Issues

• [BUG]: Test "DeviceSetOps pairs can be tuned" must initialize elements of d_block_size vector to zero

  Bug

  NVIDIA/cccl#12097 · by Filter by author oleksandr-pavlyk was closed on Oct 21, 2026
  1

  

• [BUG]: NVCC 12.9 rejects valid 16-byte types for __nv_atomic_* generic atomics on Windows

  needs triage
  Bug

  NVIDIA/cccl#12067 · by Filter by author VincentL5201 was closed on Oct 20, 2026
  5

• [BUG]: libcu++ NVRTC header parsing cost and PCH footprint

  needs triage
  Bug

  NVIDIA/cccl#12057 · Filter by author lamarrr opened on Oct 17, 2026
  2

  

• [BUG]: Using ProxyArray with cuda.compute.make_merge_sort fails to compile

  Bug

  NVIDIA/cccl#12021 · Filter by author oleksandr-pavlyk opened on Oct 15, 2026
  2

• [BUG]: Incorrect test asserts in libcudacxx/test/libcudacxx/std - they constant evaluate on Clang

  needs triage
  Bug

  NVIDIA/cccl#11989 · Filter by author seanbaxter opened on Oct 12, 2026
  2

  

• [BUG]: cuda-cccl 1.2.1 on Windows + CTK 12.4: NVVM link error, __half(__nv_bfloat16) multiply defined

  cuda.compute
  Bug

  NVIDIA/cccl#11885 · by Filter by author leofang was closed on Oct 8, 2026
  2

  

• [BUG]: Include for device segmented scan is missing from <cub/cub.cuh>

  Bug

  NVIDIA/cccl#11859 · by Filter by author nmetzinger was closed on Oct 6, 2026
  1

  

• [BUG]: reduce_by_keys with __half

  Bug

  NVIDIA/cccl#11816 · by Filter by author MarcelKoch was closed on Oct 8, 2026
  2

  

• [BUG] libcccl.c.parallel.v2.so and libnvcc.so dynamically link libcudart despite CUDA::cudart_static intent

  cuda.compute
  Bug

  NVIDIA/cccl#11735 · by Filter by author leofang was closed on Oct 1, 2026
  2

• [BUG]: Device segmented reduce fails to compile with 64 bit signed integer type and -fdevice-time-trace

  Bug

  NVIDIA/cccl#11680 · Filter by author miscco opened on Sep 28, 2026

• [BUG]: CUB BLOCK_REDUCE_WARP_REDUCTIONS_NONDETERMINISTIC specialization doesn't support custom operators

  Bug

  NVIDIA/cccl#11647 · Filter by author fbusato opened on Sep 24, 2026

• [BUG]: cuda::to_host_mdspan rejects valid zero-sized DLPack tensors with nullptr data

  needs triage
  Bug

  NVIDIA/cccl#11605 · by Filter by author dhruv0703 was closed on Sep 30, 2026
  1

  

• [BUG]: cuda.compute: gpu_struct treats NumPy field titles as separate fields

  cuda.compute
  Bug

  NVIDIA/cccl#11576 · by Filter by author HelloWorldU was closed on Oct 2, 2026
  2

• [BUG]: [CUB][NVHPC] Mixed -gpu= fatbin: kernel bounds vs host launch CC mismatch

  Bug

  NVIDIA/cccl#11572 · by Filter by author zkhatami was closed on Sep 22, 2026
  5

  

• [BUG]: submdspan is broken

  Bug

  NVIDIA/cccl#11520 · by Filter by author gevtushenko was closed on Sep 28, 2026
  2

  

• [BUG]: device_ptr incompatibility with reverse iterator

  Bug

  NVIDIA/cccl#11489 · Filter by author KyleFromNVIDIA opened on Sep 18, 2026
  1

• cuda-cccl 1.2.0 on Windows: access violation in call_build on first make_reduce_into (regression from 1.1.1)

  cuda.compute
  Bug

  NVIDIA/cccl#11472 · by Filter by author leofang was closed on Sep 22, 2026
  

• [BUG]: cuda::std::binary_search is slow

  needs triage
  Bug

  NVIDIA/cccl#11412 · Filter by author Pennycook opened on Sep 15, 2026
  29

  

• [BUG]: cuda.compute rebuilds a stateful op's kernel when a captured array's shape changes

  cuda.compute
  Bug

  NVIDIA/cccl#11407 · by Filter by author maxymnaumchyk was closed on Oct 1, 2026
  1

  

• [BUG]: cub.test.warp.reduce target fails to compile with NVHPC

  Bug

  NVIDIA/cccl#11365 · by Filter by author oleksandr-pavlyk was closed on Sep 11, 2026
  

• [BUG]: Unqualified isnan and fabs in thrust::detail::complex::cacosh

  backport branch/3.5.xbug
  Bug

  NVIDIA/cccl#11353 · by Filter by author seanbaxter was closed on Sep 11, 2026
  1

  

• [BUG]: cuda.compute: unary_transform with OpKind.IDENTITY fails to build for struct dtypes (operator emitted with an empty symbol name)

  Bug

  NVIDIA/cccl#11350 · by Filter by author NaderAlAwar was closed on Sep 17, 2026
  2

  

• [BUG]: cuda.compute: reduce_into with an iterator input fails with "error code: 999" when sizeof(h_init) != sizeof(iterator item)

  cuda.compute
  Bug

  NVIDIA/cccl#11349 · Filter by author NaderAlAwar opened on Sep 10, 2026
  5

  

• [BUG]: cuda.compute: gpu_struct / types.from_numpy_dtype discard a structured dtype's offsets and itemsize and rebuild it with align=True

  cuda.compute
  Bug

  NVIDIA/cccl#11348 · Filter by author NaderAlAwar opened on Sep 10, 2026

• [BUG]: cuda.compute: struct and ZipIterator layouts built from complex dtypes use numpy host alignment instead of cuda::std::complex device alignment

  cuda.compute
  Bug

  NVIDIA/cccl#11347 · Filter by author NaderAlAwar opened on Sep 10, 2026
  1