NVIDIA Blackwell

async_copy
clc	Cluster Launch Control (CLC) for Blackwell (SM100+) dynamic persistent kernels.
mbarrier
tma

allocate_tensor_memory	Allocate tensor memory.
fence_async_shared	Issue a fence to complete asynchronous shared memory operations.
mma_v2
tensor_memory_descriptor	Represents a tensor memory descriptor handle for Tensor Core Gen5 operations.
tensor_memory_descriptor_type
TensorMemoryLayout	Describes the layout for tensor memory in Blackwell architecture.
TensorMemoryScalesLayout	Describes the layout for tensor memory scales in Blackwell architecture.