Skip to content

[SPARK-9058][SQL] Split projectionCode if it is too large for JVM #7418

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

Closed
viirya wants to merge 5 commits into from
Closed

[SPARK-9058][SQL] Split projectionCode if it is too large for JVM #7418

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
d715fd5
Split projectionCode if it is too large for JVM.
viirya Jul 15, 2015
8775359
Add semicolon.
viirya Jul 15, 2015
7435454
Merge remote-tracking branch 'upstream/master' into fix_codegen_size
viirya Jul 16, 2015
12d3794
For comment.
viirya Jul 16, 2015
b8e274e
Fix it.
viirya Jul 16, 2015
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
27 changes: 25 additions & 2 deletions ...n/scala/org/apache/spark/sql/catalyst/expressions/codegen/GenerateMutableProjection.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ object GenerateMutableProjection extends CodeGenerator[Seq[Expression], () => Mu

protected def create(expressions: Seq[Expression]): (() => MutableProjection) = {
val ctx = newCodeGenContext()
val projectionCode = expressions.zipWithIndex.map { case (e, i) =>
val projectionCodes = expressions.zipWithIndex.map { case (e, i) =>
val evaluationCode = e.gen(ctx)
evaluationCode.code +
s"""
Expand All @@ -45,10 +45,31 @@ object GenerateMutableProjection extends CodeGenerator[Seq[Expression], () => Mu
else
${ctx.setColumn("mutableRow", e.dataType, i, evaluationCode.primitive)};
"""
}.mkString("\n")
}

val projectionCodeSegments = projectionCodes.grouped(50).toSeq.map(_.mkString("\n"))
Copy link
Member

@maropu maropu Jul 15, 2015

Choose a reason for hiding this comment

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

How did you decide this number '50' for the JVM code size limitation?


val (projectionCode, projectionFuncs) = if (projectionCodeSegments.length == 1) {
(projectionCodeSegments(0), "")
Copy link
Contributor

@chenghao-intel chenghao-intel Jul 17, 2015

Choose a reason for hiding this comment

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

Oh, right! Here. Thanks for explanation.

} else {
val pCode = (0 until projectionCodeSegments.length).map { i =>
s"projectSeg$i(i);"
}.mkString("\n")

val pFuncs = (0 until projectionCodeSegments.length).map { i =>
s"""
private final void projectSeg$i(InternalRow i) {
${projectionCodeSegments(i)}
}
"""
}.mkString("\n")
(pCode, pFuncs)
}

val mutableStates = ctx.mutableStates.map { case (javaType, variableName, initialValue) =>
s"private $javaType $variableName = $initialValue;"
}.mkString("\n ")

val code = s"""
public Object generate($exprType[] expr) {
return new SpecificProjection(expr);
Expand All @@ -75,6 +96,8 @@ object GenerateMutableProjection extends CodeGenerator[Seq[Expression], () => Mu
return (InternalRow) mutableRow;
}

$projectionFuncs

public Object apply(Object _i) {
InternalRow i = (InternalRow) _i;
$projectionCode
Expand Down