zhztheplayer commented on PR #5401:
URL: 
https://github.com/apache/incubator-gluten/pull/5401#issuecomment-2055423981

   > ===== Performance report for TPCH SF2000 with Velox backend, for reference 
only ====
   > 
   > query      log/native_5401_time.csv        
log/native_master_04_14_2024_0e3387255_time.csv difference      percentage
   > q1 40.53   38.08   -2.454  93.95%
   > q2 24.12   23.72   -0.401  98.34%
   > q3 38.22   37.90   -0.320  99.16%
   > q4 38.24   39.60   1.363   103.56%
   > q5 69.95   71.73   1.780   102.55%
   > q6 7.40    6.97    -0.437  94.10%
   > q7 82.30   84.03   1.721   102.09%
   > q8 85.06   84.99   -0.072  99.92%
   > q9 126.13  125.91  -0.222  99.82%
   > q10        47.91   44.70   -3.200  93.32%
   > q11        20.50   20.27   -0.230  98.88%
   > q12        26.89   27.63   0.737   102.74%
   > q13        48.21   55.28   7.071   114.67%
   > q14        20.30   17.73   -2.573  87.32%
   > q15        29.35   31.75   2.396   108.16%
   > q16        14.43   13.82   -0.603  95.82%
   > q17        101.30  100.77  -0.537  99.47%
   > q18        144.74  143.22  -1.518  98.95%
   > q19        15.00   15.86   0.856   105.71%
   > q20        28.59   28.08   -0.508  98.22%
   > q21        229.26  289.28  60.017  126.18%
   > q22        14.07   14.35   0.274   101.94%
   > total      1252.52 1315.66 63.139  105.04%
   
   Hi @zhli1142015 
   
   https://github.com/facebookincubator/velox/pull/9079 caused performance 
regression on hash join. Would you like to have a check?
   
   Say q21 was slowed down from 289s to 229s in the above benchmark result 
table.


-- 
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]

Reply via email to