triton.experimental.gluon.language.nvidia.blackwell.mbarrier

Functions

allocate_mbarrier	Helper function to allocate an mbarrier
arrive	Arrive at an mbarrier with a specified count.
expect	Expect a specific number of bytes being copied.
fence_init_release_cluster	Fence that makes prior mbarrier initialization visible across the CTA cluster.
init	Initialize an mbarrier with a specified count.
invalidate	Invalidate an mbarrier, resetting its state.
wait	Wait until the mbarrier object completes its current phase.

Classes

MBarrierLayout

Layout for mbarrier synchronization in Ampere and later architectures.