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

Use the new chunked API from multi-get_json_object #11289

Merged
merged 1 commit into from
Aug 5, 2024
Merged

Use the new chunked API from multi-get_json_object #11289

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
26 changes: 8 additions & 18 deletions sql-plugin/src/main/scala/com/nvidia/spark/rapids/GpuGetJsonObject.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,29 +220,19 @@ case class GpuMultiGetJsonObject(json: Expression,
case _ => None
}

var validPathsIndex = 0
val validPaths = validPathsWithIndexes.map(_._1)
withResource(new Array[ColumnVector](validPaths.length)) { validPathColumns =>
var validPathsIndex = 0
withResource(json.columnarEval(batch)) { input =>
// The get_json_object implementation will allocate an output that is as large
// as the input for each path being processed. This can cause memory to grow
// by a lot. We want to avoid this, but still try to run as many in parallel
// as we can
val p = parallel.getOrElse {
val inputSize = input.getBase.getDeviceMemorySize
// Our memory budget is 4x the target batch size. This is technically going
// to go over that, but in practice it is okay with the default settings
Math.max(Math.ceil((targetBatchSize * 4.0) / inputSize).toInt, 1)
}
validPaths.grouped(p).foreach { validPathChunk =>
withResource(JSONUtils.getJsonObjectMultiplePaths(input.getBase,
java.util.Arrays.asList(validPathChunk: _*))) { chunkedResult =>
chunkedResult.foreach { cr =>
validPathColumns(validPathsIndex) = cr.incRefCount()
validPathsIndex += 1
}
withResource(JSONUtils.getJsonObjectMultiplePaths(input.getBase,
java.util.Arrays.asList(validPaths: _*), 4 * targetBatchSize,
Copy link
Collaborator

Choose a reason for hiding this comment

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

the comment about memory budget is being removed. Consider re-adding or making 4 a mnemonic constant.

Copy link
Collaborator

Choose a reason for hiding this comment

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

nit:

throughout this file constructs like java.util.Arrays.asList(validPaths: _*)
can be replaced with validPaths.asJava
if we import scala.collection.JavaConverters._

parallel.getOrElse(-1))) { chunkedResult =>
chunkedResult.foreach { cr =>
validPathColumns(validPathsIndex) = cr.incRefCount()
validPathsIndex += 1
}
}

withResource(new Array[ColumnVector](paths.length)) { columns =>
if (nullIndexes.nonEmpty) {
val nullCol = withResource(GpuScalar.from(null, StringType)) { s =>
Expand Down