs + (__size_x / 2);
         __ym = std::lower_bound(__ys, __ye, *__xm, __comp);
     }
 
     auto __zm = __zs + (__xm - __xs) + (__ym - __ys);
 
     _PSTL_PRAGMA(omp task untied mergeable default(none)
                      firstprivate(__xs, __xm, __ys, __ym, __zs, __comp, __leaf_merge))
     __pstl::__omp_backend::__parallel_merge_body(__xm - __xs, __ym - __ys, __xs, __xm, __ys, __ym, __zs, __comp,
                                                       __leaf_merge);
 
     _PSTL_PRAGMA(omp task untied mergeable default(none)
                      firstprivate(__xm, __xe, __ym, __ye, __zm, __comp, __leaf_merge))
     __pstl::__omp_backend::__parallel_merge_body(__xe - __xm, __ye - __ym, __xm, __xe, __ym, __ye, __zm, __comp,
                                                       __leaf_merge);
 
     _PSTL_PRAGMA(omp taskwait)
 }
 
 template <class _ExecutionPolicy, typename _RandomAccessIterator1, typename _RandomAccessIterator2,
           typename _RandomAccessIterator3, typename _Compare, typename _LeafMerge>
 void
 __parallel_merge(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&& /*__exec*/, _RandomAccessIterator1 __xs,
                  _RandomAccessIterator1 __xe, _RandomAccessIterator2 __ys, _RandomAccessIterator2 __ye,
                  _RandomAccessIterator3 __zs, _Compare __comp, _LeafMerge __leaf_merge)
 
 {
     std::size_t __size_x = __xe - __xs;
     std::size_t __size_y = __ye - __ys;
 
     /*
      * Run the merge in parallel by chunking it up. Use the smaller range (if any) as the iteration range, and the
      * larger range as the search range.
      */
 
     if (omp_in_parallel())
     {
         __pstl::__omp_backend::__parallel_merge_body(__size_x, __size_y, __xs, __xe, __ys, __ye, __zs, __comp,
                                                           __leaf_merge);
     }
     else
     {
         _PSTL_PRAGMA(omp parallel)
         {
             _PSTL_PRAGMA(omp single nowait)
             __pstl::__omp_backend::__parallel_merge_body(__size_x, __size_y, __xs, __xe, __ys, __ye, __zs, __comp,
                                                               __leaf_merge);
         }
     }
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_MERGE_H

libcxx/include/__pstl/internal/omp/parallel_reduce.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_REDUCE_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_REDUCE_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <class _RandomAccessIterator, class _Value, typename _RealBody, typename _Reduction>
 _Value
 __parallel_reduce_body(_RandomAccessIterator __first, _RandomAccessIterator __last, _Value __identity,
                        _RealBody __real_body, _Reduction __reduce)
 {
     if (__should_run_serial(__first, __last))
     {
         return __real_body(__first, __last, __identity);
     }
 
     auto __middle = __first + ((__last - __first) / 2);
     _Value __v1(__identity), __v2(__identity);
     __parallel_invoke_body(
         [&]() { __v1 = __parallel_reduce_body(__first, __middle, __identity, __real_body, __reduce); },
         [&]() { __v2 = __parallel_reduce_body(__middle, __last, __identity, __real_body, __reduce); });
 
     return __reduce(__v1, __v2);
 }
 
 //------------------------------------------------------------------------
 // Notation:
 //      r(i,j,init) returns reduction of init with reduction over [i,j)
 //      c(x,y) combines values x and y that were the result of r
 //------------------------------------------------------------------------
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _Value, typename _RealBody, typename _Reduction>
 _Value
 __parallel_reduce(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&&, _RandomAccessIterator __first,
                   _RandomAccessIterator __last, _Value __identity, _RealBody __real_body, _Reduction __reduction)
 {
     // We don't create a nested parallel region in an existing parallel region:
     // just create tasks.
     if (omp_in_parallel())
     {
         return __pstl::__omp_backend::__parallel_reduce_body(__first, __last, __identity, __real_body, __reduction);
     }
 
     // In any case (nested or non-nested) one parallel region is created and only
     // one thread creates a set of tasks.
     _Value __res = __identity;
 
     _PSTL_PRAGMA(omp parallel)
     _PSTL_PRAGMA(omp single nowait)
     {
         __res = __pstl::__omp_backend::__parallel_reduce_body(__first, __last, __identity, __real_body, __reduction);
     }
 
     return __res;
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_REDUCE_H

libcxx/include/__pstl/internal/omp/parallel_scan.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_SCAN_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_SCAN_H
 
 #include "parallel_invoke.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <typename _Index>
 _Index
 __split(_Index __m)
 {
     _Index __k = 1;
     while (2 * __k < __m)
         __k *= 2;
     return __k;
 }
 
 template <typename _Index, typename _Tp, typename _Rp, typename _Cp>
 void
 __upsweep(_Index __i, _Index __m, _Index __tilesize, _Tp* __r, _Index __lastsize, _Rp __reduce, _Cp __combine)
 {
     if (__m == 1)
         __r[0] = __reduce(__i * __tilesize, __lastsize);
     else
     {
         _Index __k = __split(__m);
         __omp_backend::__parallel_invoke_body(
             [=] { __omp_backend::__upsweep(__i, __k, __tilesize, __r, __tilesize, __reduce, __combine); },
             [=] {
                 __omp_backend::__upsweep(__i + __k, __m - __k, __tilesize, __r + __k, __lastsize, __reduce, __combine);
             });
         if (__m == 2 * __k)
             __r[__m - 1] = __combine(__r[__k - 1], __r[__m - 1]);
     }
 }
 
 template <typename _Index, typename _Tp, typename _Cp, typename _Sp>
 void
 __downsweep(_Index __i, _Index __m, _Index __tilesize, _Tp* __r, _Index __lastsize, _Tp __initial, _Cp __combine,
             _Sp __scan)
 {
     if (__m == 1)
         __scan(__i * __tilesize, __lastsize, __initial);
     else
     {
         const _Index __k = __split(__m);
         __omp_backend::__parallel_invoke_body(
             [=] { __omp_backend::__downsweep(__i, __k, __tilesize, __r, __tilesize, __initial, __combine, __scan); },
             // Assumes that __combine never throws.
             // TODO: Consider adding a requirement for user functors to be constant.
             [=, &__combine]
             {
                 __omp_backend::__downsweep(__i + __k, __m - __k, __tilesize, __r + __k, __lastsize,
                                            __combine(__initial, __r[__k - 1]), __combine, __scan);
             });
     }
 }
 
 template <typename _ExecutionPolicy, typename _Index, typename _Tp, typename _Rp, typename _Cp, typename _Sp,
           typename _Ap>
 void
 __parallel_strict_scan_body(_Index __n, _Tp __initial, _Rp __reduce, _Cp __combine, _Sp __scan, _Ap __apex)
 {
     _Index __p = omp_get_num_threads();
     const _Index __slack = 4;
     _Index __tilesize = (__n - 1) / (__slack * __p) + 1;
     _Index __m = (__n - 1) / __tilesize;
     __buffer<_Tp> __buf(__m + 1);
     _Tp* __r = __buf.get();
 
     __omp_backend::__upsweep(_Index(0), _Index(__m + 1), __tilesize, __r, __n - __m * __tilesize, __reduce, __combine);
 
     std::size_t __k = __m + 1;
     _Tp __t = __r[__k - 1];
     while ((__k &= __k - 1))
     {
         __t = __combine(__r[__k - 1], __t);
     }
 
     __apex(__combine(__initial, __t));
     __omp_backend::__downsweep(_Index(0), _Index(__m + 1), __tilesize, __r, __n - __m * __tilesize, __initial,
                                __combine, __scan);
 }
 
 template <class _ExecutionPolicy, typename _Index, typename _Tp, typename _Rp, typename _Cp, typename _Sp, typename _Ap>
 void
 __parallel_strict_scan(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&&, _Index __n, _Tp __initial,
                        _Rp __reduce, _Cp __combine, _Sp __scan, _Ap __apex)
 {
     if (__n <= __default_chunk_size)
     {
         _Tp __sum = __initial;
         if (__n)
         {
             __sum = __combine(__sum, __reduce(_Index(0), __n));
         }
         __apex(__sum);
         if (__n)
         {
             __scan(_Index(0), __n, __initial);
         }
         return;
     }
 
     if (omp_in_parallel())
     {
         __pstl::__omp_backend::__parallel_strict_scan_body<_ExecutionPolicy>(__n, __initial, __reduce, __combine,
                                                                              __scan, __apex);
     }
     else
     {
         _PSTL_PRAGMA(omp parallel)
         _PSTL_PRAGMA(omp single nowait)
         {
             __pstl::__omp_backend::__parallel_strict_scan_body<_ExecutionPolicy>(__n, __initial, __reduce, __combine,
                                                                                  __scan, __apex);
         }
     }
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_SCAN_H

libcxx/include/__pstl/internal/omp/parallel_stable_partial_sort.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_STABLE_PARTIAL_SORT_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_STABLE_PARTIAL_SORT_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <typename _RandomAccessIterator, typename _Compare, typename _LeafSort>
 void
 __parallel_stable_partial_sort(__pstl::__internal::__openmp_backend_tag, _RandomAccessIterator __xs,
                                _RandomAccessIterator __xe, _Compare __comp, _LeafSort __leaf_sort,
                                std::size_t /* __nsort */)
 {
     // TODO: "Parallel partial sort needs to be implemented.");
     __leaf_sort(__xs, __xe, __comp);
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_STABLE_PARTIAL_SORT_H

libcxx/include/__pstl/internal/omp/parallel_stable_sort.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_STABLE_SORT_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_STABLE_SORT_H
 
 #include "util.h"
 #include "parallel_merge.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 namespace __sort_details
 {
 struct __move_value
 {
     template <typename _Iterator, typename _OutputIterator>
     void
     operator()(_Iterator __x, _OutputIterator __z) const
     {
         *__z = std::move(*__x);
     }
 };
 
 template <typename _RandomAccessIterator, typename _OutputIterator>
 _OutputIterator
 __parallel_move_range(_RandomAccessIterator __first1, _RandomAccessIterator __last1, _OutputIterator __d_first)
 {
     std::size_t __size = __last1 - __first1;
 
     // Perform serial moving of small chunks
 
     if (__size <= __default_chunk_size)
     {
         return std::move(__first1, __last1, __d_first);
     }
 
     // Perform parallel moving of larger chunks
     auto __policy = __pstl::__omp_backend::__chunk_partitioner(__first1, __last1);
 
     _PSTL_PRAGMA(omp taskloop)
     for (std::size_t __chunk = 0; __chunk < __policy.__n_chunks; ++__chunk)
     {
         __pstl::__omp_backend::__process_chunk(__policy, __first1, __chunk,
                                        [&](auto __chunk_first, auto __chunk_last)
                                        {
                                            auto __chunk_offset = __chunk_first - __first1;
                                            auto __output_it = __d_first + __chunk_offset;
                                            std::move(__chunk_first, __chunk_last, __output_it);
                                        });
     }
 
     return __d_first + __size;
 }
 
 struct __move_range
 {
     template <typename _RandomAccessIterator, typename _OutputIterator>
     _OutputIterator
     operator()(_RandomAccessIterator __first1, _RandomAccessIterator __last1, _OutputIterator __d_first) const
     {
         return __pstl::__omp_backend::__sort_details::__parallel_move_range(__first1, __last1, __d_first);
     }
 };
 } // namespace __sort_details
 
 template <typename _RandomAccessIterator, typename _Compare, typename _LeafSort>
 void
 __parallel_stable_sort_body(_RandomAccessIterator __xs, _RandomAccessIterator __xe, _Compare __comp,
                             _LeafSort __leaf_sort)
 {
     using _ValueType = typename std::iterator_traits<_RandomAccessIterator>::value_type;
     using _VecType = typename std::vector<_ValueType>;
     using _OutputIterator = typename _VecType::iterator;
     using _MoveValue = typename __omp_backend::__sort_details::__move_value;
     using _MoveRange = __omp_backend::__sort_details::__move_range;
 
     if (__should_run_serial(__xs, __xe))
     {
         __leaf_sort(__xs, __xe, __comp);
     }
     else
     {
         std::size_t __size = __xe - __xs;
         auto __mid = __xs + (__size / 2);
         __pstl::__omp_backend::__parallel_invoke_body(
             [&]() { __parallel_stable_sort_body(__xs, __mid, __comp, __leaf_sort); },
             [&]() { __parallel_stable_sort_body(__mid, __xe, __comp, __leaf_sort); });
 
         // Perform a parallel merge of the sorted ranges into __output_data.
         _VecType __output_data(__size);
         _MoveValue __move_value;
         _MoveRange __move_range;
         __utils::__serial_move_merge __merge(__size);
         __pstl::__omp_backend::__parallel_merge_body(
             __mid - __xs, __xe - __mid, __xs, __mid, __mid, __xe, __output_data.begin(), __comp,
             [&__merge, &__move_value, &__move_range](_RandomAccessIterator __as, _RandomAccessIterator __ae,
                                                      _RandomAccessIterator __bs, _RandomAccessIterator __be,
                                                      _OutputIterator __cs, _Compare __comp)
             { __merge(__as, __ae, __bs, __be, __cs, __comp, __move_value, __move_value, __move_range, __move_range); });
 
         // Move the values from __output_data back in the original source range.
         __pstl::__omp_backend::__sort_details::__parallel_move_range(__output_data.begin(), __output_data.end(), __xs);
     }
 }
 
 template <class _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare, typename _LeafSort>
 void
 __parallel_stable_sort(__pstl::__internal::__openmp_backend_tag __tag, _ExecutionPolicy&& /*__exec*/,
                        _RandomAccessIterator __xs, _RandomAccessIterator __xe, _Compare __comp, _LeafSort __leaf_sort,
                        std::size_t __nsort = 0)
 {
     auto __count = static_cast<std::size_t>(__xe - __xs);
     if (__count <= __default_chunk_size || __nsort < __count)
     {
         __leaf_sort(__xs, __xe, __comp);
         return;
     }
 
     // TODO: the partial sort implementation should
     // be shared with the other backends.
 
     if (omp_in_parallel())
     {
         if (__count <= __nsort)
         {
             __pstl::__omp_backend::__parallel_stable_sort_body(__xs, __xe, __comp, __leaf_sort);
         }
         else
         {
             __pstl::__omp_backend::__parallel_stable_partial_sort(__tag, __xs, __xe, __comp, __leaf_sort, __nsort);
         }
     }
     else
     {
         _PSTL_PRAGMA(omp parallel)
         _PSTL_PRAGMA(omp single nowait)
         if (__count <= __nsort)
         {
             __pstl::__omp_backend::__parallel_stable_sort_body(__xs, __xe, __comp, __leaf_sort);
         }
         else
         {
             __pstl::__omp_backend::__parallel_stable_partial_sort(__tag, __xs, __xe, __comp, __leaf_sort, __nsort);
         }
     }
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_STABLE_SORT_H

libcxx/include/__pstl/internal/omp/parallel_transform_reduce.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_TRANSFORM_REDUCE_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_TRANSFORM_REDUCE_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 //------------------------------------------------------------------------
 // parallel_transform_reduce
 //
 // Notation:
 //      r(i,j,init) returns reduction of init with reduction over [i,j)
 //      u(i) returns f(i,i+1,identity) for a hypothetical left identity element
 //      of r c(x,y) combines values x and y that were the result of r or u
 //------------------------------------------------------------------------
 
 template <class _RandomAccessIterator, class _UnaryOp, class _Value, class _Combiner, class _Reduction>
 auto
 __transform_reduce_body(_RandomAccessIterator __first, _RandomAccessIterator __last, _UnaryOp __unary_op, _Value __init,
                         _Combiner __combiner, _Reduction __reduction)
 {
     const std::size_t __num_threads = omp_get_num_threads();
     const std::size_t __size = __last - __first;
 
     // Initial partition of the iteration space into chunks. If the range is too small,
     // this will result in a nonsense policy, so we check on the size as well below.
     auto __policy = __omp_backend::__chunk_partitioner(__first + __num_threads, __last);
 
     if (__size <= __num_threads || __policy.__n_chunks < 2)
     {
         return __reduction(__first, __last, __init);
     }
 
     // Here, we cannot use OpenMP UDR because we must store the init value in
     // the combiner and it will be used several times. Although there should be
     // the only one; we manually generate the identity elements for each thread.
     std::vector<_Value> __accums;
     __accums.reserve(__num_threads);
 
     // initialize accumulators for all threads
     for (std::size_t __i = 0; __i < __num_threads; ++__i)
     {
         __accums.emplace_back(__unary_op(__first + __i));
     }
 
     // main loop
     _PSTL_PRAGMA(omp taskloop shared(__accums))
     for (std::size_t __chunk = 0; __chunk < __policy.__n_chunks; ++__chunk)
     {
         __pstl::__omp_backend::__process_chunk(__policy, __first + __num_threads, __chunk,
                                        [&](auto __chunk_first, auto __chunk_last)
                                        {
                                            auto __thread_num = omp_get_thread_num();
                                            __accums[__thread_num] =
                                                __reduction(__chunk_first, __chunk_last, __accums[__thread_num]);
                                        });
     }
 
     // combine by accumulators
     for (std::size_t __i = 0; __i < __num_threads; ++__i)
     {
         __init = __combiner(__init, __accums[__i]);
     }
 
     return __init;
 }
 
 template <class _ExecutionPolicy, class _RandomAccessIterator, class _UnaryOp, class _Value, class _Combiner,
           class _Reduction>
 _Value
 __parallel_transform_reduce(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&&, _RandomAccessIterator __first,
                             _RandomAccessIterator __last, _UnaryOp __unary_op, _Value __init, _Combiner __combiner,
                             _Reduction __reduction)
 {
     _Value __result = __init;
     if (omp_in_parallel())
     {
         // We don't create a nested parallel region in an existing parallel
         // region: just create tasks
         __result = __pstl::__omp_backend::__transform_reduce_body(__first, __last, __unary_op, __init, __combiner,
                                                                   __reduction);
     }
     else
     {
         // Create a parallel region, and a single thread will create tasks
         // for the region.
         _PSTL_PRAGMA(omp parallel)
         _PSTL_PRAGMA(omp single nowait)
         {
             __result = __pstl::__omp_backend::__transform_reduce_body(__first, __last, __unary_op, __init, __combiner,
                                                                       __reduction);
         }
     }
 
     return __result;
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_TRANSFORM_REDUCE_H

libcxx/include/__pstl/internal/omp/parallel_transform_scan.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_PARALLEL_TRANSFORM_SCAN_H
 #define _PSTL_INTERNAL_OMP_PARALLEL_TRANSFORM_SCAN_H
 
 #include "util.h"
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 template <class _ExecutionPolicy, class _Index, class _Up, class _Tp, class _Cp, class _Rp, class _Sp>
 _Tp
 __parallel_transform_scan(__pstl::__internal::__openmp_backend_tag, _ExecutionPolicy&&, _Index __n, _Up /* __u */,
                           _Tp __init, _Cp /* __combine */, _Rp /* __brick_reduce */, _Sp __scan)
 {
     // TODO: parallelize this function.
     return __scan(_Index(0), __n, __init);
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 #endif // _PSTL_INTERNAL_OMP_PARALLEL_TRANSFORM_SCAN_H

libcxx/include/__pstl/internal/omp/util.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_INTERNAL_OMP_UTIL_H
 #define _PSTL_INTERNAL_OMP_UTIL_H
 
 #include <algorithm>
 #include <atomic>
 #include <iterator>
 #include <cstddef>
 #include <cstdio>
 #include <memory>
 #include <vector>
 #include <omp.h>
 
 #include "../parallel_backend_utils.h"
 #include "../unseq_backend_simd.h"
 #include "../utils.h"
 
 // Portability "#pragma" definition
 #ifdef _MSC_VER
 #    define _PSTL_PRAGMA(x) __pragma(x)
 #else
 #    define _PSTL_PRAGMA(x) _Pragma(#    x)
 #endif
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl
 {
 namespace __omp_backend
 {
 
 //------------------------------------------------------------------------
 // use to cancel execution
 //------------------------------------------------------------------------
 inline void
 __cancel_execution()
 {
     // TODO: Figure out how to make cancelation work.
 }
 
 //------------------------------------------------------------------------
 // raw buffer
 //------------------------------------------------------------------------
 
 template <typename _Tp>
 class __buffer
 {
     std::allocator<_Tp> __allocator_;
     _Tp* __ptr_;
     const std::size_t __buf_size_;
     __buffer(const __buffer&) = delete;
     void
     operator=(const __buffer&) = delete;
 
   public:
     __buffer(std::size_t __n) : __allocator_(), __ptr_(__allocator_.allocate(__n)), __buf_size_(__n) {}
 
     operator bool() const { return __ptr_ != nullptr; }
 
     _Tp*
     get() const
     {
         return __ptr_;
     }
     ~__buffer() { __allocator_.deallocate(__ptr_, __buf_size_); }
 };
 
 // Preliminary size of each chunk: requires further discussion
 inline constexpr std::size_t __default_chunk_size = 2048;
 
 // Convenience function to determine when we should run serial.
 template <typename _Iterator, std::enable_if_t<!std::is_integral<_Iterator>::value, bool> = true>
 constexpr auto
 __should_run_serial(_Iterator __first, _Iterator __last) -> bool
 {
     using _difference_type = typename std::iterator_traits<_Iterator>::difference_type;
     auto __size = std::distance(__first, __last);
     return __size <= static_cast<_difference_type>(__default_chunk_size);
 }
 
 template <typename _Index, std::enable_if_t<std::is_integral<_Index>::value, bool> = true>
 constexpr auto
 __should_run_serial(_Index __first, _Index __last) -> bool
 {
     using _difference_type = _Index;
     auto __size = __last - __first;
     return __size <= static_cast<_difference_type>(__default_chunk_size);
 }
 
 struct __chunk_metrics
 {
     std::size_t __n_chunks;
     std::size_t __chunk_size;
     std::size_t __first_chunk_size;
 };
 
 // The iteration space partitioner according to __requested_chunk_size
 template <class _RandomAccessIterator, class _Size = std::size_t>
 auto
 __chunk_partitioner(_RandomAccessIterator __first, _RandomAccessIterator __last,
                     _Size __requested_chunk_size = __default_chunk_size) -> __chunk_metrics
 {
     /*
      * This algorithm improves distribution of elements in chunks by avoiding
      * small tail chunks. The leftover elements that do not fit neatly into
      * the chunk size are redistributed to early chunks. This improves
      * utilization of the processor's prefetch and reduces the number of
      * tasks needed by 1.
      */
 
     const _Size __n = __last - __first;
     _Size __n_chunks = 0;
     _Size __chunk_size = 0;
     _Size __first_chunk_size = 0;
     if (__n < __requested_chunk_size)
     {
         __chunk_size = __n;
         __first_chunk_size = __n;
         __n_chunks = 1;
         return __chunk_metrics{__n_chunks, __chunk_size, __first_chunk_size};
     }
 
     __n_chunks = (__n / __requested_chunk_size) + 1;
     __chunk_size = __n / __n_chunks;
     __first_chunk_size = __chunk_size;
     const _Size __n_leftover_items = __n - (__n_chunks * __chunk_size);
 
     if (__n_leftover_items == __chunk_size)
     {
         __n_chunks += 1;
         return __chunk_metrics{__n_chunks, __chunk_size, __first_chunk_size};
     }
     else if (__n_leftover_items == 0)
     {
         __first_chunk_size = __chunk_size;
         return __chunk_metrics{__n_chunks, __chunk_size, __first_chunk_size};
     }
 
     const _Size __n_extra_items_per_chunk = __n_leftover_items / __n_chunks;
     const _Size __n_final_leftover_items = __n_leftover_items - (__n_extra_items_per_chunk * __n_chunks);
 
     __chunk_size += __n_extra_items_per_chunk;
     __first_chunk_size = __chunk_size + __n_final_leftover_items;
 
     return __chunk_metrics{__n_chunks, __chunk_size, __first_chunk_size};
 }
 
 template <typename _Iterator, typename _Index, typename _Func>
 void
 __process_chunk(const __chunk_metrics& __metrics, _Iterator __base, _Index __chunk_index, _Func __f)
 {
     auto __this_chunk_size = __chunk_index == 0 ? __metrics.__first_chunk_size : __metrics.__chunk_size;
     auto __index = __chunk_index == 0 ? 0
                                       : (__chunk_index * __metrics.__chunk_size) +
                                             (__metrics.__first_chunk_size - __metrics.__chunk_size);
     auto __first = __base + __index;
     auto __last = __first + __this_chunk_size;
     __f(__first, __last);
 }
 
 } // namespace __omp_backend
 } // namespace __pstl
 
 #endif // _PSTL_INTERNAL_OMP_UTIL_H

libcxx/include/__pstl/internal/parallel_backend.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_PARALLEL_BACKEND_H
 #define _PSTL_PARALLEL_BACKEND_H
 
 #include "pstl_config.h"
 
 #if defined(_PSTL_PAR_BACKEND_SERIAL)
 #    include "parallel_backend_serial.h"
 namespace __pstl
 {
 namespace __par_backend = __serial_backend;
 }
 #elif defined(_PSTL_PAR_BACKEND_TBB)
 #    include "parallel_backend_tbb.h"
 namespace __pstl
 {
 namespace __par_backend = __tbb_backend;
 }
 #elif defined(_PSTL_PAR_BACKEND_OPENMP)
 #    include "parallel_backend_omp.h"
 namespace __pstl
 {
 namespace __par_backend = __omp_backend;
 }
 #else
 _PSTL_PRAGMA_MESSAGE("Parallel backend was not specified");
 #endif
 
 #endif /* _PSTL_PARALLEL_BACKEND_H */

libcxx/include/__pstl/internal/parallel_backend_omp.h

 // -*- C++ -*-
 // -*-===----------------------------------------------------------------------===//
 //
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_PARALLEL_BACKEND_OMP_H
 #define _PSTL_PARALLEL_BACKEND_OMP_H
 
 //------------------------------------------------------------------------
 // parallel_invoke
 //------------------------------------------------------------------------
 
 #include "./omp/parallel_invoke.h"
 
 //------------------------------------------------------------------------
 // parallel_for
 //------------------------------------------------------------------------
 
 #include "./omp/parallel_for.h"
 
 //------------------------------------------------------------------------
 // parallel_for_each
 //------------------------------------------------------------------------
 
 #include "./omp/parallel_for_each.h"
 
 //------------------------------------------------------------------------
 // parallel_reduce
 //------------------------------------------------------------------------
 
 #include "./omp/parallel_reduce.h"
 #include "./omp/parallel_transform_reduce.h"
 
 //------------------------------------------------------------------------
 // parallel_scan
 //------------------------------------------------------------------------
 
 #include "./omp/parallel_scan.h"
 #include "./omp/parallel_transform_scan.h"
 
 //------------------------------------------------------------------------
 // parallel_stable_sort
 //------------------------------------------------------------------------
 
 #include "./omp/parallel_stable_partial_sort.h"
 #include "./omp/parallel_stable_sort.h"
 
 //------------------------------------------------------------------------
 // parallel_merge
 //------------------------------------------------------------------------
 #include "./omp/parallel_merge.h"
 
 #endif //_PSTL_PARALLEL_BACKEND_OMP_H

libcxx/include/__pstl/internal/parallel_backend_serial.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_PARALLEL_BACKEND_SERIAL_H
 #define _PSTL_PARALLEL_BACKEND_SERIAL_H
 
 #include <algorithm>
 #include <cstddef>
 #include <memory>
 #include <numeric>
 #include <utility>
 
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl
 {
 namespace __serial_backend
 {
 
 template <typename _Tp>
 class __buffer
 {
     std::allocator<_Tp> __allocator_;
     _Tp* __ptr_;
     const std::size_t __buf_size_;
     __buffer(const __buffer&) = delete;
     void
     operator=(const __buffer&) = delete;
 
   public:
     __buffer(std::size_t __n) : __allocator_(), __ptr_(__allocator_.allocate(__n)), __buf_size_(__n) {}
 
     operator bool() const { return __ptr_ != nullptr; }
     _Tp*
     get() const
     {
         return __ptr_;
     }
     ~__buffer() { __allocator_.deallocate(__ptr_, __buf_size_); }
 };
 
 inline void
 __cancel_execution()
 {
 }
 
 template <class _ExecutionPolicy, class _Index, class _Fp>
 void
 __parallel_for(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last, _Fp __f)
 {
     __f(__first, __last);
 }
 
 template <class _ExecutionPolicy, class _Value, class _Index, typename _RealBody, typename _Reduction>
 _Value
 __parallel_reduce(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last,
                   const _Value& __identity, const _RealBody& __real_body, const _Reduction&)
 {
     if (__first == __last)
     {
         return __identity;
     }
     else
     {
         return __real_body(__first, __last, __identity);
     }
 }
 
 template <class _ExecutionPolicy, class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
 _Tp
 __parallel_transform_reduce(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last,
                             _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce)
 {
     return __reduce(__first, __last, __init);
 }
 
 template <class _ExecutionPolicy, typename _Index, typename _Tp, typename _Rp, typename _Cp, typename _Sp, typename _Ap>
 void
 __parallel_strict_scan(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _Index __n, _Tp __initial,
                        _Rp __reduce, _Cp __combine, _Sp __scan, _Ap __apex)
 {
     _Tp __sum = __initial;
     if (__n)
         __sum = __combine(__sum, __reduce(_Index(0), __n));
     __apex(__sum);
     if (__n)
         __scan(_Index(0), __n, __initial);
 }
 
 template <class _ExecutionPolicy, class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce, class _Scan>
 _Tp
 __parallel_transform_scan(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _Index __n, _UnaryOp,
                           _Tp __init, _BinaryOp, _Reduce, _Scan __scan)
 {
     return __scan(_Index(0), __n, __init);
 }
 
 template <class _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare, typename _LeafSort>
 void
 __parallel_stable_sort(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _RandomAccessIterator __first,
                        _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort, std::size_t = 0)
 {
     __leaf_sort(__first, __last, __comp);
 }
 
 template <class _ExecutionPolicy, typename _RandomAccessIterator1, typename _RandomAccessIterator2,
           typename _RandomAccessIterator3, typename _Compare, typename _LeafMerge>
 void
 __parallel_merge(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _RandomAccessIterator1 __first1,
                  _RandomAccessIterator1 __last1, _RandomAccessIterator2 __first2, _RandomAccessIterator2 __last2,
                  _RandomAccessIterator3 __outit, _Compare __comp, _LeafMerge __leaf_merge)
 {
     __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
 }
 
 template <class _ExecutionPolicy, typename _F1, typename _F2>
 void
 __parallel_invoke(__pstl::__internal::__serial_backend_tag, _ExecutionPolicy&&, _F1&& __f1, _F2&& __f2)
 {
     std::forward<_F1>(__f1)();
     std::forward<_F2>(__f2)();
 }
 
 } // namespace __serial_backend
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_PARALLEL_BACKEND_SERIAL_H */

libcxx/include/__pstl/internal/parallel_backend_tbb.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_PARALLEL_BACKEND_TBB_H
 #define _PSTL_PARALLEL_BACKEND_TBB_H
 
 #include <algorithm>
 #include <type_traits>
 
 #include "pstl_config.h"
 #include "parallel_backend_utils.h"
 
 // Bring in minimal required subset of Intel TBB
 #include <tbb/blocked_range.h>
 #include <tbb/parallel_for.h>
 #include <tbb/parallel_reduce.h>
 #include <tbb/parallel_scan.h>
 #include <tbb/parallel_invoke.h>
 #include <tbb/task_arena.h>
 #include <tbb/tbb_allocator.h>
 #include <tbb/task.h>
 
 #if TBB_INTERFACE_VERSION < 10000
 #    error Intel(R) Threading Building Blocks 2018 is required; older versions are not supported.
 #endif
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl
 {
 namespace __tbb_backend
 {
 
 //! Raw memory buffer with automatic freeing and no exceptions.
 /** Some of our algorithms need to start with raw memory buffer,
 not an initialize array, because initialization/destruction
 would make the span be at least O(N). */
 // tbb::allocator can improve performance in some cases.
 template <typename _Tp>
 class __buffer
 {
     tbb::tbb_allocator<_Tp> _M_allocator;
     _Tp* _M_ptr;
     const std::size_t _M_buf_size;
     __buffer(const __buffer&) = delete;
     void
     operator=(const __buffer&) = delete;
 
   public:
     //! Try to obtain buffer of given size to store objects of _Tp type
     __buffer(std::size_t n) : _M_allocator(), _M_ptr(_M_allocator.allocate(n)), _M_buf_size(n) {}
     //! True if buffer was successfully obtained, zero otherwise.
     operator bool() const { return _M_ptr != NULL; }
     //! Return pointer to buffer, or  NULL if buffer could not be obtained.
     _Tp*
     get() const
     {
         return _M_ptr;
     }
     //! Destroy buffer
     ~__buffer() { _M_allocator.deallocate(_M_ptr, _M_buf_size); }
 };
 
 // Wrapper for tbb::task
 inline void
 __cancel_execution()
 {
 #if TBB_INTERFACE_VERSION <= 12000
     tbb::task::self().group()->cancel_group_execution();
 #else
     tbb::task::current_context()->cancel_group_execution();
 #endif
 }
 
 //------------------------------------------------------------------------
 // parallel_for
 //------------------------------------------------------------------------
 
 template <class _Index, class _RealBody>
 class __parallel_for_body
 {
   public:
     __parallel_for_body(const _RealBody& __body) : _M_body(__body) {}
     __parallel_for_body(const __parallel_for_body& __body) : _M_body(__body._M_body) {}
     void
     operator()(const tbb::blocked_range<_Index>& __range) const
     {
         _M_body(__range.begin(), __range.end());
     }
 
   private:
     _RealBody _M_body;
 };
 
 //! Evaluation of brick f[i,j) for each subrange [i,j) of [first,last)
 // wrapper over tbb::parallel_for
 template <class _ExecutionPolicy, class _Index, class _Fp>
 void
 __parallel_for(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last, _Fp __f)
 {
     tbb::this_task_arena::isolate([=]() {
         tbb::parallel_for(tbb::blocked_range<_Index>(__first, __last), __parallel_for_body<_Index, _Fp>(__f));
     });
 }
 
 //! Evaluation of brick f[i,j) for each subrange [i,j) of [first,last)
 // wrapper over tbb::parallel_reduce
 template <class _ExecutionPolicy, class _Value, class _Index, typename _RealBody, typename _Reduction>
 _Value
 __parallel_reduce(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last,
                   const _Value& __identity, const _RealBody& __real_body, const _Reduction& __reduction)
 {
     return tbb::this_task_arena::isolate([__first, __last, &__identity, &__real_body, &__reduction]() -> _Value {
         return tbb::parallel_reduce(
             tbb::blocked_range<_Index>(__first, __last), __identity,
             [__real_body](const tbb::blocked_range<_Index>& __r, const _Value& __value) -> _Value {
                 return __real_body(__r.begin(), __r.end(), __value);
             },
             __reduction);
     });
 }
 
 //------------------------------------------------------------------------
 // parallel_transform_reduce
 //
 // Notation:
 //      r(i,j,init) returns reduction of init with reduction over [i,j)
 //      u(i) returns f(i,i+1,identity) for a hypothetical left identity element of r
 //      c(x,y) combines values x and y that were the result of r or u
 //------------------------------------------------------------------------
 
 template <class _Index, class _Up, class _Tp, class _Cp, class _Rp>
 struct __par_trans_red_body
 {
     alignas(_Tp) char _M_sum_storage[sizeof(_Tp)]; // Holds generalized non-commutative sum when has_sum==true
     _Rp _M_brick_reduce;                           // Most likely to have non-empty layout
     _Up _M_u;
     _Cp _M_combine;
     bool _M_has_sum; // Put last to minimize size of class
     _Tp&
     sum()
     {
         __TBB_ASSERT(_M_has_sum, "sum expected");
         return *(_Tp*)_M_sum_storage;
     }
     __par_trans_red_body(_Up __u, _Tp __init, _Cp __c, _Rp __r)
         : _M_brick_reduce(__r), _M_u(__u), _M_combine(__c), _M_has_sum(true)
     {
         new (_M_sum_storage) _Tp(__init);
     }
 
     __par_trans_red_body(__par_trans_red_body& __left, tbb::split)
         : _M_brick_reduce(__left._M_brick_reduce), _M_u(__left._M_u), _M_combine(__left._M_combine), _M_has_sum(false)
     {
     }
 
     ~__par_trans_red_body()
     {
         // 17.6.5.12 tells us to not worry about catching exceptions from destructors.
         if (_M_has_sum)
             sum().~_Tp();
     }
 
     void
     join(__par_trans_red_body& __rhs)
     {
         sum() = _M_combine(sum(), __rhs.sum());
     }
 
     void
     operator()(const tbb::blocked_range<_Index>& __range)
     {
         _Index __i = __range.begin();
         _Index __j = __range.end();
         if (!_M_has_sum)
         {
             __TBB_ASSERT(__range.size() > 1, "there should be at least 2 elements");
             new (&_M_sum_storage)
                 _Tp(_M_combine(_M_u(__i), _M_u(__i + 1))); // The condition i+1 < j is provided by the grain size of 3
             _M_has_sum = true;
             std::advance(__i, 2);
             if (__i == __j)
                 return;
         }
         sum() = _M_brick_reduce(__i, __j, sum());
     }
 };
 
 template <class _ExecutionPolicy, class _Index, class _Up, class _Tp, class _Cp, class _Rp>
 _Tp
 __parallel_transform_reduce(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _Index __first, _Index __last,
                             _Up __u, _Tp __init, _Cp __combine, _Rp __brick_reduce)
 {
     __tbb_backend::__par_trans_red_body<_Index, _Up, _Tp, _Cp, _Rp> __body(__u, __init, __combine, __brick_reduce);
     // The grain size of 3 is used in order to provide mininum 2 elements for each body
     tbb::this_task_arena::isolate(
         [__first, __last, &__body]() { tbb::parallel_reduce(tbb::blocked_range<_Index>(__first, __last, 3), __body); });
     return __body.sum();
 }
 
 //------------------------------------------------------------------------
 // parallel_scan
 //------------------------------------------------------------------------
 
 template <class _Index, class _Up, class _Tp, class _Cp, class _Rp, class _Sp>
 class __trans_scan_body
 {
     alignas(_Tp) char _M_sum_storage[sizeof(_Tp)]; // Holds generalized non-commutative sum when has_sum==true
     _Rp _M_brick_reduce;                           // Most likely to have non-empty layout
     _Up _M_u;
     _Cp _M_combine;
     _Sp _M_scan;
     bool _M_has_sum; // Put last to minimize size of class
   public:
     __trans_scan_body(_Up __u, _Tp __init, _Cp __combine, _Rp __reduce, _Sp __scan)
         : _M_brick_reduce(__reduce), _M_u(__u), _M_combine(__combine), _M_scan(__scan), _M_has_sum(true)
     {
         new (_M_sum_storage) _Tp(__init);
     }
 
     __trans_scan_body(__trans_scan_body& __b, tbb::split)
         : _M_brick_reduce(__b._M_brick_reduce), _M_u(__b._M_u), _M_combine(__b._M_combine), _M_scan(__b._M_scan),
           _M_has_sum(false)
     {
     }
 
     ~__trans_scan_body()
     {
         // 17.6.5.12 tells us to not worry about catching exceptions from destructors.
         if (_M_has_sum)
             sum().~_Tp();
     }
 
     _Tp&
     sum() const
     {
         __TBB_ASSERT(_M_has_sum, "sum expected");
         return *const_cast<_Tp*>(reinterpret_cast<_Tp const*>(_M_sum_storage));
     }
 
     void
     operator()(const tbb::blocked_range<_Index>& __range, tbb::pre_scan_tag)
     {
         _Index __i = __range.begin();
         _Index __j = __range.end();
         if (!_M_has_sum)
         {
             new (&_M_sum_storage) _Tp(_M_u(__i));
             _M_has_sum = true;
             ++__i;
             if (__i == __j)
                 return;
         }
         sum() = _M_brick_reduce(__i, __j, sum());
     }
 
     void
     operator()(const tbb::blocked_range<_Index>& __range, tbb::final_scan_tag)
     {
         sum() = _M_scan(__range.begin(), __range.end(), sum());
     }
 
     void
     reverse_join(__trans_scan_body& __a)
     {
         if (_M_has_sum)
         {
             sum() = _M_combine(__a.sum(), sum());
         }
         else
         {
             new (&_M_sum_storage) _Tp(__a.sum());
             _M_has_sum = true;
         }
     }
 
     void
     assign(__trans_scan_body& __b)
     {
         sum() = __b.sum();
     }
 };
 
 template <typename _Index>
 _Index
 __split(_Index __m)
 {
     _Index __k = 1;
     while (2 * __k < __m)
         __k *= 2;
     return __k;
 }
 
 //------------------------------------------------------------------------
 // __parallel_strict_scan
 //------------------------------------------------------------------------
 
 template <typename _Index, typename _Tp, typename _Rp, typename _Cp>
 void
 __upsweep(_Index __i, _Index __m, _Index __tilesize, _Tp* __r, _Index __lastsize, _Rp __reduce, _Cp __combine)
 {
     if (__m == 1)
         __r[0] = __reduce(__i * __tilesize, __lastsize);
     else
     {
         _Index __k = __split(__m);
         tbb::parallel_invoke(
             [=] { __tbb_backend::__upsweep(__i, __k, __tilesize, __r, __tilesize, __reduce, __combine); },
             [=] {
                 __tbb_backend::__upsweep(__i + __k, __m - __k, __tilesize, __r + __k, __lastsize, __reduce, __combine);
             });
         if (__m == 2 * __k)
             __r[__m - 1] = __combine(__r[__k - 1], __r[__m - 1]);
     }
 }
 
 template <typename _Index, typename _Tp, typename _Cp, typename _Sp>
 void
 __downsweep(_Index __i, _Index __m, _Index __tilesize, _Tp* __r, _Index __lastsize, _Tp __initial, _Cp __combine,
             _Sp __scan)
 {
     if (__m == 1)
         __scan(__i * __tilesize, __lastsize, __initial);
     else
     {
         const _Index __k = __split(__m);
         tbb::parallel_invoke(
             [=] { __tbb_backend::__downsweep(__i, __k, __tilesize, __r, __tilesize, __initial, __combine, __scan); },
             // Assumes that __combine never throws.
             //TODO: Consider adding a requirement for user functors to be constant.
             [=, &__combine] {
                 __tbb_backend::__downsweep(__i + __k, __m - __k, __tilesize, __r + __k, __lastsize,
                                            __combine(__initial, __r[__k - 1]), __combine, __scan);
             });
     }
 }
 
 // Adapted from Intel(R) Cilk(TM) version from cilkpub.
 // Let i:len denote a counted interval of length n starting at i.  s denotes a generalized-sum value.
 // Expected actions of the functors are:
 //     reduce(i,len) -> s  -- return reduction value of i:len.
 //     combine(s1,s2) -> s -- return merged sum
 //     apex(s) -- do any processing necessary between reduce and scan.
 //     scan(i,len,initial) -- perform scan over i:len starting with initial.
 // The initial range 0:n is partitioned into consecutive subranges.
 // reduce and scan are each called exactly once per subrange.
 // Thus callers can rely upon side effects in reduce.
 // combine must not throw an exception.
 // apex is called exactly once, after all calls to reduce and before all calls to scan.
 // For example, it's useful for allocating a __buffer used by scan but whose size is the sum of all reduction values.
 // T must have a trivial constructor and destructor.
 template <class _ExecutionPolicy, typename _Index, typename _Tp, typename _Rp, typename _Cp, typename _Sp, typename _Ap>
 void
 __parallel_strict_scan(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _Index __n, _Tp __initial,
                        _Rp __reduce, _Cp __combine, _Sp __scan, _Ap __apex)
 {
     tbb::this_task_arena::isolate([=, &__combine]() {
         if (__n > 1)
         {
             _Index __p = tbb::this_task_arena::max_concurrency();
             const _Index __slack = 4;
             _Index __tilesize = (__n - 1) / (__slack * __p) + 1;
             _Index __m = (__n - 1) / __tilesize;
             __buffer<_Tp> __buf(__m + 1);
             _Tp* __r = __buf.get();
             __tbb_backend::__upsweep(_Index(0), _Index(__m + 1), __tilesize, __r, __n - __m * __tilesize, __reduce,
                                      __combine);
 
             // When __apex is a no-op and __combine has no side effects, a good optimizer
             // should be able to eliminate all code between here and __apex.
             // Alternatively, provide a default value for __apex that can be
             // recognized by metaprogramming that conditionlly executes the following.
             size_t __k = __m + 1;
             _Tp __t = __r[__k - 1];
             while ((__k &= __k - 1))
                 __t = __combine(__r[__k - 1], __t);
             __apex(__combine(__initial, __t));
             __tbb_backend::__downsweep(_Index(0), _Index(__m + 1), __tilesize, __r, __n - __m * __tilesize, __initial,
                                        __combine, __scan);
             return;
         }
         // Fewer than 2 elements in sequence, or out of memory.  Handle has single block.
         _Tp __sum = __initial;
         if (__n)
             __sum = __combine(__sum, __reduce(_Index(0), __n));
         __apex(__sum);
         if (__n)
             __scan(_Index(0), __n, __initial);
     });
 }
 
 template <class _ExecutionPolicy, class _Index, class _Up, class _Tp, class _Cp, class _Rp, class _Sp>
 _Tp
 __parallel_transform_scan(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _Index __n, _Up __u, _Tp __init,
                           _Cp __combine, _Rp __brick_reduce, _Sp __scan)
 {
     __trans_scan_body<_Index, _Up, _Tp, _Cp, _Rp, _Sp> __body(__u, __init, __combine, __brick_reduce, __scan);
     auto __range = tbb::blocked_range<_Index>(0, __n);
     tbb::this_task_arena::isolate([__range, &__body]() { tbb::parallel_scan(__range, __body); });
     return __body.sum();
 }
 
 //------------------------------------------------------------------------
 // parallel_stable_sort
 //------------------------------------------------------------------------
 
 //------------------------------------------------------------------------
 // stable_sort utilities
 //
 // These are used by parallel implementations but do not depend on them.
 //------------------------------------------------------------------------
 #define _PSTL_MERGE_CUT_OFF 2000
 
 template <typename _Func>
 class __func_task;
 template <typename _Func>
 class __root_task;
 
 #if TBB_INTERFACE_VERSION <= 12000
 class __task : public tbb::task
 {
   public:
     template <typename _Fn>
     __task*
     make_continuation(_Fn&& __f)
     {
         return new (allocate_continuation()) __func_task<typename std::decay<_Fn>::type>(std::forward<_Fn>(__f));
     }
 
     template <typename _Fn>
     __task*
     make_child_of(__task* parent, _Fn&& __f)
     {
         return new (parent->allocate_child()) __func_task<typename std::decay<_Fn>::type>(std::forward<_Fn>(__f));
     }
 
     template <typename _Fn>
     __task*
     make_additional_child_of(tbb::task* parent, _Fn&& __f)
     {
         return new (tbb::task::allocate_additional_child_of(*parent))
             __func_task<typename std::decay<_Fn>::type>(std::forward<_Fn>(__f));
     }
 
     inline void
     recycle_as_continuation()
     {
         tbb::task::recycle_as_continuation();
     }
 
     inline void
     recycle_as_child_of(__task* parent)
     {
         tbb::task::recycle_as_child_of(*parent);
     }
 
     inline void
     spawn(__task* __t)
     {
         tbb::task::spawn(*__t);
     }
 
     template <typename _Fn>
     static inline void
     spawn_root_and_wait(__root_task<_Fn>& __root)
     {
         tbb::task::spawn_root_and_wait(*__root._M_task);
     }
 };
 
 template <typename _Func>
 class __func_task : public __task
 {
     _Func _M_func;
 
     tbb::task*
     execute()
     {
         return _M_func(this);
     };
 
   public:
     template <typename _Fn>
     __func_task(_Fn&& __f) : _M_func{std::forward<_Fn>(__f)}
     {
     }
 
     _Func&
     body()
     {
         return _M_func;
     }
 };
 
 template <typename _Func>
 class __root_task
 {
     tbb::task* _M_task;
 
   public:
     template <typename... Args>
     __root_task(Args&&... args)
         : _M_task{new (tbb::task::allocate_root()) __func_task<_Func>{_Func(std::forward<Args>(args)...)}}
     {
     }
 
     friend class __task;
     friend class __func_task<_Func>;
 };
 
 #else  // TBB_INTERFACE_VERSION <= 12000
 class __task : public tbb::detail::d1::task
 {
   protected:
     tbb::detail::d1::small_object_allocator _M_allocator{};
     tbb::detail::d1::execution_data* _M_execute_data{};
     __task* _M_parent{};
     std::atomic<int> _M_refcount{};
     bool _M_recycle{};
 
     template <typename _Fn>
     __task*
     allocate_func_task(_Fn&& __f)
     {
         _PSTL_ASSERT(_M_execute_data != nullptr);
         tbb::detail::d1::small_object_allocator __alloc{};
         auto __t =
             __alloc.new_object<__func_task<typename std::decay<_Fn>::type>>(*_M_execute_data, std::forward<_Fn>(__f));
         __t->_M_allocator = __alloc;
         return __t;
     }
 
   public:
     __task*
     parent()
     {
         return _M_parent;
     }
 
     void
     set_ref_count(int __n)
     {
         _M_refcount.store(__n, std::memory_order_release);
     }
 
     template <typename _Fn>
     __task*
     make_continuation(_Fn&& __f)
     {
         auto __t = allocate_func_task(std::forward<_Fn&&>(__f));
         __t->_M_parent = _M_parent;
         _M_parent = nullptr;
         return __t;
     }
 
     template <typename _Fn>
     __task*
     make_child_of(__task* __parent, _Fn&& __f)
     {
         auto __t = allocate_func_task(std::forward<_Fn&&>(__f));
         __t->_M_parent = __parent;
         return __t;
     }
 
     template <typename _Fn>
     __task*
     make_additional_child_of(__task* __parent, _Fn&& __f)
     {
         auto __t = make_child_of(__parent, std::forward<_Fn>(__f));
         _PSTL_ASSERT(__parent->_M_refcount.load(std::memory_order_relaxed) > 0);
         ++__parent->_M_refcount;
         return __t;
     }
 
     inline void
     recycle_as_continuation()
     {
         _M_recycle = true;
     }
 
     inline void
     recycle_as_child_of(__task* parent)
     {
         _M_recycle = true;
         _M_parent = parent;
     }
 
     inline void
     spawn(__task* __t)
     {
         _PSTL_ASSERT(_M_execute_data != nullptr);
         tbb::detail::d1::spawn(*__t, *_M_execute_data->context);
     }
 
     template <typename _Fn>
     static inline void
     spawn_root_and_wait(__root_task<_Fn>& __root)
     {
         tbb::detail::d1::execute_and_wait(*__root._M_func_task, __root._M_context, __root._M_wait_object,
                                           __root._M_context);
     }
 
     template <typename _Func>
     friend class __func_task;
 };
 
 template <typename _Func>
 class __func_task : public __task
 {
     _Func _M_func;
 
     __task*
     execute(tbb::detail::d1::execution_data& __ed) override
     {
         _M_execute_data = &__ed;
         _M_recycle = false;
         __task* __next = _M_func(this);
         return finalize(__next);
     };
 
     __task*
     cancel(tbb::detail::d1::execution_data& __ed) override
     {
         return finalize(nullptr);
     }
 
     __task*
     finalize(__task* __next)
     {
         bool __recycle = _M_recycle;
         _M_recycle = false;
 
         if (__recycle)
         {
             return __next;
         }
 
         auto __parent = _M_parent;
         auto __alloc = _M_allocator;
         auto __ed = _M_execute_data;
 
         this->~__func_task();
 
         _PSTL_ASSERT(__parent != nullptr);
         _PSTL_ASSERT(__parent->_M_refcount.load(std::memory_order_relaxed) > 0);
         if (--__parent->_M_refcount == 0)
         {
             _PSTL_ASSERT(__next == nullptr);
             __alloc.deallocate(this, *__ed);
             return __parent;
         }
 
         return __next;
     }
 
     friend class __root_task<_Func>;
 
   public:
     template <typename _Fn>
     __func_task(_Fn&& __f) : _M_func(std::forward<_Fn>(__f))
     {
     }
 
     _Func&
     body()
     {
         return _M_func;
     }
 };
 
 template <typename _Func>
 class __root_task : public __task
 {
     __task*
     execute(tbb::detail::d1::execution_data& __ed) override
     {
         _M_wait_object.release();
         return nullptr;
     };
 
     __task*
     cancel(tbb::detail::d1::execution_data& __ed) override
     {
         _M_wait_object.release();
         return nullptr;
     }
 
     __func_task<_Func>* _M_func_task{};
     tbb::detail::d1::wait_context _M_wait_object{0};
     tbb::task_group_context _M_context{};
 
   public:
     template <typename... Args>
     __root_task(Args&&... args) : _M_wait_object{1}
     {
         tbb::detail::d1::small_object_allocator __alloc{};
         _M_func_task = __alloc.new_object<__func_task<_Func>>(_Func(std::forward<Args>(args)...));
         _M_func_task->_M_allocator = __alloc;
         _M_func_task->_M_parent = this;
         _M_refcount.store(1, std::memory_order_relaxed);
     }
 
     friend class __task;
 };
 #endif // TBB_INTERFACE_VERSION <= 12000
 
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename _Compare, typename _Cleanup,
           typename _LeafMerge>
 class __merge_func
 {
     typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType1;
     typedef typename std::iterator_traits<_RandomAccessIterator2>::difference_type _DifferenceType2;
     typedef typename std::common_type<_DifferenceType1, _DifferenceType2>::type _SizeType;
     typedef typename std::iterator_traits<_RandomAccessIterator1>::value_type _ValueType;
 
     _RandomAccessIterator1 _M_x_beg;
     _RandomAccessIterator2 _M_z_beg;
 
     _SizeType _M_xs, _M_xe;
     _SizeType _M_ys, _M_ye;
     _SizeType _M_zs;
     _Compare _M_comp;
     _LeafMerge _M_leaf_merge;
     _SizeType _M_nsort; //number of elements to be sorted for partial_sort alforithm
 
     static const _SizeType __merge_cut_off = _PSTL_MERGE_CUT_OFF;
 
     bool _root;   //means a task is merging root task
     bool _x_orig; //"true" means X(or left ) subrange is in the original container; false - in the buffer
     bool _y_orig; //"true" means Y(or right) subrange is in the original container; false - in the buffer
     bool _split; //"true" means a merge task is a split task for parallel merging, the execution logic differs
 
     bool
     is_partial() const
     {
         return _M_nsort > 0;
     }
 
     struct __move_value
     {
         template <typename Iterator1, typename Iterator2>
         void
         operator()(Iterator1 __x, Iterator2 __z)
         {
             *__z = std::move(*__x);
         }
     };
 
     struct __move_value_construct
     {
         template <typename Iterator1, typename Iterator2>
         void
         operator()(Iterator1 __x, Iterator2 __z)
         {
             ::new (std::addressof(*__z)) _ValueType(std::move(*__x));
         }
     };
 
     struct __move_range
     {
         template <typename Iterator1, typename Iterator2>
         Iterator2
         operator()(Iterator1 __first1, Iterator1 __last1, Iterator2 __first2)
         {
             if (__last1 - __first1 < __merge_cut_off)
                 return std::move(__first1, __last1, __first2);
 
             auto __n = __last1 - __first1;
             tbb::parallel_for(tbb::blocked_range<_SizeType>(0, __n, __merge_cut_off),
                               [__first1, __first2](const tbb::blocked_range<_SizeType>& __range) {
                                   std::move(__first1 + __range.begin(), __first1 + __range.end(),
                                             __first2 + __range.begin());
                               });
             return __first2 + __n;
         }
     };
 
     struct __move_range_construct
     {
         template <typename Iterator1, typename Iterator2>
         Iterator2
         operator()(Iterator1 __first1, Iterator1 __last1, Iterator2 __first2)
         {
             if (__last1 - __first1 < __merge_cut_off)
             {
                 for (; __first1 != __last1; ++__first1, ++__first2)
                     __move_value_construct()(__first1, __first2);
                 return __first2;
             }
 
             auto __n = __last1 - __first1;
             tbb::parallel_for(tbb::blocked_range<_SizeType>(0, __n, __merge_cut_off),
                               [__first1, __first2](const tbb::blocked_range<_SizeType>& __range) {
                                   for (auto i = __range.begin(); i != __range.end(); ++i)
                                       __move_value_construct()(__first1 + i, __first2 + i);
                               });
             return __first2 + __n;
         }
     };
 
     struct __cleanup_range
     {
         template <typename Iterator>
         void
         operator()(Iterator __first, Iterator __last)
         {
             if (__last - __first < __merge_cut_off)
                 _Cleanup()(__first, __last);
             else
             {
                 auto __n = __last - __first;
                 tbb::parallel_for(tbb::blocked_range<_SizeType>(0, __n, __merge_cut_off),
                                   [__first](const tbb::blocked_range<_SizeType>& __range) {
                                       _Cleanup()(__first + __range.begin(), __first + __range.end());
                                   });
             }
         }
     };
 
   public:
     __merge_func(_SizeType __xs, _SizeType __xe, _SizeType __ys, _SizeType __ye, _SizeType __zs, _Compare __comp,
                  _Cleanup, _LeafMerge __leaf_merge, _SizeType __nsort, _RandomAccessIterator1 __x_beg,
                  _RandomAccessIterator2 __z_beg, bool __x_orig, bool __y_orig, bool __root)
         : _M_xs(__xs), _M_xe(__xe), _M_ys(__ys), _M_ye(__ye), _M_zs(__zs), _M_x_beg(__x_beg), _M_z_beg(__z_beg),
           _M_comp(__comp), _M_leaf_merge(__leaf_merge), _M_nsort(__nsort), _root(__root),
           _x_orig(__x_orig), _y_orig(__y_orig), _split(false)
     {
     }
 
     bool
     is_left(_SizeType __idx) const
     {
         return _M_xs == __idx;
     }
 
     template <typename IndexType>
     void
     set_odd(IndexType __idx, bool __on_off)
     {
         if (is_left(__idx))
             _x_orig = __on_off;
         else
             _y_orig = __on_off;
     }
 
     __task*
     operator()(__task* __self);
 
   private:
     __merge_func*
     parent_merge(__task* __self) const
     {
         return _root ? nullptr : &static_cast<__func_task<__merge_func>*>(__self->parent())->body();
     }
     bool
     x_less_y()
     {
         const auto __nx = (_M_xe - _M_xs);
         const auto __ny = (_M_ye - _M_ys);
         _PSTL_ASSERT(__nx > 0 && __ny > 0);
 
         _PSTL_ASSERT(_x_orig == _y_orig);
         _PSTL_ASSERT(!is_partial());
 
         if (_x_orig)
         {
             _PSTL_ASSERT(std::is_sorted(_M_x_beg + _M_xs, _M_x_beg + _M_xe, _M_comp));
             _PSTL_ASSERT(std::is_sorted(_M_x_beg + _M_ys, _M_x_beg + _M_ye, _M_comp));
             return !_M_comp(*(_M_x_beg + _M_ys), *(_M_x_beg + _M_xe - 1));
         }
 
         _PSTL_ASSERT(std::is_sorted(_M_z_beg + _M_xs, _M_z_beg + _M_xe, _M_comp));
         _PSTL_ASSERT(std::is_sorted(_M_z_beg + _M_ys, _M_z_beg + _M_ye, _M_comp));
         return !_M_comp(*(_M_z_beg + _M_zs + __nx), *(_M_z_beg + _M_zs + __nx - 1));
     }
     void
     move_x_range()
     {
         const auto __nx = (_M_xe - _M_xs);
         const auto __ny = (_M_ye - _M_ys);
         _PSTL_ASSERT(__nx > 0 && __ny > 0);
 
         if (_x_orig)
             __move_range_construct()(_M_x_beg + _M_xs, _M_x_beg + _M_xe, _M_z_beg + _M_zs);
         else
         {
             __move_range()(_M_z_beg + _M_zs, _M_z_beg + _M_zs + __nx, _M_x_beg + _M_xs);
             __cleanup_range()(_M_z_beg + _M_zs, _M_z_beg + _M_zs + __nx);
         }
 
         _x_orig = !_x_orig;
     }
     void
     move_y_range()
     {
         const auto __nx = (_M_xe - _M_xs);
         const auto __ny = (_M_ye - _M_ys);
 
         if (_y_orig)
             __move_range_construct()(_M_x_beg + _M_ys, _M_x_beg + _M_ye, _M_z_beg + _M_zs + __nx);
         else
         {
             __move_range()(_M_z_beg + _M_zs + __nx, _M_z_beg + _M_zs + __nx + __ny, _M_x_beg + _M_ys);
             __cleanup_range()(_M_z_beg + _M_zs + __nx, _M_z_beg + _M_zs + __nx + __ny);
         }
 
         _y_orig = !_y_orig;
     }
     __task*
     merge_ranges(__task* __self)
     {
         _PSTL_ASSERT(_x_orig == _y_orig); //two merged subrange must be lie into the same buffer
 
         const auto __nx = (_M_xe - _M_xs);
         const auto __ny = (_M_ye - _M_ys);
         const auto __n = __nx + __ny;
 
         // need to merge {x} and {y}
         if (__n > __merge_cut_off)
             return split_merging(__self);
 
         //merge to buffer
         if (_x_orig)
         {
             _M_leaf_merge(_M_x_beg + _M_xs, _M_x_beg + _M_xe, _M_x_beg + _M_ys, _M_x_beg + _M_ye, _M_z_beg + _M_zs,
                           _M_comp, __move_value_construct(), __move_value_construct(), __move_range_construct(),
                           __move_range_construct());
             _PSTL_ASSERT(parent_merge(__self)); //not root merging task
         }
         //merge to "origin"
         else
         {
             _PSTL_ASSERT(_x_orig == _y_orig);
 
             _PSTL_ASSERT(is_partial() || std::is_sorted(_M_z_beg + _M_xs, _M_z_beg + _M_xe, _M_comp));
             _PSTL_ASSERT(is_partial() || std::is_sorted(_M_z_beg + _M_ys, _M_z_beg + _M_ye, _M_comp));
 
             const auto __nx = (_M_xe - _M_xs);
             const auto __ny = (_M_ye - _M_ys);
 
             _M_leaf_merge(_M_z_beg + _M_xs, _M_z_beg + _M_xe, _M_z_beg + _M_ys, _M_z_beg + _M_ye, _M_x_beg + _M_zs,
                           _M_comp, __move_value(), __move_value(), __move_range(), __move_range());
 
             __cleanup_range()(_M_z_beg + _M_xs, _M_z_beg + _M_xe);
             __cleanup_range()(_M_z_beg + _M_ys, _M_z_beg + _M_ye);
         }
         return nullptr;
     }
 
     __task*
     process_ranges(__task* __self)
     {
         _PSTL_ASSERT(_x_orig == _y_orig);
         _PSTL_ASSERT(!_split);
 
         auto p = parent_merge(__self);
 
         if (!p)
         { //root merging task
 
             //optimization, just for sort algorithm, //{x} <= {y}
             if (!is_partial() && x_less_y()) //we have a solution
             {
                 if (!_x_orig)
                 {                   //we have to move the solution to the origin
                     move_x_range(); //parallel moving
                     move_y_range(); //parallel moving
                 }
                 return nullptr;
             }
             //else: if we have data in the origin,
             //we have to move data to the buffer for final merging into the origin.
             if (_x_orig)
             {
                 move_x_range(); //parallel moving
                 move_y_range(); //parallel moving
             }
             // need to merge {x} and {y}.
             return merge_ranges(__self);
         }
         //else: not root merging task (parent_merge() == NULL)
         //optimization, just for sort algorithm, //{x} <= {y}
         if (!is_partial() && x_less_y())
         {
             const auto id_range = _M_zs;
             p->set_odd(id_range, _x_orig);
             return nullptr;
         }
         //else: we have to revert "_x(y)_orig" flag of the parent merging task
         const auto id_range = _M_zs;
         p->set_odd(id_range, !_x_orig);
 
         return merge_ranges(__self);
     }
 
     //splitting as merge task into 2 of the same level
     __task*
     split_merging(__task* __self)
     {
         _PSTL_ASSERT(_x_orig == _y_orig);
         const auto __nx = (_M_xe - _M_xs);
         const auto __ny = (_M_ye - _M_ys);
 
         _SizeType __xm{};
         _SizeType __ym{};
         if (__nx < __ny)
         {
             __ym = _M_ys + __ny / 2;
 
             if (_x_orig)
                 __xm = std::upper_bound(_M_x_beg + _M_xs, _M_x_beg + _M_xe, *(_M_x_beg + __ym), _M_comp) - _M_x_beg;
             else
                 __xm = std::upper_bound(_M_z_beg + _M_xs, _M_z_beg + _M_xe, *(_M_z_beg + __ym), _M_comp) - _M_z_beg;
         }
         else
         {
             __xm = _M_xs + __nx / 2;
 
             if (_y_orig)
                 __ym = std::lower_bound(_M_x_beg + _M_ys, _M_x_beg + _M_ye, *(_M_x_beg + __xm), _M_comp) - _M_x_beg;
             else
                 __ym = std::lower_bound(_M_z_beg + _M_ys, _M_z_beg + _M_ye, *(_M_z_beg + __xm), _M_comp) - _M_z_beg;
         }
 
         auto __zm = _M_zs + ((__xm - _M_xs) + (__ym - _M_ys));
         __merge_func __right_func(__xm, _M_xe, __ym, _M_ye, __zm, _M_comp, _Cleanup(), _M_leaf_merge, _M_nsort,
                                   _M_x_beg, _M_z_beg, _x_orig, _y_orig, _root);
         __right_func._split = true;
         auto __merge_task = __self->make_additional_child_of(__self->parent(), std::move(__right_func));
         __self->spawn(__merge_task);
         __self->recycle_as_continuation();
 
         _M_xe = __xm;
         _M_ye = __ym;
         _split = true;
 
         return __self;
     }
 };
 
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename __M_Compare, typename _Cleanup,
           typename _LeafMerge>
 __task*
 __merge_func<_RandomAccessIterator1, _RandomAccessIterator2, __M_Compare, _Cleanup, _LeafMerge>::
 operator()(__task* __self)
 {
     //a. split merge task into 2 of the same level; the special logic,
     //without processing(process_ranges) adjacent sub-ranges x and y
     if (_split)
         return merge_ranges(__self);
 
     //b. General merging of adjacent sub-ranges x and y (with optimization in case of {x} <= {y} )
 
     //1. x and y are in the even buffer
     //2. x and y are in the odd buffer
     if (_x_orig == _y_orig)
         return process_ranges(__self);
 
     //3. x is in even buffer, y is in the odd buffer
     //4. x is in odd buffer, y is in the even buffer
     if (!parent_merge(__self))
     { //root merge task
         if (_x_orig)
             move_x_range();
         else
             move_y_range();
     }
     else
     {
         const _SizeType __nx = (_M_xe - _M_xs);
         const _SizeType __ny = (_M_ye - _M_ys);
         _PSTL_ASSERT(__nx > 0);
         _PSTL_ASSERT(__nx > 0);
 
         if (__nx < __ny)
             move_x_range();
         else
             move_y_range();
     }
 
     return process_ranges(__self);
 }
 
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename _Compare, typename _LeafSort>
 class __stable_sort_func
 {
   public:
     typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType1;
     typedef typename std::iterator_traits<_RandomAccessIterator2>::difference_type _DifferenceType2;
     typedef typename std::common_type<_DifferenceType1, _DifferenceType2>::type _SizeType;
 
   private:
     _RandomAccessIterator1 _M_xs, _M_xe, _M_x_beg;
     _RandomAccessIterator2 _M_zs, _M_z_beg;
     _Compare _M_comp;
     _LeafSort _M_leaf_sort;
     bool _M_root;
     _SizeType _M_nsort; //zero or number of elements to be sorted for partial_sort alforithm
 
   public:
     __stable_sort_func(_RandomAccessIterator1 __xs, _RandomAccessIterator1 __xe, _RandomAccessIterator2 __zs,
                        bool __root, _Compare __comp, _LeafSort __leaf_sort, _SizeType __nsort,
                        _RandomAccessIterator1 __x_beg, _RandomAccessIterator2 __z_beg)
         : _M_xs(__xs), _M_xe(__xe), _M_x_beg(__x_beg), _M_zs(__zs), _M_z_beg(__z_beg), _M_comp(__comp),
           _M_leaf_sort(__leaf_sort), _M_root(__root), _M_nsort(__nsort)
     {
     }
 
     __task*
     operator()(__task* __self);
 };
 
 #define _PSTL_STABLE_SORT_CUT_OFF 500
 
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename _Compare, typename _LeafSort>
 __task*
 __stable_sort_func<_RandomAccessIterator1, _RandomAccessIterator2, _Compare, _LeafSort>::operator()(__task* __self)
 {
     typedef __merge_func<_RandomAccessIterator1, _RandomAccessIterator2, _Compare, __utils::__serial_destroy,
                          __utils::__serial_move_merge>
         _MergeTaskType;
 
     const _SizeType __n = _M_xe - _M_xs;
     const _SizeType __nmerge = _M_nsort > 0 ? _M_nsort : __n;
     const _SizeType __sort_cut_off = _PSTL_STABLE_SORT_CUT_OFF;
     if (__n <= __sort_cut_off)
     {
         _M_leaf_sort(_M_xs, _M_xe, _M_comp);
         _PSTL_ASSERT(!_M_root);
         return nullptr;
     }
 
     const _RandomAccessIterator1 __xm = _M_xs + __n / 2;
     const _RandomAccessIterator2 __zm = _M_zs + (__xm - _M_xs);
     const _RandomAccessIterator2 __ze = _M_zs + __n;
     _MergeTaskType __m(_MergeTaskType(_M_xs - _M_x_beg, __xm - _M_x_beg, __xm - _M_x_beg, _M_xe - _M_x_beg,
                                       _M_zs - _M_z_beg, _M_comp, __utils::__serial_destroy(),
                                       __utils::__serial_move_merge(__nmerge), _M_nsort, _M_x_beg, _M_z_beg,
                                       /*x_orig*/ true, /*y_orig*/ true, /*root*/ _M_root));
     auto __parent = __self->make_continuation(std::move(__m));
     __parent->set_ref_count(2);
     auto __right = __self->make_child_of(
         __parent, __stable_sort_func(__xm, _M_xe, __zm, false, _M_comp, _M_leaf_sort, _M_nsort, _M_x_beg, _M_z_beg));
     __self->spawn(__right);
     __self->recycle_as_child_of(__parent);
     _M_root = false;
     _M_xe = __xm;
 
     return __self;
 }
 
 template <class _ExecutionPolicy, typename _RandomAccessIterator, typename _Compare, typename _LeafSort>
 void
 __parallel_stable_sort(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _RandomAccessIterator __xs,
                        _RandomAccessIterator __xe, _Compare __comp, _LeafSort __leaf_sort, std::size_t __nsort = 0)
 {
     tbb::this_task_arena::isolate([=, &__nsort]() {
         //sorting based on task tree and parallel merge
         typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _ValueType;
         typedef typename std::iterator_traits<_RandomAccessIterator>::difference_type _DifferenceType;
         const _DifferenceType __n = __xe - __xs;
         if (__nsort == __n)
             __nsort = 0; // 'partial_sort' becames 'sort'
 
         const _DifferenceType __sort_cut_off = _PSTL_STABLE_SORT_CUT_OFF;
         if (__n > __sort_cut_off)
         {
             __buffer<_ValueType> __buf(__n);
             __root_task<__stable_sort_func<_RandomAccessIterator, _ValueType*, _Compare, _LeafSort>> __root{
                 __xs, __xe, __buf.get(), true, __comp, __leaf_sort, __nsort, __xs, __buf.get()};
             __task::spawn_root_and_wait(__root);
             return;
         }
         //serial sort
         __leaf_sort(__xs, __xe, __comp);
     });
 }
 
 //------------------------------------------------------------------------
 // parallel_merge
 //------------------------------------------------------------------------
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename _RandomAccessIterator3,
           typename _Compare, typename _LeafMerge>
 class __merge_func_static
 {
     _RandomAccessIterator1 _M_xs, _M_xe;
     _RandomAccessIterator2 _M_ys, _M_ye;
     _RandomAccessIterator3 _M_zs;
     _Compare _M_comp;
     _LeafMerge _M_leaf_merge;
 
   public:
     __merge_func_static(_RandomAccessIterator1 __xs, _RandomAccessIterator1 __xe, _RandomAccessIterator2 __ys,
                         _RandomAccessIterator2 __ye, _RandomAccessIterator3 __zs, _Compare __comp,
                         _LeafMerge __leaf_merge)
         : _M_xs(__xs), _M_xe(__xe), _M_ys(__ys), _M_ye(__ye), _M_zs(__zs), _M_comp(__comp), _M_leaf_merge(__leaf_merge)
     {
     }
 
     __task*
     operator()(__task* __self);
 };
 
 //TODO: consider usage of parallel_for with a custom blocked_range
 template <typename _RandomAccessIterator1, typename _RandomAccessIterator2, typename _RandomAccessIterator3,
           typename __M_Compare, typename _LeafMerge>
 __task*
 __merge_func_static<_RandomAccessIterator1, _RandomAccessIterator2, _RandomAccessIterator3, __M_Compare, _LeafMerge>::
 operator()(__task* __self)
 {
     typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType1;
     typedef typename std::iterator_traits<_RandomAccessIterator2>::difference_type _DifferenceType2;
     typedef typename std::common_type<_DifferenceType1, _DifferenceType2>::type _SizeType;
     const _SizeType __n = (_M_xe - _M_xs) + (_M_ye - _M_ys);
     const _SizeType __merge_cut_off = _PSTL_MERGE_CUT_OFF;
     if (__n <= __merge_cut_off)
     {
         _M_leaf_merge(_M_xs, _M_xe, _M_ys, _M_ye, _M_zs, _M_comp);
         return nullptr;
     }
 
     _RandomAccessIterator1 __xm;
     _RandomAccessIterator2 __ym;
     if (_M_xe - _M_xs < _M_ye - _M_ys)
     {
         __ym = _M_ys + (_M_ye - _M_ys) / 2;
         __xm = std::upper_bound(_M_xs, _M_xe, *__ym, _M_comp);
     }
     else
     {
         __xm = _M_xs + (_M_xe - _M_xs) / 2;
         __ym = std::lower_bound(_M_ys, _M_ye, *__xm, _M_comp);
     }
     const _RandomAccessIterator3 __zm = _M_zs + ((__xm - _M_xs) + (__ym - _M_ys));
     auto __right = __self->make_additional_child_of(
         __self->parent(), __merge_func_static(__xm, _M_xe, __ym, _M_ye, __zm, _M_comp, _M_leaf_merge));
     __self->spawn(__right);
     __self->recycle_as_continuation();
     _M_xe = __xm;
     _M_ye = __ym;
 
     return __self;
 }
 
 template <class _ExecutionPolicy, typename _RandomAccessIterator1, typename _RandomAccessIterator2,
           typename _RandomAccessIterator3, typename _Compare, typename _LeafMerge>
 void
 __parallel_merge(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _RandomAccessIterator1 __xs,
                  _RandomAccessIterator1 __xe, _RandomAccessIterator2 __ys, _RandomAccessIterator2 __ye,
                  _RandomAccessIterator3 __zs, _Compare __comp, _LeafMerge __leaf_merge)
 {
     typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type _DifferenceType1;
     typedef typename std::iterator_traits<_RandomAccessIterator2>::difference_type _DifferenceType2;
     typedef typename std::common_type<_DifferenceType1, _DifferenceType2>::type _SizeType;
     const _SizeType __n = (__xe - __xs) + (__ye - __ys);
     const _SizeType __merge_cut_off = _PSTL_MERGE_CUT_OFF;
     if (__n <= __merge_cut_off)
     {
         // Fall back on serial merge
         __leaf_merge(__xs, __xe, __ys, __ye, __zs, __comp);
     }
     else
     {
         tbb::this_task_arena::isolate([=]() {
             typedef __merge_func_static<_RandomAccessIterator1, _RandomAccessIterator2, _RandomAccessIterator3,
                                         _Compare, _LeafMerge>
                 _TaskType;
             __root_task<_TaskType> __root{__xs, __xe, __ys, __ye, __zs, __comp, __leaf_merge};
             __task::spawn_root_and_wait(__root);
         });
     }
 }
 
 //------------------------------------------------------------------------
 // parallel_invoke
 //------------------------------------------------------------------------
 template <class _ExecutionPolicy, typename _F1, typename _F2>
 void
 __parallel_invoke(__pstl::__internal::__tbb_backend_tag, _ExecutionPolicy&&, _F1&& __f1, _F2&& __f2)
 {
     //TODO: a version of tbb::this_task_arena::isolate with variadic arguments pack should be added in the future
     tbb::this_task_arena::isolate([&]() { tbb::parallel_invoke(std::forward<_F1>(__f1), std::forward<_F2>(__f2)); });
 }
 
 } // namespace __tbb_backend
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_PARALLEL_BACKEND_TBB_H */

libcxx/include/__pstl/internal/parallel_backend_utils.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_PARALLEL_BACKEND_UTILS_H
 #define _PSTL_PARALLEL_BACKEND_UTILS_H
 
 #include <iterator>
 #include <utility>
 #include "utils.h"
 
 #include "pstl_config.h"
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl
 {
 
 namespace __utils
 {
 
 //! Destroy sequence [xs,xe)
 struct __serial_destroy
 {
     template <typename _RandomAccessIterator>
     void
     operator()(_RandomAccessIterator __zs, _RandomAccessIterator __ze)
     {
         typedef typename std::iterator_traits<_RandomAccessIterator>::value_type _ValueType;
         while (__zs != __ze)
         {
             --__ze;
             (*__ze).~_ValueType();
         }
     }
 };
 
 //! Merge sequences [__xs,__xe) and [__ys,__ye) to output sequence [__zs,(__xe-__xs)+(__ye-__ys)), using std::move
 struct __serial_move_merge
 {
     const std::size_t _M_nmerge;
 
     explicit __serial_move_merge(std::size_t __nmerge) : _M_nmerge(__nmerge) {}
     template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIterator3, class _Compare,
               class _MoveValueX, class _MoveValueY, class _MoveSequenceX, class _MoveSequenceY>
     void
     operator()(_RandomAccessIterator1 __xs, _RandomAccessIterator1 __xe, _RandomAccessIterator2 __ys,
                _RandomAccessIterator2 __ye, _RandomAccessIterator3 __zs, _Compare __comp, _MoveValueX __move_value_x,
                _MoveValueY __move_value_y, _MoveSequenceX __move_sequence_x, _MoveSequenceY __move_sequence_y)
     {
         constexpr bool __same_move_val = std::is_same<_MoveValueX, _MoveValueY>::value;
         constexpr bool __same_move_seq = std::is_same<_MoveSequenceX, _MoveSequenceY>::value;
 
         auto __n = _M_nmerge;
         _PSTL_ASSERT(__n > 0);
 
         auto __nx = __xe - __xs;
         //auto __ny = __ye - __ys;
         _RandomAccessIterator3 __zs_beg = __zs;
 
         if (__xs != __xe)
         {
             if (__ys != __ye)
             {
                 for (;;)
                 {
                     if (__comp(*__ys, *__xs))
                     {
                         const auto __i = __zs - __zs_beg;
                         if (__i < __nx)
                             __move_value_x(__ys, __zs);
                         else
                             __move_value_y(__ys, __zs);
                         ++__zs, --__n;
                         if (++__ys == __ye)
                         {
                             break;
                         }
                         else if (__n == 0)
                         {
                             const auto __j = __zs - __zs_beg;
                             if (__same_move_seq || __j < __nx)
                                 __zs = __move_sequence_x(__ys, __ye, __zs);
                             else
                                 __zs = __move_sequence_y(__ys, __ye, __zs);
                             break;
                         }
                     }
                     else
                     {
                         const auto __i = __zs - __zs_beg;
                         if (__same_move_val || __i < __nx)
                             __move_value_x(__xs, __zs);
                         else
                             __move_value_y(__xs, __zs);
                         ++__zs, --__n;
                         if (++__xs == __xe)
                         {
                             const auto __j = __zs - __zs_beg;
                             if (__same_move_seq || __j < __nx)
                                 __move_sequence_x(__ys, __ye, __zs);
                             else
                                 __move_sequence_y(__ys, __ye, __zs);
                             return;
                         }
                         else if (__n == 0)
                         {
                             const auto __j = __zs - __zs_beg;
                             if (__same_move_seq || __j < __nx)
                             {
                                 __zs = __move_sequence_x(__xs, __xe, __zs);
                                 __move_sequence_x(__ys, __ye, __zs);
                             }
                             else
                             {
                                 __zs = __move_sequence_y(__xs, __xe, __zs);
                                 __move_sequence_y(__ys, __ye, __zs);
                             }
                             return;
                         }
                     }
                 }
             }
             __ys = __xs;
             __ye = __xe;
         }
         const auto __i = __zs - __zs_beg;
         if (__same_move_seq || __i < __nx)
             __move_sequence_x(__ys, __ye, __zs);
         else
             __move_sequence_y(__ys, __ye, __zs);
     }
 };
 
 template <typename _ForwardIterator1, typename _ForwardIterator2, typename _OutputIterator, typename _Compare,
           typename _CopyConstructRange>
 _OutputIterator
 __set_union_construct(_ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2,
                       _ForwardIterator2 __last2, _OutputIterator __result, _Compare __comp,
                       _CopyConstructRange __cc_range)
 {
     using _Tp = typename std::iterator_traits<_OutputIterator>::value_type;
 
     for (; __first1 != __last1; ++__result)
     {
         if (__first2 == __last2)
             return __cc_range(__first1, __last1, __result);
         if (__comp(*__first2, *__first1))
         {
             ::new (std::addressof(*__result)) _Tp(*__first2);
             ++__first2;
         }
         else
         {
             ::new (std::addressof(*__result)) _Tp(*__first1);
             if (!__comp(*__first1, *__first2))
                 ++__first2;
             ++__first1;
         }
     }
     return __cc_range(__first2, __last2, __result);
 }
 
 template <typename _ForwardIterator1, typename _ForwardIterator2, typename _OutputIterator, typename _Compare>
 _OutputIterator
 __set_intersection_construct(_ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2,
                              _ForwardIterator2 __last2, _OutputIterator __result, _Compare __comp)
 {
     using _Tp = typename std::iterator_traits<_OutputIterator>::value_type;
 
     for (; __first1 != __last1 && __first2 != __last2;)
     {
         if (__comp(*__first1, *__first2))
             ++__first1;
         else
         {
             if (!__comp(*__first2, *__first1))
             {
                 ::new (std::addressof(*__result)) _Tp(*__first1);
                 ++__result;
                 ++__first1;
             }
             ++__first2;
         }
     }
     return __result;
 }
 
 template <typename _ForwardIterator1, typename _ForwardIterator2, typename _OutputIterator, typename _Compare,
           typename _CopyConstructRange>
 _OutputIterator
 __set_difference_construct(_ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2,
                            _ForwardIterator2 __last2, _OutputIterator __result, _Compare __comp,
                            _CopyConstructRange __cc_range)
 {
     using _Tp = typename std::iterator_traits<_OutputIterator>::value_type;
 
     for (; __first1 != __last1;)
     {
         if (__first2 == __last2)
             return __cc_range(__first1, __last1, __result);
 
         if (__comp(*__first1, *__first2))
         {
             ::new (std::addressof(*__result)) _Tp(*__first1);
             ++__result;
             ++__first1;
         }
         else
         {
             if (!__comp(*__first2, *__first1))
                 ++__first1;
             ++__first2;
         }
     }
     return __result;
 }
 template <typename _ForwardIterator1, typename _ForwardIterator2, typename _OutputIterator, typename _Compare,
           typename _CopyConstructRange>
 _OutputIterator
 __set_symmetric_difference_construct(_ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2,
                                      _ForwardIterator2 __last2, _OutputIterator __result, _Compare __comp,
                                      _CopyConstructRange __cc_range)
 {
     using _Tp = typename std::iterator_traits<_OutputIterator>::value_type;
 
     for (; __first1 != __last1;)
     {
         if (__first2 == __last2)
             return __cc_range(__first1, __last1, __result);
 
         if (__comp(*__first1, *__first2))
         {
             ::new (std::addressof(*__result)) _Tp(*__first1);
             ++__result;
             ++__first1;
         }
         else
         {
             if (__comp(*__first2, *__first1))
             {
                 ::new (std::addressof(*__result)) _Tp(*__first2);
                 ++__result;
             }
             else
                 ++__first1;
             ++__first2;
         }
     }
     return __cc_range(__first2, __last2, __result);
 }
 
 } // namespace __utils
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_PARALLEL_BACKEND_UTILS_H */

libcxx/include/__pstl/internal/parallel_impl.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_PARALLEL_IMPL_H
 #define _PSTL_PARALLEL_IMPL_H
 
 #include "pstl_config.h"
 
 #include <atomic>
 // This header defines the minimum set of parallel routines required to support Parallel STL,
 // implemented on top of Intel(R) Threading Building Blocks (Intel(R) TBB) library
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl
 {
 namespace __internal
 {
 
 //------------------------------------------------------------------------
 // parallel_find
 //-----------------------------------------------------------------------
 /** Return extremum value returned by brick f[i,j) for subranges [i,j) of [first,last)
 Each f[i,j) must return a value in [i,j). */
 template <class _BackendTag, class _ExecutionPolicy, class _Index, class _Brick, class _Compare>
 _Index
 __parallel_find(_BackendTag __tag, _ExecutionPolicy&& __exec, _Index __first, _Index __last, _Brick __f,
                 _Compare __comp, bool __b_first)
 {
     typedef typename std::iterator_traits<_Index>::difference_type _DifferenceType;
     const _DifferenceType __n = __last - __first;
     _DifferenceType __initial_dist = __b_first ? __n : -1;
     std::atomic<_DifferenceType> __extremum(__initial_dist);
     // TODO: find out what is better here: parallel_for or parallel_reduce
     __par_backend::__parallel_for(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __last,
                                   [__comp, __f, __first, &__extremum](_Index __i, _Index __j)
                                   {
                                       // See "Reducing Contention Through Priority Updates", PPoPP '13, for discussion of
                                       // why using a shared variable scales fairly well in this situation.
                                       if (__comp(__i - __first, __extremum))
                                       {
                                           _Index __res = __f(__i, __j);
                                           // If not '__last' returned then we found what we want so put this to extremum
                                           if (__res != __j)
                                           {
                                               const _DifferenceType __k = __res - __first;
                                               for (_DifferenceType __old = __extremum; __comp(__k, __old);
                                                    __old = __extremum)
                                               {
                                                   __extremum.compare_exchange_weak(__old, __k);
                                               }
                                           }
                                       }
                                   });
     return __extremum != __initial_dist ? __first + __extremum : __last;
 }
 
 //------------------------------------------------------------------------
 // parallel_or
 //------------------------------------------------------------------------
 //! Return true if brick f[i,j) returns true for some subrange [i,j) of [first,last)
 template <class _BackendTag, class _ExecutionPolicy, class _Index, class _Brick>
 bool
 __parallel_or(_BackendTag __tag, _ExecutionPolicy&& __exec, _Index __first, _Index __last, _Brick __f)
 {
     std::atomic<bool> __found(false);
     __par_backend::__parallel_for(__tag, std::forward<_ExecutionPolicy>(__exec), __first, __last,
                                   [__f, &__found](_Index __i, _Index __j)
                                   {
                                       if (!__found.load(std::memory_order_relaxed) && __f(__i, __j))
                                       {
                                           __found.store(true, std::memory_order_relaxed);
                                           __par_backend::__cancel_execution();
                                       }
                                   });
     return __found;
 }
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_PARALLEL_IMPL_H */

libcxx/include/__pstl/internal/pstl_config.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_CONFIG_H
 #define _PSTL_CONFIG_H
 
 // TODO: Make this a proper configuration option
 #define _PSTL_PAR_BACKEND_SERIAL
 
 // The version is XYYZ, where X is major, YY is minor, and Z is patch (i.e. X.YY.Z)
 #define _PSTL_VERSION 16000
 #define _PSTL_VERSION_MAJOR (_PSTL_VERSION / 1000)
 #define _PSTL_VERSION_MINOR ((_PSTL_VERSION % 1000) / 10)
 #define _PSTL_VERSION_PATCH (_PSTL_VERSION % 10)
 
 #if !defined(_PSTL_PAR_BACKEND_SERIAL) && !defined(_PSTL_PAR_BACKEND_TBB) && !defined(_PSTL_PAR_BACKEND_OPENMP)
 #    error "A parallel backend must be specified"
 #endif
 
 // Check the user-defined macro for warnings
 #if defined(PSTL_USAGE_WARNINGS)
 #    define _PSTL_USAGE_WARNINGS
 #endif
 
 #if defined(_LIBCPP_VERSION)
 #    include <__assert>
 #    define _PSTL_ASSERT(pred) _LIBCPP_ASSERT(pred, "")
 #elif defined(__GLIBCXX__)
 #    define _PSTL_ASSERT(pred) __glibcxx_assert(pred)
 #else
 #    include <cassert>
 #    define _PSTL_ASSERT(pred) (assert((pred)))
 #endif
 
 // Portability "#pragma" definition
 #ifdef _MSC_VER
 #    define _PSTL_PRAGMA(x) __pragma(x)
 #else
 #    define _PSTL_PRAGMA(x) _Pragma(#    x)
 #endif
 
 #define _PSTL_STRING_AUX(x) #x
 #define _PSTL_STRING(x) _PSTL_STRING_AUX(x)
 #define _PSTL_STRING_CONCAT(x, y) x #y
 
 #ifdef _PSTL_HIDE_FROM_ABI_PER_TU
 #    define _PSTL_HIDE_FROM_ABI_PUSH                                                                                   \
         _Pragma("clang attribute push(__attribute__((internal_linkage)), apply_to=any(function,record))")
 #    define _PSTL_HIDE_FROM_ABI_POP _Pragma("clang attribute pop")
 #else
 #    define _PSTL_HIDE_FROM_ABI_PUSH /* nothing */
 #    define _PSTL_HIDE_FROM_ABI_POP  /* nothing */
 #endif
 
 // note that when ICC or Clang is in use, _PSTL_GCC_VERSION might not fully match
 // the actual GCC version on the system.
 #define _PSTL_GCC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
 
 #if defined(__clang__)
 // according to clang documentation, version can be vendor specific
 #    define _PSTL_CLANG_VERSION (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__)
 #endif
 
 // Enable SIMD for compilers that support OpenMP 4.0
 #if (defined(_OPENMP) && _OPENMP >= 201307) || \
     (defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1600) || \
     (!defined(__INTEL_COMPILER) && _PSTL_GCC_VERSION >= 40900) || \
     defined(__clang__)
 #    define _PSTL_PRAGMA_SIMD _PSTL_PRAGMA(omp simd)
 #    define _PSTL_PRAGMA_DECLARE_SIMD _PSTL_PRAGMA(omp declare simd)
 #    define _PSTL_PRAGMA_SIMD_REDUCTION(PRM) _PSTL_PRAGMA(omp simd reduction(PRM))
 #elif !defined(_MSC_VER) //#pragma simd
 #    define _PSTL_PRAGMA_SIMD _PSTL_PRAGMA(simd)
 #    define _PSTL_PRAGMA_DECLARE_SIMD
 #    define _PSTL_PRAGMA_SIMD_REDUCTION(PRM) _PSTL_PRAGMA(simd reduction(PRM))
 #else //no simd
 #    define _PSTL_PRAGMA_SIMD
 #    define _PSTL_PRAGMA_DECLARE_SIMD
 #    define _PSTL_PRAGMA_SIMD_REDUCTION(PRM)
 #endif //Enable SIMD
 
 #if defined(__INTEL_COMPILER)
 #    define _PSTL_PRAGMA_FORCEINLINE _PSTL_PRAGMA(forceinline)
 #else
 #    define _PSTL_PRAGMA_FORCEINLINE
 #endif
 
 #if defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1900
 #    define _PSTL_PRAGMA_SIMD_SCAN(PRM) _PSTL_PRAGMA(omp simd reduction(inscan, PRM))
 #    define _PSTL_PRAGMA_SIMD_INCLUSIVE_SCAN(PRM) _PSTL_PRAGMA(omp scan inclusive(PRM))
 #    define _PSTL_PRAGMA_SIMD_EXCLUSIVE_SCAN(PRM) _PSTL_PRAGMA(omp scan exclusive(PRM))
 #else
 #    define _PSTL_PRAGMA_SIMD_SCAN(PRM)
 #    define _PSTL_PRAGMA_SIMD_INCLUSIVE_SCAN(PRM)
 #    define _PSTL_PRAGMA_SIMD_EXCLUSIVE_SCAN(PRM)
 #endif
 
 // Should be defined to 1 for environments with a vendor implementation of C++17 execution policies
 #define _PSTL_CPP17_EXECUTION_POLICIES_PRESENT (_MSC_VER >= 1912 && _MSVC_LANG >= 201703L) ||                          \
     (_GLIBCXX_RELEASE >= 9 && __GLIBCXX__ >= 20190503 && __cplusplus >= 201703L)
 
 #if (defined(_MSC_VER) && _MSC_VER >= 1900) || \
     __cplusplus >= 201300L || \
     __cpp_lib_robust_nonmodifying_seq_ops == 201304
 #   define _PSTL_CPP14_2RANGE_MISMATCH_EQUAL_PRESENT
 #endif
 #if (defined(_MSC_VER) && _MSC_VER >= 1900) || \
     __cplusplus >= 201402L || \
     __cpp_lib_make_reverse_iterator == 201402
 #   define _PSTL_CPP14_MAKE_REVERSE_ITERATOR_PRESENT
 #endif
 #if (defined(_MSC_VER) && _MSC_VER >= 1900) || __cplusplus >= 201402L
 #   define _PSTL_CPP14_INTEGER_SEQUENCE_PRESENT
 #endif
 #if (defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1700) || \
     (defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 190023918) || \
     __cplusplus >= 201402L
 #   define _PSTL_CPP14_VARIABLE_TEMPLATES_PRESENT
 #endif
 
 #if defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1800
 #   define _PSTL_EARLYEXIT_PRESENT
 #   define _PSTL_MONOTONIC_PRESENT
 #endif
 
 #if (defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1900) || \
     (!defined(__INTEL_COMPILER) && _PSTL_GCC_VERSION >= 40900) || \
     (defined(_OPENMP) && _OPENMP >= 201307)
 #    define _PSTL_UDR_PRESENT
 #endif
 
 #if defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1900 && __INTEL_COMPILER_BUILD_DATE >= 20180626
 #   define _PSTL_UDS_PRESENT
 #endif
 
 #if defined(_PSTL_EARLYEXIT_PRESENT)
 #    define _PSTL_PRAGMA_SIMD_EARLYEXIT _PSTL_PRAGMA(omp simd early_exit)
 #else
 #    define _PSTL_PRAGMA_SIMD_EARLYEXIT
 #endif
 
 #if defined(_PSTL_MONOTONIC_PRESENT)
 #    define _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC(PRM) _PSTL_PRAGMA(omp ordered simd monotonic(PRM))
 #    define _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC_2ARGS(PRM1, PRM2) _PSTL_PRAGMA(omp ordered simd monotonic(PRM1, PRM2))
 #else
 #    define _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC(PRM)
 #    define _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC_2ARGS(PRM1, PRM2)
 #endif
 
 // Declaration of reduction functor, where
 // NAME - the name of the functor
 // OP - type of the callable object with the reduction operation
 // omp_in - refers to the local partial result
 // omp_out - refers to the final value of the combiner operator
 // omp_priv - refers to the private copy of the initial value
 // omp_orig - refers to the original variable to be reduced
 #define _PSTL_PRAGMA_DECLARE_REDUCTION(NAME, OP)                                                                       \
     _PSTL_PRAGMA(omp declare reduction(NAME:OP : omp_out(omp_in)) initializer(omp_priv = omp_orig))
 
 #if defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1600
 #    define _PSTL_PRAGMA_VECTOR_UNALIGNED _PSTL_PRAGMA(vector unaligned)
 #else
 #    define _PSTL_PRAGMA_VECTOR_UNALIGNED
 #endif
 
 // Check the user-defined macro to use non-temporal stores
 #if defined(PSTL_USE_NONTEMPORAL_STORES) && (__INTEL_COMPILER >= 1600)
 #    define _PSTL_USE_NONTEMPORAL_STORES_IF_ALLOWED _PSTL_PRAGMA(vector nontemporal)
 #else
 #    define _PSTL_USE_NONTEMPORAL_STORES_IF_ALLOWED
 #endif
 
 #if defined(_MSC_VER) || defined(__INTEL_COMPILER) // the preprocessors don't type a message location
 #    define _PSTL_PRAGMA_LOCATION __FILE__ ":" _PSTL_STRING(__LINE__) ": [Parallel STL message]: "
 #else
 #    define _PSTL_PRAGMA_LOCATION " [Parallel STL message]: "
 #endif
 
 #define _PSTL_PRAGMA_MESSAGE_IMPL(x) _PSTL_PRAGMA(message(_PSTL_STRING_CONCAT(_PSTL_PRAGMA_LOCATION, x)))
 
 #if defined(_PSTL_USAGE_WARNINGS)
 #    define _PSTL_PRAGMA_MESSAGE(x) _PSTL_PRAGMA_MESSAGE_IMPL(x)
 #    define _PSTL_PRAGMA_MESSAGE_POLICIES(x) _PSTL_PRAGMA_MESSAGE_IMPL(x)
 #else
 #    define _PSTL_PRAGMA_MESSAGE(x)
 #    define _PSTL_PRAGMA_MESSAGE_POLICIES(x)
 #endif
 
 // broken macros
 #if (defined(__GLIBCXX__) && __GLIBCXX__ < 20150716) || \
     (defined(_MSC_VER) && _MSC_VER < 1800)
 #   define _PSTL_CPP11_STD_ROTATE_BROKEN
 #endif
 
 #if defined(__INTEL_COMPILER) && __INTEL_COMPILER == 1800
 #   define _PSTL_ICC_18_OMP_SIMD_BROKEN
 #endif
 
 #endif /* _PSTL_CONFIG_H */

libcxx/include/__pstl/internal/unseq_backend_simd.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_UNSEQ_BACKEND_SIMD_H
 #define _PSTL_UNSEQ_BACKEND_SIMD_H
 
 #include <type_traits>
 
 #include "pstl_config.h"
 #include "utils.h"
 
 // This header defines the minimum set of vector routines required
 // to support parallel STL.
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl
 {
 namespace __unseq_backend
 {
 
 // Expect vector width up to 64 (or 512 bit)
 const std::size_t __lane_size = 64;
 
 template <class _Iterator, class _DifferenceType, class _Function>
 _Iterator
 __simd_walk_1(_Iterator __first, _DifferenceType __n, _Function __f) noexcept
 {
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         __f(__first[__i]);
 
     return __first + __n;
 }
 
 template <class _Iterator1, class _DifferenceType, class _Iterator2, class _Function>
 _Iterator2
 __simd_walk_2(_Iterator1 __first1, _DifferenceType __n, _Iterator2 __first2, _Function __f) noexcept
 {
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         __f(__first1[__i], __first2[__i]);
     return __first2 + __n;
 }
 
 template <class _Iterator1, class _DifferenceType, class _Iterator2, class _Iterator3, class _Function>
 _Iterator3
 __simd_walk_3(_Iterator1 __first1, _DifferenceType __n, _Iterator2 __first2, _Iterator3 __first3,
               _Function __f) noexcept
 {
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         __f(__first1[__i], __first2[__i], __first3[__i]);
     return __first3 + __n;
 }
 
 // TODO: check whether __simd_first() can be used here
 template <class _Index, class _DifferenceType, class _Pred>
 bool
 __simd_or(_Index __first, _DifferenceType __n, _Pred __pred) noexcept
 {
 #if defined(_PSTL_EARLYEXIT_PRESENT)
     _DifferenceType __i;
     _PSTL_PRAGMA_VECTOR_UNALIGNED
     _PSTL_PRAGMA_SIMD_EARLYEXIT
     for (__i = 0; __i < __n; ++__i)
         if (__pred(__first[__i]))
             break;
     return __i < __n;
 #else
     _DifferenceType __block_size = 4 < __n ? 4 : __n;
     const _Index __last = __first + __n;
     while (__last != __first)
     {
         int32_t __flag = 1;
         _PSTL_PRAGMA_SIMD_REDUCTION(& : __flag)
         for (_DifferenceType __i = 0; __i < __block_size; ++__i)
             if (__pred(*(__first + __i)))
                 __flag = 0;
         if (!__flag)
             return true;
 
         __first += __block_size;
         if (__last - __first >= __block_size << 1)
         {
             // Double the block _Size.  Any unnecessary iterations can be amortized against work done so far.
             __block_size <<= 1;
         }
         else
         {
             __block_size = __last - __first;
         }
     }
     return false;
 #endif
 }
 
 template <class _Index, class _DifferenceType, class _Compare>
 _Index
 __simd_first(_Index __first, _DifferenceType __begin, _DifferenceType __end, _Compare __comp) noexcept
 {
 #if defined(_PSTL_EARLYEXIT_PRESENT)
     _DifferenceType __i = __begin;
     _PSTL_PRAGMA_VECTOR_UNALIGNED // Do not generate peel loop part
         _PSTL_PRAGMA_SIMD_EARLYEXIT for (; __i < __end; ++__i)
     {
         if (__comp(__first, __i))
         {
             break;
         }
     }
     return __first + __i;
 #else
     // Experiments show good block sizes like this
     const _DifferenceType __block_size = 8;
     alignas(__lane_size) _DifferenceType __lane[__block_size] = {0};
     while (__end - __begin >= __block_size)
     {
         _DifferenceType __found = 0;
         _PSTL_PRAGMA_VECTOR_UNALIGNED // Do not generate peel loop part
             _PSTL_PRAGMA_SIMD_REDUCTION(|
                                         : __found) for (_DifferenceType __i = __begin; __i < __begin + __block_size;
                                                         ++__i)
         {
             const _DifferenceType __t = __comp(__first, __i);
             __lane[__i - __begin] = __t;
             __found |= __t;
         }
         if (__found)
         {
             _DifferenceType __i;
             // This will vectorize
             for (__i = 0; __i < __block_size; ++__i)
             {
                 if (__lane[__i])
                 {
                     break;
                 }
             }
             return __first + __begin + __i;
         }
         __begin += __block_size;
     }
 
     //Keep remainder scalar
     while (__begin != __end)
     {
         if (__comp(__first, __begin))
         {
             return __first + __begin;
         }
         ++__begin;
     }
     return __first + __end;
 #endif //_PSTL_EARLYEXIT_PRESENT
 }
 
 template <class _Index1, class _DifferenceType, class _Index2, class _Pred>
 std::pair<_Index1, _Index2>
 __simd_first(_Index1 __first1, _DifferenceType __n, _Index2 __first2, _Pred __pred) noexcept
 {
 #if defined(_PSTL_EARLYEXIT_PRESENT)
     _DifferenceType __i = 0;
     _PSTL_PRAGMA_VECTOR_UNALIGNED
     _PSTL_PRAGMA_SIMD_EARLYEXIT
     for (; __i < __n; ++__i)
         if (__pred(__first1[__i], __first2[__i]))
             break;
     return std::make_pair(__first1 + __i, __first2 + __i);
 #else
     const _Index1 __last1 = __first1 + __n;
     const _Index2 __last2 = __first2 + __n;
     // Experiments show good block sizes like this
     const _DifferenceType __block_size = 8;
     alignas(__lane_size) _DifferenceType __lane[__block_size] = {0};
     while (__last1 - __first1 >= __block_size)
     {
         _DifferenceType __found = 0;
         _DifferenceType __i;
         _PSTL_PRAGMA_VECTOR_UNALIGNED // Do not generate peel loop part
             _PSTL_PRAGMA_SIMD_REDUCTION(|
                                         : __found) for (__i = 0; __i < __block_size; ++__i)
         {
             const _DifferenceType __t = __pred(__first1[__i], __first2[__i]);
             __lane[__i] = __t;
             __found |= __t;
         }
         if (__found)
         {
             _DifferenceType __i2;
             // This will vectorize
             for (__i2 = 0; __i2 < __block_size; ++__i2)
             {
                 if (__lane[__i2])
                     break;
             }
             return std::make_pair(__first1 + __i2, __first2 + __i2);
         }
         __first1 += __block_size;
         __first2 += __block_size;
     }
 
     //Keep remainder scalar
     for (; __last1 != __first1; ++__first1, ++__first2)
         if (__pred(*(__first1), *(__first2)))
             return std::make_pair(__first1, __first2);
 
     return std::make_pair(__last1, __last2);
 #endif //_PSTL_EARLYEXIT_PRESENT
 }
 
 template <class _Index, class _DifferenceType, class _Pred>
 _DifferenceType
 __simd_count(_Index __index, _DifferenceType __n, _Pred __pred) noexcept
 {
     _DifferenceType __count = 0;
     _PSTL_PRAGMA_SIMD_REDUCTION(+ : __count)
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         if (__pred(*(__index + __i)))
             ++__count;
 
     return __count;
 }
 
 template <class _InputIterator, class _DifferenceType, class _OutputIterator, class _BinaryPredicate>
 _OutputIterator
 __simd_unique_copy(_InputIterator __first, _DifferenceType __n, _OutputIterator __result,
                    _BinaryPredicate __pred) noexcept
 {
     if (__n == 0)
         return __result;
 
     _DifferenceType __cnt = 1;
     __result[0] = __first[0];
 
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 1; __i < __n; ++__i)
     {
         _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC(__cnt : 1)
         if (!__pred(__first[__i], __first[__i - 1]))
         {
             __result[__cnt] = __first[__i];
             ++__cnt;
         }
     }
     return __result + __cnt;
 }
 
 template <class _InputIterator, class _DifferenceType, class _OutputIterator, class _Assigner>
 _OutputIterator
 __simd_assign(_InputIterator __first, _DifferenceType __n, _OutputIterator __result, _Assigner __assigner) noexcept
 {
     _PSTL_USE_NONTEMPORAL_STORES_IF_ALLOWED
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         __assigner(__first + __i, __result + __i);
     return __result + __n;
 }
 
 template <class _InputIterator, class _DifferenceType, class _OutputIterator, class _UnaryPredicate>
 _OutputIterator
 __simd_copy_if(_InputIterator __first, _DifferenceType __n, _OutputIterator __result, _UnaryPredicate __pred) noexcept
 {
     _DifferenceType __cnt = 0;
 
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
     {
         _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC(__cnt : 1)
         if (__pred(__first[__i]))
         {
             __result[__cnt] = __first[__i];
             ++__cnt;
         }
     }
     return __result + __cnt;
 }
 
 template <class _InputIterator, class _DifferenceType, class _BinaryPredicate>
 _DifferenceType
 __simd_calc_mask_2(_InputIterator __first, _DifferenceType __n, bool* __mask, _BinaryPredicate __pred) noexcept
 {
     _DifferenceType __count = 0;
 
     _PSTL_PRAGMA_SIMD_REDUCTION(+ : __count)
     for (_DifferenceType __i = 0; __i < __n; ++__i)
     {
         __mask[__i] = !__pred(__first[__i], __first[__i - 1]);
         __count += __mask[__i];
     }
     return __count;
 }
 
 template <class _InputIterator, class _DifferenceType, class _UnaryPredicate>
 _DifferenceType
 __simd_calc_mask_1(_InputIterator __first, _DifferenceType __n, bool* __mask, _UnaryPredicate __pred) noexcept
 {
     _DifferenceType __count = 0;
 
     _PSTL_PRAGMA_SIMD_REDUCTION(+ : __count)
     for (_DifferenceType __i = 0; __i < __n; ++__i)
     {
         __mask[__i] = __pred(__first[__i]);
         __count += __mask[__i];
     }
     return __count;
 }
 
 template <class _InputIterator, class _DifferenceType, class _OutputIterator, class _Assigner>
 void
 __simd_copy_by_mask(_InputIterator __first, _DifferenceType __n, _OutputIterator __result, bool* __mask,
                     _Assigner __assigner) noexcept
 {
     _DifferenceType __cnt = 0;
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
     {
         if (__mask[__i])
         {
             _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC(__cnt : 1)
             {
                 __assigner(__first + __i, __result + __cnt);
                 ++__cnt;
             }
         }
     }
 }
 
 template <class _InputIterator, class _DifferenceType, class _OutputIterator1, class _OutputIterator2>
 void
 __simd_partition_by_mask(_InputIterator __first, _DifferenceType __n, _OutputIterator1 __out_true,
                          _OutputIterator2 __out_false, bool* __mask) noexcept
 {
     _DifferenceType __cnt_true = 0, __cnt_false = 0;
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
     {
         _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC_2ARGS(__cnt_true : 1, __cnt_false : 1)
         if (__mask[__i])
         {
             __out_true[__cnt_true] = __first[__i];
             ++__cnt_true;
         }
         else
         {
             __out_false[__cnt_false] = __first[__i];
             ++__cnt_false;
         }
     }
 }
 
 template <class _Index, class _DifferenceType, class _Tp>
 _Index
 __simd_fill_n(_Index __first, _DifferenceType __n, const _Tp& __value) noexcept
 {
     _PSTL_USE_NONTEMPORAL_STORES_IF_ALLOWED
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         __first[__i] = __value;
     return __first + __n;
 }
 
 template <class _Index, class _DifferenceType, class _Generator>
 _Index
 __simd_generate_n(_Index __first, _DifferenceType __size, _Generator __g) noexcept
 {
     _PSTL_USE_NONTEMPORAL_STORES_IF_ALLOWED
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __size; ++__i)
         __first[__i] = __g();
     return __first + __size;
 }
 
 template <class _Index, class _BinaryPredicate>
 _Index
 __simd_adjacent_find(_Index __first, _Index __last, _BinaryPredicate __pred, bool __or_semantic) noexcept
 {
     if (__last - __first < 2)
         return __last;
 
     typedef typename std::iterator_traits<_Index>::difference_type _DifferenceType;
     _DifferenceType __i = 0;
 
 #if defined(_PSTL_EARLYEXIT_PRESENT)
     //Some compiler versions fail to compile the following loop when iterators are used. Indices are used instead
     const _DifferenceType __n = __last - __first - 1;
     _PSTL_PRAGMA_VECTOR_UNALIGNED
     _PSTL_PRAGMA_SIMD_EARLYEXIT
     for (; __i < __n; ++__i)
         if (__pred(__first[__i], __first[__i + 1]))
             break;
 
     return __i < __n ? __first + __i : __last;
 #else
     // Experiments show good block sizes like this
     //TODO: to consider tuning block_size for various data types
     const _DifferenceType __block_size = 8;
     alignas(__lane_size) _DifferenceType __lane[__block_size] = {0};
     while (__last - __first >= __block_size)
     {
         _DifferenceType __found = 0;
         _PSTL_PRAGMA_VECTOR_UNALIGNED // Do not generate peel loop part
             _PSTL_PRAGMA_SIMD_REDUCTION(|
                                         : __found) for (__i = 0; __i < __block_size - 1; ++__i)
         {
             //TODO: to improve SIMD vectorization
             const _DifferenceType __t = __pred(*(__first + __i), *(__first + __i + 1));
             __lane[__i] = __t;
             __found |= __t;
         }
 
         //Process a pair of elements on a boundary of a data block
         if (__first + __block_size < __last && __pred(*(__first + __i), *(__first + __i + 1)))
             __lane[__i] = __found = 1;
 
         if (__found)
         {
             if (__or_semantic)
                 return __first;
 
             // This will vectorize
             for (__i = 0; __i < __block_size; ++__i)
                 if (__lane[__i])
                     break;
             return __first + __i; //As far as found is true a __result (__lane[__i] is true) is guaranteed
         }
         __first += __block_size;
     }
     //Process the rest elements
     for (; __last - __first > 1; ++__first)
         if (__pred(*__first, *(__first + 1)))
             return __first;
 
     return __last;
 #endif
 }
 
 // It was created to reduce the code inside std::enable_if
 template <typename _Tp, typename _BinaryOperation>
 using is_arithmetic_plus = std::integral_constant<bool, std::is_arithmetic<_Tp>::value &&
                                                             std::is_same<_BinaryOperation, std::plus<_Tp>>::value>;
 
 template <typename _DifferenceType, typename _Tp, typename _BinaryOperation, typename _UnaryOperation>
 typename std::enable_if<is_arithmetic_plus<_Tp, _BinaryOperation>::value, _Tp>::type
 __simd_transform_reduce(_DifferenceType __n, _Tp __init, _BinaryOperation, _UnaryOperation __f) noexcept
 {
     _PSTL_PRAGMA_SIMD_REDUCTION(+ : __init)
     for (_DifferenceType __i = 0; __i < __n; ++__i)
         __init += __f(__i);
     return __init;
 }
 
 template <typename _Size, typename _Tp, typename _BinaryOperation, typename _UnaryOperation>
 typename std::enable_if<!is_arithmetic_plus<_Tp, _BinaryOperation>::value, _Tp>::type
 __simd_transform_reduce(_Size __n, _Tp __init, _BinaryOperation __binary_op, _UnaryOperation __f) noexcept
 {
     const _Size __block_size = __lane_size / sizeof(_Tp);
     if (__n > 2 * __block_size && __block_size > 1)
     {
         alignas(__lane_size) char __lane_[__lane_size];
         _Tp* __lane = reinterpret_cast<_Tp*>(__lane_);
 
         // initializer
         _PSTL_PRAGMA_SIMD
         for (_Size __i = 0; __i < __block_size; ++__i)
         {
             ::new (__lane + __i) _Tp(__binary_op(__f(__i), __f(__block_size + __i)));
         }
         // main loop
         _Size __i = 2 * __block_size;
         const _Size last_iteration = __block_size * (__n / __block_size);
         for (; __i < last_iteration; __i += __block_size)
         {
             _PSTL_PRAGMA_SIMD
             for (_Size __j = 0; __j < __block_size; ++__j)
             {
                 __lane[__j] = __binary_op(__lane[__j], __f(__i + __j));
             }
         }
         // remainder
         _PSTL_PRAGMA_SIMD
         for (_Size __j = 0; __j < __n - last_iteration; ++__j)
         {
             __lane[__j] = __binary_op(__lane[__j], __f(last_iteration + __j));
         }
         // combiner
         for (_Size __j = 0; __j < __block_size; ++__j)
         {
             __init = __binary_op(__init, __lane[__j]);
         }
         // destroyer
         _PSTL_PRAGMA_SIMD
         for (_Size __j = 0; __j < __block_size; ++__j)
         {
             __lane[__j].~_Tp();
         }
     }
     else
     {
         for (_Size __i = 0; __i < __n; ++__i)
         {
             __init = __binary_op(__init, __f(__i));
         }
     }
     return __init;
 }
 
 // Exclusive scan for "+" and arithmetic types
 template <class _InputIterator, class _Size, class _OutputIterator, class _UnaryOperation, class _Tp,
           class _BinaryOperation>
 typename std::enable_if<is_arithmetic_plus<_Tp, _BinaryOperation>::value, std::pair<_OutputIterator, _Tp>>::type
 __simd_scan(_InputIterator __first, _Size __n, _OutputIterator __result, _UnaryOperation __unary_op, _Tp __init,
             _BinaryOperation, /*Inclusive*/ std::false_type)
 {
     _PSTL_PRAGMA_SIMD_SCAN(+ : __init)
     for (_Size __i = 0; __i < __n; ++__i)
     {
         __result[__i] = __init;
         _PSTL_PRAGMA_SIMD_EXCLUSIVE_SCAN(__init)
         __init += __unary_op(__first[__i]);
     }
     return std::make_pair(__result + __n, __init);
 }
 
 // As soon as we cannot call __binary_op in "combiner" we create a wrapper over _Tp to encapsulate __binary_op
 template <typename _Tp, typename _BinaryOp>
 struct _Combiner
 {
     _Tp __value;
     _BinaryOp* __bin_op; // Here is a pointer to function because of default ctor
 
     _Combiner() : __value{}, __bin_op(nullptr) {}
     _Combiner(const _Tp& value, const _BinaryOp* bin_op) : __value(value), __bin_op(const_cast<_BinaryOp*>(bin_op)) {}
     _Combiner(const _Combiner& __obj) : __value{}, __bin_op(__obj.__bin_op) {}
 
     void
     operator()(const _Combiner& __obj)
     {
         __value = (*__bin_op)(__value, __obj.__value);
     }
 };
 
 // Exclusive scan for other binary operations and types
 template <class _InputIterator, class _Size, class _OutputIterator, class _UnaryOperation, class _Tp,
           class _BinaryOperation>
 typename std::enable_if<!is_arithmetic_plus<_Tp, _BinaryOperation>::value, std::pair<_OutputIterator, _Tp>>::type
 __simd_scan(_InputIterator __first, _Size __n, _OutputIterator __result, _UnaryOperation __unary_op, _Tp __init,
             _BinaryOperation __binary_op, /*Inclusive*/ std::false_type)
 {
     typedef _Combiner<_Tp, _BinaryOperation> _CombinerType;
     _CombinerType __init_{__init, &__binary_op};
 
     _PSTL_PRAGMA_DECLARE_REDUCTION(__bin_op, _CombinerType)
 
     _PSTL_PRAGMA_SIMD_SCAN(__bin_op : __init_)
     for (_Size __i = 0; __i < __n; ++__i)
     {
         __result[__i] = __init_.__value;
         _PSTL_PRAGMA_SIMD_EXCLUSIVE_SCAN(__init_)
         _PSTL_PRAGMA_FORCEINLINE
         __init_.__value = __binary_op(__init_.__value, __unary_op(__first[__i]));
     }
     return std::make_pair(__result + __n, __init_.__value);
 }
 
 // Inclusive scan for "+" and arithmetic types
 template <class _InputIterator, class _Size, class _OutputIterator, class _UnaryOperation, class _Tp,
           class _BinaryOperation>
 typename std::enable_if<is_arithmetic_plus<_Tp, _BinaryOperation>::value, std::pair<_OutputIterator, _Tp>>::type
 __simd_scan(_InputIterator __first, _Size __n, _OutputIterator __result, _UnaryOperation __unary_op, _Tp __init,
             _BinaryOperation, /*Inclusive*/ std::true_type)
 {
     _PSTL_PRAGMA_SIMD_SCAN(+ : __init)
     for (_Size __i = 0; __i < __n; ++__i)
     {
         __init += __unary_op(__first[__i]);
         _PSTL_PRAGMA_SIMD_INCLUSIVE_SCAN(__init)
         __result[__i] = __init;
     }
     return std::make_pair(__result + __n, __init);
 }
 
 // Inclusive scan for other binary operations and types
 template <class _InputIterator, class _Size, class _OutputIterator, class _UnaryOperation, class _Tp,
           class _BinaryOperation>
 typename std::enable_if<!is_arithmetic_plus<_Tp, _BinaryOperation>::value, std::pair<_OutputIterator, _Tp>>::type
 __simd_scan(_InputIterator __first, _Size __n, _OutputIterator __result, _UnaryOperation __unary_op, _Tp __init,
             _BinaryOperation __binary_op, std::true_type)
 {
     typedef _Combiner<_Tp, _BinaryOperation> _CombinerType;
     _CombinerType __init_{__init, &__binary_op};
 
     _PSTL_PRAGMA_DECLARE_REDUCTION(__bin_op, _CombinerType)
 
     _PSTL_PRAGMA_SIMD_SCAN(__bin_op : __init_)
     for (_Size __i = 0; __i < __n; ++__i)
     {
         _PSTL_PRAGMA_FORCEINLINE
         __init_.__value = __binary_op(__init_.__value, __unary_op(__first[__i]));
         _PSTL_PRAGMA_SIMD_INCLUSIVE_SCAN(__init_)
         __result[__i] = __init_.__value;
     }
     return std::make_pair(__result + __n, __init_.__value);
 }
 
 // [restriction] - std::iterator_traits<_ForwardIterator>::value_type should be DefaultConstructible.
 // complexity [violation] - We will have at most (__n-1 + number_of_lanes) comparisons instead of at most __n-1.
 template <typename _ForwardIterator, typename _Size, typename _Compare>
 _ForwardIterator
 __simd_min_element(_ForwardIterator __first, _Size __n, _Compare __comp) noexcept
 {
     if (__n == 0)
     {
         return __first;
     }
 
     typedef typename std::iterator_traits<_ForwardIterator>::value_type _ValueType;
     struct _ComplexType
     {
         _ValueType __min_val;
         _Size __min_ind;
         _Compare* __min_comp;
 
         _ComplexType() : __min_val{}, __min_ind{}, __min_comp(nullptr) {}
         _ComplexType(const _ValueType& val, const _Compare* comp)
             : __min_val(val), __min_ind(0), __min_comp(const_cast<_Compare*>(comp))
         {
         }
         _ComplexType(const _ComplexType& __obj)
             : __min_val(__obj.__min_val), __min_ind(__obj.__min_ind), __min_comp(__obj.__min_comp)
         {
         }
 
         _PSTL_PRAGMA_DECLARE_SIMD
         void
         operator()(const _ComplexType& __obj)
         {
             if (!(*__min_comp)(__min_val, __obj.__min_val) &&
                 ((*__min_comp)(__obj.__min_val, __min_val) || __obj.__min_ind - __min_ind < 0))
             {
                 __min_val = __obj.__min_val;
                 __min_ind = __obj.__min_ind;
             }
         }
     };
 
     _ComplexType __init{*__first, &__comp};
 
     _PSTL_PRAGMA_DECLARE_REDUCTION(__min_func, _ComplexType)
 
     _PSTL_PRAGMA_SIMD_REDUCTION(__min_func : __init)
     for (_Size __i = 1; __i < __n; ++__i)
     {
         const _ValueType __min_val = __init.__min_val;
         const _ValueType __current = __first[__i];
         if (__comp(__current, __min_val))
         {
             __init.__min_val = __current;
             __init.__min_ind = __i;
         }
     }
     return __first + __init.__min_ind;
 }
 
 // [restriction] - std::iterator_traits<_ForwardIterator>::value_type should be DefaultConstructible.
 // complexity [violation] - We will have at most (2*(__n-1) + 4*number_of_lanes) comparisons instead of at most [1.5*(__n-1)].
 template <typename _ForwardIterator, typename _Size, typename _Compare>
 std::pair<_ForwardIterator, _ForwardIterator>
 __simd_minmax_element(_ForwardIterator __first, _Size __n, _Compare __comp) noexcept
 {
     if (__n == 0)
     {
         return std::make_pair(__first, __first);
     }
     typedef typename std::iterator_traits<_ForwardIterator>::value_type _ValueType;
 
     struct _ComplexType
     {
         _ValueType __min_val;
         _ValueType __max_val;
         _Size __min_ind;
         _Size __max_ind;
         _Compare* __minmax_comp;
 
         _ComplexType() : __min_val{}, __max_val{}, __min_ind{}, __max_ind{}, __minmax_comp(nullptr) {}
         _ComplexType(const _ValueType& min_val, const _ValueType& max_val, const _Compare* comp)
             : __min_val(min_val), __max_val(max_val), __min_ind(0), __max_ind(0),
               __minmax_comp(const_cast<_Compare*>(comp))
         {
         }
         _ComplexType(const _ComplexType& __obj)
             : __min_val(__obj.__min_val), __max_val(__obj.__max_val), __min_ind(__obj.__min_ind),
               __max_ind(__obj.__max_ind), __minmax_comp(__obj.__minmax_comp)
         {
         }
 
         void
         operator()(const _ComplexType& __obj)
         {
             // min
             if ((*__minmax_comp)(__obj.__min_val, __min_val))
             {
                 __min_val = __obj.__min_val;
                 __min_ind = __obj.__min_ind;
             }
             else if (!(*__minmax_comp)(__min_val, __obj.__min_val))
             {
                 __min_val = __obj.__min_val;
                 __min_ind = (__min_ind - __obj.__min_ind < 0) ? __min_ind : __obj.__min_ind;
             }
 
             // max
             if ((*__minmax_comp)(__max_val, __obj.__max_val))
             {
                 __max_val = __obj.__max_val;
                 __max_ind = __obj.__max_ind;
             }
             else if (!(*__minmax_comp)(__obj.__max_val, __max_val))
             {
                 __max_val = __obj.__max_val;
                 __max_ind = (__max_ind - __obj.__max_ind < 0) ? __obj.__max_ind : __max_ind;
             }
         }
     };
 
     _ComplexType __init{*__first, *__first, &__comp};
 
     _PSTL_PRAGMA_DECLARE_REDUCTION(__min_func, _ComplexType);
 
     _PSTL_PRAGMA_SIMD_REDUCTION(__min_func : __init)
     for (_Size __i = 1; __i < __n; ++__i)
     {
         auto __min_val = __init.__min_val;
         auto __max_val = __init.__max_val;
         auto __current = __first + __i;
         if (__comp(*__current, __min_val))
         {
             __init.__min_val = *__current;
             __init.__min_ind = __i;
         }
         else if (!__comp(*__current, __max_val))
         {
             __init.__max_val = *__current;
             __init.__max_ind = __i;
         }
     }
     return std::make_pair(__first + __init.__min_ind, __first + __init.__max_ind);
 }
 
 template <class _InputIterator, class _DifferenceType, class _OutputIterator1, class _OutputIterator2,
           class _UnaryPredicate>
 std::pair<_OutputIterator1, _OutputIterator2>
 __simd_partition_copy(_InputIterator __first, _DifferenceType __n, _OutputIterator1 __out_true,
                       _OutputIterator2 __out_false, _UnaryPredicate __pred) noexcept
 {
     _DifferenceType __cnt_true = 0, __cnt_false = 0;
 
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 0; __i < __n; ++__i)
     {
         _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC_2ARGS(__cnt_true : 1, __cnt_false : 1)
         if (__pred(__first[__i]))
         {
             __out_true[__cnt_true] = __first[__i];
             ++__cnt_true;
         }
         else
         {
             __out_false[__cnt_false] = __first[__i];
             ++__cnt_false;
         }
     }
     return std::make_pair(__out_true + __cnt_true, __out_false + __cnt_false);
 }
 
 template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
 _ForwardIterator1
 __simd_find_first_of(_ForwardIterator1 __first, _ForwardIterator1 __last, _ForwardIterator2 __s_first,
                      _ForwardIterator2 __s_last, _BinaryPredicate __pred) noexcept
 {
     typedef typename std::iterator_traits<_ForwardIterator1>::difference_type _DifferencType;
 
     const _DifferencType __n1 = __last - __first;
     const _DifferencType __n2 = __s_last - __s_first;
     if (__n1 == 0 || __n2 == 0)
     {
         return __last; // according to the standard
     }
 
     // Common case
     // If first sequence larger than second then we'll run simd_first with parameters of first sequence.
     // Otherwise, vice versa.
     if (__n1 < __n2)
     {
         for (; __first != __last; ++__first)
         {
             if (__unseq_backend::__simd_or(
                     __s_first, __n2,
                     __internal::__equal_value_by_pred<decltype(*__first), _BinaryPredicate>(*__first, __pred)))
             {
                 return __first;
             }
         }
     }
     else
     {
         for (; __s_first != __s_last; ++__s_first)
         {
             const auto __result = __unseq_backend::__simd_first(
                 __first, _DifferencType(0), __n1, [__s_first, &__pred](_ForwardIterator1 __it, _DifferencType __i) {
                     return __pred(__it[__i], *__s_first);
                 });
             if (__result != __last)
             {
                 return __result;
             }
         }
     }
     return __last;
 }
 
 template <class _RandomAccessIterator, class _DifferenceType, class _UnaryPredicate>
 _RandomAccessIterator
 __simd_remove_if(_RandomAccessIterator __first, _DifferenceType __n, _UnaryPredicate __pred) noexcept
 {
     // find first element we need to remove
     auto __current = __unseq_backend::__simd_first(
         __first, _DifferenceType(0), __n,
         [&__pred](_RandomAccessIterator __it, _DifferenceType __i) { return __pred(__it[__i]); });
     __n -= __current - __first;
 
     // if we have in sequence only one element that pred(__current[1]) != false we can exit the function
     if (__n < 2)
     {
         return __current;
     }
 
     _DifferenceType __cnt = 0;
     _PSTL_PRAGMA_SIMD
     for (_DifferenceType __i = 1; __i < __n; ++__i)
     {
         _PSTL_PRAGMA_SIMD_ORDERED_MONOTONIC(__cnt : 1)
         if (!__pred(__current[__i]))
         {
             __current[__cnt] = std::move(__current[__i]);
             ++__cnt;
         }
     }
     return __current + __cnt;
 }
 } // namespace __unseq_backend
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_UNSEQ_BACKEND_SIMD_H */

libcxx/include/__pstl/internal/utils.h

 // -*- C++ -*-
 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef _PSTL_UTILS_H
 #define _PSTL_UTILS_H
 
 #include <iterator>
 #include <new>
 #include <utility>
 
 _PSTL_HIDE_FROM_ABI_PUSH
 
 namespace __pstl {
 namespace __internal {
 
 template <typename _Fp>
 auto __except_handler(_Fp __f) -> decltype(__f()) {
 #ifndef _LIBCPP_HAS_NO_EXCEPTIONS
   try {
 #endif // _LIBCPP_HAS_NO_EXCEPTIONS
     return __f();
 #ifndef _LIBCPP_HAS_NO_EXCEPTIONS
   } catch (const std::bad_alloc&) {
     throw;            // re-throw bad_alloc according to the standard [algorithms.parallel.exceptions]
   } catch (...) {
     std::terminate(); // Good bye according to the standard [algorithms.parallel.exceptions]
   }
 #endif // _LIBCPP_HAS_NO_EXCEPTIONS
 }
 
 template <typename _Fp>
 void __invoke_if(std::true_type, _Fp __f) {
   __f();
 }
 
 template <typename _Fp>
 void __invoke_if(std::false_type, _Fp) {}
 
 template <typename _Fp>
 void __invoke_if_not(std::false_type, _Fp __f) {
   __f();
 }
 
 template <typename _Fp>
 void __invoke_if_not(std::true_type, _Fp) {}
 
 template <typename _F1, typename _F2>
 auto __invoke_if_else(std::true_type, _F1 __f1, _F2) -> decltype(__f1()) {
   return __f1();
 }
 
 template <typename _F1, typename _F2>
 auto __invoke_if_else(std::false_type, _F1, _F2 __f2) -> decltype(__f2()) {
   return __f2();
 }
 
 //! Unary operator that returns reference to its argument.
 struct __no_op {
   template <typename _Tp>
   _Tp&& operator()(_Tp&& __a) const {
     return std::forward<_Tp>(__a);
   }
 };
 
 template <typename _Pred>
 class __reorder_pred {
   _Pred _M_pred;
 
 public:
   explicit __reorder_pred(_Pred __pred) : _M_pred(__pred) {}
 
   template <typename _FTp, typename _STp>
   bool operator()(_FTp&& __a, _STp&& __b) {
     return _M_pred(std::forward<_STp>(__b), std::forward<_FTp>(__a));
   }
 };
 
 //! Like a polymorphic lambda for pred(...,value)
 template <typename _Tp, typename _Predicate>
 class __equal_value_by_pred {
   const _Tp& _M_value;
   _Predicate _M_pred;
 
 public:
   __equal_value_by_pred(const _Tp& __value, _Predicate __pred) : _M_value(__value), _M_pred(__pred) {}
 
   template <typename _Arg>
   bool operator()(_Arg&& __arg) {
     return _M_pred(std::forward<_Arg>(__arg), _M_value);
   }
 };
 
 //! Like a polymorphic lambda for ==value
 template <typename _Tp>
 class __equal_value {
   const _Tp& _M_value;
 
 public:
   explicit __equal_value(const _Tp& __value) : _M_value(__value) {}
 
   template <typename _Arg>
   bool operator()(_Arg&& __arg) const {
     return std::forward<_Arg>(__arg) == _M_value;
   }
 };
 
 //! Logical negation of ==value
 template <typename _Tp>
 class __not_equal_value {
   const _Tp& _M_value;
 
 public:
   explicit __not_equal_value(const _Tp& __value) : _M_value(__value) {}
 
   template <typename _Arg>
   bool operator()(_Arg&& __arg) const {
     return !(std::forward<_Arg>(__arg) == _M_value);
   }
 };
 
 template <typename _ForwardIterator, typename _Compare>
 _ForwardIterator __cmp_iterators_by_values(_ForwardIterator __a, _ForwardIterator __b, _Compare __comp) {
   if (__a < __b) { // we should return closer iterator
     return __comp(*__b, *__a) ? __b : __a;
   } else {
     return __comp(*__a, *__b) ? __a : __b;
   }
 }
 
 } // namespace __internal
 } // namespace __pstl
 
 _PSTL_HIDE_FROM_ABI_POP
 
 #endif /* _PSTL_UTILS_H */

libcxx/include/libcxx.imp

[
   { include: [ "@<__fwd/.*>", "private", "<fwd>", "public" ] },
   { include: [ "@<__ios/.*>", "private", "<ios>", "public" ] },
   { include: [ "@<__iterator/.*>", "private", "<iterator>", "public" ] },
   { include: [ "@<__locale_dir/.*>", "private", "<locale>", "public" ] },
   { include: [ "@<__memory/.*>", "private", "<memory>", "public" ] },
   { include: [ "@<__memory_resource/.*>", "private", "<memory_resource>", "public" ] },
   { include: [ "@<__mutex/.*>", "private", "<mutex>", "public" ] },
   { include: [ "@<__numeric/.*>", "private", "<numeric>", "public" ] },
+  { include: [ "@<__pstl/.*>", "private", "<pstl>", "public" ] },
   { include: [ "@<__random/.*>", "private", "<random>", "public" ] },
   { include: [ "@<__ranges/.*>", "private", "<ranges>", "public" ] },
   { include: [ "@<__string/.*>", "private", "<string>", "public" ] },
   { include: [ "@<__support/.*>", "private", "<support>", "public" ] },
   { include: [ "@<__system_error/.*>", "private", "<system_error>", "public" ] },
   { include: [ "@<__thread/.*>", "private", "<thread>", "public" ] },
   { include: [ "@<__tuple/.*>", "private", "<tuple>", "public" ] },
   { include: [ "@<__type_traits/.*>", "private", "<type_traits>", "public" ] },
   { include: [ "@<__utility/.*>", "private", "<utility>", "public" ] },
   { include: [ "@<__variant/.*>", "private", "<variant>", "public" ] },
 ]

libcxx/utils/data/ignore_format.txt

 libcxx/include/__numeric/midpoint.h
 libcxx/include/__numeric/partial_sum.h
 libcxx/include/__numeric/reduce.h
 libcxx/include/__numeric/transform_exclusive_scan.h
 libcxx/include/__numeric/transform_inclusive_scan.h
 libcxx/include/__numeric/transform_reduce.h
 libcxx/include/optional
 libcxx/include/ostream
-libcxx/include/pstl/internal/omp/parallel_for_each.h
-libcxx/include/pstl/internal/omp/parallel_for.h
-libcxx/include/pstl/internal/omp/parallel_invoke.h
-libcxx/include/pstl/internal/omp/parallel_merge.h
-libcxx/include/pstl/internal/omp/parallel_reduce.h
-libcxx/include/pstl/internal/omp/parallel_scan.h
-libcxx/include/pstl/internal/omp/parallel_stable_partial_sort.h
-libcxx/include/pstl/internal/omp/parallel_stable_sort.h
-libcxx/include/pstl/internal/omp/parallel_transform_reduce.h
-libcxx/include/pstl/internal/omp/parallel_transform_scan.h
-libcxx/include/pstl/internal/omp/util.h
-libcxx/include/pstl/internal/parallel_backend.h
-libcxx/include/pstl/internal/parallel_backend_serial.h
-libcxx/include/pstl/internal/parallel_backend_tbb.h
-libcxx/include/pstl/internal/parallel_backend_utils.h
-libcxx/include/pstl/internal/parallel_impl.h
-libcxx/include/pstl/internal/pstl_config.h
-libcxx/include/pstl/internal/unseq_backend_simd.h
+libcxx/include/__pstl/internal/omp/parallel_for_each.h
+libcxx/include/__pstl/internal/omp/parallel_for.h
+libcxx/include/__pstl/internal/omp/parallel_invoke.h
+libcxx/include/__pstl/internal/omp/parallel_merge.h
+libcxx/include/__pstl/internal/omp/parallel_reduce.h
+libcxx/include/__pstl/internal/omp/parallel_scan.h
+libcxx/include/__pstl/internal/omp/parallel_stable_partial_sort.h
+libcxx/include/__pstl/internal/omp/parallel_stable_sort.h
+libcxx/include/__pstl/internal/omp/parallel_transform_reduce.h
+libcxx/include/__pstl/internal/omp/parallel_transform_scan.h
+libcxx/include/__pstl/internal/omp/util.h
+libcxx/include/__pstl/internal/parallel_backend.h
+libcxx/include/__pstl/internal/parallel_backend_serial.h
+libcxx/include/__pstl/internal/parallel_backend_tbb.h
+libcxx/include/__pstl/internal/parallel_backend_utils.h
+libcxx/include/__pstl/internal/parallel_impl.h
+libcxx/include/__pstl/internal/pstl_config.h
+libcxx/include/__pstl/internal/unseq_backend_simd.h
 libcxx/include/queue
 libcxx/include/__random/bernoulli_distribution.h
 libcxx/include/__random/binomial_distribution.h
 libcxx/include/__random/cauchy_distribution.h
 libcxx/include/__random/chi_squared_distribution.h
 libcxx/include/__random/clamp_to_integral.h
 libcxx/include/__random/discard_block_engine.h
 libcxx/include/__random/discrete_distribution.h