Issues · NVIDIA/cccl · GitHub

Labels Milestones New issue

[BUG]: scan_by_key benchmark was not adapted to policy_selector correctly

#8537

· griwes opened

on Apr 20, 2026

[BUG]: `cooperative_groups::thread_block::sync()` produces incorrect results on SM_120 with CUDA 13.2

#8534

· PointKernel opened

on Apr 20, 2026

[BUG] CUB DeviceScanKernel out-of-bounds global write in torch.isin with BFloat16 on sm120

#8528

· bernhardmgruber opened

on Apr 20, 2026

[FEA]: Provide mappping that selects first `N` units from a group

#8520

· davebayer opened

on Apr 18, 2026

[FEA]: Implement `cudax::group_synchronizer_instance` concept

#8519

· davebayer opened

on Apr 18, 2026

[FEA]: Expose `cudax::group_mapping_result` concept

#8518

· davebayer opened

on Apr 18, 2026

[FEA]: Implement `cudax::lane_synchronizer` synchronizer

#8517

· davebayer opened

on Apr 18, 2026

[FEA]: Implement `cudax::barrier_synchronizer` synchronizer

#8516

· davebayer opened

on Apr 18, 2026

[FEA]: Implement `cudax::group_as` mapping

#8515

· davebayer opened

on Apr 18, 2026

[FEA]: Make `cudax::group` constructible from other groups

#8514

· davebayer opened

on Apr 18, 2026

[CUB] Expose tuning parameters for `WarpReduceBatched`

#8512

· pauleonix opened

on Apr 18, 2026

[CUB] Add `WarpReduceBatched` benchmarks

#8511

· pauleonix opened

on Apr 18, 2026