triton.experimental.gluon.language.nvidia.hopper.warpgroup_mma_wait

triton.experimental.gluon.language.nvidia.hopper.warpgroup_mma_wait(num_outstanding=0, deps=None, _semantic=None)

Wait until num_outstanding or less warpgroup MMA operations are in-flight.

Parameters:

• num_outstanding (int) – Number of outstanding warpgroup MMA operations to wait for. Defaults to 0.

• deps (Sequence[tensor]) – List of dependencies that need to be kept alive while the mma is unfinished.