[GitHub] [incubator-tvm] jackwish commented on a change in pull request #4329: [QNN] Use Int16 upcast in Fallback Conv2D. Fix test names.

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

jackwish commented on a change in pull request #4329: [QNN] Use Int16 upcast in Fallback Conv2D. Fix test names.
URL: https://github.com/apache/incubator-tvm/pull/4329#discussion_r346212861
 
 

 ##########
 File path: src/relay/qnn/op/convolution.cc
 ##########
 @@ -106,25 +106,26 @@ WorkloadType GetWorkload(const Array<tvm::relay::Type>& arg_types, const QnnConv
  * \brief Fallback to simpler lowering for dilation or depthwise conv.
  * \param data The input expr.
  * \param weight The weight expr.
- * \param zp_data The data zero point expr.
- * \param zp_kernel The kernel zero point expr.
  * \param param The qnn conv2d attributes.
  * \return The fallback lowered sequence of Relay expr.
  * \note In case of dilation, normal lowering would require a dilated pool.
  *       Since, we don't have dilated pool, we fallback to a simpler sequence of
  *       Relay operations. This will potentially lead to performance degradation
  *       as the convolution is called on int32 tensors instead of int8 tensors.
  */
-Expr Conv2DFallBack(const Expr& data, const Expr& weight, const Expr& zp_data,
-                    const Expr& zp_kernel, const QnnConv2DAttrs* param) {
-  auto shifted_data = data;
+Expr Conv2DFallBack(const Expr& data, const Expr& weight, const QnnConv2DAttrs* param) {
+  // Upcast the zero point to Int16.
 
 Review comment:
   This is pretty straightforward :)

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services