philo-he commented on PR #12309: URL: https://github.com/apache/gluten/pull/12309#issuecomment-4748629797
> @philo-he Thanks for your work. > > We are working on using Velox’s HashTableCache to cache pre-built hash tables [here](https://github.com/facebookincubator/velox/pull/17662/changes#diff-88ab260046edcbf89b1dde37f3132dfb40369899a99f14757900e8bd0d9121fd) so we can revert a custom [commit](https://github.com/IBM/velox/commit/55578058810edfdce25366855745d0416578a113) in our internal Velox repo. Since HashTableCache currently uses String as the cache key, can we continue using a string for our cache key. @JkSelf, thanks for the comment. The String key was previously used both as a cache key and as a type descriptor (via starts_with("BuiltBNLJBroadcastTable-")). This PR separates them — the isBhj flag is used to distinguish join operator type, so the key only needs to be a unique identifier. In theory, we can keep using string key in this PR, converted from build plan ID. Let me think about this. And even if Gluten finally uses Int key, we still can convert it to string at the Velox boundary. So I think you can keep using the string cache key in that Velox PR. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
