[SPARK-10735] [SQL] Generate aggregation w/o grouping keys [WIP] #10786
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 27 commits
January 11, 2016 17:34
Conflicts: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/BoundAttribute.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/conditionalExpressions.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/misc.scala sql/core/src/main/scala/org/apache/spark/sql/execution/basicOperators.scala
Conflicts: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/CodeGenerator.scala
Conflicts: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/CodegenFallback.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/GenerateMutableProjection.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/GenerateOrdering.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/GeneratePredicate.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/GenerateSafeProjection.scala sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/GenerateUnsafeProjection.scala sql/core/src/main/scala/org/apache/spark/sql/execution/columnar/GenerateColumnAccessor.scala
|
Test build #49526 has finished for PR 10786 at commit
|
Conflicts: sql/core/src/main/scala/org/apache/spark/sql/execution/WholeStageCodegen.scala sql/core/src/test/scala/org/apache/spark/sql/execution/BenchmarkWholeStageCodegen.scala
|
Test build #49537 has finished for PR 10786 at commit
|
|
Test build #49540 has finished for PR 10786 at commit
|
|
Test build #49546 has finished for PR 10786 at commit
|
|
Test build #49548 has finished for PR 10786 at commit
|
Contributor
Author
Contributor
There was a problem hiding this comment.
Glad to see 5x speed-up! +1 on switching to declarative on stddev (and other 2nd-order statistics). But for skewness and kurtosis, we still need to benchmark the performance to decide because their expressions are more complex.
Contributor
There was a problem hiding this comment.
I've always preferred declarative aggregate because its much easier to optimize (very cool the kind of speed ups you are getting!). As such, I'd support having all of our built in functions done this way.
@mengxr argues that its too confusing for users and that we should also support the imperative one. How high cost is this for us?
asfgit
pushed a commit
that referenced
this pull request
Jan 20, 2016
As discussed in #10786, the generated TungstenAggregate does not support imperative functions. For a query ``` sqlContext.range(10).filter("id > 1").groupBy().count() ``` The generated code will looks like: ``` /* 032 */ if (!initAgg0) { /* 033 */ initAgg0 = true; /* 034 */ /* 035 */ // initialize aggregation buffer /* 037 */ long bufValue2 = 0L; /* 038 */ /* 039 */ /* 040 */ // initialize Range /* 041 */ if (!range_initRange5) { /* 042 */ range_initRange5 = true; ... /* 071 */ } /* 072 */ /* 073 */ while (!range_overflow8 && range_number7 < range_partitionEnd6) { /* 074 */ long range_value9 = range_number7; /* 075 */ range_number7 += 1L; /* 076 */ if (range_number7 < range_value9 ^ 1L < 0) { /* 077 */ range_overflow8 = true; /* 078 */ } /* 079 */ /* 085 */ boolean primitive11 = false; /* 086 */ primitive11 = range_value9 > 1L; /* 087 */ if (!false && primitive11) { /* 092 */ // do aggregate and update aggregation buffer /* 099 */ long primitive17 = -1L; /* 100 */ primitive17 = bufValue2 + 1L; /* 101 */ bufValue2 = primitive17; /* 105 */ } /* 107 */ } /* 109 */ /* 110 */ // output the result /* 112 */ bufferHolder25.reset(); /* 114 */ rowWriter26.initialize(bufferHolder25, 1); /* 118 */ rowWriter26.write(0, bufValue2); /* 120 */ result24.pointTo(bufferHolder25.buffer, bufferHolder25.totalSize()); /* 121 */ currentRow = result24; /* 122 */ return; /* 124 */ } /* 125 */ ``` cc nongli Author: Davies Liu <[email protected]> Closes #10840 from davies/gen_agg.
Contributor
Author
|
This PR will be replaced by multiple small PRs. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The benchmark show that generated aggregation could be 6X time faster than non-generated (with generated filter/range).
It also showed that declarative function is much faster than imperative one (5X faster for stddev), we may need to switch to a declarative version. (the difference need to be measured when there are grouping keys).