Github user a-roberts commented on the issue:

    https://github.com/apache/spark/pull/15736
  
    Passed on your question to our JIT developers
    
    > The sense* of the test can have an impact of the VM interpreter 
performance, but that is not usually much of a component of actual throughput 
since important methods will be JIT'd very quickly regardless of which specific 
JVM you use. The J9 VM is capable of profiling the branch and flipping the 
sense when JITing the code
    > 
    > The sense refers to the way a code branches, so either down the equals 
branch or not equals branch
    
    Numbers for us
    
    **Refactored further as above**
    ```
    ScalaSparkPagerank 2016-11-30 14:27:49 259928115            49.841          
     5215146
    ScalaSparkPagerank 2016-11-30 14:29:52 259928115            51.310          
     5065837
    ScalaSparkPagerank 2016-11-30 14:31:59 259928115            52.086          
     4990364
    ScalaSparkPagerank 2016-11-30 14:34:05 259928115            50.667          
     5130126
    ScalaSparkPagerank 2016-11-30 14:36:04 259928115            47.096          
     5519112
    ScalaSparkPagerank 2016-11-30 14:38:04 259928115            48.244          
     5387781
    ScalaSparkPagerank 2016-11-30 14:40:10 259928115            48.734          
     5333609
    ScalaSparkPagerank 2016-11-30 14:42:12 259928115            49.295          
     5272910
    397.273 / 8 = 49.659 sec average
    ```
    **initial commit**
    ```
    ScalaSparkPagerank 2016-11-30 14:48:01 259928115            46.442          
     5596832
    ScalaSparkPagerank 2016-11-30 14:50:06 259928115            50.016          
     5196899
    ScalaSparkPagerank 2016-11-30 14:52:12 259928115            51.113          
     5085362
    ScalaSparkPagerank 2016-11-30 14:54:12 259928115            46.424          
     5599002
    ScalaSparkPagerank 2016-11-30 14:56:15 259928115            47.604          
     5460215
    ScalaSparkPagerank 2016-11-30 14:58:14 259928115            46.802          
     5553782
    ScalaSparkPagerank 2016-11-30 15:00:16 259928115            47.021          
     5527915
    ScalaSparkPagerank 2016-11-30 15:02:16 259928115            47.072          
     5521926
    382.494 / 8 = 47.811s average
    ```
    
    The first commit performs better on average, I'd like to next add the 
improved compare code as above and "push this down" into the subclasses to see 
how this performs


---
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