csullivan commented on a change in pull request #8756:
URL: https://github.com/apache/tvm/pull/8756#discussion_r692507919
##########
File path: tests/python/unittest/test_target_codegen_opencl.py
##########
@@ -119,7 +119,27 @@ def check_max(dev, n, dtype):
check_max(dev, 1, "float32")
check_max(dev, 1, "float64")
[email protected]_gpu
[email protected]_opencl
Review comment:
Thanks for adding this test! This is a bit of a meta-comment, but we are
in the process of changing our testing infra to run fewer integration tests
(e.g. running on device) to more unit tests during pre-commit. In this case,
you could consider an additional unit test that removes the call to to run the
kernel and replace it with a check for `erf` in the opencl source,
```
source_str = fun.imported_modules[0].get_source()
```
to ensure it is not `erff` as it is for cuda.
The above test would be a unit test that can run in pre-commit with the full
integration test (build + accuracy comparison) could be run as a nightly.
--
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]
