MCPcopy Create free account
hub / github.com/apache/tvm / WgmmaNamespace

Class WgmmaNamespace

python/tvm/backend/cuda/script.py:254–263  ·  view source on GitHub ↗

The WGMMA instruction submodule.

Source from the content-addressed store, hash-verified

252
253
254class WgmmaNamespace:
255 """The WGMMA instruction submodule."""
256
257 def __init__(self):
258 self.fence: Callable[..., Any] = _op_wrapper(_cuda_op.ptx_wgmma_fence)
259 self.commit_group = _op_wrapper(_cuda_op.ptx_wgmma_commit_group)
260 self.wait_group = _op_wrapper(_cuda_op.ptx_wgmma_wait_group)
261 self.noop_barrier = _op_wrapper(_cuda_op.ptx_wgmma_noop_barrier)
262 self.mma_async = WgmmaMmaAsyncNamespace()
263 self.encode_matrix_descriptor = _op_wrapper(_cuda_op.ptx_wgmma_encode_matrix_descriptor)
264
265
266class WgmmaMmaAsyncNamespace:

Callers 1

__init__Method · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected

Used in the wild real call sites across dependent graphs

searching dependent graphs…