comaniac commented on a change in pull request #4465: [AutoTVM] Tune softmax
CUDA schedule
URL: https://github.com/apache/incubator-tvm/pull/4465#discussion_r356217203
##########
File path: topi/python/topi/cuda/softmax.py
##########
@@ -52,13 +60,22 @@ def schedule_softmax(outs):
raise ValueError('Tag is expected to be softmax_output or
log_softmax_output. \
Got {0}'.format(op_tag))
+ # create tuning space
+ max_num_threads =
tvm.target.current_target(allow_none=False).max_num_threads
+ possible_num_thread = get_powers_of_two_in_range(32, max_num_threads)
+ cfg.define_knob("num_thread", possible_num_thread)
Review comment:
Right. I am also aware of the tuning space and that's why I was interesting
in more investigations. However, I don't think we should merge this PR after
the investigation because that would take too long. I'd say as long as the
performance achieved by this template is sufficient for your use case at this
momnet then it's good enough to go.
How about we keep the current solution (same thread number for each stage)
and leave a comment saying that this could be further investigated and improved
in the future?
----------------------------------------------------------------
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
