andrei5055 commented on pull request #18582:
URL: https://github.com/apache/incubator-mxnet/pull/18582#issuecomment-656967881
Regarding my previous comment...
**Please, for now ignore graphs for CPU and CPU_PINNED memory. I am not sure
why, but I don't see 2x run time improvement
andrei5055 commented on pull request #18582:
URL: https://github.com/apache/incubator-mxnet/pull/18582#issuecomment-654943614
![image](https://user-images.githubusercontent.com/7293680/86804698-fda81800-c02b-11ea-92d0-29b6bfc6ad46.png)
(Here we do have only three lines, because the
andrei5055 commented on pull request #18582:
URL: https://github.com/apache/incubator-mxnet/pull/18582#issuecomment-654937596
![image](https://user-images.githubusercontent.com/7293680/86802902-55de1a80-c02a-11ea-9c0c-b8aa6c7d9089.png)
(Here we do have only three lines, because the
andrei5055 commented on pull request #18582:
URL: https://github.com/apache/incubator-mxnet/pull/18582#issuecomment-654562254
Here is first graph for GPU Storage Managers:
![image](https://user-images.githubusercontent.com/7293680/86691969-30b3c280-bfbe-11ea-9101-0e2ef18b2210.png)
andrei5055 commented on pull request #18582:
URL: https://github.com/apache/incubator-mxnet/pull/18582#issuecomment-653164922
@szha : I think this PR is ready for review. Two non-required tests failed,
but for some reason, CI launched both of them twice, and first launches
succeeded.