Speed up ORC writer dictionary encoding for narrow tables with string columns - #23796

Open

vuule wants to merge 12 commits into

NVIDIA:mainfrom

vuule:orc-writer-dictionary-parallelism

Commits on Aug 25, 2026

Parallelize ORC writer dictionary encoding across rows
vuule
committed
93192da

Commits on Sep 11, 2026

Merge remote-tracking branch 'origin/main' into orc-writer-dictionary-parallelism
vuule
committed
f333741
Align the narrow write benchmark with the read one, trim comments
vuule
committed
8141850
Drop the param list from blocks_per_dictionary
vuule
committed
2f976d9
Make target_waves local to blocks_per_dictionary
vuule
committed
e79dbb5
Move the kernel notes down to the lines they describe
vuule
committed
73c5d5d
Assign the dictionary counts when one block owns them
vuule
committed
6a4dd58
Move the block budget rationale next to the expressions
vuule
committed
53f43f3
Trim the dictionary kernel comments
vuule
committed
920047e

Commits on Sep 17, 2026

Merge remote-tracking branch 'origin/main' into pr/vuule/23796
vuule
committed
897938f
Use grid_1d for the dictionary row indexing
vuule
committed
b45ecc4
Accumulate the dictionary counts through atomic_ref
vuule
committed
a93666b