[SPARK-34620][SQL] Code-gen broadcast nested loop join (inner/cross) #31736

c21 · 2021-03-04T12:29:01Z

What changes were proposed in this pull request?

BroadcastNestedLoopJoinExec does not have code-gen, and we can potentially boost the CPU performance for this operator if we add code-gen for it. https://databricks.com/blog/2017/02/16/processing-trillion-rows-per-second-single-machine-can-nested-loop-joins-fast.html also showed the evidence in one fork.

The codegen for BroadcastNestedLoopJoinExec shared some code with HashJoin, and the interface JoinCodegenSupport is created to hold those common logic. This PR is only supporting inner and cross join. Other join types will be added later in followup PRs.

Example query and generated code:

val df1 = spark.range(4).select($"id".as("k1"))
val df2 = spark.range(3).select($"id".as("k2"))
df1.join(df2, $"k1" + 1 =!= $"k2").explain("codegen")

== Subtree 2 / 2 (maxMethodCodeSize:282; maxConstantPoolSize:203(0.31% used); numInnerClasses:0) ==
*(2) BroadcastNestedLoopJoin BuildRight, Inner, NOT ((k1#2L + 1) = k2#6L)
:- *(2) Project [id#0L AS k1#2L]
:  +- *(2) Range (0, 4, step=1, splits=2)
+- BroadcastExchange IdentityBroadcastMode, [id=#22]
   +- *(1) Project [id#4L AS k2#6L]
      +- *(1) Range (0, 3, step=1, splits=2)

Generated code:
/* 001 */ public Object generate(Object[] references) {
/* 002 */   return new GeneratedIteratorForCodegenStage2(references);
/* 003 */ }
/* 004 */
/* 005 */ // codegenStageId=2
/* 006 */ final class GeneratedIteratorForCodegenStage2 extends org.apache.spark.sql.execution.BufferedRowIterator {
/* 007 */   private Object[] references;
/* 008 */   private scala.collection.Iterator[] inputs;
/* 009 */   private boolean range_initRange_0;
/* 010 */   private long range_nextIndex_0;
/* 011 */   private TaskContext range_taskContext_0;
/* 012 */   private InputMetrics range_inputMetrics_0;
/* 013 */   private long range_batchEnd_0;
/* 014 */   private long range_numElementsTodo_0;
/* 015 */   private InternalRow[] bnlj_buildRowArray_0;
/* 016 */   private org.apache.spark.sql.catalyst.expressions.codegen.UnsafeRowWriter[] range_mutableStateArray_0 = new org.apache.spark.sql.catalyst.expressions.codegen.UnsafeRowWriter[4];
/* 017 */
/* 018 */   public GeneratedIteratorForCodegenStage2(Object[] references) {
/* 019 */     this.references = references;
/* 020 */   }
/* 021 */
/* 022 */   public void init(int index, scala.collection.Iterator[] inputs) {
/* 023 */     partitionIndex = index;
/* 024 */     this.inputs = inputs;
/* 025 */
/* 026 */     range_taskContext_0 = TaskContext.get();
/* 027 */     range_inputMetrics_0 = range_taskContext_0.taskMetrics().inputMetrics();
/* 028 */     range_mutableStateArray_0[0] = new org.apache.spark.sql.catalyst.expressions.codegen.UnsafeRowWriter(1, 0);
/* 029 */     range_mutableStateArray_0[1] = new org.apache.spark.sql.catalyst.expressions.codegen.UnsafeRowWriter(1, 0);
/* 030 */     range_mutableStateArray_0[2] = new org.apache.spark.sql.catalyst.expressions.codegen.UnsafeRowWriter(1, 0);
/* 031 */     bnlj_buildRowArray_0 = (InternalRow[]) ((org.apache.spark.broadcast.TorrentBroadcast) references[1] /* broadcastTerm */).value();
/* 032 */     range_mutableStateArray_0[3] = new org.apache.spark.sql.catalyst.expressions.codegen.UnsafeRowWriter(2, 0);
/* 033 */
/* 034 */   }
/* 035 */
/* 036 */   private void bnlj_doConsume_0(long bnlj_expr_0_0) throws java.io.IOException {
/* 037 */     for (int bnlj_arrayIndex_0 = 0; bnlj_arrayIndex_0 < bnlj_buildRowArray_0.length; bnlj_arrayIndex_0++) {
/* 038 */       UnsafeRow bnlj_buildRow_0 = (UnsafeRow) bnlj_buildRowArray_0[bnlj_arrayIndex_0];
/* 039 */
/* 040 */       long bnlj_value_1 = bnlj_buildRow_0.getLong(0);
/* 041 */
/* 042 */       long bnlj_value_4 = -1L;
/* 043 */
/* 044 */       bnlj_value_4 = bnlj_expr_0_0 + 1L;
/* 045 */
/* 046 */       boolean bnlj_value_3 = false;
/* 047 */       bnlj_value_3 = bnlj_value_4 == bnlj_value_1;
/* 048 */       boolean bnlj_value_2 = false;
/* 049 */       bnlj_value_2 = !(bnlj_value_3);
/* 050 */       if (!(false || !bnlj_value_2))
/* 051 */       {
/* 052 */         ((org.apache.spark.sql.execution.metric.SQLMetric) references[2] /* numOutputRows */).add(1);
/* 053 */
/* 054 */         range_mutableStateArray_0[3].reset();
/* 055 */
/* 056 */         range_mutableStateArray_0[3].write(0, bnlj_expr_0_0);
/* 057 */
/* 058 */         range_mutableStateArray_0[3].write(1, bnlj_value_1);
/* 059 */         append((range_mutableStateArray_0[3].getRow()).copy());
/* 060 */
/* 061 */       }
/* 062 */     }
/* 063 */
/* 064 */   }
/* 065 */
/* 066 */   private void initRange(int idx) {
/* 067 */     java.math.BigInteger index = java.math.BigInteger.valueOf(idx);
/* 068 */     java.math.BigInteger numSlice = java.math.BigInteger.valueOf(2L);
/* 069 */     java.math.BigInteger numElement = java.math.BigInteger.valueOf(4L);
/* 070 */     java.math.BigInteger step = java.math.BigInteger.valueOf(1L);
/* 071 */     java.math.BigInteger start = java.math.BigInteger.valueOf(0L);
/* 072 */     long partitionEnd;
/* 073 */
/* 074 */     java.math.BigInteger st = index.multiply(numElement).divide(numSlice).multiply(step).add(start);
/* 075 */     if (st.compareTo(java.math.BigInteger.valueOf(Long.MAX_VALUE)) > 0) {
/* 076 */       range_nextIndex_0 = Long.MAX_VALUE;
/* 077 */     } else if (st.compareTo(java.math.BigInteger.valueOf(Long.MIN_VALUE)) < 0) {
/* 078 */       range_nextIndex_0 = Long.MIN_VALUE;
/* 079 */     } else {
/* 080 */       range_nextIndex_0 = st.longValue();
/* 081 */     }
/* 082 */     range_batchEnd_0 = range_nextIndex_0;
/* 083 */
/* 084 */     java.math.BigInteger end = index.add(java.math.BigInteger.ONE).multiply(numElement).divide(numSlice)
/* 085 */     .multiply(step).add(start);
/* 086 */     if (end.compareTo(java.math.BigInteger.valueOf(Long.MAX_VALUE)) > 0) {
/* 087 */       partitionEnd = Long.MAX_VALUE;
/* 088 */     } else if (end.compareTo(java.math.BigInteger.valueOf(Long.MIN_VALUE)) < 0) {
/* 089 */       partitionEnd = Long.MIN_VALUE;
/* 090 */     } else {
/* 091 */       partitionEnd = end.longValue();
/* 092 */     }
/* 093 */
/* 094 */     java.math.BigInteger startToEnd = java.math.BigInteger.valueOf(partitionEnd).subtract(
/* 095 */       java.math.BigInteger.valueOf(range_nextIndex_0));
/* 096 */     range_numElementsTodo_0  = startToEnd.divide(step).longValue();
/* 097 */     if (range_numElementsTodo_0 < 0) {
/* 098 */       range_numElementsTodo_0 = 0;
/* 099 */     } else if (startToEnd.remainder(step).compareTo(java.math.BigInteger.valueOf(0L)) != 0) {
/* 100 */       range_numElementsTodo_0++;
/* 101 */     }
/* 102 */   }
/* 103 */
/* 104 */   protected void processNext() throws java.io.IOException {
/* 105 */     // initialize Range
/* 106 */     if (!range_initRange_0) {
/* 107 */       range_initRange_0 = true;
/* 108 */       initRange(partitionIndex);
/* 109 */     }
/* 110 */
/* 111 */     while (true) {
/* 112 */       if (range_nextIndex_0 == range_batchEnd_0) {
/* 113 */         long range_nextBatchTodo_0;
/* 114 */         if (range_numElementsTodo_0 > 1000L) {
/* 115 */           range_nextBatchTodo_0 = 1000L;
/* 116 */           range_numElementsTodo_0 -= 1000L;
/* 117 */         } else {
/* 118 */           range_nextBatchTodo_0 = range_numElementsTodo_0;
/* 119 */           range_numElementsTodo_0 = 0;
/* 120 */           if (range_nextBatchTodo_0 == 0) break;
/* 121 */         }
/* 122 */         range_batchEnd_0 += range_nextBatchTodo_0 * 1L;
/* 123 */       }
/* 124 */
/* 125 */       int range_localEnd_0 = (int)((range_batchEnd_0 - range_nextIndex_0) / 1L);
/* 126 */       for (int range_localIdx_0 = 0; range_localIdx_0 < range_localEnd_0; range_localIdx_0++) {
/* 127 */         long range_value_0 = ((long)range_localIdx_0 * 1L) + range_nextIndex_0;
/* 128 */
/* 129 */         // common sub-expressions
/* 130 */
/* 131 */         bnlj_doConsume_0(range_value_0);
/* 132 */
/* 133 */         if (shouldStop()) {
/* 134 */           range_nextIndex_0 = range_value_0 + 1L;
/* 135 */           ((org.apache.spark.sql.execution.metric.SQLMetric) references[0] /* numOutputRows */).add(range_localIdx_0 + 1);
/* 136 */           range_inputMetrics_0.incRecordsRead(range_localIdx_0 + 1);
/* 137 */           return;
/* 138 */         }
/* 139 */
/* 140 */       }
/* 141 */       range_nextIndex_0 = range_batchEnd_0;
/* 142 */       ((org.apache.spark.sql.execution.metric.SQLMetric) references[0] /* numOutputRows */).add(range_localEnd_0);
/* 143 */       range_inputMetrics_0.incRecordsRead(range_localEnd_0);
/* 144 */       range_taskContext_0.killTaskIfInterrupted();
/* 145 */     }
/* 146 */   }
/* 147 */
/* 148 */ }

Why are the changes needed?

Improve query CPU performance. Added a micro benchmark query in JoinBenchmark.scala.
Saw 1x of run time improvement:

OpenJDK 64-Bit Server VM 11.0.9+11-LTS on Linux 4.14.219-161.340.amzn2.x86_64
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
broadcast nested loop join:                Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-------------------------------------------------------------------------------------------------------------------------
broadcast nested loop join wholestage off          62922          63052         184          0.3        3000.3       1.0X
broadcast nested loop join wholestage on           30946          30972          26          0.7        1475.6       2.0X

Does this PR introduce any user-facing change?

No.

How was this patch tested?

Added unit test in WholeStageCodegenSuite.scala, and existing unit tests for BroadcastNestedLoopJoinExec.
Updated golden files for several TCPDS query plans, as whole stage code-gen for BroadcastNestedLoopJoinExec is triggered.
Updated JoinBenchmark-jdk11-results.txt and JoinBenchmark-results.txt with new benchmark result. Followed previous benchmark PRs - [SPARK-30409][SPARK-29173][SQL][TESTS] Use NoOp datasource in SQL benchmarks #27078 and [SPARK-29320][TESTS] Compare sql/core module in JDK8/11 (Part 1) #26003 to use same type of machine:

Amazon AWS EC2
type: r3.xlarge
region: us-west-2 (Oregon)
OS: Linux

c21 · 2021-03-04T12:30:00Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

+   * Generate the (non-equi) condition used to filter joined rows. This is used in Inner, Left Semi
+   * and Left Anti joins.
+   */
+  protected def getJoinCondition(


This is copied from HashJoin.scala: getJoinCondition().

c21 · 2021-03-04T12:30:10Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

+  /**
+   * Generates the code for variable of build side.
+   */
+  protected def genBuildSideVars(


This is copied from HashJoin.scala: genBuildSideVars().

c21 · 2021-03-04T12:30:47Z

cc @cloud-fan and @maropu if you have time to take a look, thanks.

wangyum · 2021-03-04T13:45:49Z

sql/core/src/test/scala/org/apache/spark/sql/execution/benchmark/JoinBenchmark.scala

@@ -178,6 +191,7 @@ object JoinBenchmark extends SqlBasedBenchmark {
      sortMergeJoin()
      sortMergeJoinWithDuplicates()
      shuffleHashJoin()
+      broadcastNestedLoopJoin()


Could we update the benchmark results(JoinBenchmark-results.txt and JoinBenchmark-jdk11-results.txt)?

Generate result:

SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain org.apache.spark.sql.execution.benchmark.JoinBenchmark"

@wangyum - sure, I noticed the original benchmark was running on OpenJDK 64-Bit Server VM 1.8.0_232-8u232-b09-0ubuntu1~18.04.1-b09 on Linux 4.15.0-1044-aws in result file. Should I try to run benchmark in the same type of machine? or this is not a hard requirement. cc @maropu as well.

@wangyum, @maropu - updated the join benchmark results with same type machines as previous PRs. The description of PR has more details, thanks.

maropu · 2021-03-04T13:07:43Z

sql/core/src/test/scala/org/apache/spark/sql/execution/benchmark/JoinBenchmark.scala

+    val M = 1 << 4
+
+    val dim = broadcast(spark.range(M).selectExpr("id as k", "cast(id as string) as v"))
+    codegenBenchmark("broadcast nested loop join", N) {


Could you update benchmarks/JoinBenchmark-results.txt, too?

@maropu - sure, replied back above.

maropu · 2021-03-04T13:18:39Z

sql/core/src/test/scala/org/apache/spark/sql/execution/WholeStageCodegenSuite.scala

+    assert(oneJoinDF.queryExecution.executedPlan.collect {
+      case WholeStageCodegenExec(_ : BroadcastNestedLoopJoinExec) => true
+    }.size === 1)
+    checkAnswer(oneJoinDF,


For better test coverage, could you run tests with codegen enabled/disabled?

@maropu - sure, updated.

maropu · 2021-03-04T13:49:26Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoinExec.scala

+    streamed.asInstanceOf[CodegenSupport].inputRDDs()
+  }
+
+  override def needCopyResult: Boolean = true


always true? Is there a trade-off between the overheads of uniqueness checks and result copys?

@maropu - in case of inner/cross broadcast nested loop join, one input row can potentially have multiple output rows, so I am following the comment here to set it true. btw sort merge join has same behavior. I think if we want to improve it, I can do in another followup PR. I am not very familiar with this part and I can check closer later.

This is not a hash join and we don't know the uniqueness. I think it needs to be always true.

maropu · 2021-03-04T13:50:39Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoinExec.scala

+      case _: InnerLike => codegenInner(ctx, input)
+      case x =>
+        throw new IllegalArgumentException(
+          s"BroadcastNestedLoopJoin code-gen should not take $x as the JoinType")


nit:

case _ => throw new IllegalArgumentException( s"BroadcastNestedLoopJoin code-gen should not take $joinType as the JoinType")

?

@maropu - updated.

maropu · 2021-03-04T13:59:30Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoinExec.scala

+  }
+
+  private def codegenInner(ctx: CodegenContext, input: Seq[ExprCode]): String = {
+    val arrayTerm = prepareBroadcast(ctx)


nit: arrayTerm -> buildRowArrayTerm, buildRowsTerm, ...?

@maropu - updated to buildRowArrayTerm.

maropu · 2021-03-04T14:03:15Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoinExec.scala

+    val broadcastTerm = ctx.addReferenceObj("broadcastTerm", broadcastArray)
+
+    // Inline mutable state since not many join operations in a task
+    ctx.addMutableState("InternalRow[]", "broadcastArray",


nit: broadcastArray -> buildRows, buidlRowArray, ...?

@maropu - updated to buildRowArray.

SparkQA · 2021-03-04T14:06:13Z

Kubernetes integration test starting
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40336/

SparkQA · 2021-03-04T14:42:25Z

Kubernetes integration test status success
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40336/

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

SparkQA · 2021-03-05T09:01:25Z

Kubernetes integration test starting
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40384/

cloud-fan · 2021-03-05T09:01:40Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

+
+  /**
+   * Generate the (non-equi) condition used to filter joined rows. This is used in Inner, Left Semi
+   * and Left Anti joins.


Since it's in a base trait now, can we add more doc to explain the return type?

@cloud-fan - makes sense, updated.

SparkQA · 2021-03-05T09:06:32Z

Kubernetes integration test status failure
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40384/

cloud-fan · 2021-03-05T09:10:14Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoinExec.scala

+    s"""
+       |int $arrayIndex = 0;
+       |UnsafeRow $buildRow;
+       |while ($arrayIndex < $buildRowArrayTerm.length) {


Let's make it more java style

for (int $arrayIndex = 0; $arrayIndex < $buildRowArrayTerm.length; $arrayIndex++) { UnsafeRow $buildRow = (UnsafeRow) $buildRowArrayTerm[$arrayIndex]; ... }

I vaguely remembered some places in one PR that I saw while is preferred over for (or my memory can be wrong and it's opposite). @cloud-fan - do you remember related discussion?

while is preferred over for in Scala. I don't think it's true for Java, at least from my experience of Java development in the past.

@cloud-fan - updated.

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

c21 · 2021-03-06T09:53:37Z

Addressed all comments, and the PR is ready to review again, thanks @cloud-fan , @maropu and @wangyum .

SparkQA · 2021-03-06T11:11:35Z

Kubernetes integration test starting
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40408/

SparkQA · 2021-03-06T11:46:40Z

Kubernetes integration test status success
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40408/

maropu · 2021-03-06T13:34:16Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

+   * Generate the (non-equi) condition used to filter joined rows.
+   * This is used in Inner, Left Semi and Left Anti joins.
+   *
+   * @return Variable name for row of build side.


nit: How about this instead?

* @return Tuple of variable name for row of build side, generated code for condition, * and generated code for variables of build side.

@maropu - sure, updated.

SparkQA · 2021-03-06T14:46:38Z

Test build #135826 has finished for PR 31736 at commit 4e5341f.

This patch passes all tests.
This patch merges cleanly.
This patch adds no public classes.

SparkQA · 2021-03-07T22:22:44Z

Kubernetes integration test starting
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40429/

SparkQA · 2021-03-07T22:53:52Z

Kubernetes integration test status success
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40429/

SparkQA · 2021-03-08T01:53:06Z

Test build #135847 has finished for PR 31736 at commit 4d26e49.

This patch passes all tests.
This patch merges cleanly.
This patch adds no public classes.

cloud-fan · 2021-03-08T09:09:18Z

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala

+   */
+  protected def getJoinCondition(
+      ctx: CodegenContext,
+      input: Seq[ExprCode],


nit: streamVars?

@cloud-fan - updated.

c21 · 2021-03-09T06:32:56Z

Addressed all comments for now, and the PR is ready for check again, thanks.

c21 · 2021-03-09T07:59:21Z

Close & reopen PR to trigger test rerun as some transient unit test failure happened.

SparkQA · 2021-03-09T08:18:58Z

Kubernetes integration test starting
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40483/

SparkQA · 2021-03-09T08:45:57Z

Kubernetes integration test status success
URL: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder-K8s/40483/

SparkQA · 2021-03-09T11:25:49Z

Test build #135901 has finished for PR 31736 at commit 6f6d511.

This patch passes all tests.
This patch merges cleanly.
This patch adds no public classes.

cloud-fan · 2021-03-09T11:45:37Z

thanks, merging to master!

c21 · 2021-03-09T19:34:07Z

Thank you @cloud-fan , @maropu , @viirya and @wangyum for review!

Codegen inner/cross broadcast nested loop join

f5a06ba

c21 commented Mar 4, 2021

View reviewed changes

github-actions bot added the SQL label Mar 4, 2021

wangyum reviewed Mar 4, 2021

View reviewed changes

maropu reviewed Mar 4, 2021

View reviewed changes

Address comments and try to fix unit test failure

07ed459

cloud-fan reviewed Mar 5, 2021

View reviewed changes

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala Show resolved Hide resolved

cloud-fan reviewed Mar 5, 2021

View reviewed changes

sql/core/src/main/scala/org/apache/spark/sql/execution/joins/JoinCodegenSupport.scala Show resolved Hide resolved

Address all comments and update benchmark result

4e5341f

maropu reviewed Mar 6, 2021

View reviewed changes

maropu approved these changes Mar 6, 2021

View reviewed changes

Address comment for documentation

4d26e49

cloud-fan reviewed Mar 8, 2021

View reviewed changes

cloud-fan approved these changes Mar 8, 2021

View reviewed changes

Revert change for getJoinCondition and address nit comment

6f6d511

c21 closed this Mar 9, 2021

c21 reopened this Mar 9, 2021

cloud-fan closed this in b5b1985 Mar 9, 2021

c21 deleted the nested-join-exec branch March 9, 2021 19:34

c21 mentioned this pull request Mar 10, 2021

[SPARK-34702][SQL] Avoid unnecessary code generation in JoinCodegenSupport.genBuildSideVars #31802

Closed

[SPARK-34620][SQL] Code-gen broadcast nested loop join (inner/cross) #31736

[SPARK-34620][SQL] Code-gen broadcast nested loop join (inner/cross) #31736

Conversation

c21 commented Mar 4, 2021 • edited Loading

What changes were proposed in this pull request?

Why are the changes needed?

Does this PR introduce any user-facing change?

How was this patch tested?

Choose a reason for hiding this comment

Choose a reason for hiding this comment

c21 commented Mar 4, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SparkQA commented Mar 4, 2021

SparkQA commented Mar 4, 2021

SparkQA commented Mar 5, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SparkQA commented Mar 5, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

c21 commented Mar 6, 2021

SparkQA commented Mar 6, 2021

SparkQA commented Mar 6, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SparkQA commented Mar 6, 2021

SparkQA commented Mar 7, 2021

SparkQA commented Mar 7, 2021

SparkQA commented Mar 8, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

c21 commented Mar 9, 2021

c21 commented Mar 9, 2021

SparkQA commented Mar 9, 2021

SparkQA commented Mar 9, 2021

SparkQA commented Mar 9, 2021

cloud-fan commented Mar 9, 2021

c21 commented Mar 9, 2021

c21 commented Mar 4, 2021 •

edited

Loading