Add CodSpeed continuous performance testing (#2919)

This PR integrates [CodSpeed](https://codspeed.io) for continuous
performance monitoring of xtensor's existing benchmark suite.

## What changed

**`benchmark/CMakeLists.txt`** -- The Google Benchmark dependency is
replaced with [CodSpeed's compatibility
layer](https://github.com/CodSpeedHQ/codspeed-cpp), which wraps Google
Benchmark with CodSpeed instrumentation. When `CODSPEED_MODE` is not set
(or set to `off`), the build behaves identically to upstream Google
Benchmark. Two additional changes are made for CodSpeed compatibility:
- `-march=native` is conditionally skipped in simulation mode, since
valgrind does not support all native instruction sets.
- The build type is set to `RelWithDebInfo` (instead of `Release`) when
CodSpeed is active, to include debug symbols for flame graph profiling.

**`.github/workflows/codspeed.yml`** -- A new GitHub Actions workflow
that builds and runs the full benchmark suite (160+ benchmarks) under
CodSpeed's CPU simulation instrument on every push to `master` and on
pull requests. The workflow uses OIDC authentication and follows the
same dependency setup (micromamba, xsimd) as the existing
`benchmarks.yml`.

**`README.md`** -- Added the CodSpeed badge.

## How it works

CodSpeed's simulation mode runs each benchmark once through a CPU
instruction counter (valgrind-based), producing deterministic
measurements that are independent of CI environment load. This makes
performance regressions detectable on standard `ubuntu-latest` runners
without dedicated hardware.

## Next steps

1. Install the [CodSpeed GitHub App](https://github.com/apps/codspeed)
on this repository to enable performance reports on pull requests.
2. After merging, CodSpeed will automatically create a performance
baseline from the `master` branch.
3. Subsequent PRs will receive performance comparisons showing
regressions and improvements.

---------

Co-authored-by: codspeed-hq[bot] <117304815+codspeed-hq[bot]@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Alex-PLACET <2400067+Alex-PLACET@users.noreply.github.com>

Author: 3 people

.github/workflows/codspeed.yml

@@ -0,0 +1,51 @@
+name: CodSpeed
+
+on:
+  push:
+    branches: [master]
+  pull_request:
+  # Allow CodSpeed to trigger backtest performance analysis
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  id-token: write
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.job }}-${{ github.ref }}
+  cancel-in-progress: true
+
+defaults:
+  run:
+    shell: bash -e -l {0}
+
+jobs:
+  benchmarks:
+    name: Run benchmarks
+    runs-on: ubuntu-latest
+    steps:
+    - name: Checkout code
+      uses: actions/checkout@v6
+
+    - name: Set conda environment
+      uses: mamba-org/setup-micromamba@main
+      with:
+        environment-name: myenv
+        environment-file: environment-dev.yml
+        init-shell: bash
+        cache-downloads: true
+
+    - name: Build benchmarks
+      run: |
+        cmake -G Ninja -Bbuild \
+          -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX \
+          -DBUILD_BENCHMARK=ON \
+          -DXTENSOR_USE_XSIMD=ON \
+          -DCODSPEED_MODE=simulation
+        cmake --build build --target benchmark_xtensor --parallel 8
+
+    - name: Run benchmarks
+      uses: CodSpeedHQ/action@v4
+      with:
+        mode: simulation
+        run: ./build/benchmark/benchmark_xtensor

README.md

@@ -7,6 +7,7 @@
 [![Doxygen -> gh-pages](https://github.com/xtensor-stack/xtensor/workflows/gh-pages/badge.svg)](https://xtensor-stack.github.io/xtensor)
 [![Binder](https://mybinder.org/badge.svg)](https://mybinder.org/v2/gh/xtensor-stack/xtensor/stable?filepath=notebooks%2Fxtensor.ipynb)
 [![Zulip](https://img.shields.io/badge/social_chat-zulip-blue.svg)](https://xtensor.zulipchat.com/#narrow/channel/539553-Ask-anything)
+[![CodSpeed](https://img.shields.io/endpoint?url=https://codspeed.io/badge.json)](https://codspeed.io/xtensor-stack/xtensor?utm_source=badge)
 
 Multi-dimensional arrays with broadcasting and lazy computing.
 

benchmark/CMakeLists.txt

@@ -16,17 +16,26 @@ if (CMAKE_CURRENT_SOURCE_DIR STREQUAL CMAKE_SOURCE_DIR)
     set(XTENSOR_INCLUDE_DIR ${xtensor_INCLUDE_DIRS})
 endif ()
 
-message(STATUS "Forcing tests build type to Release")
-set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build." FORCE)
+if(CODSPEED_MODE AND NOT CODSPEED_MODE STREQUAL "off")
+    message(STATUS "CodSpeed mode enabled (${CODSPEED_MODE}) - using RelWithDebInfo build type")
+    set(CMAKE_BUILD_TYPE RelWithDebInfo CACHE STRING "Choose the type of build." FORCE)
+else()
+    message(STATUS "Forcing tests build type to Release")
+    set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build." FORCE)
+endif()
 
 include(CheckCXXCompilerFlag)
 
 string(TOUPPER "${CMAKE_BUILD_TYPE}" U_CMAKE_BUILD_TYPE)
 
 if (CMAKE_CXX_COMPILER_ID MATCHES "Clang" OR CMAKE_CXX_COMPILER_ID MATCHES "GNU" OR CMAKE_CXX_COMPILER_ID MATCHES "Intel")
-    CHECK_CXX_COMPILER_FLAG(-march=native arch_native_supported)
-    if(arch_native_supported AND NOT CMAKE_CXX_FLAGS MATCHES "-march")
-        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=native")
+    # Skip -march=native for CodSpeed simulation mode (valgrind does not
+    # support all native instruction sets)
+    if(NOT (CODSPEED_MODE STREQUAL "simulation"))
+        CHECK_CXX_COMPILER_FLAG(-march=native arch_native_supported)
+        if(arch_native_supported AND NOT CMAKE_CXX_FLAGS MATCHES "-march")
+            set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=native")
+        endif()
     endif()
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3 -g -Wunused-parameter -Wextra -Wreorder")
 
@@ -80,13 +89,14 @@ if(DOWNLOAD_GBENCHMARK OR GBENCHMARK_SRC_DIR)
             GIT_TAG main)
 
     FetchContent_Declare(googlebenchmark
-            GIT_REPOSITORY https://github.com/google/benchmark.git
-            GIT_TAG main) # need main for benchmark::benchmark
+            GIT_REPOSITORY https://github.com/CodSpeedHQ/codspeed-cpp.git
+            GIT_TAG main
+            SOURCE_SUBDIR google_benchmark)
 
     FetchContent_MakeAvailable(
             googletest
             googlebenchmark)
-    set(GBENCHMARK_INCLUDE_DIRS "${googlebenchmark_SOURCE_DIR}/include")
+    set(GBENCHMARK_INCLUDE_DIRS "${googlebenchmark_SOURCE_DIR}/google_benchmark/include")
     set(GBENCHMARK_LIBRARIES benchmark)
 else()
     find_package(benchmark REQUIRED)