diff options
author | Ye Xianjin <advancedxy@gmail.com> | 2015-05-02 23:08:09 +0100 |
---|---|---|
committer | Sean Owen <sowen@cloudera.com> | 2015-05-02 23:08:09 +0100 |
commit | bfcd528d6f5a5ebe61e0fcca890143e9a3c7f7f9 (patch) | |
tree | f2e273d87bfab66702f02e0ea32b794fd743254c /core/src/test | |
parent | da303526e54e9a0adfedb49417f383cde7870a69 (diff) | |
download | spark-bfcd528d6f5a5ebe61e0fcca890143e9a3c7f7f9.tar.gz spark-bfcd528d6f5a5ebe61e0fcca890143e9a3c7f7f9.tar.bz2 spark-bfcd528d6f5a5ebe61e0fcca890143e9a3c7f7f9.zip |
[SPARK-6030] [CORE] Using simulated field layout method to compute class shellSize
SizeEstimator gives wrong result for Integer on 64bit JVM with UseCompressedOops on, this pr fixes that. For more details, please refer [SPARK-6030](https://issues.apache.org/jira/browse/SPARK-6030)
sryza, I noticed there is a pr to expose SizeEstimator, maybe that should be waited by this pr get merged if we confirm this problem.
And shivaram would you mind to review this pr since you contribute related code. Also cc to srowen and mateiz
Author: Ye Xianjin <advancedxy@gmail.com>
Closes #4783 from advancedxy/SPARK-6030 and squashes the following commits:
c4dcb41 [Ye Xianjin] Add super.beforeEach in the beforeEach method to make the trait stackable.. Remove useless leading whitespace.
3f80640 [Ye Xianjin] The size of Integer class changes from 24 to 16 on a 64-bit JVM with -UseCompressedOops flag on after the fix. I don't how 100000 was originally calculated, It looks like 100000 is the magic number which makes sure spilling. Because of the size change, It fails because there is no spilling at all. Change the number to a slightly larger number fixes that.
e849d2d [Ye Xianjin] Merge two shellSize assignments into one. Add some explanation to alignSizeUp method.
85a0b51 [Ye Xianjin] Fix typos and update wording in comments. Using alignSizeUp to compute alignSize.
d27eb77 [Ye Xianjin] Add some detailed comments in the code. Add some test cases. It's very difficult to design test cases as the final object alignment will hide a lot of filed layout details if we just considering the whole size.
842aed1 [Ye Xianjin] primitiveSize(cls) can just return Int. Use a simplified class field layout method to calculate class instance size. Will add more documents and test cases. Add a new alignSizeUp function which uses bitwise operators to speedup.
62e8ab4 [Ye Xianjin] Don't alignSize for objects' shellSize, alignSize when added to state.size. Add some primitive wrapper objects size tests.
Diffstat (limited to 'core/src/test')
-rw-r--r-- | core/src/test/scala/org/apache/spark/util/SizeEstimatorSuite.scala | 47 | ||||
-rw-r--r-- | core/src/test/scala/org/apache/spark/util/collection/ExternalSorterSuite.scala | 10 |
2 files changed, 49 insertions, 8 deletions
diff --git a/core/src/test/scala/org/apache/spark/util/SizeEstimatorSuite.scala b/core/src/test/scala/org/apache/spark/util/SizeEstimatorSuite.scala index 28915bd533..133a76f28e 100644 --- a/core/src/test/scala/org/apache/spark/util/SizeEstimatorSuite.scala +++ b/core/src/test/scala/org/apache/spark/util/SizeEstimatorSuite.scala @@ -36,6 +36,15 @@ class DummyClass4(val d: DummyClass3) { val x: Int = 0 } +// dummy class to show class field blocks alignment. +class DummyClass5 extends DummyClass1 { + val x: Boolean = true +} + +class DummyClass6 extends DummyClass5 { + val y: Boolean = true +} + object DummyString { def apply(str: String) : DummyString = new DummyString(str.toArray) } @@ -50,6 +59,7 @@ class SizeEstimatorSuite override def beforeEach() { // Set the arch to 64-bit and compressedOops to true to get a deterministic test-case + super.beforeEach() System.setProperty("os.arch", "amd64") System.setProperty("spark.test.useCompressedOops", "true") } @@ -62,6 +72,22 @@ class SizeEstimatorSuite assertResult(48)(SizeEstimator.estimate(new DummyClass4(new DummyClass3))) } + test("primitive wrapper objects") { + assertResult(16)(SizeEstimator.estimate(new java.lang.Boolean(true))) + assertResult(16)(SizeEstimator.estimate(new java.lang.Byte("1"))) + assertResult(16)(SizeEstimator.estimate(new java.lang.Character('1'))) + assertResult(16)(SizeEstimator.estimate(new java.lang.Short("1"))) + assertResult(16)(SizeEstimator.estimate(new java.lang.Integer(1))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Long(1))) + assertResult(16)(SizeEstimator.estimate(new java.lang.Float(1.0))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Double(1.0d))) + } + + test("class field blocks rounding") { + assertResult(16)(SizeEstimator.estimate(new DummyClass5)) + assertResult(24)(SizeEstimator.estimate(new DummyClass6)) + } + // NOTE: The String class definition varies across JDK versions (1.6 vs. 1.7) and vendors // (Sun vs IBM). Use a DummyString class to make tests deterministic. test("strings") { @@ -102,18 +128,18 @@ class SizeEstimatorSuite val arr = new Array[Char](100000) assertResult(200016)(SizeEstimator.estimate(arr)) assertResult(480032)(SizeEstimator.estimate(Array.fill(10000)(new DummyString(arr)))) - + val buf = new ArrayBuffer[DummyString]() for (i <- 0 until 5000) { buf.append(new DummyString(new Array[Char](10))) } assertResult(340016)(SizeEstimator.estimate(buf.toArray)) - + for (i <- 0 until 5000) { buf.append(new DummyString(arr)) } assertResult(683912)(SizeEstimator.estimate(buf.toArray)) - + // If an array contains the *same* element many times, we should only count it once. val d1 = new DummyClass1 // 10 pointers plus 8-byte object @@ -155,5 +181,20 @@ class SizeEstimatorSuite assertResult(64)(SizeEstimator.estimate(DummyString("a"))) assertResult(64)(SizeEstimator.estimate(DummyString("ab"))) assertResult(72)(SizeEstimator.estimate(DummyString("abcdefgh"))) + + // primitive wrapper classes + assertResult(24)(SizeEstimator.estimate(new java.lang.Boolean(true))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Byte("1"))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Character('1'))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Short("1"))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Integer(1))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Long(1))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Float(1.0))) + assertResult(24)(SizeEstimator.estimate(new java.lang.Double(1.0d))) + } + + test("class field blocks rounding on 64-bit VM without useCompressedOops") { + assertResult(24)(SizeEstimator.estimate(new DummyClass5)) + assertResult(32)(SizeEstimator.estimate(new DummyClass6)) } } diff --git a/core/src/test/scala/org/apache/spark/util/collection/ExternalSorterSuite.scala b/core/src/test/scala/org/apache/spark/util/collection/ExternalSorterSuite.scala index 20fd22b78e..7a98723bc6 100644 --- a/core/src/test/scala/org/apache/spark/util/collection/ExternalSorterSuite.scala +++ b/core/src/test/scala/org/apache/spark/util/collection/ExternalSorterSuite.scala @@ -377,7 +377,7 @@ class ExternalSorterSuite extends FunSuite with LocalSparkContext with PrivateMe val sorter = new ExternalSorter[Int, Int, Int]( None, Some(new HashPartitioner(3)), Some(ord), None) assertDidNotBypassMergeSort(sorter) - sorter.insertAll((0 until 100000).iterator.map(i => (i, i))) + sorter.insertAll((0 until 120000).iterator.map(i => (i, i))) assert(diskBlockManager.getAllFiles().length > 0) sorter.stop() assert(diskBlockManager.getAllBlocks().length === 0) @@ -385,9 +385,9 @@ class ExternalSorterSuite extends FunSuite with LocalSparkContext with PrivateMe val sorter2 = new ExternalSorter[Int, Int, Int]( None, Some(new HashPartitioner(3)), Some(ord), None) assertDidNotBypassMergeSort(sorter2) - sorter2.insertAll((0 until 100000).iterator.map(i => (i, i))) + sorter2.insertAll((0 until 120000).iterator.map(i => (i, i))) assert(diskBlockManager.getAllFiles().length > 0) - assert(sorter2.iterator.toSet === (0 until 100000).map(i => (i, i)).toSet) + assert(sorter2.iterator.toSet === (0 until 120000).map(i => (i, i)).toSet) sorter2.stop() assert(diskBlockManager.getAllBlocks().length === 0) } @@ -428,8 +428,8 @@ class ExternalSorterSuite extends FunSuite with LocalSparkContext with PrivateMe None, Some(new HashPartitioner(3)), Some(ord), None) assertDidNotBypassMergeSort(sorter) intercept[SparkException] { - sorter.insertAll((0 until 100000).iterator.map(i => { - if (i == 99990) { + sorter.insertAll((0 until 120000).iterator.map(i => { + if (i == 119990) { throw new SparkException("Intentional failure") } (i, i) |