How to add custom node operation in tvm using cpp

Application
#1

I was able to use custom node using following python function. However, I was not able to figure out doing the same in cpp. Could anyone point out me implementation of same thing in cpp?

@tvm.register_func(“tvm.accel.accel_fused”) def accel_fused(*tensors): some operation return tvm.nd.array

#2

Not sure I understand the question. Are you referring to TVM_REGISTER_GLOBAL?

1 Like