atlas-jmh
API
Generated with
atlas-jmh/com.netflix.atlas/com.netflix.atlas.core/com.netflix.atlas.core.util

com.netflix.atlas.core.util

package com.netflix.atlas.core.util

Members list

Type members

Classlikes

class BoxesRuntime 
Benchmark                         Mode  Cnt          Score         Error  Units
BoxesRuntime.javaEquals          thrpt   10  112495954.095 ± 2795842.625  ops/s
BoxesRuntime.scalaEquals         thrpt   10   82963668.047 ± 6225820.287  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class IntIntMap

Sanity check for integer hash set.

Sanity check for integer hash set.

> run -wi 10 -i 10 -f1 -t1 .*IntIntMap.*
...
load factor of 0.7

[info] Benchmark                        Mode  Cnt      Score     Error  Units
[info] IntIntMap.testIntIntHashMap800  thrpt   10  10374.640 ± 782.568  ops/s
[info] IntIntMap.testIntIntHashMap8k   thrpt   10    589.537 ±   8.124  ops/s
[info] IntIntMap.testJavaHashMap800    thrpt   10   4051.768 ± 499.362  ops/s
[info] IntIntMap.testJavaHashMap8k     thrpt   10   2846.103 ±  14.858  ops/s

load factor of 0.5

[info] Benchmark                        Mode  Cnt      Score     Error  Units
[info] IntIntMap.testIntIntHashMap800  thrpt   10  11713.871 ± 271.158  ops/s
[info] IntIntMap.testIntIntHashMap8k   thrpt   10   3963.805 ±  59.359  ops/s
[info] IntIntMap.testJavaHashMap800    thrpt   10   4371.380 ±  93.346  ops/s
[info] IntIntMap.testJavaHashMap8k     thrpt   10   2709.591 ±  71.376  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class IntSet

Sanity check for integer hash set.

Sanity check for integer hash set.

> run -wi 10 -i 10 -f1 -t1 .*IntSet.*
...
[info] Benchmark                    Mode  Cnt       Score       Error  Units
[info] IntSet.testIntHashSet       thrpt   10  478541.525 ± 29615.965  ops/s
[info] IntSet.testIntHashSet10k    thrpt   10    1531.614 ±    73.903  ops/s
[info] IntSet.testJavaHashSet      thrpt   10  376161.309 ±  5242.341  ops/s
[info] IntSet.testJavaHashSet10k   thrpt   10    2169.164 ±   126.964  ops/s
[info] IntSet.testTroveHashSet     thrpt   10  416539.087 ± 29554.649  ops/s
[info] IntSet.testTroveHashSet10k  thrpt   10    1813.617 ±    30.182  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class ListMerge

Sanity check performance of utility for merging sorted lists.

Sanity check performance of utility for merging sorted lists.

> jmh:run -wi 10 -i 10 -f1 -t1 .*ListMerge.*
...
Benchmark                         Mode  Cnt   Score   Error  Units
ListMerge.hashSet                thrpt   10   0.386 ± 0.021  ops/s
ListMerge.merge                  thrpt   10  42.918 ± 2.044  ops/s
ListMerge.sortedSet              thrpt   10   0.238 ± 0.007  ops/s
ListMerge.treeSet                thrpt   10   0.316 ± 0.016  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class ListMergeCommonPrefix

Sanity check performance of utility for merging sorted lists.

Sanity check performance of utility for merging sorted lists.

> jmh:run -prof jmh.extras.JFR -wi 10 -i 10 -f1 -t1 .*ListMergeCommonPrefix.*
...
Benchmark                         Mode  Cnt   Score   Error  Units
ListMergeCommonPrefix.merge      thrpt   10   8.700 ± 0.160  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class MathMax

There was an old suggestion that max via bit manip would be faster and avoid branch instructions. That doesn't appear to be the case:

There was an old suggestion that max via bit manip would be faster and avoid branch instructions. That doesn't appear to be the case:

> run -wi 10 -i 10 -f1 -t1 .*MathMax.*
...
[info] Benchmark                                  Mode  Samples           Score   Score error  Units
[info] c.n.a.c.u.MathMax.testMaxWithBitManip     thrpt       10   849801537.629  13395055.478  ops/s
[info] c.n.a.c.u.MathMax.testMaxWithIf           thrpt       10  1015950530.655  14758001.309  ops/s
[info] c.n.a.c.u.MathMax.testMaxWithScalaMath    thrpt       10  1021526747.880   8691594.381  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class SmallHashMapEntrySet

Check the overhead of java wrapper for use-cases that iterate over the entry set.

Check the overhead of java wrapper for use-cases that iterate over the entry set.

> jmh:run -prof gc -wi 10 -i 10 -f1 -t1 .*SmallHashMapEntrySet.*
...
Benchmark                      Mode  Cnt         Score        Error   Units
customEntrySet                thrpt   10  13480190.254 ± 351370.866   ops/s
scalaEntrySet                 thrpt   10   7782130.178 ± 514660.491   ops/s

customEntrySet   gc.alloc.rate.norm   10        24.000 ±      0.001    B/op
scalaEntrySet    gc.alloc.rate.norm   10       272.000 ±      0.001    B/op

Attributes

Supertypes
class Object
trait Matchable
class Any
class SmallHashMapEquals

Check performance of comparing small hash maps for equality.

Check performance of comparing small hash maps for equality.

> jmh:run -prof jmh.extras.JFR -wi 10 -i 10 -f1 -t1 .*SmallHashMapEquals.*
...
[info] Benchmark                     Mode  Cnt          Score          Error  Units

[info] currentEquals                thrpt   10   47004215.059 ±  2291847.719  ops/s
[info] inheritedEquals              thrpt   10    2134457.383 ±   111821.520  ops/s
[info] dataEquals                   thrpt   10   51326103.645 ±   758297.787  ops/s
[info] selfEquals                   thrpt   10  351279563.043 ± 18578115.816  ops/s

[info] currentEqualsNot             thrpt   10  273893522.221 ±  7980051.600  ops/s
[info] inheritedEqualsNot           thrpt   10    2341207.187 ±   206356.584  ops/s
[info] dataEqualsNot                thrpt   10   32392263.165 ±  1289262.059  ops/s

[info] currentEqualHashCodes        thrpt   10   14601802.119 ±   360902.793  ops/s
[info] inheritedEqualHashCodes      thrpt   10     483515.784 ±    10044.781  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class SmallHashMapHashCode

Check performance of computing the hash code for a small hash map. Note it will cache after the first run, see test with caching. However, the murmur3* tests are more interesting as they indicate the performance for the initial computation which is often in the critical path for new tag maps.

Check performance of computing the hash code for a small hash map. Note it will cache after the first run, see test with caching. However, the murmur3* tests are more interesting as they indicate the performance for the initial computation which is often in the critical path for new tag maps.

> jmh:run -prof jmh.extras.JFR -wi 10 -i 10 -f1 -t1 .*SmallHashMapHashCode.*
...
[info] Benchmark              Mode  Cnt          Score          Error  Units
[info] computeHashCode       thrpt   10   33962020.269 ±   883664.164  ops/s
[info] currentHashCode       thrpt   10  360180789.347 ± 10164654.707  ops/s
[info] murmur3arrayHash      thrpt   10   13861013.249 ±  3160191.522  ops/s
[info] murmur3mapHash        thrpt   10    4067194.458 ±    78185.171  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class SmallHashMapJavaGet

Check the overhead of java wrapper for use-cases that perform a lot of get calls.

Check the overhead of java wrapper for use-cases that perform a lot of get calls.

> jmh:run -prof gc -wi 10 -i 10 -f1 -t1 .*SmallHashMapJavaGet.*
...
[info] Benchmark               Mode  Cnt         Score         Error  Units
[info] customGetFound         thrpt   10  94270900.203 ± 5825997.538  ops/s
[info] customGetNotFound      thrpt   10   6799704.339 ±  462769.738  ops/s
[info] scalaGetFound          thrpt   10  85325015.251 ± 4367808.653  ops/s
[info] scalaGetNotFound       thrpt   10   6385962.734 ±  318520.923  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class SmallHashMapModify

Check performance of creating a copy of the map when adding/removing a single pair.

Check performance of creating a copy of the map when adding/removing a single pair.

> jmh:run -prof gc -wi 10 -i 10 -f1 -t1 .*SmallHashMapModify.*
...
[info] Benchmark                           Mode  Cnt        Score       Error  Units
[info] SmallHashMapModify.addPair         thrpt    5  2541571.202 ± 239240.513   ops/s
[info] SmallHashMapModify.removePair      thrpt    5  3840622.669 ± 439259.261   ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class StringIntern

Results:

> jmh:run -wi 10 -i 10 -f1 -t1 .*StringIntern.*

Results:

Benchmark           Mode  Cnt     Score     Error  Units
caffeineIntern100  thrpt   10  5030.208 ± 179.907  ops/s
caffeineIntern50   thrpt   10   689.302 ±  36.348  ops/s
customIntern       thrpt   10  1249.873 ±  32.805  ops/s
javaIntern         thrpt   10   664.655 ±  40.109  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class StringMatching

There was an old suggestion that max via bit manip would be faster and avoid branch instructions. That doesn't appear to be the case:

There was an old suggestion that max via bit manip would be faster and avoid branch instructions. That doesn't appear to be the case:

> run -wi 10 -i 10 -f1 -t1 .*StringMatching.*

Attributes

Supertypes
class Object
trait Matchable
class Any
class StringMatchingOr

Regex OR queries have terrible performance. In many cases these can be rewritten to a set of startsWith or indexOf patterns in the string matcher. On some of the real data and queries this was up to 20 times faster. For the quick sample data in this benchmark it is 8 to 15 times faster.

Regex OR queries have terrible performance. In many cases these can be rewritten to a set of startsWith or indexOf patterns in the string matcher. On some of the real data and queries this was up to 20 times faster. For the quick sample data in this benchmark it is 8 to 15 times faster.

> run -wi 10 -i 10 -f1 -t1 .*StringMatchingOr.*
[info] Benchmark                              Mode  Cnt    Score    Error  Units
[info] StringMatchingOr.testOrMatcher08      thrpt   10  953.969 ± 57.275  ops/s
[info] StringMatchingOr.testOrMatcher32      thrpt   10  437.536 ± 32.653  ops/s
[info] StringMatchingOr.testRegex08          thrpt   10  117.730 ± 18.946  ops/s
[info] StringMatchingOr.testRegex32          thrpt   10   29.162 ±  1.151  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class StringMatchingPrefix

Sanity check prefix matching and the benefits of short ciruiting with a startsWith check before falling back to the regex matcher.

Sanity check prefix matching and the benefits of short ciruiting with a startsWith check before falling back to the regex matcher.

> run -wi 10 -i 10 -f1 -t1 .*StringMatchingPrefix.*
[info] Benchmark                      Mode  Cnt      Score     Error  Units
[info] StringMatchingPrefix.matcher  thrpt   10  10314.028 ± 191.544  ops/s
[info] StringMatchingPrefix.regex    thrpt   10   3431.614 ± 350.880  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class StringSub

Quick sanity check on substitute changes. Main goal is to reduce some of the allocations seen in batch use-cases.

Quick sanity check on substitute changes. Main goal is to reduce some of the allocations seen in batch use-cases.

> run -wi 10 -i 10 -f1 -t1 .*StringSub.*

Initial results:

[info] Benchmark                     Mode  Cnt        Score       Error  Units
[info] StringSub.testParsingByHand  thrpt   10  2223516.436 ± 43697.844  ops/s
[info] StringSub.testUsingRegex     thrpt   10   267252.094 ±  6066.379  ops/s

Attributes

Supertypes
class Object
trait Matchable
class Any
class TimeWaveCalc 
> run -bm all -wi 10 -i 10 -f1 -t1 .*TimeWaveCalc.*
...
[info] Benchmark                         Mode     Cnt         Score        Error  Units
[info] TimeWaveCalc.testMathSin         thrpt      10  11484308.329 ± 729466.777  ops/s
[info] TimeWaveCalc.testPrecomputeSin   thrpt      10  39806386.423 ± 868184.835  ops/s
[info] TimeWaveCalc.testMathSin          avgt      10        ≈ 10⁻⁷                s/op
[info] TimeWaveCalc.testPrecomputeSin    avgt      10        ≈ 10⁻⁸                s/op
[info] TimeWaveCalc.testMathSin        sample  110059        ≈ 10⁻⁷                s/op
[info] TimeWaveCalc.testPrecomputeSin  sample  163664        ≈ 10⁻⁷                s/op
[info] TimeWaveCalc.testMathSin            ss      10        ≈ 10⁻⁶                s/op
[info] TimeWaveCalc.testPrecomputeSin      ss      10        ≈ 10⁻⁶                s/op

Attributes

Supertypes
class Object
trait Matchable
class Any
class ZeroPad

Check use of String.format against just padding with our custom Strings.zeroPad.

Check use of String.format against just padding with our custom Strings.zeroPad.

> jmh:run -prof gc -wi 10 -i 10 -f1 -t1 .*ZeroPad.*

Throughput

Benchmark                         Mode  Cnt        Score        Error   Units
hashArrayPad                     thrpt   10  4721308.898 ± 151781.009   ops/s
hashBigIntPad                    thrpt   10   867143.537 ±  19475.920   ops/s
hashFormat                       thrpt   10   519479.340 ±  36074.274   ops/s
oneArrayPad                      thrpt   10  7666430.180 ± 727283.845   ops/s
oneBigIntPad                     thrpt   10  3212304.227 ± 235176.496   ops/s
oneFormat                        thrpt   10   964090.075 ±  39087.882   ops/s

Allocations

hashArrayPad        gc.alloc.rate.norm   10      240.000 ±      0.001    B/op
hashBigIntPad       gc.alloc.rate.norm   10     1824.000 ±      0.001    B/op
hashFormat          gc.alloc.rate.norm   10     3344.000 ±      0.001    B/op
oneArrayPad         gc.alloc.rate.norm   10      240.000 ±      0.001    B/op
oneBigIntPad        gc.alloc.rate.norm   10      664.000 ±      0.001    B/op
oneFormat           gc.alloc.rate.norm   10     2176.000 ±      0.001    B/op

Attributes

Supertypes
class Object
trait Matchable
class Any
Generated with