Script.tcgen05.commit

Script.tcgen05.commit

Script.tcgen05.commit(mbarrier, cta_group=1, multicast_mask=None)[source]

Commit pending tcgen05 async operations and signal an mbarrier.

Groups all prior uncommitted tcgen05 async operations (e.g., copy, mma) and signals the specified mbarrier upon completion. The mbarrier’s tx-count will be decreased when the operations finish.

Parameters:

  • mbarrier (Expr | RegisterTensor) – The memory barrier to signal upon completion.

  • cta_group (int) – The CTA group size. Must be 1 or 2.

  • multicast_mask (Optional[int]) – If provided, signals mbarriers on multiple CTAs in the cluster specified by the bitmask.

Return type:

None

Notes

  • Thread group: Must be executed by a single warp (use self.single_warp()).

  • Hardware: Requires compute capability 10.0+ (sm_100).

  • PTX: tcgen05.commit

Contents