Github user VinceShieh commented on the issue:
https://github.com/apache/spark/pull/18936
@srowen currently, what we see is, with default thread setting(take up all
computation resource available) for native blas, the No. 1 hot spot (with 95%+
self time) is scala.concurrent.impl.Promise$DefaultPromise.tryAwait.
The performance hit might due to, the native takes up all computation cores
and doesn't release in time, so the processes on JVM are unable to get enough
resource to continue the following tasks. Just a bold guess. We will try to dig
deeper to see if we can find more.
We only made changes to the places that take effect in SVM because that's
we benchmark on. I will make the changes to the rest of the place once we have
more benchmark data on other methods/algorithms affected.
Thanks.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
