FranklandJack commented on PR #14519: URL: https://github.com/apache/tvm/pull/14519#issuecomment-1513236253
> > Yeah, although I'm not sure where a test for something like this would live. We would need to test that vectorization happens as expected, so some kind of lit test based on tir? > > I believe that a test in [test_conv2d.py](https://github.com/apache/tvm/blob/main/tests/python/relay/strategy/arm_cpu/test_conv2d.py) might be enough. Probably you can also add a test for tir or codegen to see that vectorization was applied but in this case in codegen test it will be necessary to call an existing schedule instead of writing a new one. Sorry I've been on annual leave and just got back to reviewing this. I took a look at [test_conv2d.py](https://github.com/apache/tvm/blob/main/tests/python/relay/strategy/arm_cpu/test_conv2d.py), although would a more appropriate place for this not be [test_depthwise_conv2d.py](https://github.com/apache/tvm/blob/main/tests/python/relay/strategy/arm_cpu/test_depthwise_conv2d.py). I think at present though we don't do any testing for Arm Cortex A targets (see [here](https://github.com/apache/tvm/blob/main/tests/python/relay/strategy/arm_cpu/test_generalized_conv2d.py#L149)), so perhaps this is outside the scope of this PR? -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
