Rich-T-kid commented on PR #27252: URL: https://github.com/apache/datafusion/pull/27252#issuecomment-6729004639
## Arrow main comparison RLE dictionary reads are enabled on both sides. Released Arrow (`036f7377e`) -> live Arrow main (`fa57639f4`): - TPCH: `778.54 -> 764.36 ms` (`-1.82%`) - TPCDS: `9418.63 -> 9256.50 ms` (`-1.72%`) - ClickBench partitioned: `41774.32 -> 39337.68 ms` (`-5.83%`); wall time `210 -> 200s` The ClickBench gain is concentrated in Q33/Q34, followed by Q13, Q5, Q12, and Q14. It is real progress, but does not remove the original RLE-on regression. Chaining the paired ratios from the earlier RLE off -> on run with this released Arrow -> live main run gives the directional residual versus RLE off: - TPCH: about `+6.71%` (was `+8.69%`) - TPCDS: about `-1.35%` (was `+0.38%`) - ClickBench partitioned: about `+97.30%` (was `+109.52%`) The direct commit-to-commit deltas above are the primary result; the chained figures are directional because they combine two separate hosted runs. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
