giuseros commented on a change in pull request #5980: URL: https://github.com/apache/incubator-tvm/pull/5980#discussion_r454477447
########## File path: topi/python/topi/arm_cpu/injective.py ########## @@ -62,9 +62,13 @@ def schedule_injective(outs): outs = [outs] if isinstance(outs, te.tensor.Tensor) else outs s = te.create_schedule([x.op for x in outs]) x = outs[0] + ins = x.op.input_tensors + dtype = ins[0].dtype if len(ins) > 0 else x.dtype + max_vlen = 4 if dtype == 'int32' else 8 Review comment: I will set at 4, as a good enough value. The point is that in a network this doesn't matter, as it will get fused to the convolution op (which will decide the level of vectorization) ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org