Class CUtensorMap

examples/matmul/this-sm90/test_v9.py:7–8

import subprocess, ctypes, torch, os, sys

class CUtensorMap(ctypes.Structure):
    _fields_ = [("data", ctypes.c_ubyte * 128)]

BM, BN, BK = 128, 256, 64
libcuda = ctypes.CDLL("libcuda.so.1")