autoware_cuda_utils

Purpose

This package contains a library of common functions related to CUDA.

Changelog for package autoware_cuda_utils

0.53.0 (2026-09-29)

• Merge remote-tracking branch 'origin/main' into prepare-0.53.0-changelog

• fix(autoware_cuda_utils): export CCCL for host compilation with CUDA 13 (#13302)

  * fix(autoware_cuda_utils): export CCCL for host compilation with CUDA 13 CUDA 13 moved the Thrust, CUB and libcu++ headers to include/cccl/. nvcc adds that directory by itself. The host compiler does not, so g++ fails on any host file that reaches a CCCL header. The installed thrust_utils.hpp uses Thrust, so the include path belongs in the exported interface of autoware_cuda_utils. Export an INTERFACE target that carries the package headers and CCCL::CCCL, and add cmake/find_cccl.cmake to locate the toolkit CCCL config outside the default CMake search path. Consumers that already depend on autoware_cuda_utils need no change. Link the target in autoware_ptv3, whose CUDA library is not built through ament_target_dependencies.

  * fix(autoware_cuda_utils): mark the CCCL include as system, export CUDA::cudart The CCCL config creates non-imported targets, so the cccl include directory reached consumers as -I in 79 flag files across 25 packages, and a future CCCL header warning would break builds under -Werror. A small imported marker target, autoware_cuda_utils::cccl_system, now carries the same directory as a system include. CMake unions system include sets across the link closure, and the explicit property survives the CMAKE_NO_SYSTEM_FROM_IMPORTED that mrt_cmake_modules leaks into consumer scopes. No NVIDIA target is modified. The exported interface also carries CUDA::cudart: the installed thrust_utils.hpp includes cuda_runtime_api.h, which the package and CCCL include directories alone cannot resolve. Also state above the autoware_ptv3 link why it is load-bearing: nvcc supplies CCCL to the .cu files, and detection3d_postprocess_test host-compiles Thrust without ament_target_dependencies.

• Contributors: Mete Fatih Cırıt, Ryohsuke Mitsudome

0.52.0 (2026-06-30)

• Merge remote-tracking branch 'origin/main' into tmp/bot/bump_version_base

• feat: polar voxel noise filter (#12496)

  • feat: polar voxel noise filter
  • chore: added missing destructor and mutex
  • chore: adressing comments about depercated functions
  • chore: adressing comments about unused items and readme
  • chore: adressing comments about pointcloud format and prefix
  • style(pre-commit): autofix
  • chore: adressing new parameter in bounds checking and suffix renamed variables
  • style(pre-commit): autofix
  • chore: adding unit test for the polar voxel filter
  • style(pre-commit): autofix
  • chore: added test on cmakelist all points primary return default
  • style(pre-commit): autofix
  • chore: handling case when pointcloud without return type information

  * chore: pointcloud msg format validation only once to avoid redundancy ---------Co-authored-by: Yoshi Ri <<yoshiyoshidetteiu@gmail.com>> Co-authored-by: pre-commit-ci-lite[bot] <117423508+pre-commit-ci-lite[bot]@users.noreply.github.com> Co-authored-by: Amadeusz Szymko <<amadeusz.szymko.2@tier4.jp>>

• Contributors: SergioReyesSan, github-actions

0.51.0 (2026-05-01)

0.50.0 (2026-02-14)

0.49.0 (2025-12-30)

0.48.0 (2025-11-18)

• Merge remote-tracking branch 'origin/main' into humble

• feat(autoware_cuda_pointcloud_preprocessor): cuda/polar voxel filter (#11122)

  • feat(cuda_utils): support device memory allocation from memory pool

  * feat(cuda_pointcloud_preprocessor): support polar_voxel_outlier_filter WIP: use cuda::std::optional. compile passed wip: version 1 wip: update wip: update launcher

  • feat(cuda_pointcloud_preprocessor): add a flag to enable/disable ring outlier filter in cuda_pointcloud_preprocessor
  • chore: clean up the code
  • docs: update documents briefly

File truncated at 100 lines see the full file