zhouyuan commented on code in PR #6475:
URL: https://github.com/apache/incubator-gluten/pull/6475#discussion_r1687370806
##########
cpp/core/operators/serializer/ColumnarBatchSerializer.h:
##########
@@ -25,7 +25,7 @@ namespace gluten {
class ColumnarBatchSerializer {
public:
- ColumnarBatchSerializer(arrow::MemoryPool* arrowPool, struct ArrowSchema*
cSchema) : arrowPool_(arrowPool) {}
+ ColumnarBatchSerializer(arrow::MemoryPool* arrowPool) :
arrowPool_(arrowPool) {}
Review Comment:
good cleanup
##########
cpp/velox/shuffle/VeloxSortShuffleWriter.cc:
##########
@@ -0,0 +1,315 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "shuffle/VeloxSortShuffleWriter.h"
+
+#include <arrow/io/memory.h>
+
+#include "memory/ArrowMemory.h"
+#include "memory/VeloxColumnarBatch.h"
+#include "utils/Common.h"
+#include "utils/Timer.h"
+
+namespace gluten {
+
+namespace {
+constexpr uint32_t kMaskLower27Bits = (1 << 27) - 1;
+constexpr uint64_t kMaskLower40Bits = (1UL << 40) - 1;
+
+uint64_t toCompactRowId(uint32_t partitionId, uint32_t pageNumber, uint32_t
offsetInPage) {
+ // |63 partitionId(24) |39 inputIndex(13) |26 rowIndex(27) |
+ return (uint64_t)partitionId << 40 | (uint64_t)pageNumber << 27 |
offsetInPage;
+}
+
+uint32_t extractPartitionId(uint64_t compactRowId) {
+ return (uint32_t)(compactRowId >> 40);
+}
+
+std::pair<uint32_t, uint32_t> extractPageNumberAndOffset(uint64_t
compactRowId) {
+ return {(compactRowId & kMaskLower40Bits) >> 27, compactRowId &
kMaskLower27Bits};
+}
+} // namespace
+
+arrow::Result<std::shared_ptr<VeloxShuffleWriter>>
VeloxSortShuffleWriter::create(
+ uint32_t numPartitions,
+ std::unique_ptr<PartitionWriter> partitionWriter,
+ ShuffleWriterOptions options,
+ std::shared_ptr<facebook::velox::memory::MemoryPool> veloxPool,
+ arrow::MemoryPool* arrowPool) {
+ std::shared_ptr<VeloxSortShuffleWriter> writer(new VeloxSortShuffleWriter(
+ numPartitions, std::move(partitionWriter), std::move(options),
std::move(veloxPool), arrowPool));
+ RETURN_NOT_OK(writer->init());
+ return writer;
+}
+
+VeloxSortShuffleWriter::VeloxSortShuffleWriter(
+ uint32_t numPartitions,
+ std::unique_ptr<PartitionWriter> partitionWriter,
+ ShuffleWriterOptions options,
+ std::shared_ptr<facebook::velox::memory::MemoryPool> veloxPool,
+ arrow::MemoryPool* pool)
+ : VeloxShuffleWriter(numPartitions, std::move(partitionWriter),
std::move(options), std::move(veloxPool), pool),
+
allocator_{std::make_unique<facebook::velox::HashStringAllocator>(veloxPool_.get())},
+ array_{SortArray{Allocator(allocator_.get())}} {}
+
+arrow::Status VeloxSortShuffleWriter::write(std::shared_ptr<ColumnarBatch> cb,
int64_t memLimit) {
+ ARROW_ASSIGN_OR_RAISE(auto rv, getPeeledRowVector(cb));
+ initRowType(rv);
+ RETURN_NOT_OK(insert(rv, memLimit));
+ return arrow::Status::OK();
+}
+
+arrow::Status VeloxSortShuffleWriter::stop() {
+ ARROW_RETURN_IF(evictState_ == EvictState::kUnevictable,
arrow::Status::Invalid("Unevictable state in stop."));
+
+ EvictGuard evictGuard{evictState_};
+
+ stopped_ = true;
+ RETURN_NOT_OK(evictAllPartitions());
+ RETURN_NOT_OK(partitionWriter_->stop(&metrics_));
+ return arrow::Status::OK();
+}
+
+arrow::Status VeloxSortShuffleWriter::reclaimFixedSize(int64_t size, int64_t*
actual) {
+ if (evictState_ == EvictState::kUnevictable || offset_ == 0) {
+ *actual = 0;
+ return arrow::Status::OK();
+ }
+ EvictGuard evictGuard{evictState_};
+ auto beforeReclaim = veloxPool_->usedBytes();
+ RETURN_NOT_OK(evictAllPartitions());
+ *actual = beforeReclaim - veloxPool_->usedBytes();
+ return arrow::Status::OK();
+}
+
+arrow::Status VeloxSortShuffleWriter::init() {
+ ARROW_RETURN_IF(
+ options_.partitioning == Partitioning::kSingle,
+ arrow::Status::Invalid("VeloxSortShuffleWriter doesn't support single
partition."));
+ partition2RowCount_.resize(numPartitions_, 0);
+ array_.resize(initialSize_);
+ return arrow::Status::OK();
+}
+
+void VeloxSortShuffleWriter::initRowType(const facebook::velox::RowVectorPtr&
rv) {
+ if (UNLIKELY(!rowType_)) {
+ rowType_ = facebook::velox::asRowType(rv->type());
+ fixedRowSize_ = facebook::velox::row::CompactRow::fixedRowSize(rowType_);
+ }
+}
+
+arrow::Result<facebook::velox::RowVectorPtr>
VeloxSortShuffleWriter::getPeeledRowVector(
+ const std::shared_ptr<ColumnarBatch>& cb) {
+ if (options_.partitioning == Partitioning::kRange) {
+ auto compositeBatch =
std::dynamic_pointer_cast<CompositeColumnarBatch>(cb);
+ VELOX_CHECK_NOT_NULL(compositeBatch);
+ auto batches = compositeBatch->getBatches();
+ VELOX_CHECK_EQ(batches.size(), 2);
+
+ auto pidBatch = VeloxColumnarBatch::from(veloxPool_.get(), batches[0]);
+ auto pidArr = getFirstColumn(*(pidBatch->getRowVector()));
+ RETURN_NOT_OK(partitioner_->compute(pidArr, pidBatch->numRows(),
row2Partition_, partition2RowCount_));
+
+ auto rvBatch = VeloxColumnarBatch::from(veloxPool_.get(), batches[1]);
+ return rvBatch->getFlattenedRowVector();
+ } else {
+ auto veloxColumnBatch = VeloxColumnarBatch::from(veloxPool_.get(), cb);
+ VELOX_CHECK_NOT_NULL(veloxColumnBatch);
+ auto rv = veloxColumnBatch->getFlattenedRowVector();
+ if (partitioner_->hasPid()) {
+ auto pidArr = getFirstColumn(*rv);
+ RETURN_NOT_OK(partitioner_->compute(pidArr, rv->size(), row2Partition_,
partition2RowCount_));
+ return getStrippedRowVector(*rv);
+ } else {
+ RETURN_NOT_OK(partitioner_->compute(nullptr, rv->size(), row2Partition_,
partition2RowCount_));
+ return rv;
+ }
+ }
+}
+
+arrow::Status VeloxSortShuffleWriter::insert(const
facebook::velox::RowVectorPtr& vector, int64_t memLimit) {
+ ScopedTimer timer(&c2rTime_);
+ auto inputRows = vector->size();
+ VELOX_DCHECK_GT(inputRows, 0);
+
+ facebook::velox::row::CompactRow row(vector);
+
+ if (!fixedRowSize_) {
+ rowSizes_.resize(inputRows + 1);
+ rowSizes_[0] = 0;
+ for (auto i = 0; i < inputRows; ++i) {
+ rowSizes_[i + 1] = rowSizes_[i] + row.rowSize(i);
+ }
+ }
+
+ uint32_t rowOffset = 0;
+ while (rowOffset < inputRows) {
+ auto remainingRows = inputRows - rowOffset;
+ auto rows = maxRowsToInsert(rowOffset, remainingRows);
+ if (rows == 0) {
+ auto minSizeRequired = fixedRowSize_ ? fixedRowSize_.value() :
rowSizes_[rowOffset + 1] - rowSizes_[rowOffset];
+ acquireNewBuffer(memLimit, minSizeRequired);
+ rows = maxRowsToInsert(rowOffset, remainingRows);
+ ARROW_RETURN_IF(
+ rows == 0, arrow::Status::Invalid("Failed to insert rows. Remaining
rows: " + std::to_string(remainingRows)));
+ }
+ RETURN_NOT_OK(maybeSpill(rows));
+ insertRows(row, rowOffset, rows);
+ rowOffset += rows;
+ }
+ return arrow::Status::OK();
+}
+
+void VeloxSortShuffleWriter::insertRows(facebook::velox::row::CompactRow& row,
uint32_t offset, uint32_t rows) {
+ // Allocate newArray can trigger spill.
+ growArrayIfNecessary(rows);
+ for (auto i = offset; i < offset + rows; ++i) {
+ auto size = row.serialize(i, currentPage_ + pageCursor_);
+ array_[offset_++] = {toCompactRowId(row2Partition_[i], pageNumber_,
pageCursor_), size};
+ pageCursor_ += size;
+ }
+}
+
+arrow::Status VeloxSortShuffleWriter::maybeSpill(int32_t nextRows) {
+ if ((uint64_t)offset_ + nextRows > std::numeric_limits<uint32_t>::max()) {
+ EvictGuard evictGuard{evictState_};
+ RETURN_NOT_OK(evictAllPartitions());
+ }
+ return arrow::Status::OK();
+}
+
+arrow::Status VeloxSortShuffleWriter::evictAllPartitions() {
+ {
+ ScopedTimer timer(&sortTime_);
+ // TODO: Add radix sort to align with Spark.
+ std::sort(array_.begin(), array_.begin() + offset_);
Review Comment:
can we use the prefix sort in Velox here?
##########
cpp/core/shuffle/HashPartitioner.cc:
##########
@@ -43,7 +43,6 @@ arrow::Status gluten::HashPartitioner::compute(
std::vector<uint32_t>& row2partition,
std::vector<uint32_t>& partition2RowCount) {
row2partition.resize(numRows);
- std::fill(std::begin(partition2RowCount), std::end(partition2RowCount), 0);
Review Comment:
it seems this fill operation is all removed, will this impact the
performance?
##########
backends-velox/src/main/scala/org/apache/gluten/backendsapi/velox/VeloxMetricsApi.scala:
##########
@@ -258,27 +258,33 @@ class VeloxMetricsApi extends MetricsApi with Logging {
sparkContext: SparkContext,
isSort: Boolean): Map[String, SQLMetric] = {
val baseMetrics = Map(
- "dataSize" -> SQLMetrics.createSizeMetric(sparkContext, "data size"),
"numPartitions" -> SQLMetrics.createMetric(sparkContext, "number of
partitions"),
+ "dataSize" -> SQLMetrics.createSizeMetric(sparkContext, "data size"),
"bytesSpilled" -> SQLMetrics.createSizeMetric(sparkContext, "shuffle
bytes spilled"),
- "splitBufferSize" -> SQLMetrics.createSizeMetric(sparkContext, "split
buffer size"),
- "splitTime" -> SQLMetrics.createNanoTimingMetric(sparkContext, "time to
split"),
- "spillTime" -> SQLMetrics.createNanoTimingMetric(sparkContext, "time to
spill"),
- "deserializeTime" -> SQLMetrics.createNanoTimingMetric(sparkContext,
"time to deserialize"),
"avgReadBatchNumRows" -> SQLMetrics
.createAverageMetric(sparkContext, "avg read batch num rows"),
"numInputRows" -> SQLMetrics.createMetric(sparkContext, "number of input
rows"),
"numOutputRows" -> SQLMetrics
.createMetric(sparkContext, "number of output rows"),
"inputBatches" -> SQLMetrics
- .createMetric(sparkContext, "number of input batches")
+ .createMetric(sparkContext, "number of input batches"),
+ "spillTime" -> SQLMetrics.createNanoTimingMetric(sparkContext, "time to
spill"),
+ "compressTime" -> SQLMetrics.createNanoTimingMetric(sparkContext, "time
to compress"),
+ "decompressTime" -> SQLMetrics.createNanoTimingMetric(sparkContext,
"time to decompress"),
+ "deserializeTime" -> SQLMetrics.createNanoTimingMetric(sparkContext,
"time to deserialize"),
+ // For hash shuffle writer, the peak bytes represents the maximum split
buffer size.
+ // For sort shuffle writer, the peak bytes represents the maximum
+ // row buffer + sort buffer size.
+ "peakBytes" -> SQLMetrics.createSizeMetric(sparkContext, "peak bytes
allocated")
)
if (isSort) {
- baseMetrics
+ baseMetrics ++ Map(
+ "sortTime" -> SQLMetrics.createNanoTimingMetric(sparkContext, "time to
shuffle sort"),
+ "c2rTime" -> SQLMetrics.createNanoTimingMetric(sparkContext, "time to
shuffle c2r")
Review Comment:
I understand this is newly added for sort based shuffle, it's not available
in hash based shuffle, however this may bring some trouble for analysis script
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
