NVIDIA / NVIDIA/cudf

Enable cudf-polars parquet metadata prefetching by default on-prem

Open
#22,826 1 comment 1 reaction 1 assignee Claimed by @TomAugspurger View on GitHub
cudf-polars feature request Performance
Dominant language
C++
Stars
9.8k
Forks
1.1k
Avg merge
3d 6m
Merged PRs (30d)
278

Description

https://github.com/rapidsai/cudf/pull/22700 changes cudf-polars execution to add an explicit metadata prefetching stage to query execution, just after lowering and just before executing the actual query nodes.

Currently it is off by default because we've noticed performance regressions with it on, on-prem. This issue is for tracking and fixing the performance regressions.

The following issues are related, but may not fix the root-cause of the perf regression.
- #23526
- #23131

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.