Search results

[FEA]: Collect gaps in support matrix between CuPy and cuda.compute

Status: Open.

#9083 In NVIDIA/cccl;

· gevtushenko opened on May 20, 2026

[FEA]: Expose CUB init-less reduction algorithms

cubFor all items related to CUBFor all items related to CUB

Status: Open.

#9080 In NVIDIA/cccl;

· Jacobfaib opened on May 20, 2026

[RFC] Is the purpose of *_api.cu tests in CUB to only test API examples?

Status: Open.

#9078 In NVIDIA/cccl;

· bernhardmgruber opened on May 20, 2026

[FEA]: PoC segmented TopK cluster optimization

Status: Open.

#9077 In NVIDIA/cccl;

· gevtushenko opened on May 20, 2026

[FEA]: Optimize segmented TopK with clusters for sm100f

Status: Open.

#9075 In NVIDIA/cccl;

· gevtushenko opened on May 20, 2026

thrust::tabulate is much slower than equivalent transform/for_each for wide row functor

Status: Open.

#9070 In NVIDIA/cccl;

· bdice opened on May 19, 2026

cuda::experimental::coop::reduce with partial block support

Status: Open.

#9062 In NVIDIA/cccl;

· jrhemstad opened on May 18, 2026

[Tile] Implement RMS Norm

Status: Open.

Task

#9060 In NVIDIA/cccl;

· fbusato opened on May 18, 2026

Use BlockLoadToShared in transpose versions of BlockLoad

Status: Open.

#9057 In NVIDIA/cccl;

· bernhardmgruber opened on May 18, 2026

Expose unrolling control in cub::DeviceMergeSort to tuning

Status: Open.

#9052 In NVIDIA/cccl;

· bernhardmgruber opened on May 18, 2026

The cub.bench.segmented_radix_sort.keys.base benchmark covers too many combinations

Status: Open.

#9041 In NVIDIA/cccl;

· bernhardmgruber opened on May 15, 2026

Functional/Non-Functional C++ Tile Evaluation

Status: Open.

Task

#9038 In NVIDIA/cccl;

· fbusato opened on May 15, 2026