The GitHub Actions job "Nightly Docker Update" on tvm.git/main has failed.
Run started by GitHub user areusch (triggered by areusch).

Head commit for run:
cfadc1aff28631f8b1c02935280032e9979c90ac / senhtry <[email protected]>
[CUDA][FFI] Extend kernel launch config to support Programmatic Dependent 
Launch and cuLaunchCooperativeKernel (#18604)

This patch adds support for **Programmatic Dependent Kernel Launch
(PDL)** in the TVM CUDA FFI layer. PDL enables launching dependent
kernels on the GPU without host intervention, improving performance and
expressiveness for dynamic CUDA workloads.

Refer to NVIDIA documentation for PDL semantics:  

https://docs.nvidia.com/cuda/cuda-programming-guide/03-advanced/advanced-host-programming.html#programmatic-dependent-kernel-launch

In addition, this patch extends the CUDA FFI layer to support
cooperative kernel launches via cuLaunchCooperativeKernel. Cooperative
kernels allow grid-wide synchronization and are required for certain
multi-stage or producer–consumer GPU workloads. When a kernel is marked
for cooperative launch, it will be dispatched using
cuLaunchCooperativeKernel instead of the standard cuLaunchKernel.

Refer to NVIDIA documentation for cooperative kernel usage:

https://docs.nvidia.com/cuda/cuda-programming-guide/04-special-topics/cooperative-groups.html#when-to-use-cudalaunchcooperativekernel

Report URL: https://github.com/apache/tvm/actions/runs/20546073186

With regards,
GitHub Actions via GitBox


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to