triton.experimental.gluon.language.nvidia.rubin.TensorMemoryScalesLayout

class triton.experimental.gluon.language.nvidia.rubin.TensorMemoryScalesLayout(self, cga_layout: List[List[int]] = <factory>, block_rep_order: str = 'mnThenK')

Describes the layout for tensor memory scales in Rubin architecture.

Parameters:

• cga_layout (Optional[List[List[int]]]) – CGA layout bases. Defaults to [].

• block_rep_order (str) – Order of repeated scale blocks. Must be either "mnThenK" or "kThenMn". Defaults to "mnThenK".

__init__(self, cga_layout: List[List[int]] = <factory>, block_rep_order: str = 'mnThenK') → None

Methods

__init__(self, cga_layout, block_rep_order)
mangle(self)

Attributes

block_rep_order
cga_layout