mingmwang commented on PR #6003: URL: https://github.com/apache/arrow-datafusion/pull/6003#issuecomment-1512476474
@yjshen @mustafasrepo @Dandandan I did some experiment by removing the Scalar values and using the rust native types to update the Accumulator state directly. It get some performance gain but not that obvious(reduced 50ms~ on my Mac). In this PR, I will continue on the Scalar values update approach. And I think after this PR, the major bottleneck of the high cardinality aggregations on DataFusion should be the Cache miss and Memory stall and our CPU instructions per cycle is much lower than DuckDB's (on tpch-q17). -- 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]
