triton.experimental.gluon.language.NVMMADistributedLayout

class triton.experimental.gluon.language.NVMMADistributedLayout(self, version: ~typing.List[int], warps_per_cta: ~typing.List[int], instr_shape: ~typing.List[int], cga_layout: ~typing.List[~typing.List[int]] = <factory>)

Represents a layout for NVIDIA MMA (tensor core) operations.

Parameters:

• version (List[int]) – Version identifier for the MMA instruction.

• warps_per_cta (List[int]) – Number of warps per CTA.

• instr_shape (List[int]) – Instruction shape for MMA.

• cga_layout (Optional[List[List[int]]]) – Bases describing CTA tiling.

__init__(self, version: ~typing.List[int], warps_per_cta: ~typing.List[int], instr_shape: ~typing.List[int], cga_layout: ~typing.List[~typing.List[int]] = <factory>) → None

Methods

__init__(self, version, warps_per_cta, ...)
format_hardware_view(self, shape)
format_tensor_view(self, shape)
mangle(self)

Attributes

rank
type
version
warps_per_cta
instr_shape
cga_layout