| // Benchmarks for matrix exponential. |
| // Critical for Sophus Lie group operations (SLAM, visual odometry). |
| // SPDX-FileCopyrightText: The Eigen Authors |
| // SPDX-License-Identifier: MPL-2.0 |
| |
| #include <benchmark/benchmark.h> |
| #include <Eigen/Core> |
| #include <contrib/Eigen/MatrixFunctions> |
| |
| using namespace Eigen; |
| |
| #ifndef SCALAR |
| #define SCALAR double |
| #endif |
| |
| typedef SCALAR Scalar; |
| |
| static void BM_MatrixExp(benchmark::State& state) { |
| int n = state.range(0); |
| Scalar scale = Scalar(state.range(1)); |
| typedef Matrix<Scalar, Dynamic, Dynamic> MatrixType; |
| |
| // The larger scale exercises the scale-and-square path, including the |
| // coefficient-wise exact power-of-two scale-down. |
| MatrixType A = scale * MatrixType::Random(n, n) / Scalar(n); |
| MatrixType result(n, n); |
| |
| for (auto _ : state) { |
| result = A.exp(); |
| benchmark::DoNotOptimize(result.data()); |
| benchmark::ClobberMemory(); |
| } |
| } |
| |
| // Fixed-size specializations for Lie group sizes. |
| template <int N> |
| static void BM_MatrixExp_Fixed(benchmark::State& state) { |
| typedef Matrix<Scalar, N, N> MatrixType; |
| |
| MatrixType A = MatrixType::Random() / Scalar(N); |
| MatrixType result; |
| |
| for (auto _ : state) { |
| result = A.exp(); |
| benchmark::DoNotOptimize(result.data()); |
| benchmark::ClobberMemory(); |
| } |
| } |
| |
| // Dynamic sizes: Lie groups (2,3,4) plus larger, both well-scaled and through |
| // the scale-and-square path. |
| BENCHMARK(BM_MatrixExp)->ArgsProduct({{2, 3, 4, 8, 16, 32, 64, 128}, {1, 64}}); |
| |
| // Fixed-size Lie group dimensions. |
| BENCHMARK(BM_MatrixExp_Fixed<2>); |
| BENCHMARK(BM_MatrixExp_Fixed<3>); |
| BENCHMARK(BM_MatrixExp_Fixed<4>); |