Description
Is your feature request related to a problem or challenge?
Part of #16065
Note: some steps are not related spilling execution, but they're very related general optimization, so I also mentioned them here.
Background
In sort queries, when the sort key is multiple columns, a row format can be used to accelerate comparison.
See more about the background in:
https://arrow.apache.org/blog/2022/11/07/multi-column-sorts-in-arrow-rust-part-1/
https://duckdb.org/pdf/ICDE2023-kuiper-muehleisen-sorting.pdf
Current status
Row format is used in the sort-preserving merge executor to speed up comparison among multiple merge inputs, but not used in the sort executor.
Note: probably its better to help get #15380 merged first, then start the following steps.
Purposed next steps
-
Use row format by default in the sort executor. This may temporarily slow down end-to-end benchmarks, but performance can be recovered through future optimizations (see the next point). This change also resolves issue #14748.
-
A sort query executed across multiple partitions currently goes through two stages of SPM:
Local Sort -> SPM1 -> SPM2
. At each stage, the Arrow Row Format array is repeatedly converted without reuse. We should pass the converted rows downstream to enable reuse and reduce overhead. -
For further optimization in workloads with nearly sorted input, the converted rows can be used to compute each local sort batch’s min/max. This can help concatenate batches without gaps and reduce the degree of merging needed.
-
For further optimization in memory-limited sorts, the converted rows can be written to the spill file and reused when reading back, improving efficiency.
Describe the solution you'd like
No response
Describe alternatives you've considered
No response
Additional context
No response