[GitHub] [tvm] ashutosh-arm commented on a diff in pull request #11359: [CMSIS-NN] Aligned buffer sizes for Conv2D post CMSIS-NN SHA update

[GitBox <gi...@apache.org>]

ashutosh-arm commented on code in PR #11359:
URL: https://github.com/apache/tvm/pull/11359#discussion_r879363410


##########
tests/python/relay/aot/test_crt_aot.py:
##########
@@ -994,7 +994,7 @@ def test_workspace_calculation_cmsis_nn():
     ):
         lib = tvm.relay.build(mod, target, executor=executor, runtime=runtime, params=params)
     mlf_memory_map = mlf._build_function_memory_map(lib.function_metadata)
-    assert mlf_memory_map["main"][0]["workspace_size_bytes"] == 9904
+    assert mlf_memory_map["main"][0]["workspace_size_bytes"] == 14384

Review Comment:
   It was raise by @Mousius as well offline. There is a type of Conv2D layer in this network for which CMSIS-NN didn't allocate a buffer, but now it does. It looks like some sort of additional optimization that has come in as part of new CMSIS-NN.



-- 
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: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org