Commit e67e9d3

committed

Consolidate the TMA examples around the libcudacxx wrappers.

Keep the example surface smaller and closer to CUDA C++ by showing barrier/TMA helpers and replace_address() in one place instead of duplicating raw PTX snippets. Made-with: Cursor

1 parent 358d975 commit e67e9d3Copy full SHA for e67e9d3

2 files changed

cuda_core/examples
- tma_replace_address.py
- tma_tensor_map.py

`‎cuda_core/examples/tma_replace_address.py‎`

This file was deleted.

Comments

(0)