Support Thread Block Clusters for CUDA 12 + Hopper #8780

Activity

1. kmaehashi

   added

   cat:featureNew features/APIsNew features/APIs

   on Dec 2, 2024
2. kmaehashi

   added

   prio:high

   on Dec 18, 2024
3. leofang commented on Dec 26, 2024

   

   leofang

   on Dec 26, 2024

   Member

   More actions

   FYI, as a stop-gap solution, users can use cuda.core to launch kernels that require the thread block cluster support:
   https://github.com/NVIDIA/cuda-python/blob/main/cuda_core/examples/thread_block_cluster.py
4. kmaehashi

   marked How to launch dynamic thread block clusters with CuPy? #8775 as a duplicate of this issue on Apr 16, 2025
5. leofang

   added

   prio:low

   and removed

   prio:high

   on Jun 8, 2026
6. leofang commented on Sep 12, 2026

   

   leofang

   on Sep 12, 2026

   Member

   More actions

   With cuda.core becoming stable (v1.0+), users should go and use official CUDA tooling. We shouldn't duplicate the work and expand RawKernel.
7. leofang

   closed this as not plannedon Sep 12, 2026