npawar opened a new pull request #7493: URL: https://github.com/apache/pinot/pull/7493
1) Add more info to FetchContext 2) Make AcquireReleaseOperator take the plan instead of the operator for the child, to ensure acquire can happen before buffers accessed in planning 3) Move acquire release to BaseCombineOperator level, to better ensure release only happens after buffer access are done To ensure clean acquire/release semantics, dataBuffers can no longer cache DirectByteBuffers. Removing it from BitmapInvertedIndexReader, RangeInvertedIndexReader and NullValueVectorReaderImpl. Included jmh benchmarks in the change, which shows 1. while cached bitmaps outperform in a setup with large heap and high repeatability, it performs significantly worse in a setup with tight heap and low repeatability, compared to always building the bitmap. 2. in the case where alwaysBuild performed slower, the overhead we're talking about here is <1 microseconds per read. In the case where alwaysBuild performed better, it outdid by 4.6 microseconds per read. 3. These results remain consistent regardless of low/high cardinality -- 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: commits-unsubscr...@pinot.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org For additional commands, e-mail: commits-h...@pinot.apache.org
