triton.experimental.gluon.language.nvidia.rubin.tcgen05_mma_barrier_count

triton.experimental.gluon.language.nvidia.rubin.tcgen05_mma_barrier_count()

Calculate the number of CTAs that will commit the tcgen05 MMA instruction.

Parameters:

• smems (Sequence[shared_memory_descriptor]) – Shared memory descriptors used in the tcgen05 instruction.

• multicast (bool) – Whether the tcgen05 instruction is multicast.

• two_ctas (bool) – Whether the tcgen05 instruction uses cta_group::2.

Returns:

The number of CTAs that will commit the tcgen05 MMA instruction.

Return type:

int