yongfeng-nv commented on a change in pull request #4885: Split node min range
is not stringent.
URL: https://github.com/apache/incubator-tvm/pull/4885#discussion_r379973859
##########
File path: tests/python/unittest/test_schedule_tensor_core.py
##########
@@ -339,17 +339,13 @@ def test_tensor_core_batch_conv():
ty, yo = s[AS].split(xo, nparts=block_col_warps)
t = s[AS].fuse(nn, ii)
to, ti = s[AS].split(t, factor=warp_size)
- s[AS].bind(tx, thread_y)
Review comment:
I have similar questions, especially echo on your first one. I didn't know
how to schedule your example properly until I read your code. After my change,
I don't know how to do that. That's why I wonder whether this behavior of
split is by design or just a work around. How about bringing this discussion
to the forum to get insight from @tqchen and more people.
My intuition to your second question is that my PR is neutral. While it
disallows certain schedules, it also enables/improves others. When we think
about auto scheduling, it may be more helpful. Fundamentally, we need to
address the first question. AutoTVM will benefit from this PR, once we can
express more requests about thread binding and infer bounds more accurately at
the same time.
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
