[FEA] Replace cuco::static_multimap
by cuco::static_map
in semi-anti-join
#11313
Labels
0 - Backlog
In queue waiting for assignment
feature request
New feature or request
libcudf
Affects libcudf (C++/CUDA) code.
non-breaking
Non-breaking change
Performance
Performance related issue
Spark
Functionality that helps Spark RAPIDS
The implementation of semi-anti-join was refactored in #11100. One of the changes was to use
cuco::static_multimap
, which was later discovered that it has performance issue when the input tables have too many duplicate rows (#11299).We should use
cuco::static_map
to avoid the performance issue. However, this is not just a simple change in implementation but needs a new FEA from cuco which addspair_contains
intostatic_map
: NVIDIA/cuCollections#191.The text was updated successfully, but these errors were encountered: