Releases: rapidsai/rapids-cmake
Releases · rapidsai/rapids-cmake
v24.10.00
🚨 Breaking Changes
- update fmt (to 11.0.2) and spdlog (to 1.14.1) (#689) @jameslamb
- Remove 24.10 deprecated commands (#665) @robertmaynard
🐛 Bug Fixes
- Add workaround for gcc bug in cuco (#685) @miscco
- Revert "Update to CPM 0.40.2 to fix CMake 3.30 deprecation warnings (#678)" (#679) @jameslamb
- Update to CPM 0.40.2 to fix CMake 3.30 deprecation warnings (#678) @robertmaynard
- Bump cuco version to fetch several bug fixes (#677) @PointKernel
- rapids_cpm_nvcomp has consistent behavior on every cmake execution (#676) @robertmaynard
- rapids-cmake generated C++ files have current copyright year (#674) @robertmaynard
write_language
now properly walks up theadd_subdirectory
call stack (#671) @robertmaynard
📖 Documentation
- Update docs for overriding rapids-cmake (#681) @KyleFromNVIDIA
- add ASSOCIATED_TARGETS to function signature in rapids_cython_create_modules() docs (#670) @jameslamb
🚀 New Features
- Reduce cpm tests network usage (#683) @robertmaynard
- rapids-cmake tests no longer download CPM when
NO_CPM_CACHE
is set. (#682) @robertmaynard - Remove deprecated rapids_export_find_package_* signatures (#666) @robertmaynard
- Upgrade nvcomp to 4.0.1 (#633) @vuule
🛠️ Improvements
- Add INSTALL_TARGET argument to rapids_add_test() (#692) @KyleFromNVIDIA
- Use CI workflow branch 'branch-24.10' again (#691) @jameslamb
- update fmt (to 11.0.2) and spdlog (to 1.14.1) (#689) @jameslamb
- Add support for Python 3.12 (#688) @jameslamb
- Update rapidsai/pre-commit-hooks (#686) @KyleFromNVIDIA
- Drop Python 3.9 support (#684) @jameslamb
- Allow nested lib location for nvcomp (#675) @KyleFromNVIDIA
- Update pre-commit hooks (#669) @KyleFromNVIDIA
- Improve update-version.sh (#668) @bdice
- Remove 24.10 deprecated commands (#665) @robertmaynard
[NIGHTLY] v24.12.00
🔗 Links
🐛 Bug Fixes
- Search for the normalized package name when patching (#715) @vyasr
- rapids_export_write_language no longer generates cmake dev warnings (#712) @robertmaynard
- Update rapids-cmake to support cccl 2.8 new install rules (#711) @robertmaynard
- gtest_discover_tests doesn't crash rapids_test_install_relocatable (#697) @robertmaynard
📖 Documentation
🚀 New Features
- Verifies that current_json_dir placeholder works with normalized package names (#716) @robertmaynard
- Allow for cpm override via file and direct calls to rapids_cpm_package_override (#714) @robertmaynard
- Upgrade nvcomp to 4.1.0.6 (#709) @bdice
- Warn when override of a default package uses the incorrect case (#708) @robertmaynard
- Allow CCCL 2.7 to be used in a rapids-cmake override (#706) @robertmaynard
- version.json patch command supports required field (#700) @robertmaynard
- Bump cuco version to fetch the Hyperloglog renaming (#698) @PointKernel
🛠️ Improvements
- Bump cuco version to fetch the new multiset retrieve (#707) @PointKernel
v24.08.00
🚨 Breaking Changes
- Move required CMake version to 3.26.4 (#627) @robertmaynard
- Removes legacy rapids-cmake cython implementations as it is deprecated in 24.08 (#614) @robertmaynard
- Update CCCL to v2.5.0 (#607) @trxcllnt
🐛 Bug Fixes
- bs_thread_pool uses C++ 17 (#662) @KyleFromNVIDIA
- Use
CMAKE_CUDA_ARCHITECTURES
value when ENV{CUDAARCHS} is set (#659) @robertmaynard - Pass
GLOBAL_TARGETS
torapids_cpm_find()
inrapids_cpm_bs_thread_pool()
(#655) @KyleFromNVIDIA - Add rapids_cpm_nvtx3 to cmake-format-rapids-cmake.json (#652) @KyleFromNVIDIA
- generate_resource_spec uses the enabled languages to determine compiler (#645) @robertmaynard
- Set CUDA_RUNTIME_LIBRARY to documented case style (#641) @KyleFromNVIDIA
- rapids_test_install_relocatable now handles SOVERSION target properties (#636) @robertmaynard
- Eval CMAKE_CUDA_ARCHITECTURES before ENV{CUDAARCHS} (#624) @robertmaynard
📖 Documentation
- Fix CPM package docs. (#637) @bdice
- expand rapids-cmake cpm docs (#613) @robertmaynard
🚀 New Features
- Add rapids_cpm_bs_thread_pool() (#651) @KyleFromNVIDIA
- proprietary_binary_cuda_version_mapping allows for better CUDA version mapping (#648) @robertmaynard
- use latest cuco with insert_or_apply CAS fix (#646) @srinivasyadav18
- Fetch the latest cuco with CAS fixes (#644) @PointKernel
- Update to CPM v0.40.0 (#642) @robertmaynard
- Remove CCCL patch for PR 211. (#640) @bdice
- Try any failing tests up to 3 times to guard against network issues (#639) @robertmaynard
- rapids_cmake_support_conda_env adds
-O0
to debug compile lines (#635) @robertmaynard - Update cuco git tag to fetch new multiset data structure (#628) @PointKernel
- Move required CMake version to 3.26.4 (#627) @robertmaynard
- Removes legacy rapids-cmake cython implementations as it is deprecated in 24.08 (#614) @robertmaynard
- Update CCCL to v2.5.0 (#607) @trxcllnt
🛠️ Improvements
- Use workflow branch 24.08 again (#647) @KyleFromNVIDIA
- Build and test with CUDA 12.5.1 (#643) @KyleFromNVIDIA
- Bump CCCL version to include cuda::std::span fix (#631) @sleeepyjack
- Adopt CI/packaging codeowners (#629) @bdice
v24.06.00
🐛 Bug Fixes
- Only output CUDA architectures we are building for once (#621) @robertmaynard
- set_architectures output correct CUDA arch values for RAPIDS mode (#619) @robertmaynard
- Always offer the install target names for nvtx3 (#617) @robertmaynard
- Support CMAKE_INSTALL_MESSAGE + rapids_test_install_relocatable (#604) @robertmaynard
- Ensure nvcomps build and install layouts are consistent (#602) @robertmaynard
- Correctly set the install location for nvcomp when using the proprietary binary (#597) @vyasr
- Ensure support_conda_env uses
isystem
(#588) @robertmaynard - Update rapids_test_install_relocatable to be aware of CMake 3.29 (#586) @robertmaynard
- rapids_cpm_gtest(BUILD_STATIC) now doesn't search for a local version (#585) @robertmaynard
- Add new patch to hide more CCCL APIs (#580) @vyasr
- Forward-merge branch-24.04 into branch-24.06 [skip ci] (#565) @rapids-bot[bot]
📖 Documentation
- use inline code formatting in docs for variables, functions, and modules (#591) @jameslamb
- clarify language around how FetchContent is used in RAPIDS.cmake (#590) @jameslamb
- fix docs for rapids_export_ functions (#589) @jameslamb
- Better explain OVERRIDE option (#578) @robertmaynard
🚀 New Features
- Output what cuda archs rapids-cmake is building for (#609) @robertmaynard
- Add rapids_cpm_nvtx3. (#606) @bdice
- Refactor the common
verify
cpm pin test logic to a single source (#601) @robertmaynard - rapids-cmake allow GENERATE_PINNED_VERSIONS via CMake variable (#600) @robertmaynard
- Allow for cpm default and override files via variables (#596) @robertmaynard
- Expand rapids_cpm_init to support custom default version files (#595) @robertmaynard
- Bump NVBench version for new
main
hooks. (#584) @alliepiper rapids_cython_create_modules()
: Generate Cython dependency file (#579) @Jacobfaib- rapids_cpm_gtest adds support for BUILD_STATIC (#576) @robertmaynard
🛠️ Improvements
- Fix
nvtx3
build export (#615) @trxcllnt - limit pinning tests to CPM-downloaded projects (#599) @jameslamb
- Migrate to
{{ stdlib("c") }}
(#594) @hcho3 - resolve 'file_key' deprecation warning from rapids-dependency-file-generator (#593) @jameslamb
- Prevent path conflict in builds (#571) @AyodeAwe
- Bump cuco version to fetch the lastest set retrieve API (#569) @PointKernel
- Forward-merge branch-24.04 to branch-24.06 (#563) @bdice
v24.04.00
🐛 Bug Fixes
- nvcomp try proprietary binary when 'always_download' is on (#570) @robertmaynard
- Update pre-commit-hooks to v0.0.3 (#566) @KyleFromNVIDIA
- Always link against cudart_static (#564) @KyleFromNVIDIA
- Use CUDA compiler if available for generate_resource_spec (#561) @KyleFromNVIDIA
- rapids-cmake support empty patches array in json (#559) @robertmaynard
- Handle CMake 3.28 new EXCLUDE_FROM_ALL option of
FetchContent
(#557) @robertmaynard - Add rapids_cuda_set_runtime to default includes of cuda (#538) @robertmaynard
- add rapids-dependency-file-generator pre-commit hook (#531) @jameslamb
🚀 New Features
- Deprecate rapids_cpm_libcudacxx and rapids_cpm_thrust. (#560) @bdice
- rapids_cpm_package_details now validates required entries exist (#558) @robertmaynard
- Support getting rapids-cmake via git clone (#555) @robertmaynard
- Bump nvbench version for faster benchmark runs (#549) @PointKernel
- Remove unneeded whitespace from json (#544) @robertmaynard
- Officially support env var expansion in version.json (#540) @robertmaynard
- rapids-cmake can generate pinned versions file (#530) @robertmaynard
- Fetch the latest cuco and remove outdated patches (#526) @PointKernel
- Support CUDA 12.2 (#521) @jameslamb
🛠️ Improvements
- Use
conda env create --yes
instead of--force
(#573) @bdice - Replace local copyright check with pre-commit-hooks verify-copyright (#556) @KyleFromNVIDIA
- Add patch to fix fmt
v10.1.1
version (#548) @trxcllnt - Add support for Python 3.11 (#547) @jameslamb
- Forward-merge branch-24.02 to branch-24.04 (#545) @bdice
- target branch-24.04 for GitHub Actions workflows (#541) @jameslamb
- Build generate_ctest_json in try_compile() (#537) @KyleFromNVIDIA
- Ensure that
ctest
is called with--no-tests=error
. (#535) @bdice - Update ops-bot.yaml (#532) @AyodeAwe
v24.02.01
🚨 Breaking Changes
🐛 Bug Fixes
- Bump to nvcomp 3.0.6. (#542) @bdice
- Error out if generate_ctest_json fails to build or run (#533) @KyleFromNVIDIA
- rapids_cpm_cccl now works as expected when given
DOWNLOAD_ONLY ON
(#527) @robertmaynard - Always download repos when they are being patched (#525) @vyasr
- Mark all cccl and cuco kernels with hidden visibility (#523) @robertmaynard
- Fix message context (#520) @vyasr
- Generate template copyright year at build time. (#325) (#519) @KyleFromNVIDIA
- Add link libraries to generate_resource_spec cmake (#516) @davidwendt
- rapids_cpm_cccl preserve install location details from first invocation. (#513) @robertmaynard
- Only apply install rules for CCCL if we actually downloaded (#507) @bdice
- Mark flaky test as serial (#506) @vyasr
- Manually invoke install rules for components (#505) @vyasr
- multiple entry overrides now sets FetchContent for all entries (#494) @robertmaynard
- Remove deprecated function usages (#484) @robertmaynard
📖 Documentation
- Fix docs references to API sections. (#509) @bdice
- fix typo in README (#501) @jameslamb
- Fix indentation typo (#497) @vyasr
🚀 New Features
- rapids cpm patches now support differences in white space. (#515) @robertmaynard
- Upgrade nvCOMP to 3.0.5 (#498) @davidwendt
- Move to latest nvbench which has nvml+static support (#488) @robertmaynard
- Update to spdlog 1.12 and fmt 10.1.1 (#473) @kkraus14
- Support scikit-build-core (#433) @vyasr
🛠️ Improvements
- Remove usages of rapids-env-update (#524) @KyleFromNVIDIA
- refactor CUDA versions in dependencies.yaml (#517) @jameslamb
- Remove scikit-build from dependency list (#512) @vyasr
- Add patch reverting CCCL PR 211. (#511) @bdice
- Update cuCollections for CCCL 2.2.0 support. (#510) @bdice
- Disable NVBench CUPTI support by default. (#504) @bdice
- Remove CCCL::Thrust from GLOBAL_TARGETS. (#500) @bdice
- Add missing nvcomp targets (#496) @vyasr
- Add rapids_cpm_cccl feature. (#495) @bdice
- Drop Pascal architecture (60). (#482) @bdice
v24.02.00
🚨 Breaking Changes
🐛 Bug Fixes
- Error out if generate_ctest_json fails to build or run (#533) @KyleFromNVIDIA
- rapids_cpm_cccl now works as expected when given
DOWNLOAD_ONLY ON
(#527) @robertmaynard - Always download repos when they are being patched (#525) @vyasr
- Mark all cccl and cuco kernels with hidden visibility (#523) @robertmaynard
- Fix message context (#520) @vyasr
- Generate template copyright year at build time. (#325) (#519) @KyleFromNVIDIA
- Add link libraries to generate_resource_spec cmake (#516) @davidwendt
- rapids_cpm_cccl preserve install location details from first invocation. (#513) @robertmaynard
- Only apply install rules for CCCL if we actually downloaded (#507) @bdice
- Mark flaky test as serial (#506) @vyasr
- Manually invoke install rules for components (#505) @vyasr
- multiple entry overrides now sets FetchContent for all entries (#494) @robertmaynard
- Remove deprecated function usages (#484) @robertmaynard
📖 Documentation
- Fix docs references to API sections. (#509) @bdice
- fix typo in README (#501) @jameslamb
- Fix indentation typo (#497) @vyasr
🚀 New Features
- rapids cpm patches now support differences in white space. (#515) @robertmaynard
- Upgrade nvCOMP to 3.0.5 (#498) @davidwendt
- Move to latest nvbench which has nvml+static support (#488) @robertmaynard
- Update to spdlog 1.12 and fmt 10.1.1 (#473) @kkraus14
- Support scikit-build-core (#433) @vyasr
🛠️ Improvements
- Remove usages of rapids-env-update (#524) @KyleFromNVIDIA
- refactor CUDA versions in dependencies.yaml (#517) @jameslamb
- Remove scikit-build from dependency list (#512) @vyasr
- Add patch reverting CCCL PR 211. (#511) @bdice
- Update cuCollections for CCCL 2.2.0 support. (#510) @bdice
- Disable NVBench CUPTI support by default. (#504) @bdice
- Remove CCCL::Thrust from GLOBAL_TARGETS. (#500) @bdice
- Add missing nvcomp targets (#496) @vyasr
- Add rapids_cpm_cccl feature. (#495) @bdice
- Drop Pascal architecture (60). (#482) @bdice
v23.12.00
🚨 Breaking Changes
🐛 Bug Fixes
- Ensure nvbench initializes nvml context when built statically (#486) @robertmaynard
- Remove invalid argument to find_package_root (#483) @robertmaynard
- target from write_git_revision now works with export sets (#474) @robertmaynard
🚀 New Features
- Support static builds of gbench and nvbench. (#481) @robertmaynard
- Allow rapids_test to be used without CUDAToolkit (#480) @robertmaynard
- Update cuco git tag (#479) @sleeepyjack
- GTest will always be PIC enabled when built (#477) @robertmaynard
- Thrust when exported now automatically calls
thrust_create_target
(#467) @robertmaynard - Upgrade nvCOMP to 3.0.4 (#451) @vuule
🛠️ Improvements
- Build concurrency for nightly and merge triggers (#490) @bdice
- Add patch for libcudacxx memory resource. (#476) @bdice
- Use branch-23.12 workflows. (#472) @bdice
- Express Python version in dependencies.yaml. (#470) @bdice
- Build CUDA 12.0 ARM conda packages. (#468) @bdice
- Update libcudacxx to 2.1.0 (#464) @bdice
v23.10.00
🐛 Bug Fixes
- Quote the list of patch files in case they have spaces in their paths (#463) @ericniebler
- cpm overrides don't occur when
CPM_<pkg>_SOURCE
exists (#458) @robertmaynard - Use
conda mambabuild
notmamba mambabuild
(#457) @bdice - Support fmt use in debug builds (#456) @robertmaynard
📖 Documentation
- Move rapids_cpm_package_override to CPM section of docs (#462) @robertmaynard
- Improve docs around fetch content and rapids-cmake overrides (#444) @robertmaynard
🚀 New Features
- Bump cuco version (#452) @PointKernel
🛠️ Improvements
v23.08.00
🐛 Bug Fixes
- Use < gcc-11 with cuda 11.5 to avoid nvbench compile failures (#448) @robertmaynard
- Ensure tests the modify same git repo don't execute at the same time (#446) @robertmaynard
- Fix CUDA 11.5 tests by adding dependencies entries. (#443) @bdice
- Remove trailing comma and add pre-commit hook for JSON validation. (#440) @bdice
- When nvcomp is found locally print where it is on disk (#434) @robertmaynard
- Correct two issues found when testing CMake 3.27 rc2 (#432) @robertmaynard
- Correct re-root controls from conda-forge with thrust/cub/etc (#431) @robertmaynard
- Bug/proprietary binary obeys
always_download
(#430) @robertmaynard - Correct install_relocatable issues found by libcudf (#423) @robertmaynard
- test_install_relocatable correct run_gpu_test.cmake location (#420) @robertmaynard
- Fea/move to latest nvbench (#417) @robertmaynard
- Use @loader_path instead of $ORIGIN on MacOS (#403) @manopapad
- Make NAMESPACE property truly optional in rapids_export (#358) @agirault
🚀 New Features
- Update rapids-cmake ci to support conda-forge CUDA 12 (#437) @robertmaynard
- Bump cuco version (#435) @PointKernel
- Add rapids_cuda_set_runtime (#429) @robertmaynard
- support_conda_env support host and build CTK 12 locations (#428) @robertmaynard
- rapids_find_generate_module Support user code blocks (#415) @robertmaynard
- Rewrite of rapids_test_install_relocatable to support genex expressions (#410) @robertmaynard
🛠️ Improvements
- Conditionally modify envvar vs. global CMAKE_PREFIX_PATH in
rapids_cmake_support_conda_env
(#439) @trxcllnt - Migrate to updated shared-action-workflows name for CUDA 12 CI (#438) @bdice
- Fix google benchmark name and update version (#425) @vyasr
- use rapids-upload-docs script (#419) @AyodeAwe
- Remove documentation build scripts for Jenkins (#418) @ajschmidt8
- Upload conda packages for rapids_core_dependencies. (#414) @bdice