diff options
author | Adrian Vogelsgesang <avogelsgesang@tableau.com> | 2022-08-04 15:21:27 -0700 |
---|---|---|
committer | Adrian Vogelsgesang <avogelsgesang@salesforce.com> | 2023-02-12 14:51:08 -0800 |
commit | 2a06757a200cc8dd4c3aeca98509d50d75bb4a27 (patch) | |
tree | 140e7a9d72c815e222a83ccb3662d7329afbf0b6 /libcxx/benchmarks | |
parent | 2e6430666caf303b84dd281442533d2f3b6ad1b2 (diff) | |
download | llvm-2a06757a200cc8dd4c3aeca98509d50d75bb4a27.tar.gz |
[libc++][spaceship] Implement `lexicographical_compare_three_way`
The implementation makes use of the freedom added by LWG 3410. We have
two variants of this algorithm:
* a fast path for random access iterators: This fast path computes the
maximum number of loop iterations up-front and does not compare the
iterators against their limits on every loop iteration.
* A basic implementation for all other iterators: This implementation
compares the iterators against their limits in every loop iteration.
However, it still takes advantage of the freedom added by LWG 3410 to
avoid unnecessary additional iterator comparisons, as originally
specified by P1614R2.
https://godbolt.org/z/7xbMEen5e shows the benefit of the fast path:
The hot loop generated of `lexicographical_compare_three_way3` is
more tight than for `lexicographical_compare_three_way1`. The added
benchmark illustrates how this leads to a 30% - 50% performance
improvement on integer vectors.
Implements part of P1614R2 "The Mothership has Landed"
Fixes LWG 3410 and LWG 3350
Differential Revision: https://reviews.llvm.org/D131395
Diffstat (limited to 'libcxx/benchmarks')
-rw-r--r-- | libcxx/benchmarks/CMakeLists.txt | 1 | ||||
-rw-r--r-- | libcxx/benchmarks/lexicographical_compare_three_way.bench.cpp | 96 |
2 files changed, 97 insertions, 0 deletions
diff --git a/libcxx/benchmarks/CMakeLists.txt b/libcxx/benchmarks/CMakeLists.txt index 7eb76ac6370f..2c3c39ef5345 100644 --- a/libcxx/benchmarks/CMakeLists.txt +++ b/libcxx/benchmarks/CMakeLists.txt @@ -186,6 +186,7 @@ set(BENCHMARK_TESTS formatter_int.bench.cpp function.bench.cpp join_view.bench.cpp + lexicographical_compare_three_way.bench.cpp map.bench.cpp monotonic_buffer.bench.cpp ordered_set.bench.cpp diff --git a/libcxx/benchmarks/lexicographical_compare_three_way.bench.cpp b/libcxx/benchmarks/lexicographical_compare_three_way.bench.cpp new file mode 100644 index 000000000000..e58134adaf21 --- /dev/null +++ b/libcxx/benchmarks/lexicographical_compare_three_way.bench.cpp @@ -0,0 +1,96 @@ +//===----------------------------------------------------------------------===// +// 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 +// +//===----------------------------------------------------------------------===// + +#include <algorithm> + +#include "benchmark/benchmark.h" +#include "test_iterators.h" + +static void BM_lexicographical_compare_three_way_slow_path(benchmark::State& state) { + auto size = state.range(0); + std::vector<int> v1; + v1.resize(size); + // v2 is identical except for the last value. + // This means, that `lexicographical_compare_three_way` actually has to + // compare the complete vector and cannot bail out early. + std::vector<int> v2 = v1; + v2.back() += 1; + int* b1 = v1.data(); + int* e1 = b1 + v1.size(); + int* b2 = v2.data(); + int* e2 = b2 + v2.size(); + + for (auto _ : state) { + auto cmp = std::compare_three_way(); + benchmark::DoNotOptimize(std::__lexicographical_compare_three_way_slow_path(b1, e1, b2, e2, cmp)); + } +} + +BENCHMARK(BM_lexicographical_compare_three_way_slow_path)->RangeMultiplier(4)->Range(1, 1 << 20); + +static void BM_lexicographical_compare_three_way_fast_path(benchmark::State& state) { + auto size = state.range(0); + std::vector<int> v1; + v1.resize(size); + // v2 is identical except for the last value. + // This means, that `lexicographical_compare_three_way` actually has to + // compare the complete vector and cannot bail out early. + std::vector<int> v2 = v1; + v2.back() += 1; + int* b1 = v1.data(); + int* e1 = b1 + v1.size(); + int* b2 = v2.data(); + int* e2 = b2 + v2.size(); + + for (auto _ : state) { + auto cmp = std::compare_three_way(); + benchmark::DoNotOptimize(std::__lexicographical_compare_three_way_fast_path(b1, e1, b2, e2, cmp)); + } +} + +BENCHMARK(BM_lexicographical_compare_three_way_fast_path)->RangeMultiplier(4)->Range(1, 1 << 20); + +template <class IteratorT> +static void BM_lexicographical_compare_three_way(benchmark::State& state) { + auto size = state.range(0); + std::vector<int> v1; + v1.resize(size); + // v2 is identical except for the last value. + // This means, that `lexicographical_compare_three_way` actually has to + // compare the complete vector and cannot bail out early. + std::vector<int> v2 = v1; + v2.back() += 1; + auto b1 = IteratorT{v1.data()}; + auto e1 = IteratorT{v1.data() + v1.size()}; + auto b2 = IteratorT{v2.data()}; + auto e2 = IteratorT{v2.data() + v2.size()}; + + for (auto _ : state) { + benchmark::DoNotOptimize(std::lexicographical_compare_three_way(b1, e1, b2, e2)); + } +} + +// Type alias to make sure the `*` does not appear in the benchmark name. +// A `*` would confuse the Python test runner running this google benchmark. +using IntPtr = int*; + +// `lexicographical_compare_three_way` has a fast path for random access iterators. +BENCHMARK_TEMPLATE(BM_lexicographical_compare_three_way, IntPtr)->RangeMultiplier(4)->Range(1, 1 << 20); +BENCHMARK_TEMPLATE(BM_lexicographical_compare_three_way, random_access_iterator<IntPtr>) + ->RangeMultiplier(4) + ->Range(1, 1 << 20); +BENCHMARK_TEMPLATE(BM_lexicographical_compare_three_way, cpp17_input_iterator<IntPtr>) + ->RangeMultiplier(4) + ->Range(1, 1 << 20); + +int main(int argc, char** argv) { + benchmark::Initialize(&argc, argv); + if (benchmark::ReportUnrecognizedArguments(argc, argv)) + return 1; + + benchmark::RunSpecifiedBenchmarks(); +} |