| // Ceres Solver - A fast non-linear least squares minimizer |
| // Copyright 2023 Google Inc. All rights reserved. |
| // http://ceres-solver.org/ |
| // |
| // Redistribution and use in source and binary forms, with or without |
| // modification, are permitted provided that the following conditions are met: |
| // |
| // * Redistributions of source code must retain the above copyright notice, |
| // this list of conditions and the following disclaimer. |
| // * Redistributions in binary form must reproduce the above copyright notice, |
| // this list of conditions and the following disclaimer in the documentation |
| // and/or other materials provided with the distribution. |
| // * Neither the name of Google Inc. nor the names of its contributors may be |
| // used to endorse or promote products derived from this software without |
| // specific prior written permission. |
| // |
| // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
| // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
| // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
| // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
| // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
| // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
| // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| // POSSIBILITY OF SUCH DAMAGE. |
| // |
| // Authors: sameeragarwal@google.com (Sameer Agarwal) |
| |
| #include <iostream> |
| |
| #include "Eigen/Dense" |
| #include "benchmark/benchmark.h" |
| #include "ceres/small_blas.h" |
| |
| namespace ceres::internal { |
| |
| // Benchmarking matrix-matrix multiply routines and optimizing memory |
| // access requires that we make sure that they are not just sitting in |
| // the cache. So, as the benchmarking routine iterates, we need to |
| // multiply new/different matrice. Allocating/creating these objects |
| // in the benchmarking loop is too heavy duty, so we create them |
| // before hand and cycle through them in the benchmark. This class, |
| // given the size of the matrices creates such objects for use in the |
| // benchmark. |
| class MatrixMatrixMultiplyData { |
| public: |
| MatrixMatrixMultiplyData( |
| int a_rows, int a_cols, int b_rows, int b_cols, int c_rows, int c_cols) |
| : num_elements_(1000), |
| a_size_(num_elements_ * a_rows * a_cols), |
| b_size_(b_rows * b_cols), |
| c_size_(c_rows * c_cols), |
| a_(num_elements_ * a_size_, 1.00001), |
| b_(num_elements_ * b_size_, 0.5), |
| c_(num_elements_ * c_size_, -1.1) {} |
| |
| int num_elements() const { return num_elements_; } |
| double* GetA(int i) { return &a_[i * a_size_]; } |
| double* GetB(int i) { return &b_[i * b_size_]; } |
| double* GetC(int i) { return &c_[i * c_size_]; } |
| |
| private: |
| int num_elements_; |
| int a_size_; |
| int b_size_; |
| int c_size_; |
| std::vector<double> a_; |
| std::vector<double> b_; |
| std::vector<double> c_; |
| }; |
| |
| #define GEMM_KIND_EQ 0 |
| #define GEMM_KIND_ADD 1 |
| #define GEMM_KIND_SUB -1 |
| |
| #define BENCHMARK_MM_FN(FN, M, N, K, NAME, MT, NT, KT) \ |
| void static BM_##FN##_##NAME##_##M##x##N##x##K(benchmark::State& state) { \ |
| const int b_rows = M; \ |
| const int b_cols = N; \ |
| const int c_rows = b_cols; \ |
| const int c_cols = K; \ |
| const int a_rows = b_rows; \ |
| const int a_cols = c_cols; \ |
| MatrixMatrixMultiplyData data( \ |
| a_rows, a_cols, b_rows, b_cols, c_rows, c_cols); \ |
| const int num_elements = data.num_elements(); \ |
| int iter = 0; \ |
| for (auto _ : state) { \ |
| FN<MT, KT, KT, NT, GEMM_KIND_ADD>(data.GetB(iter), \ |
| b_rows, \ |
| b_cols, \ |
| data.GetC(iter), \ |
| c_rows, \ |
| c_cols, \ |
| data.GetA(iter), \ |
| 512, \ |
| 512, \ |
| a_rows, \ |
| a_cols); \ |
| iter = (iter + 1) % num_elements; \ |
| } \ |
| } \ |
| BENCHMARK(BM_##FN##_##NAME##_##M##x##N##x##K); |
| |
| #define BENCHMARK_STATIC_MM_FN(FN, M, N, K) \ |
| BENCHMARK_MM_FN(FN, M, N, K, Static, M, N, K) |
| #define BENCHMARK_DYNAMIC_MM_FN(FN, M, N, K) \ |
| BENCHMARK_MM_FN( \ |
| FN, M, N, K, Dynamic, Eigen::Dynamic, Eigen::Dynamic, Eigen::Dynamic) |
| |
| #define BENCHMARK_MTM_FN(FN, M, N, K, NAME, MT, NT, KT) \ |
| void static BM_##FN##_##NAME##_##M##x##N##x##K(benchmark::State& state) { \ |
| const int b_rows = M; \ |
| const int b_cols = N; \ |
| const int c_rows = b_rows; \ |
| const int c_cols = K; \ |
| const int a_rows = b_cols; \ |
| const int a_cols = c_cols; \ |
| MatrixMatrixMultiplyData data( \ |
| a_rows, a_cols, b_rows, b_cols, c_rows, c_cols); \ |
| const int num_elements = data.num_elements(); \ |
| int iter = 0; \ |
| for (auto _ : state) { \ |
| FN<KT, MT, KT, NT, GEMM_KIND_ADD>(data.GetB(iter), \ |
| b_rows, \ |
| b_cols, \ |
| data.GetC(iter), \ |
| c_rows, \ |
| c_cols, \ |
| data.GetA(iter), \ |
| 0, \ |
| 0, \ |
| a_rows, \ |
| a_cols); \ |
| iter = (iter + 1) % num_elements; \ |
| } \ |
| } \ |
| BENCHMARK(BM_##FN##_##NAME##_##M##x##N##x##K); |
| |
| #define BENCHMARK_STATIC_MMT_FN(FN, M, N, K) \ |
| BENCHMARK_MTM_FN(FN, M, N, K, Static, M, N, K) |
| #define BENCHMARK_DYNAMIC_MMT_FN(FN, M, N, K) \ |
| BENCHMARK_MTM_FN( \ |
| FN, M, N, K, Dynamic, Eigen::Dynamic, Eigen::Dynamic, Eigen::Dynamic) |
| |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 2, 3, 4) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 3, 3, 3) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 4, 4, 4) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 8, 8, 8) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 9, 9, 3) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 9, 3, 3) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyEigen, 3, 9, 9) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 2, 3, 4) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 3, 3, 3) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 4, 4, 4) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 8, 8, 8) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 9, 9, 3) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 9, 3, 3) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyEigen, 3, 9, 9) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 2, 3, 4) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 3, 3, 3) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 4, 4, 4) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 8, 8, 8) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 9, 9, 3) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 9, 3, 3) |
| BENCHMARK_STATIC_MM_FN(MatrixMatrixMultiplyNaive, 3, 9, 9) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 2, 3, 4) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 3, 3, 3) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 4, 4, 4) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 8, 8, 8) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 9, 9, 3) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 9, 3, 3) |
| BENCHMARK_DYNAMIC_MM_FN(MatrixMatrixMultiplyNaive, 3, 9, 9) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 2, 3, 4) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 3, 3, 3) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 4, 4, 4) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 8, 8, 8) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 9, 9, 3) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 9, 3, 3) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 3, 9, 9) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 2, 3, 4) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 3, 3, 3) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 4, 4, 4) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 8, 8, 8) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 9, 9, 3) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 9, 3, 3) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyEigen, 3, 9, 9) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 2, 3, 4) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 3, 3, 3) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 4, 4, 4) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 8, 8, 8) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 9, 9, 3) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 9, 3, 3) |
| BENCHMARK_STATIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 3, 9, 9) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 2, 3, 4) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 3, 3, 3) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 4, 4, 4) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 8, 8, 8) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 9, 9, 3) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 9, 3, 3) |
| BENCHMARK_DYNAMIC_MMT_FN(MatrixTransposeMatrixMultiplyNaive, 3, 9, 9) |
| |
| #undef GEMM_KIND_EQ |
| #undef GEMM_KIND_ADD |
| #undef GEMM_KIND_SUB |
| #undef BENCHMARK_MM_FN |
| #undef BENCHMARK_STATIC_MM_FN |
| #undef BENCHMARK_DYNAMIC_MM_FN |
| #undef BENCHMARK_MTM_FN |
| #undef BENCHMARK_DYNAMIC_MMT_FN |
| #undef BENCHMARK_STATIC_MMT_FN |
| |
| } // namespace ceres::internal |
| |
| BENCHMARK_MAIN(); |
