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

v23.06.00

🚨 Breaking Changes

• Using deprecated CUDA_ARCHITECTURE values now produces an error. (#397) @robertmaynard
• rapids_cpm cccl packages cmake files are now relocated to not clash with upstream (#393) @robertmaynard

🐛 Bug Fixes

• Revert "Define Cython language_level explicitly. (#394)" (#396) @vyasr
• rapids_cpm cccl packages cmake files are now relocated to not clash with upstream (#393) @robertmaynard

📖 Documentation

• Correct basics to api cross refs (#405) @robertmaynard

🚀 New Features

• Update cuco git tag to support cuco::static_set (#407) @PointKernel
• Upgrade GTest version to 1.13 (#401) @robertmaynard
• Using deprecated CUDA_ARCHITECTURE values now produces an error. (#397) @robertmaynard

🛠️ Improvements

• run docs nightly too (#413) @AyodeAwe
• Update cuco git tag to fetch several bug fixes (#412) @PointKernel
• Revert shared workflows branch (#406) @ajschmidt8
• Upgrade to Python 3.9 (drop Python 3.9) (#404) @shwina
• Remove usage of rapids-get-rapids-version-from-git (#402) @jjacobelli
• Update clang-format (#398) @bdice
• Define Cython language_level explicitly. (#394) @bdice

v23.04.00

🐛 Bug Fixes

• install_relocatable only installs files that exist (#392) @robertmaynard
• Revert "install tests environment properties (#390)" (#391) @robertmaynard
• Add COMPONENT arguments for rapids_export to formatting file. (#389) @robertmaynard
• install_relocatable generate correct installed RESOURCE_SPEC_FILE (#386) @robertmaynard
• support_conda_env only add rpath-link flag to linkers that support it. (#384) @robertmaynard
• rapids_cpm_nvbench properly specify usage of external fmt library (#376) @robertmaynard
• rapids_cpm_spdlog properly specify usage of external fmt library (#375) @robertmaynard
• Patch nvbench to allow usage of external fmt (#373) @robertmaynard
• Support static builds of fmt (#372) @robertmaynard
• Update to latest nvbench (#371) @vyasr

📖 Documentation

• Fix misspelling of rapids_cpm_init (#385) @dagardner-nv

🚀 New Features

• rapids_test_install_relocatable tracks tests environment properties (#390) @robertmaynard
• rapids_test_install_relocatable EXCLUDE_FROM_ALL is now the default (#388) @robertmaynard
• Support downloading nvcomp CTK 11 or 12 binaries (#381) @robertmaynard
• Introduce clang-format to rapids-cmake to format C++ code examples (#378) @robertmaynard
• proprietary_binary now supports cuda toolkit version placeholders (#377) @robertmaynard
• Add rapids_test allowing projects to run gpu tests in parallel (#328) @robertmaynard
• Extend rapids_export to support the concept of optional COMPONENTS (#154) @robertmaynard

🛠️ Improvements

• Update to GCC 11 (#382) @bdice
• Make docs builds less verbose (#380) @AyodeAwe
• Update GHAs Workflows (#374) @ajschmidt8
• Use trap to handle errors in test scripts (#370) @AjayThorve
• Bump spdlog to 1.11, add fmt as dependency for spdlog (#368) @kkraus14
• Clean up and sort CPM packages. (#366) @bdice
• Update shared workflow branches (#365) @ajschmidt8
• Add fmt 9.1.0 (#364) @kkraus14
• Move date to build string in conda recipe (#359) @ajschmidt8
• Add docs build job (#347) @AyodeAwe

v23.02.00

🐛 Bug Fixes

• Remove incorrect deprecation for CMAKE_CUDA_ARCHITECTURES="NATIVE" (#355) @robertmaynard
• cpm: always_download now considers patches json entry (#353) @robertmaynard
• Use string literals for policy test messages so no escaping needed (#351) @robertmaynard
• Revert "Update spdlog to 1.11 ( latest version ) (#342)" (#346) @bdice
• Revert update of libcudacxx 1.9 (#337) @robertmaynard
• rapids_cuda_patch_toolkit: Better handle non-standard toolkits (#324) @robertmaynard
• Revert "Upgrade spdlog to 1.10.0 (#312)" (#323) @bdice
• rapids_cuda_init_architectures now supports CUDAARCHS env variable (#322) @robertmaynard
• Remove usage of FetchContent from tests to improve perf (#303) @robertmaynard

🚀 New Features

• Update nvCOMP version to 2.6.1 (#360) @vuule
• cpm: Rework always_download rules to be smarter (#348) @robertmaynard
• Add deprecation notice to passing "" to CMAKE_CUDA_ARCHITECTURES (#345) @robertmaynard
• Update to libcudacxx 1.9.1 to have a version >= CUDA Toolkit 12 (#343) @robertmaynard
• Update spdlog to 1.11 ( latest version ) (#342) @robertmaynard
• Update to nvcomp 2.6 (#341) @robertmaynard
• Add deprecation warnings for usage of ALL (#339) @robertmaynard
• rapids-cmake now errors out when CPM can't be downloaded (#335) @robertmaynard
• Update to nvcomp 2.5 (#333) @robertmaynard
• Update to libcudacxx 1.9 to match version found in CUDA Toolkit 12 (#332) @robertmaynard
• Update cuco git tag to fetch bug fixes and cleanups (#329) @PointKernel
• Fea/support cmake cuda architectures rapids value (#327) @robertmaynard
• Upgrade spdlog to 1.10.0 (#312) @kkraus14

🛠️ Improvements

• Update shared workflow branches (#361) @ajschmidt8
• Build against CUDA 11.8 (#344) @ajschmidt8
• Make generated find module targets global (#340) @vyasr
• Add codespell and whitespace linters to pre-commit hooks. (#338) @bdice
• Use pre-commit for style checks (#336) @bdice
• Branch 23.02 merge 22.12 (#331) @vyasr
• Update conda recipes. (#330) @bdice
• Fix typo. (#311) @vyasr

[NIGHTLY] v22.12.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• Don't use CMake 3.25.0 as it has a show stopping FindCUDAToolkit bug (#308) @robertmaynard
• Add missing CPM_ARGS to gbench (#294) @vyasr
• Patch results are only displayed once per invocation of CMake (#292) @robertmaynard
• Add thrust output iterator fix to rapids-cmake thrust patches (#291) @robertmaynard

📖 Documentation

• Update pull request template to match rest of RAPIDS (#280) @robertmaynard
• Clarify rapids_cuda_init_architectures behavior (#279) @robertmaynard

🚀 New Features

• Update cuco git tag (#302) @PointKernel
• Remove old CI files (#300) @robertmaynard
• Update cuco to version that supports Ada and Hopper (#299) @robertmaynard
• Move libcudacxx 1.8.1 so we support sm90 (#296) @robertmaynard
• Add ability to specify library directories for target rpaths (#295) @vyasr
• Add support for cloning Google benchmark (#293) @vyasr
• Add current_json_dir placeholder in json patch file values (#289) @robertmaynard
• Add sm90 ( Hopper ) to rapids-cmake "ALL" mode (#285) @robertmaynard
• Enable copy_prs ops-bot config (#284) @robertmaynard
• Add GitHub action workflow to rapids-cmake (#283) @robertmaynard
• Create conda package of patched dependencies (#275) @robertmaynard
• Switch thrust over to use rapids-cmake patches (#265) @robertmaynard

🛠️ Improvements

• Remove rapids-dependency-file-generator FIXME (#305) @ajschmidt8
• Add ninja as build dependency (#301) @ajschmidt8
• Forward merge 22.10 into 22.12 (#297) @vyasr

[NIGHTLY] v22.10.00

🔗 Links

• Development Branch
• Compare with main branch

🚨 Breaking Changes

• Update rapids-cmake to require cmake 3.23.1 (#227) @robertmaynard
• put $PREFIX before $BUILD_PREFIX in conda build (#182) @kkraus14

🐛 Bug Fixes

• Update to nvcomp 2.4.1 to fix zstd decompression (#286) @robertmaynard
• Restore rapids_cython_create_modules output variable name (#276) @robertmaynard
• rapids_cuda_init_architectures now obeys CUDAARCHS env variable (#270) @robertmaynard
• Update to Thrust 1.17.2 to fix cub ODR issues (#269) @robertmaynard
• conda_env: pass conda prefix as a rpath-link directory (#263) @robertmaynard
• Update cuCollections to fix issue with INSTALL_CUCO set to OFF. (#261) @bdice
• rapids_cpm_libcudacxx correct location of libcudacxx-config (#258) @robertmaynard
• Update rapids_find_generate_module to cmake 3.23 (#256) @robertmaynard
• Handle reconfiguring with USE_PROPRIETARY_BINARY value differing (#255) @robertmaynard
• rapids_cpm_thrust record build directory location of thrust-config (#254) @robertmaynard
• disable cuco install rules when no INSTALL_EXPORT_SET (#250) @robertmaynard
• Patch thrust and cub install rules to have proper header searches (#244) @robertmaynard
• Ensure that we install Thrust and Cub correctly. (#243) @robertmaynard
• Revert "Update to CPM v0.35.4 for URL downloads... (#236)" (#242) @robertmaynard
• put $PREFIX before $BUILD_PREFIX in conda build (#182) @kkraus14

📖 Documentation

• Correct broken patch_toolkit API docs, and CMake API cross references (#271) @robertmaynard
• Provide suggestions when encountering an incomplete GTest package (#247) @robertmaynard
• Docs: RAPIDS.cmake should be placed in current bin dir (#241) @robertmaynard
• Remove incorrect install location note on rapids_export (#232) @robertmaynard

🚀 New Features

• Update to CPM 0.35.6 as it has needed changes for cpm patching support. (#273) @robertmaynard
• Update to nvcomp 2.4 which now offers aarch64 binaries! (#272) @robertmaynard
• Support the concept of a patches to apply to a project built via CPM (#264) @robertmaynard
• Branch 22.10 merge 22.08 (#262) @robertmaynard
• Introduce rapids_cuda_patch_toolkit (#260) @robertmaynard
• Update libcudacxx to 1.8 (#253) @robertmaynard
• Update to CPM version 0.35.5 (#249) @robertmaynard
• Update to CPM v0.35.4 for URL downloads match the download time (#236) @robertmaynard
• rapids-cmake dependency tracking now understands COMPONENTS (#234) @robertmaynard
• Update to thrust 1.17 (#231) @robertmaynard
• Update to CPM v0.35.3 to support symlink build directories (#230) @robertmaynard
• Update rapids-cmake to require cmake 3.23.1 (#227) @robertmaynard
• Improve GPU detection by doing less subsequent executions (#222) @robertmaynard

🛠️ Improvements

• Fix typo in rapids-cmake-url (#267) @trxcllnt
• Ensure <pkg>_FOUND is set in the generated Find<pkg>.cmake file (#266) @trxcllnt
• Set CUDA_USE_STATIC_CUDA_RUNTIME to control legacy FindCUDA.cmakebehavior (#259) @trxcllnt
• Use the GitHub .zip URI instead of GIT_REPOSITORY and GIT_BRANCH (#257) @trxcllnt
• Update nvcomp to 2.3.3 (#221) @vyasr

[NIGHTLY] v22.08.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• json exclude flag behaves as expected libcudacx//thrust/nvcomp (#223) @robertmaynard
• Remove nvcomp dependency on CUDA::cudart_static (#218) @robertmaynard
• Timestamps for URL downloads match the download time (#215) @robertmaynard
• Revert "Update nvcomp to 2.3.2 (#209)" (#210) @vyasr
• rapids-cmake won't ever use an existing variable starting with RAPIDS_ (#203) @robertmaynard

📖 Documentation

• Docs now provide rapids_find_package examples (#220) @robertmaynard
• Minor typo fix in api.rst (#207) @vyasr
• rapids_cpm_<pkgs> document handling of unparsed args (#206) @robertmaynard
• Docs/remove doc warnings (#205) @robertmaynard
• Fix docs: default behavior is to use a shallow git clone. (#204) @bdice
• Add rapids_cython to the html docs (#197) @robertmaynard

🚀 New Features

• More robust solution of CMake policy 135 (#224) @robertmaynard
• Update cuco git tag (#213) @PointKernel
• Revert "Revert "Update nvcomp to 2.3.2 (#209)" (#210)" (#211) @vyasr
• Update nvcomp to 2.3.2 (#209) @robertmaynard
• rapids_cpm_rmm no longer install when no INSTALL_EXPORT_SET listed (#202) @robertmaynard
• Adds support for pulling cuCollections using rapids-cmake (#201) @vyasr
• Add support for a prefix in Cython module targets (#198) @vyasr

🛠️ Improvements

• rapids_find_package() called with explicit version and REQUIRED should fail (#214) @trxcllnt

[NIGHTLY] v22.06.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• nvcomp install rules need to match the pre-built layout (#194) @robertmaynard
• Use target name variable. (#187) @bdice
• Remove uneeded message from rapids_export_package (#183) @robertmaynard
• rapids_cpm_thrust: Correctly find version 1.15.0 (#181) @robertmaynard
• rapids_cpm_thrust: Correctly find version 1.15.0 (#180) @robertmaynard

📖 Documentation

• Correct spelling mistake in cpm package docs (#188) @robertmaynard

🚀 New Features

• Add rapids_cpm_nvcomp with prebuilt binary support (#190) @robertmaynard
• Default Cython module RUNPATH to $ORIGIN and return the list of created targets (#189) @vyasr
• Add rapids-cython component for scikit-build based Python package builds (#184) @vyasr
• Add more exhaustive set of tests are version values of 0 (#178) @robertmaynard
• rapids_cpm_package_override now hooks into FetchContent (#164) @robertmaynard

🛠️ Improvements

• Update nvbench tag (#193) @PointKernel

[NIGHTLY] v22.04.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• rapids_export now handles explicit version values of 0 correctly (#174) @robertmaynard
• rapids_export now internally uses better named variables (#172) @robertmaynard
• rapids_cpm_gtest will properly find GTest 1.10 packages (#168) @robertmaynard
• CMAKE_CUDA_ARCHITECTURES ALL will not insert 62 or 72 (#161) @robertmaynard
• Tracked package versions are now not required, but preferred. (#160) @robertmaynard
• cpm_thrust would fail when provided only an install export set (#155) @robertmaynard
• rapids_export generated config.cmake no longer leaks variables (#149) @robertmaynard

📖 Documentation

• Docs use intersphinx correctly to link to CMake command docs (#159) @robertmaynard
• Example explains when you should use rapids_find_generate_module (#153) @robertmaynard
• Add CMake intersphinx support (#147) @bdice

🚀 New Features

• Bump CPM 0.35 for per package CPM_DOWNLOAD controls (#158) @robertmaynard
• Track package versions to the generated find_dependency calls (#156) @robertmaynard
• Update to latest nvbench (#150) @robertmaynard

🛠️ Improvements

• Temporarily disable new ops-bot functionality (#170) @ajschmidt8
• Use exact gtest version (#165) @trxcllnt
• Add .github/ops-bot.yaml config file (#163) @ajschmidt8

[NIGHTLY] v22.02.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• Ensure that nvbench doesn't require nvml when CUDA::nvml doesn't exist (#146) @robertmaynard
• rapids_cpm_libcudacxx handle CPM already finding libcudacxx before being called (#130) @robertmaynard

📖 Documentation

• Fix typos (#142) @ajschmidt8
• Fix type-o in docs <PackageName>_BINARY_DIR instead of <PackageName>_BINAR_DIR (#140) @dagardner-nv
• Set the always_download value in versions.json to the common case (#135) @robertmaynard
• Update Changelog to capture all 21.08 and 21.10 changes (#134) @robertmaynard
• Correct minor formatting issues (#132) @robertmaynard
• Document how to control the git rep/tag that RAPIDS.cmake uses (#131) @robertmaynard

🚀 New Features

• rapids-cmake now supports an empty package entry in the override file (#145) @robertmaynard
• Update NVBench for 22.02 to be the latest version (#144) @robertmaynard
• Update rapids-cmake packages to libcudacxx 1.7 (#143) @robertmaynard
• Update rapids-cmake packages to Thrust 1.15 (#138) @robertmaynard
• add exclude_from_all flag to version.json (#137) @robertmaynard
• Add PREFIX option to write_version_file / write_git_revision_file (#118) @robertmaynard

🛠️ Improvements

• Remove rapids_cmake_install_lib_dir unstable side effect checks (#136) @robertmaynard