xtensor-stack · JohanMabille · Jun 10, 2025 · May 30, 2025
diff --git a/.github/workflows/benchmarks.yml b/.github/workflows/benchmarks.yml
@@ -0,0 +1,73 @@
+name: benchmarks
+on:
+  workflow_dispatch:
+  pull_request:
+  push:
+    branches: [master]
+concurrency:
+  group: ${{ github.workflow }}-${{ github.job }}-${{ github.ref }}
+  cancel-in-progress: true
+defaults:
+  run:
+    shell: bash -e -l {0}
+jobs:
+  build:
+    runs-on: ubuntu-24.04
+    name: ${{ matrix.sys.compiler }} ${{ matrix.sys.version }} - ${{ matrix.sys.name }}
+    strategy:
+      fail-fast: false
+      matrix:
+        sys:
+        - {compiler: clang, version: '20', name: xsimd, flags: -DXTENSOR_USE_XSIMD=ON}
+        - {compiler: clang, version: '20', name: xsimd-tbb, flags: -DXTENSOR_USE_XSIMD=ON -DXTENSOR_USE_TBB=ON}
+        - {compiler: gcc, version: '14', name: xsimd, flags: -DXTENSOR_USE_XSIMD=ON}
+        - {compiler: gcc, version: '14', name: xsimd-tbb, flags: -DXTENSOR_USE_XSIMD=ON -DXTENSOR_USE_TBB=ON}
+
+    steps:
+    - name: Install GCC
+      if: matrix.sys.compiler == 'gcc'
+      uses: egor-tensin/setup-gcc@v1
+      with:
+        version: ${{matrix.sys.version}}
+        platform: x64
+
+    - name: Install LLVM and Clang
+      if: matrix.sys.compiler == 'clang'
+      run: |
+        wget https://apt.llvm.org/llvm.sh
+        chmod +x llvm.sh
+        sudo ./llvm.sh ${{matrix.sys.version}}
+        sudo apt-get install -y clang-tools-${{matrix.sys.version}}
+        sudo update-alternatives --install /usr/bin/clang++                 clang++                     /usr/bin/clang++-${{matrix.sys.version}} 200
+        sudo update-alternatives --install /usr/bin/clang                       clang                         /usr/bin/clang-${{matrix.sys.version}} 200
+        sudo update-alternatives --install /usr/bin/clang-scan-deps   clang-scan-deps     /usr/bin/clang-scan-deps-${{matrix.sys.version}} 200
+        sudo update-alternatives --set clang /usr/bin/clang-${{matrix.sys.version}}
+        sudo update-alternatives --set clang++ /usr/bin/clang++-${{matrix.sys.version}}
+        sudo update-alternatives --set clang-scan-deps /usr/bin/clang-scan-deps-${{matrix.sys.version}}
+
+    - name: Checkout code
+      uses: actions/checkout@v3
+
+    - name: Set conda environment
+      uses: mamba-org/setup-micromamba@main
+      with:
+        environment-name: myenv
+        environment-file: environment-dev.yml
+        init-shell: bash
+        cache-downloads: true
+        create-args: |
+          ${{ (matrix.sys.name == 'tbb' || matrix.sys.name == 'xsimd-tbb' ) && 'tbb-devel' || '' }}
+
+    - name: Configure using CMake
+      run: |
+        if [[ "${{matrix.sys.compiler}}" = "gcc" ]]; then export CC=gcc-${{matrix.sys.version}}; export CXX=g++-${{matrix.sys.version}}; else export CC=clang; export CXX=clang++; fi
+        cmake -G Ninja -Bbuild -DCMAKE_C_COMPILER=$CC -DCMAKE_CXX_COMPILER=$CXX -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX -DBUILD_BENCHMARK=ON ${{ matrix.sys.flags }}
+
+    - name: Build
+      working-directory: build
+      run: cmake --build . --target benchmark_xtensor --parallel 8
+
+    - name: Run benchmark
+      timeout-minutes: 10 # Consider increasing timeout
+      working-directory: build/benchmark
+      run: ./benchmark_xtensor
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -7,7 +7,7 @@
 # The full license is in the file LICENSE, distributed with this software. #
 ############################################################################
 
-cmake_minimum_required(VERSION 3.29)
+cmake_minimum_required(VERSION 3.22)
 project(xtensor CXX)
 
 set(XTENSOR_INCLUDE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/include)

diff --git a/benchmark/CMakeLists.txt b/benchmark/CMakeLists.txt
@@ -6,7 +6,8 @@
 # The full license is in the file LICENSE, distributed with this software. #
 ############################################################################
 
-cmake_minimum_required(VERSION 3.5)
+cmake_minimum_required(VERSION 3.22)
+include(FetchContent)
 
 if (CMAKE_CURRENT_SOURCE_DIR STREQUAL CMAKE_SOURCE_DIR)
     project(xtensor-benchmark)
@@ -30,11 +31,11 @@ if (CMAKE_CXX_COMPILER_ID MATCHES "Clang" OR CMAKE_CXX_COMPILER_ID MATCHES "GNU"
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3 -g -Wunused-parameter -Wextra -Wreorder")
 
     if(NOT "${CMAKE_CXX_SIMULATE_ID}" STREQUAL "MSVC")
-        CHECK_CXX_COMPILER_FLAG("-std=c++14" HAS_CPP14_FLAG)
-        if (HAS_CPP14_FLAG)
-            set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
+        CHECK_CXX_COMPILER_FLAG("-std=c++20" HAS_CPP20_FLAG)
+        if (HAS_CPP20_FLAG)
+            set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++20")
         else()
-            message(FATAL_ERROR "Unsupported compiler -- xtensor requires C++14 support!")
+            message(FATAL_ERROR "Unsupported compiler -- xtensor requires C++17 support!")
         endif()
     endif()
 
@@ -74,31 +75,17 @@ endif()
 
 
 if(DOWNLOAD_GBENCHMARK OR GBENCHMARK_SRC_DIR)
-    if(DOWNLOAD_GBENCHMARK)
-        # Download and unpack googlebenchmark at configure time
-        configure_file(downloadGBenchmark.cmake.in googlebenchmark-download/CMakeLists.txt)
-    else()
-        # Copy local source of googlebenchmark at configure time
-        configure_file(copyGBenchmark.cmake.in googlebenchmark-download/CMakeLists.txt)
-    endif()
-    execute_process(COMMAND ${CMAKE_COMMAND} -G "${CMAKE_GENERATOR}" .
-                    RESULT_VARIABLE result
-                    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/googlebenchmark-download )
-    if(result)
-        message(FATAL_ERROR "CMake step for googlebenchmark failed: ${result}")
-    endif()
-    execute_process(COMMAND ${CMAKE_COMMAND} --build .
-                    RESULT_VARIABLE result
-                    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/googlebenchmark-download )
-    if(result)
-        message(FATAL_ERROR "Build step for googlebenchmark failed: ${result}")
-    endif()
+    FetchContent_Declare(googletest
+            GIT_REPOSITORY https://github.com/google/googletest.git
+            GIT_TAG main)
 
-    # Add googlebenchmark directly to our build. This defines
-    # the gtest and gtest_main targets.
-    add_subdirectory(${CMAKE_CURRENT_BINARY_DIR}/googlebenchmark-src
-                     ${CMAKE_CURRENT_BINARY_DIR}/googlebenchmark-build)
+    FetchContent_Declare(googlebenchmark
+            GIT_REPOSITORY https://github.com/google/benchmark.git
+            GIT_TAG main) # need main for benchmark::benchmark
 
+    FetchContent_MakeAvailable(
+            googletest
+            googlebenchmark)
     set(GBENCHMARK_INCLUDE_DIRS "${googlebenchmark_SOURCE_DIR}/include")
     set(GBENCHMARK_LIBRARIES benchmark)
 else()
@@ -129,9 +116,11 @@ set(XTENSOR_BENCHMARK
     benchmark_view_access.cpp
     benchmark_view_assignment.cpp
     benchmark_view_adapt.cpp
+    benchmark_stl.cpp
     main.cpp
 )
 
+
 set(XTENSOR_BENCHMARK_TARGET benchmark_xtensor)
 add_executable(${XTENSOR_BENCHMARK_TARGET} EXCLUDE_FROM_ALL ${XTENSOR_BENCHMARK} ${XTENSOR_HEADERS})
 target_link_libraries(${XTENSOR_BENCHMARK_TARGET} PUBLIC xtensor ${GBENCHMARK_LIBRARIES})

diff --git a/benchmark/benchmark_stl.cpp b/benchmark/benchmark_stl.cpp
@@ -0,0 +1,157 @@
+/***************************************************************************
+ * Copyright (c) 2016, Johan Mabille, Sylvain Corlay and Wolf Vollprecht    *
+ *                                                                          *
+ * Distributed under the terms of the BSD 3-Clause License.                 *
+ *                                                                          *
+ * The full license is in the file LICENSE, distributed with this software. *
+ ****************************************************************************/
+
+#include <benchmark/benchmark.h>
+
+#include "xtensor/containers/xtensor.hpp"
+#include "xtensor/core/xmath.hpp"
+#include "xtensor/generators/xrandom.hpp"
+
+namespace xt
+{
+    namespace
+    {
+        constexpr std::array<size_t, 2> cContainerAssignShape{2000, 2000};
+
+        template <class Shape>
+        auto generateRandomInt16From0To100(Shape&& x)
+        {
+            return xt::random::randint(x, 0, 100);
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_DivideBy2_StdTransform(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            std::transform(
+                vInput.begin(),
+                vInput.end(),
+                vOutput.begin(),
+                [](auto&& aInputValue)
+                {
+                    return aInputValue / 2;
+                }
+            );
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_DivideBy2_Xtensor(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            vOutput = vInput / 2;
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_DivideBy2Double_StdTransform(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            std::transform(
+                vInput.begin(),
+                vInput.end(),
+                vOutput.begin(),
+                [](auto&& aInputValue)
+                {
+                    return aInputValue / 2.0;
+                }
+            );
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_DivideBy2Double_Xtensor(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            vOutput = vInput / 2.0;
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_MultiplyBy2_StdTransform(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            std::transform(
+                vInput.begin(),
+                vInput.end(),
+                vOutput.begin(),
+                [](auto&& aInputValue)
+                {
+                    return aInputValue * 2;
+                }
+            );
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_MultiplyBy2_Xtensor(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            vOutput = vInput * 2;
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_Maximum_StdTransform(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput1 = generateRandomInt16From0To100(cContainerAssignShape);
+        xt::xtensor<uint16_t, 2> vInput2 = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            auto vInput2It = vInput2.begin();
+            std::transform(
+                vInput1.begin(),
+                vInput1.end(),
+                vOutput.begin(),
+                [&vInput2It](auto&& aInput1Value)
+                {
+                    return std::max(aInput1Value, *vInput2It++);
+                }
+            );
+        }
+    }
+
+    static void Xtensor_Uint16_2000x2000_Maximum_Xtensor(benchmark::State& aState)
+    {
+        xt::xtensor<uint16_t, 2> vInput1 = generateRandomInt16From0To100(cContainerAssignShape);
+        xt::xtensor<uint16_t, 2> vInput2 = generateRandomInt16From0To100(cContainerAssignShape);
+        auto vOutput = xt::xtensor<uint16_t, 2>::from_shape(cContainerAssignShape);
+
+        for (auto _ : aState)
+        {
+            vOutput = xt::maximum(vInput1, vInput2);
+        }
+    }
+
+    BENCHMARK(Xtensor_Uint16_2000x2000_Maximum_Xtensor);
+    BENCHMARK(Xtensor_Uint16_2000x2000_Maximum_StdTransform);
+    BENCHMARK(Xtensor_Uint16_2000x2000_MultiplyBy2_Xtensor);
+    BENCHMARK(Xtensor_Uint16_2000x2000_MultiplyBy2_StdTransform);
+    BENCHMARK(Xtensor_Uint16_2000x2000_DivideBy2Double_Xtensor);
+    BENCHMARK(Xtensor_Uint16_2000x2000_DivideBy2Double_StdTransform);
+}
diff --git a/benchmark/benchmark_view_assignment.cpp b/benchmark/benchmark_view_assignment.cpp
@@ -155,6 +155,6 @@ namespace xt
     //    BENCHMARK(assign_create_strided_view);
     BENCHMARK(assign_create_view);
     BENCHMARK(assign_create_manual_view);
-    BENCHMARK(data_offset);
+    // BENCHMARK(data_offset);
     BENCHMARK(data_offset_view);
 }
diff --git a/benchmark/copyGBenchmark.cmake.in b/benchmark/copyGBenchmark.cmake.in
@@ -6,7 +6,7 @@
 # The full license is in the file LICENSE, distributed with this software. #
 ############################################################################
 
-cmake_minimum_required(VERSION 2.8.2)
+cmake_minimum_required(VERSION 3.5)
 
 project(googlebenchmark-download NONE)
 

diff --git a/benchmark/downloadGBenchmark.cmake.in b/benchmark/downloadGBenchmark.cmake.in
@@ -6,18 +6,19 @@
 # The full license is in the file LICENSE, distributed with this software. #
 ############################################################################
 
-cmake_minimum_required(VERSION 2.8.2)
+cmake_minimum_required(VERSION 3.5)
 
 project(googlebenchmark-download NONE)
 
 include(ExternalProject)
 ExternalProject_Add(googlebenchmark
     GIT_REPOSITORY    https://github.com/google/benchmark.git
-    GIT_TAG           v1.3.0
+    GIT_TAG           v1.9.4
     SOURCE_DIR        "${CMAKE_CURRENT_BINARY_DIR}/googlebenchmark-src"
     BINARY_DIR        "${CMAKE_CURRENT_BINARY_DIR}/googlebenchmark-build"
     CONFIGURE_COMMAND ""
     BUILD_COMMAND     ""
+    CMAKE_ARGS        "BENCHMARK_DOWNLOAD_DEPENDENCIES=TRUE"
     INSTALL_COMMAND   ""
     TEST_COMMAND      ""
 )