Github user davies commented on the pull request:

    https://github.com/apache/spark/pull/11018#issuecomment-179417364
  
    @mengxr Here are four runs for BroadcastHashJoin:
    
    ```
    Running benchmark: BroadcastHashJoin
      Running case: BroadcastHashJoin codegen=false
    ArrayBuffer(   3951.07,    4019.88,    4526.42,    6504.16,    9585.88)
      Running case: BroadcastHashJoin codegen=true
    ArrayBuffer(   1857.93,    1944.03,    1961.13,    2142.73,    2223.96)
    
    Intel(R) Core(TM) i7-4558U CPU @ 2.80GHz
    BroadcastHashJoin:                   Median Time(ms)    Rate(M/s)   Per 
Row(ns)   Relative
    
-------------------------------------------------------------------------------------
    BroadcastHashJoin codegen=false              4526.42         4.63        
215.84     1.00 X
    BroadcastHashJoin codegen=true               1961.13        10.69         
93.51     2.31 X
    
    
    Running benchmark: BroadcastHashJoin
      Running case: BroadcastHashJoin codegen=false
    ArrayBuffer(   3670.16,    3766.32,    6600.27,    6629.90,    6976.39)
      Running case: BroadcastHashJoin codegen=true
    ArrayBuffer(   1866.42,    1899.43,    1973.72,    2012.05,    2026.64)
    
    Intel(R) Core(TM) i7-4558U CPU @ 2.80GHz
    BroadcastHashJoin:                   Median Time(ms)    Rate(M/s)   Per 
Row(ns)   Relative
    
-------------------------------------------------------------------------------------
    BroadcastHashJoin codegen=false              6600.27         3.18        
314.73     1.00 X
    BroadcastHashJoin codegen=true               1973.72        10.63         
94.11     3.34 X
    
    Running benchmark: BroadcastHashJoin
      Running case: BroadcastHashJoin codegen=false
    ArrayBuffer(   3790.12,    4326.11,    6543.06,    6890.16,    7029.33)
      Running case: BroadcastHashJoin codegen=true
    ArrayBuffer(   1869.67,    1921.94,    1938.57,    1939.92,    2099.17)
    
    Intel(R) Core(TM) i7-4558U CPU @ 2.80GHz
    BroadcastHashJoin:                   Median Time(ms)    Rate(M/s)   Per 
Row(ns)   Relative
    
-------------------------------------------------------------------------------------
    BroadcastHashJoin codegen=false              6543.06         3.21        
312.00     1.00 X
    BroadcastHashJoin codegen=true               1938.57        10.82         
92.44     3.38 X
    
    Running benchmark: BroadcastHashJoin
      Running case: BroadcastHashJoin codegen=false
    ArrayBuffer(   3783.44,    3826.02,    4032.82,    6582.35,    6972.17)
      Running case: BroadcastHashJoin codegen=true
    ArrayBuffer(   1868.17,    1907.98,    2004.32,    2013.73,    2027.36)
    
    Intel(R) Core(TM) i7-4558U CPU @ 2.80GHz
    BroadcastHashJoin:                   Median Time(ms)    Rate(M/s)   Per 
Row(ns)   Relative
    
-------------------------------------------------------------------------------------
    BroadcastHashJoin codegen=false              4032.82         5.20        
192.30     1.00 X
    BroadcastHashJoin codegen=true               2004.32        10.46         
95.57     2.01 X
    ```
    
    With median time, the improvements are 2.31X, 3.34X, 3.38X, 2.01X
    
    With best time, they will be 2.12X, 1.97X, 2.0X, 2.0X, they are much stable 
than those using median time.
    
    @mengxr @nongli So I still think that we should use best time here. Also, 
keep only one digits after dot.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to