hhhizzz commented on PR #10852:
URL: https://github.com/apache/arrow-rs/pull/10852#issuecomment-5435965297

   I'm not sure how to run whole bench in this PR, So I run them on my dev 
server, here's the main results:
   
   ### End-to-end benchmark results
   
   I ran three cyclic-order rounds against Arrow main using the same DataFusion 
integration harness. The first iteration of each query was excluded, and the 
table reports the cross-round median of steady-state query times.
   
   | Suite | Main OFF | Main ON | Candidate ON | Candidate vs Main ON | Query 
geomean |
   |---|---:|---:|---:|---:|---:|
   | ClickBench | 41.394 s | 35.932 s | 36.280 s | +0.97% | -4.76% |
   | TPC-H SF10 | 9.549 s | 8.265 s | 7.992 s | -3.31% | -5.02% |
   | TPC-DS SF10 | 73.218 s | 46.051 s | 44.450 s | -3.48% | -4.12% |
   
   Selected queries:
   
   | Suite | Query | Main OFF | Main ON | Candidate ON | Candidate vs Main ON | 
Candidate vs OFF |
   |---|---:|---:|---:|---:|---:|---:|
   | ClickBench | Q25 | 180.4 ms | 283.5 ms | 172.8 ms | **-39.0%** | **-4.2%** 
|
   | ClickBench | Q12 | 552.0 ms | 671.4 ms | 567.9 ms | **-15.4%** | +2.9% |
   | ClickBench | Q23 | 7,738.4 ms | 857.3 ms | 853.4 ms | -0.5% | **-89.0%** |
   | ClickBench | Q21 | 834.5 ms | 1,592.3 ms | 1,570.2 ms | -1.4% | +88.2% |
   | TPC-H | Q6 | 122.7 ms | 324.0 ms | 239.0 ms | **-26.2%** | +94.8% |
   | TPC-H | Q18 | 3,136.8 ms | 871.1 ms | 859.1 ms | -1.4% | **-72.6%** |
   | TPC-DS | Q9 | 279.3 ms | 1,198.3 ms | 468.6 ms | **-60.9%** | +67.8% |
   | TPC-DS | Q72 | 31,044.1 ms | 1,376.0 ms | 1,255.3 ms | **-8.8%** | 
**-96.0%** |
   | TPC-DS | Q14 | 1,777.4 ms | 1,622.4 ms | 1,801.8 ms | +11.1% | +1.4% |
   | TPC-DS | Q50 | 280.9 ms | 500.2 ms | 553.4 ms | +10.6% | +97.1% |
   
   The candidate fixes the target Q25 regression, preserves the major pushdown 
wins such as ClickBench Q23, TPC-H Q18, and TPC-DS Q72, and reduces several 
other Main-ON regressions. It is not a pure monotonic improvement: regressions 
remain on queries such as ClickBench Q21 and TPC-DS Q14/Q50, and pushdown is 
still not universally faster than OFF. All query row counts matched across 
every arm and round.


-- 
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]

Reply via email to