CCCL 2.2.0

(Note that these release notes are not yet finalized. They do not reflect any PRs that were merged to Thrust/CUB/libcudacxx before migrating to the nvidia/cccl repo).

What's Changed

• Add axis for docker builds by @raydouglass in #1
• Docker: Add support for ICPC and NVC++, install newer CMake, and add curl by @brycelelbach in #4
• Update excludes by @raydouglass in #5
• Docker: OS and CUDA upgrades, support for additional configurations by @brycelelbach in #9
• Docker: Add Thrust/CUB documentation toolchain to Ubuntu docker images by @brycelelbach in #15
• Re-enable CentOS images. by @allisonvacanti in #16
• Add sccache to dockerfile by @msadang in #17
• Update base containers. by @allisonvacanti in #18
• Update sccache version by @ajschmidt8 in #19
• Build 11.5.1 containers by @ajschmidt8 in #20
• Add ops-bot.yaml by @jrhemstad in #80
• Monorepo workflow by @jrhemstad in #99
• Add devcontainers by @jrhemstad in #105
• Update the libcu++ submodule by @miscco in #109
• Update libcudaxx again by @miscco in #110
• Remove submodules from CI workflow by @jrhemstad in #115
• Fix CUB CI by @senior-zero in #114
• Fix async scan / counting iterator tests by @senior-zero in #118
• Make sccache work locally by @jrhemstad in #113
• Fix compilation of thrust and cub by @miscco in #120
• Fix segfault in cub::CachingDeviceAllocator by @senior-zero in #119
• Initial GH Infra Setup by @jarmak-nv in #23
• Visualize variant space coverage by @senior-zero in #125
• Fix broken issue templates by @jarmak-nv in #124
• Tune scan by key for SM90 by @senior-zero in #121
• Update PR template to more explicitly prompt for a linked issue closed by the PR by @jrhemstad in #134
• Change component section to more general "area" by @jrhemstad in #132
• Try and fix CI for old CTK by @miscco in #116
• Fix tuple_cat for std:: qualified types by @miscco in #144
• Add ccache to lit invocation by @miscco in #147
• Benchmark batched memcpy by @senior-zero in #136
• Properly querry CMAKE_CUDA_COMPILER_LAUNCHER for ccache support by @miscco in #152
• Implement Three-Way Partition Tuning / Benchmark by @senior-zero in #155
• Port three-way partition to use Catch2 by @senior-zero in #156
• Add gcc-6 to the test matrix by @miscco in #160
• Tune reduce / unique by key for SM90 by @senior-zero in #163
• Remove unused folders by @miscco in #145
• Fix documentation of atomic_ref by @miscco in #164
• New iterator traits by @miscco in #158
• Improve implementation of destructible by @miscco in #157
• Build script improvements by @jrhemstad in #149
• Fix icpc / denormals by @senior-zero in #185
• Enable tests by @jrhemstad in #167
• Monorepo by @jrhemstad in #194
• Multi-benchmark tuning by @senior-zero in #208
• Fixes universal_vector test failure on CTK 11.1 & gcc-6 by @elstehle in #209
• Delete several directories for older CI infra. by @wmaxey in #218
• Memory-safe radix sort test by @senior-zero in #222
• [FEA] Implement iter_move CPO by @miscco in #197
• Build cub benchmarks in build_cub.sh by @jrhemstad in #216
• [skip-tests] Do not run tests when skip-tests is part of the latest commit message by @miscco in #224
• Factor out build job logic into a "run-as-coder" reusable workflow. by @jrhemstad in #205
• Fix instances of 'scan' copy-pasted into reduction documentation by @milesvant in #221
• Add clangd to devcontainer by @senior-zero in #225
• Add initial CODEOWNERS file by @jrhemstad in #226
• Attempt to fix codeowners by @jrhemstad in #231
• Make libcudacxx respect CMake options for CUDA archs. by @wmaxey in #235
• Optimize Three-Way Partition by @senior-zero in #228
• [BUG] Rework how we handle feature test macros by @miscco in #195
• Enable use of cudaMemcpyAsync for thrust::copy by @miscco in #211
• Enable additional arguments in build_common.sh by @wmaxey in #236
• [BUG] Properly uglify all qualifiers in product headers by @miscco in #201
• Port cub::Device{Select, Partition} tests to catch2 by @miscco in #229
• Fix CUB tests / MSVC 2022 by @senior-zero in #255
• Ensure that any CMake re-rooting doesn't break our find_file by @miscco in #257
• [BUG] Fix compilation issues with MSVC 2017 by @miscco in #196
• Implement iterator concepts by @miscco in #223
• Tune Histogram on H100 by @senior-zero in #266
• Add WarpExchangeAlgorithm customization for WarpExchange class by @pb-dseifert in #256
• [BUG]: Avoid deprecation warning for std::aligned_storage when building with c++23 by @miscco in #258
• Port cub::DeviceReduce tests to catch2 by @elstehle in #267
• Add support for nvcc-specific matrix. by @jrhemstad in #243
• Fix anchor link to cooperative groups in CUDA programming guide by @wence- in #274
• Fix BibTeX syntax in CITATION.md [skip-tests] by @wence- in #276
• Enforce C++17 for benches by @senior-zero in #275
• Project Automation: Move PR and Linked Issues to In Progress by @jarmak-nv in #170
• Update to 23.08 devcontainers and CUDA 12.2 by @jrhemstad in #270
• [skip-tests] CTK 12.2 tuning image by @senior-zero in #282
• Fix single-thread block reduction by @senior-zero in #287
• Tune Select and Partition on A100 by @senior-zero in #289
• Fix CUB tests / MSVC by @senior-zero in #292
• Allow building CUB tests without cuRand by @senior-zero in #250
• Fixup to CUB build - s/curand/cudart/ by @wmaxey in #301
• Fix OOB in cub::DeviceRunLengthEncode::NonTrivialRuns by @senior-zero in #294
• Tune RLE on A100 by @senior-zero in #295
• Tune scan on A100 by @senior-zero in #302
• Add new CCCL:: CMake targets by @allisonvacanti in #244
• Fix cudacc and nvcc mixup. by @wmaxey in #329
• [skip-tests] Use builtin for destructible concept on MSVC by @miscco in #333
• Fix merge conflict from two inflight PRs by @miscco in #338

New Contributors

• @raydouglass made their first contribution in #1
• @brycelelbach made their first contribution in #4
• @msadang made their first contribution in #17
• @wmaxey made their first contribution in #218
• @milesvant made their first contribution in #221
• @pb-dseifert made their first contribution in #256
• @wence- made their first contribution in #274

Full Changelog: https://github.com/NVIDIA/cccl/commits/v2.2.0