Improve 3 tests
Most of the tests are too basic (a = [0, 1, 2, 3], a = [1, 1, 1, 1]) which causes some wrong solutions to pass.
For example, it's intended that `a` gets read from, but if one peeks at the expected result, they can avoid reading altogether by writing `out[_thread_index_]` = _thread_index_.
To try and improve the first 3 tests, I used a numpy random integer generator, populating `a` with values 0-9.
If this is a good proposal and is accepted, I would be happy to update the other tests.
Thanks for making a great intro to GPU programming :)
合并状态:未合并 0 条评论