Github user ilooner commented on the issue:
https://github.com/apache/drill/pull/1057
For posterity I will post some of the benchmarking Paul did. I will then do
some of my own micro benchmarking in a later comment.
++ Code Gen
- For a query which has an execution time of **5578 ms** code generating
the copiers takes about **2159 ms** (I don't have the original query). If we
use Generic copiers we save all of this time.
++ Execution
- The execution time of code generated copiers for the query of interest
was **1302 ms / run**.
- The execution time of generic copiers for the query of interest was
**1186 ms / run**.
++ Conclusion
Generic copiers are faster than code generated copiers. The JVM has made
big strides in the past few years and can now optimize code at runtime much
better than us.
---
