yongwww commented on a change in pull request #6108:
URL: https://github.com/apache/incubator-tvm/pull/6108#discussion_r460383539
##########
File path: topi/python/topi/cuda/nms.py
##########
@@ -93,44 +67,41 @@ def get_valid_counts_ir(data, valid_count, out, out_indices,
valid_count = ib.buffer_ptr(valid_count)
out = ib.buffer_ptr(out)
out_indices = ib.buffer_ptr(out_indices)
- atomic_add_return = ib.allocate(
- valid_count.dtype, (1,), name='atomic_add_return', scope='local')
one_count = tvm.tir.const(1, dtype=valid_count.dtype)
one = tvm.tir.const(1, dtype=out.dtype)
score_threshold = tvm.ir.make_node(
"FloatImm", dtype="float32", value=score_threshold)
id_index = tvm.ir.make_node("IntImm", dtype="int32", value=id_index)
score_index = tvm.ir.make_node("IntImm", dtype="int32", value=score_index)
- max_threads = int(tvm.target.Target.current(
- allow_none=False).max_num_threads)
- nthread_tx = max_threads
- nthread_bx = batch_size * num_anchors // max_threads + 1
Review comment:
Would like to know more about the reason behind the change, perhaps
share some benchmark numbers? In some scenarios, like TF MaskRCNN, a large
number of boxes (`num_anchors` > 20000) are in one batch, multiple threads here
might provide performance improvement.
----------------------------------------------------------------
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]
