Merge pull request #1505 from rapidsai/branch-24.04

[RELEASE] rmm v24.04

.clang-format

@@ -71,8 +71,20 @@ ForEachMacros:
   - foreach
   - Q_FOREACH
   - BOOST_FOREACH
-IncludeBlocks: Preserve
-IncludeIsMainRegex: '([-_](test|unittest))?$'
+IncludeBlocks: Regroup
+IncludeCategories:
+  - Regex:           '^"' # quoted includes
+    Priority:        1
+  - Regex:           '^<rmm/' # RMM includes
+    Priority:        2
+  - Regex:           '^<(thrust|cub|cuda)/' # CCCL includes
+    Priority:        3
+  - Regex:           '^<(cooperative_groups|cuco|cuda.h|cuda_runtime|device_types|math_constants|nvtx3)' # CUDA includes
+    Priority:        3
+  - Regex:           '^<.*\..*' # other system includes (e.g. with a '.')
+    Priority:        4
+  - Regex:           '^<[^.]+' # STL includes (no '.')
+    Priority:        5
 IndentCaseLabels: true
 IndentPPDirectives: None
 IndentWidth:     2

.devcontainer/cuda11.8-conda/devcontainer.json

@@ -5,12 +5,12 @@
     "args": {
       "CUDA": "11.8",
       "PYTHON_PACKAGE_MANAGER": "conda",
-      "BASE": "rapidsai/devcontainers:24.02-cpp-cuda11.8-mambaforge-ubuntu22.04"
+      "BASE": "rapidsai/devcontainers:24.04-cpp-cuda11.8-mambaforge-ubuntu22.04"
     }
   },
   "hostRequirements": {"gpu": "optional"},
   "features": {
-    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.2": {}
+    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.4": {}
   },
   "overrideFeatureInstallOrder": [
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils"

.devcontainer/cuda11.8-pip/devcontainer.json

@@ -5,12 +5,12 @@
     "args": {
       "CUDA": "11.8",
       "PYTHON_PACKAGE_MANAGER": "pip",
-      "BASE": "rapidsai/devcontainers:24.02-cpp-llvm16-cuda11.8-ubuntu22.04"
+      "BASE": "rapidsai/devcontainers:24.04-cpp-cuda11.8-ubuntu22.04"
     }
   },
   "hostRequirements": {"gpu": "optional"},
   "features": {
-    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.2": {}
+    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.4": {}
   },
   "overrideFeatureInstallOrder": [
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils"

.devcontainer/cuda12.0-conda/devcontainer.json → .devcontainer/cuda12.2-conda/devcontainer.json

@@ -3,19 +3,19 @@
     "context": "${localWorkspaceFolder}/.devcontainer",
     "dockerfile": "${localWorkspaceFolder}/.devcontainer/Dockerfile",
     "args": {
-      "CUDA": "12.0",
+      "CUDA": "12.2",
       "PYTHON_PACKAGE_MANAGER": "conda",
-      "BASE": "rapidsai/devcontainers:24.02-cpp-mambaforge-ubuntu22.04"
+      "BASE": "rapidsai/devcontainers:24.04-cpp-mambaforge-ubuntu22.04"
     }
   },
   "hostRequirements": {"gpu": "optional"},
   "features": {
-    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.2": {}
+    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.4": {}
   },
   "overrideFeatureInstallOrder": [
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils"
   ],
-  "initializeCommand": ["/bin/bash", "-c", "mkdir -m 0755 -p ${localWorkspaceFolder}/../.{aws,cache,config,conda/pkgs,conda/${localWorkspaceFolderBasename}-cuda12.0-envs}"],
+  "initializeCommand": ["/bin/bash", "-c", "mkdir -m 0755 -p ${localWorkspaceFolder}/../.{aws,cache,config,conda/pkgs,conda/${localWorkspaceFolderBasename}-cuda12.2-envs}"],
   "postAttachCommand": ["/bin/bash", "-c", "if [ ${CODESPACES:-false} = 'true' ]; then . devcontainer-utils-post-attach-command; . rapids-post-attach-command; fi"],
   "workspaceFolder": "/home/coder",
   "workspaceMount": "source=${localWorkspaceFolder},target=/home/coder/rmm,type=bind,consistency=consistent",
@@ -24,7 +24,7 @@
     "source=${localWorkspaceFolder}/../.cache,target=/home/coder/.cache,type=bind,consistency=consistent",
     "source=${localWorkspaceFolder}/../.config,target=/home/coder/.config,type=bind,consistency=consistent",
     "source=${localWorkspaceFolder}/../.conda/pkgs,target=/home/coder/.conda/pkgs,type=bind,consistency=consistent",
-    "source=${localWorkspaceFolder}/../.conda/${localWorkspaceFolderBasename}-cuda12.0-envs,target=/home/coder/.conda/envs,type=bind,consistency=consistent"
+    "source=${localWorkspaceFolder}/../.conda/${localWorkspaceFolderBasename}-cuda12.2-envs,target=/home/coder/.conda/envs,type=bind,consistency=consistent"
   ],
   "customizations": {
     "vscode": {

.devcontainer/cuda12.0-pip/devcontainer.json → .devcontainer/cuda12.2-pip/devcontainer.json

@@ -3,27 +3,27 @@
     "context": "${localWorkspaceFolder}/.devcontainer",
     "dockerfile": "${localWorkspaceFolder}/.devcontainer/Dockerfile",
     "args": {
-      "CUDA": "12.0",
+      "CUDA": "12.2",
       "PYTHON_PACKAGE_MANAGER": "pip",
-      "BASE": "rapidsai/devcontainers:24.02-cpp-llvm16-cuda12.0-ubuntu22.04"
+      "BASE": "rapidsai/devcontainers:24.04-cpp-cuda12.2-ubuntu22.04"
     }
   },
   "hostRequirements": {"gpu": "optional"},
   "features": {
-    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.2": {}
+    "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.4": {}
   },
   "overrideFeatureInstallOrder": [
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils"
   ],
-  "initializeCommand": ["/bin/bash", "-c", "mkdir -m 0755 -p ${localWorkspaceFolder}/../.{aws,cache,config/pip,local/share/${localWorkspaceFolderBasename}-cuda12.0-venvs}"],
+  "initializeCommand": ["/bin/bash", "-c", "mkdir -m 0755 -p ${localWorkspaceFolder}/../.{aws,cache,config/pip,local/share/${localWorkspaceFolderBasename}-cuda12.2-venvs}"],
   "postAttachCommand": ["/bin/bash", "-c", "if [ ${CODESPACES:-false} = 'true' ]; then . devcontainer-utils-post-attach-command; . rapids-post-attach-command; fi"],
   "workspaceFolder": "/home/coder",
   "workspaceMount": "source=${localWorkspaceFolder},target=/home/coder/rmm,type=bind,consistency=consistent",
   "mounts": [
     "source=${localWorkspaceFolder}/../.aws,target=/home/coder/.aws,type=bind,consistency=consistent",
     "source=${localWorkspaceFolder}/../.cache,target=/home/coder/.cache,type=bind,consistency=consistent",
     "source=${localWorkspaceFolder}/../.config,target=/home/coder/.config,type=bind,consistency=consistent",
-    "source=${localWorkspaceFolder}/../.local/share/${localWorkspaceFolderBasename}-cuda12.0-venvs,target=/home/coder/.local/share/venvs,type=bind,consistency=consistent"
+    "source=${localWorkspaceFolder}/../.local/share/${localWorkspaceFolderBasename}-cuda12.2-venvs,target=/home/coder/.local/share/venvs,type=bind,consistency=consistent"
   ],
   "customizations": {
     "vscode": {

.github/ops-bot.yaml

@@ -6,3 +6,4 @@ branch_checker: true
 label_checker: true
 release_drafter: true
 recently_updated: true
+forward_merger: true

.github/workflows/build.yaml

@@ -28,7 +28,7 @@ concurrency:
 jobs:
   cpp-build:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.04
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -37,7 +37,7 @@ jobs:
   python-build:
     needs: [cpp-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.04
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -46,7 +46,7 @@ jobs:
   upload-conda:
     needs: [cpp-build, python-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-upload-packages.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-upload-packages.yaml@branch-24.04
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -56,7 +56,7 @@ jobs:
     if: github.ref_type == 'branch'
     needs: python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.04
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -68,7 +68,7 @@ jobs:
       run_script: "ci/build_docs.sh"
   wheel-build:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.04
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -78,7 +78,7 @@ jobs:
   wheel-publish:
     needs: wheel-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@branch-24.04
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}

.github/workflows/pr.yaml

@@ -22,40 +22,40 @@ jobs:
       - wheel-tests
       - devcontainer
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@branch-24.04
   checks:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@branch-24.04
     with:
       enable_check_generated_files: false
   conda-cpp-build:
     needs: checks
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.04
     with:
       build_type: pull-request
   conda-cpp-tests:
     needs: conda-cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.04
     with:
       build_type: pull-request
   conda-python-build:
     needs: conda-cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.04
     with:
       build_type: pull-request
   conda-python-tests:
     needs: conda-python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.04
     with:
       build_type: pull-request
   docs-build:
     needs: conda-python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.04
     with:
       build_type: pull-request
       node_type: "gpu-v100-latest-1"
@@ -65,21 +65,23 @@ jobs:
   wheel-build:
     needs: checks
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.04
     with:
       build_type: pull-request
       script: ci/build_wheel.sh
   wheel-tests:
     needs: wheel-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.04
     with:
       build_type: pull-request
       script: ci/test_wheel.sh
   devcontainer:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/build-in-devcontainer.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/build-in-devcontainer.yaml@fix/devcontainer-json-location
     with:
+      arch: '["amd64"]'
+      cuda: '["12.2"]'
       build_command: |
         sccache -z;
         build-all -DBUILD_BENCHMARKS=ON --verbose;

.github/workflows/test.yaml

@@ -16,23 +16,23 @@ on:
 jobs:
   cpp-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.04
     with:
       build_type: nightly
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
   python-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.04
     with:
       build_type: nightly
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
   wheel-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.02
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.04
     with:
       build_type: nightly
       branch: ${{ inputs.branch }}

CHANGELOG.md

@@ -1,3 +1,57 @@
+# RMM 24.04.00 (10 Apr 2024)
+
+## 🚨 Breaking Changes
+
+- Accept stream argument in DeviceMemoryResource allocate/deallocate ([#1494](https://github.com/rapidsai/rmm/pull/1494)) [@wence-](https://github.com/wence-)
+- Replace all internal usage of `get_upstream` with `get_upstream_resource` ([#1491](https://github.com/rapidsai/rmm/pull/1491)) [@miscco](https://github.com/miscco)
+- Deprecate rmm::mr::device_memory_resource::supports_streams() ([#1452](https://github.com/rapidsai/rmm/pull/1452)) [@harrism](https://github.com/harrism)
+- Remove deprecated rmm::detail::available_device_memory ([#1438](https://github.com/rapidsai/rmm/pull/1438)) [@harrism](https://github.com/harrism)
+- Make device_memory_resource::supports_streams() not pure virtual. Remove derived implementations and calls in RMM ([#1437](https://github.com/rapidsai/rmm/pull/1437)) [@harrism](https://github.com/harrism)
+- Deprecate rmm::mr::device_memory_resource::get_mem_info() and supports_get_mem_info(). ([#1436](https://github.com/rapidsai/rmm/pull/1436)) [@harrism](https://github.com/harrism)
+
+## 🐛 Bug Fixes
+
+- Fix search path for torch allocator in editable installs and ensure CUDA support is available ([#1498](https://github.com/rapidsai/rmm/pull/1498)) [@vyasr](https://github.com/vyasr)
+- Accept stream argument in DeviceMemoryResource allocate/deallocate ([#1494](https://github.com/rapidsai/rmm/pull/1494)) [@wence-](https://github.com/wence-)
+- Run STATISTICS_TEST and TRACKING_TEST in serial to avoid OOM errors. ([#1487](https://github.com/rapidsai/rmm/pull/1487)) [@bdice](https://github.com/bdice)
+
+## 📖 Documentation
+
+- Pin to recent breathe, to prevent getting an unsupported sphinx version. ([#1495](https://github.com/rapidsai/rmm/pull/1495)) [@bdice](https://github.com/bdice)
+
+## 🚀 New Features
+
+- Replace all internal usage of `get_upstream` with `get_upstream_resource` ([#1491](https://github.com/rapidsai/rmm/pull/1491)) [@miscco](https://github.com/miscco)
+- Add complete set of resource ref aliases ([#1479](https://github.com/rapidsai/rmm/pull/1479)) [@nvdbaranec](https://github.com/nvdbaranec)
+- Automate include grouping using clang-format ([#1463](https://github.com/rapidsai/rmm/pull/1463)) [@harrism](https://github.com/harrism)
+- Add `get_upstream_resource` to resource adaptors ([#1456](https://github.com/rapidsai/rmm/pull/1456)) [@miscco](https://github.com/miscco)
+- Deprecate rmm::mr::device_memory_resource::supports_streams() ([#1452](https://github.com/rapidsai/rmm/pull/1452)) [@harrism](https://github.com/harrism)
+- Remove duplicated memory_resource_tests ([#1451](https://github.com/rapidsai/rmm/pull/1451)) [@miscco](https://github.com/miscco)
+- Change `rmm::exec_policy` to take `async_resource_ref` ([#1449](https://github.com/rapidsai/rmm/pull/1449)) [@miscco](https://github.com/miscco)
+- Change `device_scalar` to take `async_resource_ref` ([#1447](https://github.com/rapidsai/rmm/pull/1447)) [@miscco](https://github.com/miscco)
+- Add device_async_resource_ref convenience alias ([#1441](https://github.com/rapidsai/rmm/pull/1441)) [@harrism](https://github.com/harrism)
+- Remove deprecated rmm::detail::available_device_memory ([#1438](https://github.com/rapidsai/rmm/pull/1438)) [@harrism](https://github.com/harrism)
+- Make device_memory_resource::supports_streams() not pure virtual. Remove derived implementations and calls in RMM ([#1437](https://github.com/rapidsai/rmm/pull/1437)) [@harrism](https://github.com/harrism)
+- Deprecate rmm::mr::device_memory_resource::get_mem_info() and supports_get_mem_info(). ([#1436](https://github.com/rapidsai/rmm/pull/1436)) [@harrism](https://github.com/harrism)
+- Support CUDA 12.2 ([#1419](https://github.com/rapidsai/rmm/pull/1419)) [@jameslamb](https://github.com/jameslamb)
+
+## 🛠️ Improvements
+
+- Use `conda env create --yes` instead of `--force` ([#1509](https://github.com/rapidsai/rmm/pull/1509)) [@bdice](https://github.com/bdice)
+- Add upper bound to prevent usage of NumPy 2 ([#1501](https://github.com/rapidsai/rmm/pull/1501)) [@bdice](https://github.com/bdice)
+- Remove hard-coding of RAPIDS version where possible ([#1496](https://github.com/rapidsai/rmm/pull/1496)) [@KyleFromNVIDIA](https://github.com/KyleFromNVIDIA)
+- Requre NumPy 1.23+ ([#1488](https://github.com/rapidsai/rmm/pull/1488)) [@jakirkham](https://github.com/jakirkham)
+- Use `rmm::device_async_resource_ref` in multi_stream_allocation benchmark ([#1482](https://github.com/rapidsai/rmm/pull/1482)) [@miscco](https://github.com/miscco)
+- Update devcontainers to CUDA Toolkit 12.2 ([#1470](https://github.com/rapidsai/rmm/pull/1470)) [@trxcllnt](https://github.com/trxcllnt)
+- Add support for Python 3.11 ([#1469](https://github.com/rapidsai/rmm/pull/1469)) [@jameslamb](https://github.com/jameslamb)
+- target branch-24.04 for GitHub Actions workflows ([#1468](https://github.com/rapidsai/rmm/pull/1468)) [@jameslamb](https://github.com/jameslamb)
+- [FEA]: Use `std::optional` instead of `thrust::optional` ([#1464](https://github.com/rapidsai/rmm/pull/1464)) [@miscco](https://github.com/miscco)
+- Add environment-agnostic scripts for running ctests and pytests ([#1462](https://github.com/rapidsai/rmm/pull/1462)) [@trxcllnt](https://github.com/trxcllnt)
+- Ensure that `ctest` is called with `--no-tests=error`. ([#1460](https://github.com/rapidsai/rmm/pull/1460)) [@bdice](https://github.com/bdice)
+- Update ops-bot.yaml ([#1458](https://github.com/rapidsai/rmm/pull/1458)) [@AyodeAwe](https://github.com/AyodeAwe)
+- Adopt the `rmm::device_async_resource_ref` alias ([#1454](https://github.com/rapidsai/rmm/pull/1454)) [@miscco](https://github.com/miscco)
+- Refactor error.hpp out of detail ([#1439](https://github.com/rapidsai/rmm/pull/1439)) [@lamarrr](https://github.com/lamarrr)
+
 # RMM 24.02.00 (12 Feb 2024)
 
 ## 🚨 Breaking Changes

CMakeLists.txt

@@ -14,7 +14,7 @@
 
 cmake_minimum_required(VERSION 3.26.4 FATAL_ERROR)
 
-include(fetch_rapids.cmake)
+include(rapids_config.cmake)
 
 include(rapids-cmake)
 include(rapids-cpm)
@@ -23,7 +23,7 @@ include(rapids-find)
 
 project(
   RMM
-  VERSION 24.02.00
+  VERSION "${RAPIDS_VERSION}"
   LANGUAGES CXX)
 
 # Write the version header
@@ -173,7 +173,8 @@ rapids_export(
 add_custom_command(
   OUTPUT RMM_DOXYGEN
   WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/doxygen
-  COMMAND doxygen Doxyfile
+  COMMAND ${CMAKE_COMMAND} -E env "RAPIDS_VERSION=${RAPIDS_VERSION}"
+          "RAPIDS_VERSION_MAJOR_MINOR=${RAPIDS_VERSION_MAJOR_MINOR}" doxygen Doxyfile
   VERBATIM
   COMMENT "Custom command for RMM doxygen docs")
 

README.md

@@ -43,7 +43,7 @@ conda install -c rapidsai -c conda-forge -c nvidia rmm cuda-version=12.0
 We also provide [nightly Conda packages](https://anaconda.org/rapidsai-nightly) built from the HEAD
 of our latest development branch.
 
-Note: RMM is supported only on Linux, and only tested with Python versions 3.9 and 3.10.
+Note: RMM is supported only on Linux, and only tested with Python versions 3.9, 3.10, and 3.11.
 
 Note: The RMM package from Conda requires building with GCC 9 or later. Otherwise, your application may fail to build.
 

VERSION

@@ -1 +1 @@
-24.02.00
+24.04.00

benchmarks/device_uvector/device_uvector_bench.cu

@@ -26,13 +26,12 @@
 #include <rmm/mr/device/per_device_resource.hpp>
 #include <rmm/mr/device/pool_memory_resource.hpp>
 
-#include <benchmark/benchmark.h>
-
 #include <cuda_runtime_api.h>
-
 #include <thrust/device_vector.h>
 #include <thrust/memory.h>
 
+#include <benchmark/benchmark.h>
+
 #include <cstdio>
 #include <type_traits>