Github user hvanhovell commented on the pull request:
https://github.com/apache/spark/pull/12397#issuecomment-213588703
@robbinspg I have been been looking at the assembler this generates for
`VectorizedPlainValuesReader.readInteger`. I can confirm that C2 compilation
eleminates the entire `if (bigEndian) { ... }` block.
One remark regarding benchmarking. If my logs are correct and nothing funny
happens to std.out, then I can say we need to be a bit more carefull about
benchmarking. The C1/C2 phases only kicked after the after the first two
benchmarks. So some warm-up would nice in the future.
Otherwise LGTM.
For future reference. This is what I did to print the compiled bytecode:
```
java -XX:+UnlockDiagnosticVMOptions
-XX:CompileCommand=print,*VectorizedPlainValuesReader.readInteger -cp
./dist/jars/*:./sql/core/target/scala-2.11/spark-sql_2.11-2.0.0-SNAPSHOT-tests.jar
org.apache.spark.sql.execution.datasources.parquet.ParquetReadBenchmark
```
See http://psy-lob-saw.blogspot.nl/2013/01/java-print-assembly.html for
more details.
---
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]
