[
https://issues.apache.org/jira/browse/SPARK-21870?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Takeshi Yamamuro updated SPARK-21870:
-------------------------------------
Description:
In SPARK-21603, we got performance regression if the HotSpot didn't compile too
long functions (the limit is 80000 in bytecode size).
I checked and I found the codegen of `HashAggregateExec` frequently goes over
the limit, for example:
{code}
spark.range(10000000).selectExpr("id % 1024 AS a", "id AS
b").write.saveAsTable("t")
sql("SELECT a, KURTOSIS(b)FROM t GROUP BY a")
{code}
This query goes over the limit and the actual bytecode size is `12356`.
So, it might be better to split the aggregation code into piecies.
was:
In SPARK-21603, we got performance regression if the HotSpot didn't compile too
long functions (the limit is 80000 in bytecode size).
I checked and I found the codegen of `HashAggregateExec` frequently goes over
the limit, for example:
```
spark.range(10000000).selectExpr("id % 1024 AS a", "id AS
b").write.saveAsTable("t")
sql("SELECT a, KURTOSIS(b)FROM t GROUP BY a")
```
This query goes over the limit and the actual bytecode size is `12356`.
So, it might be better to split the aggregation code into piecies.
> Split codegen'd aggregation code into small functions for the HotSpot
> ---------------------------------------------------------------------
>
> Key: SPARK-21870
> URL: https://issues.apache.org/jira/browse/SPARK-21870
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 2.2.0
> Reporter: Takeshi Yamamuro
> Priority: Minor
>
> In SPARK-21603, we got performance regression if the HotSpot didn't compile
> too long functions (the limit is 80000 in bytecode size).
> I checked and I found the codegen of `HashAggregateExec` frequently goes over
> the limit, for example:
> {code}
> spark.range(10000000).selectExpr("id % 1024 AS a", "id AS
> b").write.saveAsTable("t")
> sql("SELECT a, KURTOSIS(b)FROM t GROUP BY a")
> {code}
> This query goes over the limit and the actual bytecode size is `12356`.
> So, it might be better to split the aggregation code into piecies.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
