ZhennanQin commented on a change in pull request #13697: [MKLDNN] Enable signed
int8 support for convolution.
URL: https://github.com/apache/incubator-mxnet/pull/13697#discussion_r244627472
##########
File path: include/mxnet/ndarray.h
##########
@@ -694,9 +694,13 @@ class NDArray {
/*
* Create NDArray from mkldnn memory.
* mkldnn_mem The mkldnn memory to be managed.
- * static_data If true, mkldnn memory won't be freed on destruction.
*/
- explicit NDArray(const mkldnn::memory *mkldnn_mem, bool static_data = true);
+ explicit NDArray(const std::shared_ptr<mkldnn::memory> &mkldnn_mem);
Review comment:
Yes, but it's designed as an internal API, and it's enabled for mkldnn build
only. Can we seperate internal API from user API?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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
