Github user mpjlu commented on the issue:
https://github.com/apache/spark/pull/18624
Hi @holdenk, this is the PR we have discussed in Strata conference.
I have thought about the code again, for the performance, we can continue
to optimize the code. Because we can merge the block matrices before shuffle.
For only code simplification, I didn't get a good method.
Any change needs many tests about the performance, and now I am in a DL
team, and I will not have the test environment soon.
I will be glad if anyone wants to continue this work.
Thanks.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
