2021-06-17 23:30:11 +08:00
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
|
|
|
// 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 _LIBCPP___ALGORITHM_SORT_H
|
|
|
|
#define _LIBCPP___ALGORITHM_SORT_H
|
|
|
|
|
|
|
|
#include <__algorithm/comp.h>
|
|
|
|
#include <__algorithm/comp_ref_type.h>
|
2022-07-18 09:11:51 +08:00
|
|
|
#include <__algorithm/iterator_operations.h>
|
2021-06-17 23:30:11 +08:00
|
|
|
#include <__algorithm/min_element.h>
|
|
|
|
#include <__algorithm/partial_sort.h>
|
|
|
|
#include <__algorithm/unwrap_iter.h>
|
2022-05-19 17:06:11 +08:00
|
|
|
#include <__bits>
|
2022-01-07 22:45:05 +08:00
|
|
|
#include <__config>
|
[libc++] Make the Debug mode a configuration-time only option
The debug mode has been broken pretty much ever since it was shipped
because it was possible to enable the debug mode in user code without
actually enabling it in the dylib, leading to ODR violations that
caused various kinds of failures.
This commit makes the debug mode a knob that is configured when
building the library and which can't be changed afterwards. This is
less flexible for users, however it will actually work as intended
and it will allow us, in the future, to add various kinds of checks
that do not assume the same ABI as the normal library. Furthermore,
this will make the debug mode more robust, which means that vendors
might be more tempted to support it properly, which hasn't been the
case with the current debug mode.
This patch shouldn't break any user code, except folks who are building
against a library that doesn't have the debug mode enabled and who try
to enable the debug mode in their code. Such users will get a compile-time
error explaining that this configuration isn't supported anymore.
In the future, we should further increase the granularity of the debug
mode checks so that we can cherry-pick which checks to enable, like we
do for unspecified behavior randomization.
Differential Revision: https://reviews.llvm.org/D122941
2022-04-02 04:38:30 +08:00
|
|
|
#include <__debug>
|
2022-07-03 22:52:22 +08:00
|
|
|
#include <__debug_utils/randomize_range.h>
|
2022-05-21 06:45:51 +08:00
|
|
|
#include <__functional/operations.h>
|
2022-06-17 06:20:53 +08:00
|
|
|
#include <__functional/ranges_operations.h>
|
2022-06-11 01:53:10 +08:00
|
|
|
#include <__iterator/iterator_traits.h>
|
2022-05-19 17:06:11 +08:00
|
|
|
#include <climits>
|
2021-06-17 23:30:11 +08:00
|
|
|
#include <memory>
|
|
|
|
|
|
|
|
#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
|
2022-02-02 09:16:40 +08:00
|
|
|
# pragma GCC system_header
|
2021-06-17 23:30:11 +08:00
|
|
|
#endif
|
|
|
|
|
|
|
|
_LIBCPP_BEGIN_NAMESPACE_STD
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
// Wraps an algorithm policy tag and a comparator in a single struct, used to pass the policy tag around without
|
|
|
|
// changing the number of template arguments (to keep the ABI stable). This is only used for the "range" policy tag.
|
|
|
|
//
|
|
|
|
// To create an object of this type, use `_WrapAlgPolicy<T, C>::type` -- see the specialization below for the rationale.
|
|
|
|
template <class _PolicyT, class _CompT, class = void>
|
|
|
|
struct _WrapAlgPolicy {
|
|
|
|
using type = _WrapAlgPolicy;
|
|
|
|
|
|
|
|
using _AlgPolicy = _PolicyT;
|
|
|
|
using _Comp = _CompT;
|
|
|
|
_Comp& __comp;
|
|
|
|
|
|
|
|
_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX11
|
|
|
|
_WrapAlgPolicy(_Comp& __c) : __comp(__c) {}
|
|
|
|
};
|
|
|
|
|
|
|
|
// Specialization for the "classic" policy tag that avoids creating a struct and simply defines an alias for the
|
|
|
|
// comparator. When unwrapping, a pristine comparator is always considered to have the "classic" tag attached. Passing
|
|
|
|
// the pristine comparator where possible allows using template instantiations from the dylib.
|
|
|
|
template <class _PolicyT, class _CompT>
|
|
|
|
struct _WrapAlgPolicy<_PolicyT, _CompT, __enable_if_t<std::is_same<_PolicyT, _ClassicAlgPolicy>::value> > {
|
|
|
|
using type = _CompT;
|
|
|
|
};
|
|
|
|
|
|
|
|
// Unwraps a pristine functor (e.g. `std::less`) as if it were wrapped using `_WrapAlgPolicy`. The policy tag is always
|
|
|
|
// set to "classic".
|
|
|
|
template <class _CompT>
|
|
|
|
struct _UnwrapAlgPolicy {
|
|
|
|
using _AlgPolicy = _ClassicAlgPolicy;
|
|
|
|
using _Comp = _CompT;
|
|
|
|
|
|
|
|
_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX11 static
|
|
|
|
_Comp __get_comp(_Comp __comp) { return __comp; }
|
|
|
|
};
|
|
|
|
|
|
|
|
// Unwraps a `_WrapAlgPolicy` struct.
|
|
|
|
template <class... _Ts>
|
|
|
|
struct _UnwrapAlgPolicy<_WrapAlgPolicy<_Ts...> > {
|
|
|
|
using _Wrapped = _WrapAlgPolicy<_Ts...>;
|
|
|
|
using _AlgPolicy = typename _Wrapped::_AlgPolicy;
|
|
|
|
using _Comp = typename _Wrapped::_Comp;
|
|
|
|
|
|
|
|
_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX11 static
|
|
|
|
_Comp __get_comp(_Wrapped& __w) { return __w.__comp; }
|
|
|
|
};
|
|
|
|
|
2021-06-17 23:30:11 +08:00
|
|
|
// stable, 2-3 compares, 0-2 swaps
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _ForwardIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
_LIBCPP_CONSTEXPR_AFTER_CXX11 unsigned __sort3(_ForwardIterator __x, _ForwardIterator __y, _ForwardIterator __z,
|
|
|
|
_Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
2022-04-09 05:02:25 +08:00
|
|
|
unsigned __r = 0;
|
|
|
|
if (!__c(*__y, *__x)) // if x <= y
|
|
|
|
{
|
|
|
|
if (!__c(*__z, *__y)) // if y <= z
|
|
|
|
return __r; // x <= y && y <= z
|
|
|
|
// x <= y && y > z
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__y, __z); // x <= z && y < z
|
2022-04-09 05:02:25 +08:00
|
|
|
__r = 1;
|
|
|
|
if (__c(*__y, *__x)) // if x > y
|
2021-06-17 23:30:11 +08:00
|
|
|
{
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x, __y); // x < y && y <= z
|
2022-04-09 05:02:25 +08:00
|
|
|
__r = 2;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
return __r; // x <= y && y < z
|
|
|
|
}
|
|
|
|
if (__c(*__z, *__y)) // x > y, if y > z
|
|
|
|
{
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x, __z); // x < y && y < z
|
2022-04-09 05:02:25 +08:00
|
|
|
__r = 1;
|
2021-06-17 23:30:11 +08:00
|
|
|
return __r;
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x, __y); // x > y && y <= z
|
2022-04-09 05:02:25 +08:00
|
|
|
__r = 1; // x < y && x <= z
|
|
|
|
if (__c(*__z, *__y)) // if y > z
|
|
|
|
{
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__y, __z); // x <= y && y < z
|
2022-04-09 05:02:25 +08:00
|
|
|
__r = 2;
|
|
|
|
}
|
|
|
|
return __r;
|
|
|
|
} // x <= y && y <= z
|
2021-06-17 23:30:11 +08:00
|
|
|
|
|
|
|
// stable, 3-6 compares, 0-5 swaps
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _ForwardIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
unsigned __sort4(_ForwardIterator __x1, _ForwardIterator __x2, _ForwardIterator __x3, _ForwardIterator __x4,
|
|
|
|
_Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
|
|
|
unsigned __r = std::__sort3<_AlgPolicy, _Compare>(__x1, __x2, __x3, __c);
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__c(*__x4, *__x3)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x3, __x4);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__r;
|
|
|
|
if (__c(*__x3, *__x2)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x2, __x3);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__r;
|
|
|
|
if (__c(*__x2, *__x1)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x1, __x2);
|
2021-06-17 23:30:11 +08:00
|
|
|
++__r;
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
|
|
|
return __r;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
// stable, 4-10 compares, 0-9 swaps
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _WrappedComp, class _ForwardIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
_LIBCPP_HIDDEN unsigned __sort5(_ForwardIterator __x1, _ForwardIterator __x2, _ForwardIterator __x3,
|
2022-07-18 09:11:51 +08:00
|
|
|
_ForwardIterator __x4, _ForwardIterator __x5, _WrappedComp __wrapped_comp) {
|
|
|
|
using _Unwrap = _UnwrapAlgPolicy<_WrappedComp>;
|
|
|
|
using _AlgPolicy = typename _Unwrap::_AlgPolicy;
|
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
|
|
|
using _Compare = typename _Unwrap::_Comp;
|
|
|
|
_Compare __c = _Unwrap::__get_comp(__wrapped_comp);
|
|
|
|
|
|
|
|
unsigned __r = std::__sort4<_AlgPolicy, _Compare>(__x1, __x2, __x3, __x4, __c);
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__c(*__x5, *__x4)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x4, __x5);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__r;
|
|
|
|
if (__c(*__x4, *__x3)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x3, __x4);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__r;
|
|
|
|
if (__c(*__x3, *__x2)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x2, __x3);
|
2021-06-17 23:30:11 +08:00
|
|
|
++__r;
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__c(*__x2, *__x1)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__x1, __x2);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__r;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
|
|
|
return __r;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _ForwardIterator>
|
|
|
|
_LIBCPP_HIDDEN unsigned __sort5_wrap_policy(
|
|
|
|
_ForwardIterator __x1, _ForwardIterator __x2, _ForwardIterator __x3, _ForwardIterator __x4, _ForwardIterator __x5,
|
|
|
|
_Compare __c) {
|
|
|
|
using _WrappedComp = typename _WrapAlgPolicy<_AlgPolicy, _Compare>::type;
|
|
|
|
_WrappedComp __wrapped_comp(__c);
|
|
|
|
return std::__sort5<_WrappedComp>(
|
|
|
|
std::move(__x1), std::move(__x2), std::move(__x3), std::move(__x4), std::move(__x5), __wrapped_comp);
|
|
|
|
}
|
|
|
|
|
2022-06-17 06:20:53 +08:00
|
|
|
// The comparator being simple is a prerequisite for using the branchless optimization.
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
template <class _Tp>
|
|
|
|
struct __is_simple_comparator : false_type {};
|
|
|
|
template <class _Tp>
|
|
|
|
struct __is_simple_comparator<__less<_Tp>&> : true_type {};
|
|
|
|
template <class _Tp>
|
|
|
|
struct __is_simple_comparator<less<_Tp>&> : true_type {};
|
|
|
|
template <class _Tp>
|
|
|
|
struct __is_simple_comparator<greater<_Tp>&> : true_type {};
|
2022-06-17 06:20:53 +08:00
|
|
|
#if _LIBCPP_STD_VER > 17
|
|
|
|
template <>
|
|
|
|
struct __is_simple_comparator<ranges::less&> : true_type {};
|
|
|
|
template <>
|
|
|
|
struct __is_simple_comparator<ranges::greater&> : true_type {};
|
|
|
|
#endif
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
|
|
|
|
template <class _Compare, class _Iter, class _Tp = typename iterator_traits<_Iter>::value_type>
|
|
|
|
using __use_branchless_sort =
|
|
|
|
integral_constant<bool, __is_cpp17_contiguous_iterator<_Iter>::value && sizeof(_Tp) <= sizeof(void*) &&
|
|
|
|
is_arithmetic<_Tp>::value && __is_simple_comparator<_Compare>::value>;
|
|
|
|
|
|
|
|
// Ensures that __c(*__x, *__y) is true by swapping *__x and *__y if necessary.
|
|
|
|
template <class _Compare, class _RandomAccessIterator>
|
|
|
|
inline _LIBCPP_HIDE_FROM_ABI void __cond_swap(_RandomAccessIterator __x, _RandomAccessIterator __y, _Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
// Note: this function behaves correctly even with proxy iterators (because it relies on `value_type`).
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
using value_type = typename iterator_traits<_RandomAccessIterator>::value_type;
|
|
|
|
bool __r = __c(*__x, *__y);
|
|
|
|
value_type __tmp = __r ? *__x : *__y;
|
|
|
|
*__y = __r ? *__y : *__x;
|
|
|
|
*__x = __tmp;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Ensures that *__x, *__y and *__z are ordered according to the comparator __c,
|
|
|
|
// under the assumption that *__y and *__z are already ordered.
|
|
|
|
template <class _Compare, class _RandomAccessIterator>
|
|
|
|
inline _LIBCPP_HIDE_FROM_ABI void __partially_sorted_swap(_RandomAccessIterator __x, _RandomAccessIterator __y,
|
|
|
|
_RandomAccessIterator __z, _Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
// Note: this function behaves correctly even with proxy iterators (because it relies on `value_type`).
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
using value_type = typename iterator_traits<_RandomAccessIterator>::value_type;
|
|
|
|
bool __r = __c(*__z, *__x);
|
|
|
|
value_type __tmp = __r ? *__z : *__x;
|
|
|
|
*__z = __r ? *__x : *__z;
|
|
|
|
__r = __c(__tmp, *__y);
|
|
|
|
*__x = __r ? *__x : *__y;
|
|
|
|
*__y = __r ? *__y : __tmp;
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class, class _Compare, class _RandomAccessIterator>
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI __enable_if_t<__use_branchless_sort<_Compare, _RandomAccessIterator>::value, void>
|
|
|
|
__sort3_maybe_branchless(_RandomAccessIterator __x1, _RandomAccessIterator __x2, _RandomAccessIterator __x3,
|
|
|
|
_Compare __c) {
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x2, __x3, __c);
|
|
|
|
_VSTD::__partially_sorted_swap<_Compare>(__x1, __x2, __x3, __c);
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _RandomAccessIterator>
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI __enable_if_t<!__use_branchless_sort<_Compare, _RandomAccessIterator>::value, void>
|
|
|
|
__sort3_maybe_branchless(_RandomAccessIterator __x1, _RandomAccessIterator __x2, _RandomAccessIterator __x3,
|
|
|
|
_Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort3<_AlgPolicy, _Compare>(__x1, __x2, __x3, __c);
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class, class _Compare, class _RandomAccessIterator>
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI __enable_if_t<__use_branchless_sort<_Compare, _RandomAccessIterator>::value, void>
|
|
|
|
__sort4_maybe_branchless(_RandomAccessIterator __x1, _RandomAccessIterator __x2, _RandomAccessIterator __x3,
|
|
|
|
_RandomAccessIterator __x4, _Compare __c) {
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x1, __x3, __c);
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x2, __x4, __c);
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x1, __x2, __c);
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x3, __x4, __c);
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x2, __x3, __c);
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _RandomAccessIterator>
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI __enable_if_t<!__use_branchless_sort<_Compare, _RandomAccessIterator>::value, void>
|
|
|
|
__sort4_maybe_branchless(_RandomAccessIterator __x1, _RandomAccessIterator __x2, _RandomAccessIterator __x3,
|
|
|
|
_RandomAccessIterator __x4, _Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort4<_AlgPolicy, _Compare>(__x1, __x2, __x3, __x4, __c);
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class, class _Compare, class _RandomAccessIterator>
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI __enable_if_t<__use_branchless_sort<_Compare, _RandomAccessIterator>::value, void>
|
|
|
|
__sort5_maybe_branchless(_RandomAccessIterator __x1, _RandomAccessIterator __x2, _RandomAccessIterator __x3,
|
|
|
|
_RandomAccessIterator __x4, _RandomAccessIterator __x5, _Compare __c) {
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x1, __x2, __c);
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x4, __x5, __c);
|
|
|
|
_VSTD::__partially_sorted_swap<_Compare>(__x3, __x4, __x5, __c);
|
|
|
|
_VSTD::__cond_swap<_Compare>(__x2, __x5, __c);
|
|
|
|
_VSTD::__partially_sorted_swap<_Compare>(__x1, __x3, __x4, __c);
|
|
|
|
_VSTD::__partially_sorted_swap<_Compare>(__x2, __x3, __x4, __c);
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _RandomAccessIterator>
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI __enable_if_t<!__use_branchless_sort<_Compare, _RandomAccessIterator>::value, void>
|
|
|
|
__sort5_maybe_branchless(_RandomAccessIterator __x1, _RandomAccessIterator __x2, _RandomAccessIterator __x3,
|
|
|
|
_RandomAccessIterator __x4, _RandomAccessIterator __x5, _Compare __c) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort5_wrap_policy<_AlgPolicy, _Compare>(__x1, __x2, __x3, __x4, __x5, __c);
|
Introduce branchless sorting functions for sort3, sort4 and sort5.
We are introducing branchless variants for sort3, sort4 and sort5.
These sorting functions have been generated using Reinforcement
Learning and aim to replace __sort3, __sort4 and __sort5 variants
for integral types.
The libc++ benchmarks were run on isolated machines for Skylake, ARM and
AMD architectures and achieve statistically significant improvement in
sorting random integers on test cases from sort1 to sort262144 for
uint32 and uint64.
A full performance overview for Intel Skylake, AMD and Arm can be
found here: https://bit.ly/3AtesYf
Reviewed By: ldionne, #libc, philnik
Spies: daniel.mankowitz, mgrang, Quuxplusone, andreamichi, philnik, libcxx-commits, nilayvaish, kristof.beyls
Differential Revision: https://reviews.llvm.org/D118029
2022-04-08 14:58:48 +08:00
|
|
|
}
|
|
|
|
|
2021-06-17 23:30:11 +08:00
|
|
|
// Assumes size > 0
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _BidirectionalIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
_LIBCPP_CONSTEXPR_AFTER_CXX11 void __selection_sort(_BidirectionalIterator __first, _BidirectionalIterator __last,
|
|
|
|
_Compare __comp) {
|
|
|
|
_BidirectionalIterator __lm1 = __last;
|
|
|
|
for (--__lm1; __first != __lm1; ++__first) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_BidirectionalIterator __i = std::__min_element<_Compare>(__first, __last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__i != __first)
|
2022-07-18 09:11:51 +08:00
|
|
|
_IterOps<_AlgPolicy>::iter_swap(__first, __i);
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _BidirectionalIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
void __insertion_sort(_BidirectionalIterator __first, _BidirectionalIterator __last, _Compare __comp) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
2022-04-09 05:02:25 +08:00
|
|
|
typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
|
|
|
|
if (__first != __last) {
|
|
|
|
_BidirectionalIterator __i = __first;
|
|
|
|
for (++__i; __i != __last; ++__i) {
|
|
|
|
_BidirectionalIterator __j = __i;
|
2022-07-18 09:11:51 +08:00
|
|
|
value_type __t(_Ops::__iter_move(__j));
|
2022-04-09 05:02:25 +08:00
|
|
|
for (_BidirectionalIterator __k = __i; __k != __first && __comp(__t, *--__k); --__j)
|
2022-07-18 09:11:51 +08:00
|
|
|
*__j = _Ops::__iter_move(__k);
|
2022-04-09 05:02:25 +08:00
|
|
|
*__j = _VSTD::move(__t);
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _RandomAccessIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
void __insertion_sort_3(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
2022-04-09 05:02:25 +08:00
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
|
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
|
|
|
|
_RandomAccessIterator __j = __first + difference_type(2);
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort3_maybe_branchless<_AlgPolicy, _Compare>(__first, __first + difference_type(1), __j, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
for (_RandomAccessIterator __i = __j + difference_type(1); __i != __last; ++__i) {
|
|
|
|
if (__comp(*__i, *__j)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
value_type __t(_Ops::__iter_move(__i));
|
2022-04-09 05:02:25 +08:00
|
|
|
_RandomAccessIterator __k = __j;
|
|
|
|
__j = __i;
|
|
|
|
do {
|
2022-07-18 09:11:51 +08:00
|
|
|
*__j = _Ops::__iter_move(__k);
|
2022-04-09 05:02:25 +08:00
|
|
|
__j = __k;
|
|
|
|
} while (__j != __first && __comp(__t, *--__k));
|
|
|
|
*__j = _VSTD::move(__t);
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
__j = __i;
|
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _WrappedComp, class _RandomAccessIterator>
|
|
|
|
bool __insertion_sort_incomplete(
|
|
|
|
_RandomAccessIterator __first, _RandomAccessIterator __last, _WrappedComp __wrapped_comp) {
|
|
|
|
using _Unwrap = _UnwrapAlgPolicy<_WrappedComp>;
|
|
|
|
using _AlgPolicy = typename _Unwrap::_AlgPolicy;
|
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
|
|
|
using _Compare = typename _Unwrap::_Comp;
|
|
|
|
_Compare __comp = _Unwrap::__get_comp(__wrapped_comp);
|
|
|
|
|
2022-04-09 05:02:25 +08:00
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
|
|
|
|
switch (__last - __first) {
|
|
|
|
case 0:
|
|
|
|
case 1:
|
|
|
|
return true;
|
|
|
|
case 2:
|
|
|
|
if (__comp(*--__last, *__first))
|
2022-07-18 09:11:51 +08:00
|
|
|
_IterOps<_AlgPolicy>::iter_swap(__first, __last);
|
2022-04-09 05:02:25 +08:00
|
|
|
return true;
|
|
|
|
case 3:
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort3_maybe_branchless<_AlgPolicy, _Compare>(__first, __first + difference_type(1), --__last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return true;
|
|
|
|
case 4:
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort4_maybe_branchless<_AlgPolicy, _Compare>(
|
|
|
|
__first, __first + difference_type(1), __first + difference_type(2), --__last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return true;
|
|
|
|
case 5:
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort5_maybe_branchless<_AlgPolicy, _Compare>(
|
|
|
|
__first, __first + difference_type(1), __first + difference_type(2), __first + difference_type(3),
|
|
|
|
--__last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
|
|
|
|
_RandomAccessIterator __j = __first + difference_type(2);
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort3_maybe_branchless<_AlgPolicy, _Compare>(__first, __first + difference_type(1), __j, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
const unsigned __limit = 8;
|
|
|
|
unsigned __count = 0;
|
|
|
|
for (_RandomAccessIterator __i = __j + difference_type(1); __i != __last; ++__i) {
|
|
|
|
if (__comp(*__i, *__j)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
value_type __t(_Ops::__iter_move(__i));
|
2022-04-09 05:02:25 +08:00
|
|
|
_RandomAccessIterator __k = __j;
|
|
|
|
__j = __i;
|
|
|
|
do {
|
2022-07-18 09:11:51 +08:00
|
|
|
*__j = _Ops::__iter_move(__k);
|
2022-04-09 05:02:25 +08:00
|
|
|
__j = __k;
|
|
|
|
} while (__j != __first && __comp(__t, *--__k));
|
|
|
|
*__j = _VSTD::move(__t);
|
|
|
|
if (++__count == __limit)
|
|
|
|
return ++__i == __last;
|
|
|
|
}
|
|
|
|
__j = __i;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _BidirectionalIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
void __insertion_sort_move(_BidirectionalIterator __first1, _BidirectionalIterator __last1,
|
|
|
|
typename iterator_traits<_BidirectionalIterator>::value_type* __first2, _Compare __comp) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
2022-04-09 05:02:25 +08:00
|
|
|
typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
|
|
|
|
if (__first1 != __last1) {
|
|
|
|
__destruct_n __d(0);
|
|
|
|
unique_ptr<value_type, __destruct_n&> __h(__first2, __d);
|
|
|
|
value_type* __last2 = __first2;
|
2022-07-18 09:11:51 +08:00
|
|
|
::new ((void*)__last2) value_type(_Ops::__iter_move(__first1));
|
2022-04-09 05:02:25 +08:00
|
|
|
__d.template __incr<value_type>();
|
|
|
|
for (++__last2; ++__first1 != __last1; ++__last2) {
|
|
|
|
value_type* __j2 = __last2;
|
|
|
|
value_type* __i2 = __j2;
|
|
|
|
if (__comp(*__first1, *--__i2)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
::new ((void*)__j2) value_type(std::move(*__i2));
|
2022-04-09 05:02:25 +08:00
|
|
|
__d.template __incr<value_type>();
|
|
|
|
for (--__j2; __i2 != __first2 && __comp(*__first1, *--__i2); --__j2)
|
2022-07-18 09:11:51 +08:00
|
|
|
*__j2 = std::move(*__i2);
|
|
|
|
*__j2 = _Ops::__iter_move(__first1);
|
2022-04-09 05:02:25 +08:00
|
|
|
} else {
|
2022-07-18 09:11:51 +08:00
|
|
|
::new ((void*)__j2) value_type(_Ops::__iter_move(__first1));
|
2022-04-09 05:02:25 +08:00
|
|
|
__d.template __incr<value_type>();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
__h.release();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _Compare, class _RandomAccessIterator>
|
2022-04-09 05:02:25 +08:00
|
|
|
void __introsort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp,
|
|
|
|
typename iterator_traits<_RandomAccessIterator>::difference_type __depth) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _Ops = _IterOps<_AlgPolicy>;
|
|
|
|
|
2022-04-09 05:02:25 +08:00
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
|
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
|
|
|
|
const difference_type __limit =
|
|
|
|
is_trivially_copy_constructible<value_type>::value && is_trivially_copy_assignable<value_type>::value ? 30 : 6;
|
|
|
|
while (true) {
|
|
|
|
__restart:
|
|
|
|
difference_type __len = __last - __first;
|
|
|
|
switch (__len) {
|
2021-06-17 23:30:11 +08:00
|
|
|
case 0:
|
|
|
|
case 1:
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
2021-06-17 23:30:11 +08:00
|
|
|
case 2:
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__comp(*--__last, *__first))
|
2022-07-18 09:11:51 +08:00
|
|
|
_IterOps<_AlgPolicy>::iter_swap(__first, __last);
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
2021-06-17 23:30:11 +08:00
|
|
|
case 3:
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort3_maybe_branchless<_AlgPolicy, _Compare>(__first, __first + difference_type(1), --__last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
2021-06-17 23:30:11 +08:00
|
|
|
case 4:
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort4_maybe_branchless<_AlgPolicy, _Compare>(
|
|
|
|
__first, __first + difference_type(1), __first + difference_type(2), --__last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
2021-06-17 23:30:11 +08:00
|
|
|
case 5:
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort5_maybe_branchless<_AlgPolicy, _Compare>(
|
|
|
|
__first, __first + difference_type(1), __first + difference_type(2), __first + difference_type(3),
|
|
|
|
--__last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__len <= __limit) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__insertion_sort_3<_AlgPolicy, _Compare>(__first, __last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
// __len > 5
|
|
|
|
if (__depth == 0) {
|
|
|
|
// Fallback to heap sort as Introsort suggests.
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__partial_sort<_AlgPolicy, _Compare>(__first, __last, __last, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
--__depth;
|
|
|
|
_RandomAccessIterator __m = __first;
|
|
|
|
_RandomAccessIterator __lm1 = __last;
|
|
|
|
--__lm1;
|
|
|
|
unsigned __n_swaps;
|
2021-06-17 23:30:11 +08:00
|
|
|
{
|
2022-04-09 05:02:25 +08:00
|
|
|
difference_type __delta;
|
|
|
|
if (__len >= 1000) {
|
|
|
|
__delta = __len / 2;
|
|
|
|
__m += __delta;
|
|
|
|
__delta /= 2;
|
2022-07-18 09:11:51 +08:00
|
|
|
__n_swaps = std::__sort5_wrap_policy<_AlgPolicy, _Compare>(
|
|
|
|
__first, __first + __delta, __m, __m + __delta, __lm1, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
} else {
|
|
|
|
__delta = __len / 2;
|
|
|
|
__m += __delta;
|
2022-07-18 09:11:51 +08:00
|
|
|
__n_swaps = std::__sort3<_AlgPolicy, _Compare>(__first, __m, __lm1, __comp);
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
// *__m is median
|
|
|
|
// partition [__first, __m) < *__m and *__m <= [__m, __last)
|
|
|
|
// (this inhibits tossing elements equivalent to __m around unnecessarily)
|
|
|
|
_RandomAccessIterator __i = __first;
|
|
|
|
_RandomAccessIterator __j = __lm1;
|
|
|
|
// j points beyond range to be tested, *__m is known to be <= *__lm1
|
|
|
|
// The search going up is known to be guarded but the search coming down isn't.
|
|
|
|
// Prime the downward search with a guard.
|
|
|
|
if (!__comp(*__i, *__m)) // if *__first == *__m
|
2021-06-17 23:30:11 +08:00
|
|
|
{
|
2022-04-09 05:02:25 +08:00
|
|
|
// *__first == *__m, *__first doesn't go in first part
|
|
|
|
// manually guard downward moving __j against __i
|
|
|
|
while (true) {
|
|
|
|
if (__i == --__j) {
|
|
|
|
// *__first == *__m, *__m <= all other elements
|
|
|
|
// Parition instead into [__first, __i) == *__first and *__first < [__i, __last)
|
|
|
|
++__i; // __first + 1
|
|
|
|
__j = __last;
|
|
|
|
if (!__comp(*__first, *--__j)) // we need a guard if *__first == *(__last-1)
|
|
|
|
{
|
|
|
|
while (true) {
|
|
|
|
if (__i == __j)
|
|
|
|
return; // [__first, __last) all equivalent elements
|
|
|
|
if (__comp(*__first, *__i)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__i, __j);
|
2021-06-17 23:30:11 +08:00
|
|
|
++__n_swaps;
|
|
|
|
++__i;
|
2022-04-09 05:02:25 +08:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
++__i;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
|
|
|
// [__first, __i) == *__first and *__first < [__j, __last) and __j == __last - 1
|
|
|
|
if (__i == __j)
|
|
|
|
return;
|
|
|
|
while (true) {
|
|
|
|
while (!__comp(*__first, *__i))
|
|
|
|
++__i;
|
|
|
|
while (__comp(*__first, *--__j))
|
|
|
|
;
|
|
|
|
if (__i >= __j)
|
|
|
|
break;
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__i, __j);
|
2021-06-17 23:30:11 +08:00
|
|
|
++__n_swaps;
|
2022-04-09 05:02:25 +08:00
|
|
|
++__i;
|
|
|
|
}
|
|
|
|
// [__first, __i) == *__first and *__first < [__i, __last)
|
|
|
|
// The first part is sorted, sort the second part
|
|
|
|
// _VSTD::__sort<_Compare>(__i, __last, __comp);
|
|
|
|
__first = __i;
|
|
|
|
goto __restart;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__comp(*__j, *__m)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__i, __j);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__n_swaps;
|
|
|
|
break; // found guard for downward moving __j, now use unguarded partition
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
// It is known that *__i < *__m
|
|
|
|
++__i;
|
|
|
|
// j points beyond range to be tested, *__m is known to be <= *__lm1
|
|
|
|
// if not yet partitioned...
|
|
|
|
if (__i < __j) {
|
|
|
|
// known that *(__i - 1) < *__m
|
|
|
|
// known that __i <= __m
|
|
|
|
while (true) {
|
|
|
|
// __m still guards upward moving __i
|
|
|
|
while (__comp(*__i, *__m))
|
|
|
|
++__i;
|
|
|
|
// It is now known that a guard exists for downward moving __j
|
|
|
|
while (!__comp(*--__j, *__m))
|
|
|
|
;
|
|
|
|
if (__i > __j)
|
|
|
|
break;
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__i, __j);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__n_swaps;
|
|
|
|
// It is known that __m != __j
|
|
|
|
// If __m just moved, follow it
|
|
|
|
if (__m == __i)
|
|
|
|
__m = __j;
|
|
|
|
++__i;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
// [__first, __i) < *__m and *__m <= [__i, __last)
|
|
|
|
if (__i != __m && __comp(*__m, *__i)) {
|
2022-07-18 09:11:51 +08:00
|
|
|
_Ops::iter_swap(__i, __m);
|
2022-04-09 05:02:25 +08:00
|
|
|
++__n_swaps;
|
|
|
|
}
|
|
|
|
// [__first, __i) < *__i and *__i <= [__i+1, __last)
|
|
|
|
// If we were given a perfect partition, see if insertion sort is quick...
|
|
|
|
if (__n_swaps == 0) {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _WrappedComp = typename _WrapAlgPolicy<_AlgPolicy, _Compare>::type;
|
|
|
|
_WrappedComp __wrapped_comp(__comp);
|
|
|
|
bool __fs = std::__insertion_sort_incomplete<_WrappedComp>(__first, __i, __wrapped_comp);
|
|
|
|
if (std::__insertion_sort_incomplete<_WrappedComp>(__i + difference_type(1), __last, __wrapped_comp)) {
|
2022-04-09 05:02:25 +08:00
|
|
|
if (__fs)
|
|
|
|
return;
|
|
|
|
__last = __i;
|
|
|
|
continue;
|
|
|
|
} else {
|
|
|
|
if (__fs) {
|
2021-11-17 00:37:55 +08:00
|
|
|
__first = ++__i;
|
2022-04-09 05:02:25 +08:00
|
|
|
continue;
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
2022-04-09 05:02:25 +08:00
|
|
|
// sort smaller range with recursive call and larger with tail recursion elimination
|
|
|
|
if (__i - __first < __last - __i) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__introsort<_AlgPolicy, _Compare>(__first, __i, __comp, __depth);
|
2022-04-09 05:02:25 +08:00
|
|
|
__first = ++__i;
|
|
|
|
} else {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__introsort<_AlgPolicy, _Compare>(__i + difference_type(1), __last, __comp, __depth);
|
2022-04-09 05:02:25 +08:00
|
|
|
__last = __i;
|
|
|
|
}
|
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2021-11-17 00:37:55 +08:00
|
|
|
template <typename _Number>
|
|
|
|
inline _LIBCPP_HIDE_FROM_ABI _Number __log2i(_Number __n) {
|
2022-05-19 17:06:11 +08:00
|
|
|
if (__n == 0)
|
|
|
|
return 0;
|
|
|
|
if (sizeof(__n) <= sizeof(unsigned))
|
|
|
|
return sizeof(unsigned) * CHAR_BIT - 1 - __libcpp_clz(static_cast<unsigned>(__n));
|
|
|
|
if (sizeof(__n) <= sizeof(unsigned long))
|
|
|
|
return sizeof(unsigned long) * CHAR_BIT - 1 - __libcpp_clz(static_cast<unsigned long>(__n));
|
|
|
|
if (sizeof(__n) <= sizeof(unsigned long long))
|
|
|
|
return sizeof(unsigned long long) * CHAR_BIT - 1 - __libcpp_clz(static_cast<unsigned long long>(__n));
|
|
|
|
|
2021-11-17 00:37:55 +08:00
|
|
|
_Number __log2 = 0;
|
|
|
|
while (__n > 1) {
|
|
|
|
__log2++;
|
|
|
|
__n >>= 1;
|
|
|
|
}
|
|
|
|
return __log2;
|
|
|
|
}
|
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _WrappedComp, class _RandomAccessIterator>
|
|
|
|
void __sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _WrappedComp __wrapped_comp) {
|
2021-11-17 00:37:55 +08:00
|
|
|
typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
|
|
|
|
difference_type __depth_limit = 2 * __log2i(__last - __first);
|
2022-07-18 09:11:51 +08:00
|
|
|
|
|
|
|
using _Unwrap = _UnwrapAlgPolicy<_WrappedComp>;
|
|
|
|
using _AlgPolicy = typename _Unwrap::_AlgPolicy;
|
|
|
|
using _Compare = typename _Unwrap::_Comp;
|
|
|
|
_Compare __comp = _Unwrap::__get_comp(__wrapped_comp);
|
|
|
|
std::__introsort<_AlgPolicy, _Compare>(__first, __last, __comp, __depth_limit);
|
2021-11-17 00:37:55 +08:00
|
|
|
}
|
|
|
|
|
2021-06-17 23:30:11 +08:00
|
|
|
template <class _Compare, class _Tp>
|
2022-04-09 05:02:25 +08:00
|
|
|
inline _LIBCPP_INLINE_VISIBILITY void __sort(_Tp** __first, _Tp** __last, __less<_Tp*>&) {
|
|
|
|
__less<uintptr_t> __comp;
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort<__less<uintptr_t>&, uintptr_t*>((uintptr_t*)__first, (uintptr_t*)__last, __comp);
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2021-06-09 05:25:08 +08:00
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<char>&, char*>(char*, char*, __less<char>&);
|
2021-08-24 03:32:36 +08:00
|
|
|
#ifndef _LIBCPP_HAS_NO_WIDE_CHARACTERS
|
2021-06-09 05:25:08 +08:00
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<wchar_t>&, wchar_t*>(wchar_t*, wchar_t*, __less<wchar_t>&);
|
2021-08-24 03:32:36 +08:00
|
|
|
#endif
|
2021-06-09 05:25:08 +08:00
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<signed char>&, signed char*>(signed char*, signed char*, __less<signed char>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<unsigned char>&, unsigned char*>(unsigned char*, unsigned char*, __less<unsigned char>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<short>&, short*>(short*, short*, __less<short>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<unsigned short>&, unsigned short*>(unsigned short*, unsigned short*, __less<unsigned short>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<int>&, int*>(int*, int*, __less<int>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<unsigned>&, unsigned*>(unsigned*, unsigned*, __less<unsigned>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<long>&, long*>(long*, long*, __less<long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<unsigned long>&, unsigned long*>(unsigned long*, unsigned long*, __less<unsigned long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<long long>&, long long*>(long long*, long long*, __less<long long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<unsigned long long>&, unsigned long long*>(unsigned long long*, unsigned long long*, __less<unsigned long long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<float>&, float*>(float*, float*, __less<float>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<double>&, double*>(double*, double*, __less<double>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS void __sort<__less<long double>&, long double*>(long double*, long double*, __less<long double>&);
|
|
|
|
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<char>&, char*>(char*, char*, __less<char>&);
|
2021-08-24 03:32:36 +08:00
|
|
|
#ifndef _LIBCPP_HAS_NO_WIDE_CHARACTERS
|
2021-06-09 05:25:08 +08:00
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<wchar_t>&, wchar_t*>(wchar_t*, wchar_t*, __less<wchar_t>&);
|
2021-08-24 03:32:36 +08:00
|
|
|
#endif
|
2021-06-09 05:25:08 +08:00
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<signed char>&, signed char*>(signed char*, signed char*, __less<signed char>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned char>&, unsigned char*>(unsigned char*, unsigned char*, __less<unsigned char>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<short>&, short*>(short*, short*, __less<short>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned short>&, unsigned short*>(unsigned short*, unsigned short*, __less<unsigned short>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<int>&, int*>(int*, int*, __less<int>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned>&, unsigned*>(unsigned*, unsigned*, __less<unsigned>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<long>&, long*>(long*, long*, __less<long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned long>&, unsigned long*>(unsigned long*, unsigned long*, __less<unsigned long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<long long>&, long long*>(long long*, long long*, __less<long long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned long long>&, unsigned long long*>(unsigned long long*, unsigned long long*, __less<unsigned long long>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<float>&, float*>(float*, float*, __less<float>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<double>&, double*>(double*, double*, __less<double>&);
|
|
|
|
extern template _LIBCPP_FUNC_VIS bool __insertion_sort_incomplete<__less<long double>&, long double*>(long double*, long double*, __less<long double>&);
|
|
|
|
|
|
|
|
extern template _LIBCPP_FUNC_VIS unsigned __sort5<__less<long double>&, long double*>(long double*, long double*, long double*, long double*, long double*, __less<long double>&);
|
2021-06-17 23:30:11 +08:00
|
|
|
|
2022-07-18 09:11:51 +08:00
|
|
|
template <class _AlgPolicy, class _RandomAccessIterator, class _Comp>
|
2022-06-17 06:20:53 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17
|
|
|
|
void __sort_impl(_RandomAccessIterator __first, _RandomAccessIterator __last, _Comp& __comp) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__debug_randomize_range<_AlgPolicy>(__first, __last);
|
|
|
|
|
2022-06-17 06:20:53 +08:00
|
|
|
using _Comp_ref = typename __comp_ref_type<_Comp>::type;
|
2021-11-17 04:48:59 +08:00
|
|
|
if (__libcpp_is_constant_evaluated()) {
|
2022-07-20 11:10:02 +08:00
|
|
|
std::__partial_sort<_AlgPolicy>(__first, __last, __last, __comp);
|
2022-07-18 09:11:51 +08:00
|
|
|
|
2021-11-17 04:48:59 +08:00
|
|
|
} else {
|
2022-07-18 09:11:51 +08:00
|
|
|
using _WrappedComp = typename _WrapAlgPolicy<_AlgPolicy, _Comp_ref>::type;
|
|
|
|
_Comp_ref __comp_ref(__comp);
|
|
|
|
_WrappedComp __wrapped_comp(__comp_ref);
|
|
|
|
std::__sort<_WrappedComp>(std::__unwrap_iter(__first), std::__unwrap_iter(__last), __wrapped_comp);
|
2021-11-17 04:48:59 +08:00
|
|
|
}
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
2022-06-17 06:20:53 +08:00
|
|
|
template <class _RandomAccessIterator, class _Comp>
|
|
|
|
inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17
|
|
|
|
void sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp) {
|
2022-07-18 09:11:51 +08:00
|
|
|
std::__sort_impl<_ClassicAlgPolicy>(std::move(__first), std::move(__last), __comp);
|
2022-06-17 06:20:53 +08:00
|
|
|
}
|
|
|
|
|
2021-06-17 23:30:11 +08:00
|
|
|
template <class _RandomAccessIterator>
|
2022-06-17 06:20:53 +08:00
|
|
|
inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17
|
|
|
|
void sort(_RandomAccessIterator __first, _RandomAccessIterator __last) {
|
|
|
|
std::sort(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
|
2021-06-17 23:30:11 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
_LIBCPP_END_NAMESPACE_STD
|
|
|
|
|
|
|
|
#endif // _LIBCPP___ALGORITHM_SORT_H
|