Details
-
New Feature
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
-
ghx-label-11
Description
Impala traditionally has been a leader in low latency queries. However, it is reported that for sub second queries codegen could take the majority of the time, for example, codegen alone can take 300ms for a sub second query.
This task aims to add support of memory caches for codegen functions to improve the performance of sub second queries by reducing the codegen generation time. Users can specify the total cache size for all of the codegen functions and which query is going to use the cache.