Merge main to feature/flat_set

.github/workflows/update-status-chart.yml

@@ -6,6 +6,8 @@ on:
   schedule:
     - cron: "0 7 * * *"
   workflow_dispatch:
+permissions:
+  contents: write
 jobs:
   build:
     if: ${{ github.repository == 'microsoft/STL' }}

benchmarks/CMakeLists.txt

@@ -61,6 +61,9 @@ endif()
 
 set(CMAKE_BUILD_TYPE RelWithDebInfo)
 
+# /utf-8 affects <format>.
+add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/diagnostics:caret;/W4;/WX;/w14265;/w15038;/w15262;/utf-8>")
+
 if(NOT EXISTS "${CMAKE_CURRENT_LIST_DIR}/google-benchmark/.git")
     message(FATAL_ERROR "google-benchmark is not checked out; make sure to run\n    git submodule update --init benchmarks/google-benchmark")
 endif()
@@ -107,6 +110,7 @@ endfunction()
 
 add_benchmark(bitset_to_string src/bitset_to_string.cpp)
 add_benchmark(locale_classic src/locale_classic.cpp)
+add_benchmark(minmax_element src/minmax_element.cpp)
 add_benchmark(path_lexically_normal src/path_lexically_normal.cpp)
 add_benchmark(priority_queue_push_range src/priority_queue_push_range.cpp)
 add_benchmark(random_integer_generation src/random_integer_generation.cpp)

benchmarks/inc/udt.hpp

@@ -3,18 +3,18 @@
 
 #pragma once
 
-template <typename Contained>
+template <typename Data>
 struct aggregate {
-    Contained c;
+    Data c;
 
     friend bool operator==(const aggregate&, const aggregate&) = default;
 };
 
-template <typename Contained>
+template <typename Data>
 struct non_trivial {
-    Contained c;
+    Data c;
     non_trivial() : c() {}
-    non_trivial(const Contained& src) : c(src) {}
+    non_trivial(const Data& src) : c(src) {}
     non_trivial(const non_trivial& other) : c(other.c) {}
     non_trivial& operator=(const non_trivial& other) {
         c = other.c;

benchmarks/inc/utility.hpp

@@ -18,6 +18,15 @@ std::vector<Contained> random_vector(size_t n) {
     xoshiro256ss prng{id64(rd), id64(rd), id64(rd), id64(rd)};
 
     std::vector<Contained> res(n);
+
+// Here, the type Contained can be char, int, aggregate<Data>, or non_trivial<Data> where Data is char or int.
+// (aggregate<Data> and non_trivial<Data> are defined in udt.hpp.)
+// static_cast<Contained> silences truncation warnings when Contained is directly char or int,
+// but is insufficient for aggregate<Data> or non_trivial<Data>.
+#pragma warning(push)
+#pragma warning(disable : 4244) // warning C4244: conversion from 'uint64_t' to 'Data', possible loss of data
     std::generate(res.begin(), res.end(), [&prng] { return static_cast<Contained>(prng.next()); });
+#pragma warning(pop)
+
     return res;
 }

benchmarks/src/minmax_element.cpp

@@ -0,0 +1,86 @@
+// Copyright (c) Microsoft Corporation.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <algorithm>
+#include <benchmark/benchmark.h>
+#include <cstddef>
+#include <cstdint>
+#include <random>
+#include <ranges>
+#include <type_traits>
+
+enum class Op {
+    Min,
+    Max,
+    Both,
+};
+
+using namespace std;
+
+template <class T, size_t Size, Op Operation>
+void bm(benchmark::State& state) {
+    T a[Size];
+
+    mt19937 gen(84710);
+
+    if constexpr (is_floating_point_v<T>) {
+        normal_distribution<T> dis(0, 10000.0);
+        ranges::generate(a, [&] { return dis(gen); });
+    } else {
+        uniform_int_distribution<conditional_t<sizeof(T) != 1, T, int>> dis(1, 20);
+        ranges::generate(a, [&] { return static_cast<T>(dis(gen)); });
+    }
+
+    for (auto _ : state) {
+        if constexpr (Operation == Op::Min) {
+            benchmark::DoNotOptimize(ranges::min_element(a));
+        } else if constexpr (Operation == Op::Max) {
+            benchmark::DoNotOptimize(ranges::max_element(a));
+        } else if constexpr (Operation == Op::Both) {
+            benchmark::DoNotOptimize(ranges::minmax_element(a));
+        }
+    }
+}
+
+BENCHMARK(bm<uint8_t, 8021, Op::Min>);
+BENCHMARK(bm<uint8_t, 8021, Op::Max>);
+BENCHMARK(bm<uint8_t, 8021, Op::Both>);
+
+BENCHMARK(bm<uint16_t, 8021, Op::Min>);
+BENCHMARK(bm<uint16_t, 8021, Op::Max>);
+BENCHMARK(bm<uint16_t, 8021, Op::Both>);
+
+BENCHMARK(bm<uint32_t, 8021, Op::Min>);
+BENCHMARK(bm<uint32_t, 8021, Op::Max>);
+BENCHMARK(bm<uint32_t, 8021, Op::Both>);
+
+BENCHMARK(bm<uint64_t, 8021, Op::Min>);
+BENCHMARK(bm<uint64_t, 8021, Op::Max>);
+BENCHMARK(bm<uint64_t, 8021, Op::Both>);
+
+BENCHMARK(bm<int8_t, 8021, Op::Min>);
+BENCHMARK(bm<int8_t, 8021, Op::Max>);
+BENCHMARK(bm<int8_t, 8021, Op::Both>);
+
+BENCHMARK(bm<int16_t, 8021, Op::Min>);
+BENCHMARK(bm<int16_t, 8021, Op::Max>);
+BENCHMARK(bm<int16_t, 8021, Op::Both>);
+
+BENCHMARK(bm<int32_t, 8021, Op::Min>);
+BENCHMARK(bm<int32_t, 8021, Op::Max>);
+BENCHMARK(bm<int32_t, 8021, Op::Both>);
+
+BENCHMARK(bm<int64_t, 8021, Op::Min>);
+BENCHMARK(bm<int64_t, 8021, Op::Max>);
+BENCHMARK(bm<int64_t, 8021, Op::Both>);
+
+BENCHMARK(bm<float, 8021, Op::Min>);
+BENCHMARK(bm<float, 8021, Op::Max>);
+BENCHMARK(bm<float, 8021, Op::Both>);
+
+BENCHMARK(bm<double, 8021, Op::Min>);
+BENCHMARK(bm<double, 8021, Op::Max>);
+BENCHMARK(bm<double, 8021, Op::Both>);
+
+
+BENCHMARK_MAIN();

benchmarks/src/std/containers/sequences/vector.bool/copy/test.cpp

@@ -4,6 +4,7 @@
 #include <benchmark/benchmark.h>
 //
 #include <algorithm>
+#include <cstddef>
 #include <random>
 #include <vector>
 
@@ -17,7 +18,7 @@ static vector<bool> createRandomVector(const size_t size) {
 }
 
 static void copy_block_aligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -27,7 +28,7 @@ static void copy_block_aligned(benchmark::State& state) {
 }
 
 static void copy_source_misaligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -37,7 +38,7 @@ static void copy_source_misaligned(benchmark::State& state) {
 }
 
 static void copy_dest_misaligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -48,7 +49,7 @@ static void copy_dest_misaligned(benchmark::State& state) {
 
 // Special benchmark for matching char alignment
 static void copy_matching_alignment(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 

benchmarks/src/std/containers/sequences/vector.bool/copy_n/test.cpp

@@ -4,6 +4,7 @@
 #include <benchmark/benchmark.h>
 //
 #include <algorithm>
+#include <cstddef>
 #include <random>
 #include <vector>
 
@@ -17,7 +18,7 @@ static vector<bool> createRandomVector(const size_t size) {
 }
 
 static void copy_n_block_aligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -27,7 +28,7 @@ static void copy_n_block_aligned(benchmark::State& state) {
 }
 
 static void copy_n_source_misaligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -37,7 +38,7 @@ static void copy_n_source_misaligned(benchmark::State& state) {
 }
 
 static void copy_n_dest_misaligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -48,7 +49,7 @@ static void copy_n_dest_misaligned(benchmark::State& state) {
 
 // Special benchmark for matching char alignment
 static void copy_n_matching_alignment(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 

benchmarks/src/std/containers/sequences/vector.bool/move/test.cpp

@@ -4,6 +4,7 @@
 #include <benchmark/benchmark.h>
 //
 #include <algorithm>
+#include <cstddef>
 #include <random>
 #include <vector>
 
@@ -17,7 +18,7 @@ static vector<bool> createRandomVector(const size_t size) {
 }
 
 static void move_block_aligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -27,7 +28,7 @@ static void move_block_aligned(benchmark::State& state) {
 }
 
 static void move_source_misaligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -37,7 +38,7 @@ static void move_source_misaligned(benchmark::State& state) {
 }
 
 static void move_dest_misaligned(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 
@@ -48,7 +49,7 @@ static void move_dest_misaligned(benchmark::State& state) {
 
 // Special benchmark for matching char alignment
 static void move_matching_alignment(benchmark::State& state) {
-    const auto size           = state.range(0);
+    const auto size           = static_cast<size_t>(state.range(0));
     const vector<bool> source = createRandomVector(size);
     vector<bool> dest(size, false);
 

stl/inc/__msvc_bit_utils.hpp

@@ -177,7 +177,7 @@ _NODISCARD int _Checked_arm_arm64_countl_zero(const _Ty _Val) noexcept {
     } else {
         return static_cast<int>(_CountLeadingZeros64(_Val));
     }
-#endif // TRANSITION, GH-1586
+#endif // ^^^ no workaround ^^^
 }
 #endif // defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC) || defined(_M_HYBRID_X86_ARM64)
 #endif // _HAS_COUNTL_ZERO_INTRINSICS
@@ -376,19 +376,19 @@ constexpr decltype(auto) _Select_countr_zero_impl(_Fn _Callback) {
 #if _HAS_TZCNT_BSF_INTRINSICS && _HAS_CXX20
     if (!_STD is_constant_evaluated()) {
 #ifdef __AVX2__
-        return _Callback([](_Ty _Val) { return _Countr_zero_tzcnt(_Val); });
+        return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Countr_zero_tzcnt(_Val); });
 #else // ^^^ AVX2 / not AVX2 vvv
         const bool _Definitely_have_tzcnt = __isa_available >= _Stl_isa_available_avx2;
         if (_Definitely_have_tzcnt) {
-            return _Callback([](_Ty _Val) { return _Countr_zero_tzcnt(_Val); });
+            return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Countr_zero_tzcnt(_Val); });
         } else {
-            return _Callback([](_Ty _Val) { return _Countr_zero_bsf(_Val); });
+            return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Countr_zero_bsf(_Val); });
         }
 #endif // ^^^ not AVX2 ^^^
     }
 #endif // ^^^ _HAS_TZCNT_BSF_INTRINSICS && _HAS_CXX20 ^^^
     // C++17 constexpr gcd() calls this function, so it should be constexpr unless we detect runtime evaluation.
-    return _Callback([](_Ty _Val) { return _Countr_zero_fallback(_Val); });
+    return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Countr_zero_fallback(_Val); });
 }
 
 template <class _Ty, enable_if_t<_Is_standard_unsigned_integer<_Ty>, int> = 0>
@@ -420,16 +420,16 @@ _CONSTEXPR20 decltype(auto) _Select_popcount_impl(_Fn _Callback) {
 #ifndef __AVX__
         const bool _Definitely_have_popcnt = __isa_available >= _Stl_isa_available_sse42;
         if (!_Definitely_have_popcnt) {
-            return _Callback([](_Ty _Val) { return _Popcount_fallback(_Val); });
+            return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Popcount_fallback(_Val); });
         }
 #endif // !defined(__AVX__)
-        return _Callback([](_Ty _Val) { return _Unchecked_x86_x64_popcount(_Val); });
+        return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Unchecked_x86_x64_popcount(_Val); });
 #elif _HAS_NEON_INTRINSICS // ^^^ x86/x64 intrinsics available / ARM64 intrinsics available vvv
-        return _Callback([](_Ty _Val) { return _Arm64_popcount(_Val); });
+        return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Arm64_popcount(_Val); });
 #endif // ^^^ ARM64 intrinsics available ^^^
     }
 #endif // ^^^ any intrinsics available ^^^
-    return _Callback([](_Ty _Val) { return _Popcount_fallback(_Val); });
+    return _Callback([](_Ty _Val) _STATIC_CALL_OPERATOR { return _Popcount_fallback(_Val); });
 }
 
 #undef _HAS_POPCNT_INTRINSICS

stl/inc/__msvc_chrono.hpp

@@ -405,7 +405,7 @@ namespace chrono {
         return !(_Left < _Right);
     }
 
-#ifdef __cpp_lib_concepts
+#if _HAS_CXX20
     _EXPORT_STD template <class _Rep1, class _Period1, class _Rep2, class _Period2>
         requires three_way_comparable<typename common_type_t<duration<_Rep1, _Period1>, duration<_Rep2, _Period2>>::rep>
     _NODISCARD constexpr auto operator<=>(const duration<_Rep1, _Period1>& _Left,
@@ -414,7 +414,7 @@ namespace chrono {
         using _CT = common_type_t<duration<_Rep1, _Period1>, duration<_Rep2, _Period2>>;
         return _CT(_Left).count() <=> _CT(_Right).count();
     }
-#endif // defined(__cpp_lib_concepts)
+#endif // _HAS_CXX20
 
     _EXPORT_STD template <class _To, class _Rep, class _Period, enable_if_t<_Is_duration_v<_To>, int> /* = 0 */>
     _NODISCARD constexpr _To duration_cast(const duration<_Rep, _Period>& _Dur) noexcept(
@@ -593,14 +593,14 @@ namespace chrono {
         return !(_Left < _Right);
     }
 
-#ifdef __cpp_lib_concepts
+#if _HAS_CXX20
     _EXPORT_STD template <class _Clock, class _Duration1, three_way_comparable_with<_Duration1> _Duration2>
     _NODISCARD constexpr auto
         operator<=>(const time_point<_Clock, _Duration1>& _Left, const time_point<_Clock, _Duration2>& _Right) noexcept(
             is_arithmetic_v<typename _Duration1::rep> && is_arithmetic_v<typename _Duration2::rep>) /* strengthened */ {
         return _Left.time_since_epoch() <=> _Right.time_since_epoch();
     }
-#endif // defined(__cpp_lib_concepts)
+#endif // _HAS_CXX20
 
     _EXPORT_STD template <class _To, class _Clock, class _Duration, enable_if_t<_Is_duration_v<_To>, int> = 0>
     _NODISCARD constexpr time_point<_Clock, _To> time_point_cast(const time_point<_Clock, _Duration>& _Time) noexcept(

stl/inc/__msvc_filebuf.hpp

@@ -383,7 +383,7 @@ class basic_filebuf : public basic_streambuf<_Elem, _Traits> { // stream buffer
         }
     }
 
-#if defined(__cpp_lib_print) && defined(_CPPRTTI)
+#if _HAS_CXX23 && defined(_CPPRTTI)
     template <class _Filebuf_type>
     friend ios_base::iostate _Print_noformat_unicode(ostream&, string_view);
 #endif

stl/inc/__msvc_formatter.hpp

@@ -39,9 +39,9 @@
 #include <yvals_core.h>
 #if _STL_COMPILER_PREPROCESSOR
 
-#if !_HAS_CXX20 || !defined(__cpp_lib_concepts) // TRANSITION, GH-395
+#if !_HAS_CXX20
 #error The contents of <format> are only available with C++20. (Also, you should not include this internal header.)
-#endif // !_HAS_CXX20 || !defined(__cpp_lib_concepts)
+#endif // !_HAS_CXX20
 
 #include <concepts>
 #include <cstddef>