Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
-
ghx-label-2
Description
When there are lots of delete records IcebergDeleteBuilder can become a bottleneck. Since the left side of the JOIN is blocked on the build side any improvement we make here significantly improves Iceberg V2 table scanning.
- std vector growths need lots of re-allocations and memory copies
- inserting into the position vectors one-by-one is inefficient
- sorting the position vectors could be parallelized (especially in case of MT_DOP when there are MT_DOP threads blocked on the build)