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

Jump to bottom

Use new list reduction API to improve performance #4941

Merged
merged 1 commit into from
Mar 15, 2022
Merged

Use new list reduction API to improve performance #4941

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
73 changes: 61 additions & 12 deletions sql-plugin/src/main/scala/org/apache/spark/sql/rapids/collectionOperations.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ import java.util.Optional
import scala.collection.mutable.ArrayBuffer

import ai.rapids.cudf
import ai.rapids.cudf.{BinaryOp, ColumnVector, ColumnView, DType, GroupByAggregation, GroupByOptions, Scalar, Table}
import ai.rapids.cudf.{BinaryOp, ColumnVector, ColumnView, DType, GroupByAggregation, GroupByOptions, Scalar, SegmentedReductionAggregation, Table}
import com.nvidia.spark.rapids._
import com.nvidia.spark.rapids.ArrayIndexUtils.firstIndexAndNumElementUnchecked
import com.nvidia.spark.rapids.BoolUtils.isAllValidTrue
Expand Down Expand Up @@ -391,14 +391,11 @@ case class GpuSortArray(base: Expression, ascendingOrder: Expression)
}
}

trait GpuBaseArrayAgg extends GpuUnaryExpression {

protected def agg: GroupByAggregation

override protected def doColumnar(input: GpuColumnVector): cudf.ColumnVector = {
// TODO switch over to array aggregations once
// https://github.com/rapidsai/cudf/issues/10417 is done
object SlowGpuArrayAgg extends Arm{
def reallySlow(input: GpuColumnVector, agg: GroupByAggregation): cudf.ColumnVector = {
val baseInput = input.getBase
// TODO switch over to array aggregations once
// https://github.com/rapidsai/cudf/issues/9135 is done
val inputTab = withResource(Scalar.fromInt(0)) { zero =>
withResource(cudf.ColumnVector.sequence(zero, input.getRowCount.toInt)) { rowNums =>
new cudf.Table(rowNums, baseInput)
Expand All @@ -422,9 +419,29 @@ trait GpuBaseArrayAgg extends GpuUnaryExpression {
retTab.getColumn(1).incRefCount()
}
}

def bitCastDecimal(
input: GpuColumnVector,
agg: SegmentedReductionAggregation,
tmpType: DType,
resultType: DType): cudf.ColumnVector = {
val base = input.getBase
val tmpResult = withResource(base.getChildColumnView(0)) { dataCol =>
withResource(dataCol.bitCastTo(tmpType)) { castDataCol =>
withResource(base.replaceListChild(castDataCol)) { bitCastInput =>
bitCastInput.listReduce(agg)
}
}
}
withResource(tmpResult) { tmpResult =>
withResource(tmpResult.bitCastTo(resultType)) { bitCastResult =>
bitCastResult.copyToColumnVector()
}
}
}
}

case class GpuArrayMin(child: Expression) extends GpuBaseArrayAgg with ImplicitCastInputTypes {
case class GpuArrayMin(child: Expression) extends GpuUnaryExpression with ImplicitCastInputTypes {

override def nullable: Boolean = true

Expand All @@ -437,10 +454,26 @@ case class GpuArrayMin(child: Expression) extends GpuBaseArrayAgg with ImplicitC

override def prettyName: String = "array_min"

override protected def agg: GroupByAggregation = GroupByAggregation.min()
override protected def doColumnar(input: GpuColumnVector): cudf.ColumnVector =
dataType match {
case StringType =>
SlowGpuArrayAgg.reallySlow(input, GroupByAggregation.min())
case dt: DecimalType =>
if (dt.precision > Decimal.MAX_LONG_DIGITS) {
SlowGpuArrayAgg.reallySlow(input, GroupByAggregation.min())
} else if (dt.precision > Decimal.MAX_INT_DIGITS) {
SlowGpuArrayAgg.bitCastDecimal(input, SegmentedReductionAggregation.min(),
DType.INT64, GpuColumnVector.getNonNestedRapidsType(dataType))
} else {
SlowGpuArrayAgg.bitCastDecimal(input, SegmentedReductionAggregation.min(),
DType.INT32, GpuColumnVector.getNonNestedRapidsType(dataType))
}
case _ =>
input.getBase.listReduce(SegmentedReductionAggregation.min())
}
}

case class GpuArrayMax(child: Expression) extends GpuBaseArrayAgg with ImplicitCastInputTypes {
case class GpuArrayMax(child: Expression) extends GpuUnaryExpression with ImplicitCastInputTypes {

override def nullable: Boolean = true

Expand All @@ -453,7 +486,23 @@ case class GpuArrayMax(child: Expression) extends GpuBaseArrayAgg with ImplicitC

override def prettyName: String = "array_max"

override protected def agg: GroupByAggregation = GroupByAggregation.max()
override protected def doColumnar(input: GpuColumnVector): cudf.ColumnVector =
dataType match {
case StringType =>
SlowGpuArrayAgg.reallySlow(input, GroupByAggregation.max())
case dt: DecimalType =>
if (dt.precision > Decimal.MAX_LONG_DIGITS) {
SlowGpuArrayAgg.reallySlow(input, GroupByAggregation.max())
} else if (dt.precision > Decimal.MAX_INT_DIGITS) {
SlowGpuArrayAgg.bitCastDecimal(input, SegmentedReductionAggregation.max(),
DType.INT64, GpuColumnVector.getNonNestedRapidsType(dataType))
} else {
SlowGpuArrayAgg.bitCastDecimal(input, SegmentedReductionAggregation.max(),
DType.INT32, GpuColumnVector.getNonNestedRapidsType(dataType))
}
case _ =>
input.getBase.listReduce(SegmentedReductionAggregation.max())
}
}

class GpuSequenceMeta(
Expand Down