triton.experimental.gluon.language.nvidia.blackwell.allocate_tensor_memory

triton.experimental.gluon.language.nvidia.blackwell.allocate_tensor_memory(element_ty, shape, layout, value=None, _semantic=None)

Allocate tensor memory.

Parameters:
  • element_ty (dtype) – The element data type.

  • shape (Sequence[int]) – The descriptor shape.

  • layout (TensorMemoryLayout) – The layout of the tensor memory.

  • value (tensor, optional) – Initial tensor to copy. Defaults to None.

Returns:

Descriptor for the allocated memory.

Return type:

tensor_memory_descriptor