NVIDIA/cccl python-0.3.0

CCCL Python Libraries (v0.3.0)

on GitHub

These are the release notes for the cuda-cccl Python package version 0.3.0, dated October 2nd, 2025. The previous release was v0.2.1.

cuda.cccl is in "experimental" status, meaning that its API and feature set can change quite rapidly.

Installation

Please refer to the install instructions here

Features and improvements

• ARM64 wheel and conda package support: Installation via pip and conda now supported on ARM64 (aarch64) architecture.

• New algorithm: three-way partitioning: The three_way_partition algorithm enables partitioning an array (or iterator) into three partitions, given two selection operators.

• Improved scan performance: The inclusive_scan and exclusive_scan APIs provide improved performance by automatically selecting the optimal tuning for the input data types and device architecture.

Bug Fixes

None.

Breaking Changes

None.