[GitHub] [tvm] jverma-quic commented on a diff in pull request #12606: [TOPI][Hexagon] Implement quantized elementwise for hexagon

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

jverma-quic commented on code in PR #12606:
URL: https://github.com/apache/tvm/pull/12606#discussion_r960961455


##########
python/tvm/topi/hexagon/qnn/qadd_qsub_qmul.py:
##########
@@ -0,0 +1,235 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+# pylint: disable=invalid-name
+
+"""Compute and schedule for quantized add, multiply, subtract op
+
+Please note the following assumptions made by the implementation:
+
+1) The inputs will be multiple of crouton layout except for the axis that needs broadcasting."""
+
+from tvm import te
+from tvm import tir
+from ..utils import get_layout_transform_fn, get_fixed_point_value
+
+
+def broadcast_axis(tensor_A, tensor_B):
+    """Find out the indices that will have broadcasting"""
+    A_broadcast = []
+    B_broadcast = []
+
+    for i in range(len(tensor_A.shape)):
+        if tensor_A.shape[i] == tensor_B.shape[i]:
+            A_broadcast.append(1)
+            B_broadcast.append(1)
+        elif tensor_A.shape[i] == 1:
+            A_broadcast.append(0)
+            B_broadcast.append(1)
+        elif tensor_B.shape[i] == 1:
+            A_broadcast.append(1)
+            B_broadcast.append(0)
+    return A_broadcast, B_broadcast
+
+
+def saturate(x: te.Tensor, dtype: str):
+    """Saturate value for the specified data type"""
+    return te.max(te.min_value(dtype), te.min(x, te.max_value(dtype)))
+
+
+def get_int_scale(scale_A, scale_B, scale_M, zero_point_A, zero_point_B, zero_point_M, op):
+    """Get fixed-point number"""

Review Comment:
   Can you please add some more comment to explain what's going on in this function? 



-- 
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