triton.experimental.gluon.language.nvidia.ampere.mbarrier.allocate_mbarrier

triton.experimental.gluon.language.nvidia.ampere.mbarrier.allocate_mbarrier(batch: constexpr = None, two_ctas: constexpr = False)

Helper function to allocate an mbarrier

Parameters:

two_ctas (bool) – Whether the barrier should synchronize every other CTA