Github user viirya commented on the issue:
https://github.com/apache/spark/pull/13439
@viirya , thank you for preparing benchmark results. Let me clarify this
implementation by two questions.
1. To reduce memory foot print, does this PR allocate an array for each
column with the number of row instead of ``` DEFAULT_BATCH_SIZE```?
2. Why can this PR improve performance? When I see ```getInt()``` both
for ```OnHeapColumnVector.java``` and ```OnHeapConstantColumnVector.java```,
they are almost similar (```intData[rowId]``` and ```intData[0]```). Does the
performance improvement come from possible elimination of index bound check due
to constant index access? Correct?
If so, we could do the similar thing by using ```Platform.getInt()``` in
```OnHeapColumnVector.java```.
---
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]
