Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[SPARK-23312][SQL][followup] add a config to turn off vectorized cache reader #20513

Closed
wants to merge 2 commits into from
Closed

[SPARK-23312][SQL][followup] add a config to turn off vectorized cache reader #20513

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8525b2c
followup
cloud-fan Feb 6, 2018
880412f
fix test
cloud-fan Feb 6, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/columnar/InMemoryTableScanExec.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,9 @@ case class InMemoryTableScanExec(
}) && !WholeStageCodegenExec.isTooManyFields(conf, relation.schema)
}

// TODO: revisit this. Shall we always turn off whole stage codegen if the output data are rows?
override def supportCodegen: Boolean = supportsBatch
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In 2.4 we should look into this. My gut feeling is we don't need to enable whole stage codegen for scan nodes that output data as rows.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, we can do more perf measurement after 2.3 release

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is safe to keep the same behavior of 2.2.

I'm not sure if enabling whole stage codegen can hurt performance for scan nodes, btw. We can revisit this, of course.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We confirmed that we got performance improvements in several cases.
I will revisit this with more cases after 2.3 release.


override protected def needsUnsafeRowConversion: Boolean = false

private val columnIndices =
Expand Down
3 changes: 2 additions & 1 deletion sql/core/src/test/scala/org/apache/spark/sql/CachedTableSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -787,7 +787,8 @@ class CachedTableSuite extends QueryTest with SQLTestUtils with SharedSQLContext
withSQLConf(SQLConf.CACHE_VECTORIZED_READER_ENABLED.key -> vectorized.toString) {
val df = spark.range(10).cache()
df.queryExecution.executedPlan.foreach {
case i: InMemoryTableScanExec => assert(i.supportsBatch == vectorized)
case i: InMemoryTableScanExec =>
assert(i.supportsBatch == vectorized && i.supportCodegen == vectorized)
case _ =>
}
}
Expand Down