parallel-streams: JMH benchmarks and demos for when parallel streams are slower

N x Q model, source splittability, boxing, ordered operations, common-pool
starvation, and parallel-stream correctness bugs. Captured output in output/.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_017ic4b6hCTuFrQT8xKmmMrF
This commit is contained in:
Claude
2026-11-17 23:23:21 +05:30
parent f8c9e244a3
commit 4511cc6e6b
22 changed files with 1714 additions and 0 deletions
+1
View File
@@ -33,6 +33,7 @@ article; each module's own README has that article's version table, quickstart,
| [`copying`](copying/) | Deep Copy vs Shallow Copy in Java: clone(), Copy Constructors, Serialization and Records |
| [`datetimeformatter`](datetimeformatter/) | DateTimeFormatter Cheat Sheet: Patterns, the yyyy vs YYYY Bug, Locales and ISO Formats |
| [`bufferedreader`](bufferedreader/) | Java BufferedReader: Read a File Line by Line (vs Scanner, Files.lines and readAllLines) |
| [`parallel-streams`](parallel-streams/) | Java Parallel Streams: When They're Slower (JMH Benchmarks and the Common Pool) |
## License
+53
View File
@@ -0,0 +1,53 @@
# parallel-streams
Companion code for the ankurm.com post *"Java Parallel Streams: When They're Slower (JMH Benchmarks and the Common Pool)."*
Module in `java-core-examples`, the Java-core series. There is deliberately no `docs/` folder: the
deeper material lives in collapsible sections inside the post itself.
## Versions this was built and tested against
| Component | Version | Notes |
|---|---|---|
| JDK | 25.0.5+8 (Temurin, LTS) | |
| JMH | 1.37 | Latest GA on Maven Central; released 3 Aug 2023. |
| JUnit Jupiter | 5.11.0 | |
| Maven | 3.9.11 | |
| Host | 2 vCPU shared sandbox | Default common-pool parallelism is therefore 1 (see `output/00`). Speed-ups cap near 2x. Read the shape of the numbers, not the third significant digit. |
## Quickstart
```bash
export JAVA_HOME=/path/to/jdk-25
mvn -q package # builds target/benchmarks.jar and runs the tests
java -Djava.util.concurrent.ForkJoinPool.common.parallelism=3 \
-cp target/classes com.ankurm.parallelstreams.CorrectnessBugsDemo
java -Djava.util.concurrent.ForkJoinPool.common.parallelism=3 \
-cp target/classes com.ankurm.parallelstreams.CommonPoolStarvationDemo
java -jar target/benchmarks.jar NqBenchmark -f 1 -wi 3 -w 1s -i 5 -r 1s
```
`scripts/run-all.sh` regenerates every file in `output/` (about 6 minutes on 2 vCPUs).
`scripts/run.sh <ClassName>` runs one demo class ad hoc.
## What's in here
| File | What it shows |
|---|---|
| `NqBenchmark.java` | N elements x Q units of CPU per element, sequential vs parallel, at three sizes and two costs. Where parallel starts to pay. |
| `SourceBenchmark.java` | The same work over `int[]`, `ArrayList`, `LinkedList` and `Stream.iterate`: how splittable the source is decides the outcome. |
| `BoxingBenchmark.java` | A cheap sum over 1,000,000 elements, primitive vs boxed, with `-prof gc` allocation per call. |
| `OrderedOpsBenchmark.java` | `limit` and `distinct`, ordered vs `unordered()`, against a sequential baseline. |
| `CommonPoolStarvationDemo.java` | A neighbour that blocks the common pool slows an unrelated parallel stream 4x; a private `ForkJoinPool` and virtual threads as the fixes. |
| `CorrectnessBugsDemo.java` | `forEach(list::add)`, a non-identity `reduce` identity, a non-associative accumulator, and a racy counter. |
| `ParallelStreamsTest.java` | Pins the deterministic claims. No timing assertions. |
| `output/00-07` | Captured runs of all of the above. |
## Notes worth knowing before reading the post
- **JMH settings are light** (`-f 1 -wi 3 -i 5`, one second each) to keep the whole capture to minutes. Error bars are reported in the output; the post draws conclusions only from differences clearly larger than them.
- **The demos pin `common.parallelism=3`** so they behave the same on any machine. The JMH runs do not: they use the default for the host, which is 1 on 2 CPUs.
- **`BoxingBenchmark`'s first draft compared two different stream sources** (`Stream.iterate` for the sequential boxed case, `range().boxed()` for the parallel one). That is not a fair pair; both now use `IntStream.range(..).boxed()`.
## License
MIT - see the [repo-wide LICENSE](../LICENSE).
@@ -0,0 +1,4 @@
availableProcessors = 2 (default commonPool parallelism = 1)
openjdk version "25.0.5" 2026-10-20 LTS
OpenJDK Runtime Environment Temurin-25.0.5+8 (build 25.0.5+8-LTS)
OpenJDK 64-Bit Server VM Temurin-25.0.5+8 (build 25.0.5+8-LTS, mixed mode, sharing)
@@ -0,0 +1,15 @@
1. forEach(list::add), n = 100000, ten runs (-1 = exception):
sizes: 90126 86309 22551 86194 -1 22732 87868 10443 15012 77596
collect(toList()) size: 100000
2. reduce(10, Integer::sum) over 1..100
sequential: 5060
parallel : 5210
3. reduce(0, (a, b) -> a - b) over 1..1000
sequential: -500500
parallel : 0
4. counter[0]++ in forEach, n = 100000
racy : 99997
atomic : 100000
@@ -0,0 +1,15 @@
availableProcessors = 2
commonPool parallelism = 3
1. Victim alone
victim took 201 ms
2. Victim while a neighbour BLOCKS the common pool (8 tasks x 2000 ms)
victim took 803 ms
3. Same neighbour, but it runs in its OWN ForkJoinPool(3)
victim took 200 ms
4. 20 blocking tasks x 200 ms: parallel stream vs virtual threads
parallel stream : 1002 ms
virtual threads : 216 ms
+399
View File
@@ -0,0 +1,399 @@
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.parallel
# Parameters: (n = 100, q = 1)
# Run progress: 0.00% complete, ETA 00:01:36
# Fork: 1 of 1
# Warmup Iteration 1: 2.554 us/op
# Warmup Iteration 2: 2.826 us/op
# Warmup Iteration 3: 3.246 us/op
Iteration 1: 3.302 us/op
Iteration 2: 3.168 us/op
Iteration 3: 3.191 us/op
Iteration 4: 3.197 us/op
Iteration 5: 3.276 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.parallel":
3.227 ±(99.9%) 0.227 us/op [Average]
(min, avg, max) = (3.168, 3.227, 3.302), stdev = 0.059
CI (99.9%): [3.000, 3.453] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.parallel
# Parameters: (n = 100, q = 100)
# Run progress: 8.33% complete, ETA 00:01:30
# Fork: 1 of 1
# Warmup Iteration 1: 17.115 us/op
# Warmup Iteration 2: 19.112 us/op
# Warmup Iteration 3: 17.243 us/op
Iteration 1: 17.042 us/op
Iteration 2: 18.133 us/op
Iteration 3: 18.558 us/op
Iteration 4: 16.680 us/op
Iteration 5: 14.709 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.parallel":
17.024 ±(99.9%) 5.796 us/op [Average]
(min, avg, max) = (14.709, 17.024, 18.558), stdev = 1.505
CI (99.9%): [11.228, 22.821] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.parallel
# Parameters: (n = 10000, q = 1)
# Run progress: 16.67% complete, ETA 00:01:22
# Fork: 1 of 1
# Warmup Iteration 1: 41.559 us/op
# Warmup Iteration 2: 41.886 us/op
# Warmup Iteration 3: 40.756 us/op
Iteration 1: 45.860 us/op
Iteration 2: 45.214 us/op
Iteration 3: 45.519 us/op
Iteration 4: 45.168 us/op
Iteration 5: 44.723 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.parallel":
45.297 ±(99.9%) 1.633 us/op [Average]
(min, avg, max) = (44.723, 45.297, 45.860), stdev = 0.424
CI (99.9%): [43.664, 46.929] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.parallel
# Parameters: (n = 10000, q = 100)
# Run progress: 25.00% complete, ETA 00:01:14
# Fork: 1 of 1
# Warmup Iteration 1: 818.776 us/op
# Warmup Iteration 2: 723.115 us/op
# Warmup Iteration 3: 683.672 us/op
Iteration 1: 695.548 us/op
Iteration 2: 677.217 us/op
Iteration 3: 702.129 us/op
Iteration 4: 691.356 us/op
Iteration 5: 678.731 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.parallel":
688.996 ±(99.9%) 41.521 us/op [Average]
(min, avg, max) = (677.217, 688.996, 702.129), stdev = 10.783
CI (99.9%): [647.475, 730.518] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.parallel
# Parameters: (n = 1000000, q = 1)
# Run progress: 33.33% complete, ETA 00:01:06
# Fork: 1 of 1
# Warmup Iteration 1: 2051.571 us/op
# Warmup Iteration 2: 1906.834 us/op
# Warmup Iteration 3: 2024.894 us/op
Iteration 1: 1801.248 us/op
Iteration 2: 1801.422 us/op
Iteration 3: 1950.267 us/op
Iteration 4: 2295.446 us/op
Iteration 5: 1913.766 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.parallel":
1952.430 ±(99.9%) 781.640 us/op [Average]
(min, avg, max) = (1801.248, 1952.430, 2295.446), stdev = 202.989
CI (99.9%): [1170.790, 2734.070] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.parallel
# Parameters: (n = 1000000, q = 100)
# Run progress: 41.67% complete, ETA 00:00:57
# Fork: 1 of 1
# Warmup Iteration 1: 67554.950 us/op
# Warmup Iteration 2: 63880.972 us/op
# Warmup Iteration 3: 67555.514 us/op
Iteration 1: 66500.964 us/op
Iteration 2: 66413.244 us/op
Iteration 3: 65635.051 us/op
Iteration 4: 68084.734 us/op
Iteration 5: 64267.643 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.parallel":
66180.327 ±(99.9%) 5356.241 us/op [Average]
(min, avg, max) = (64267.643, 66180.327, 68084.734), stdev = 1390.999
CI (99.9%): [60824.086, 71536.569] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.sequential
# Parameters: (n = 100, q = 1)
# Run progress: 50.00% complete, ETA 00:00:49
# Fork: 1 of 1
# Warmup Iteration 1: 0.385 us/op
# Warmup Iteration 2: 0.359 us/op
# Warmup Iteration 3: 0.337 us/op
Iteration 1: 0.338 us/op
Iteration 2: 0.343 us/op
Iteration 3: 0.352 us/op
Iteration 4: 0.346 us/op
Iteration 5: 0.344 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.sequential":
0.345 ±(99.9%) 0.020 us/op [Average]
(min, avg, max) = (0.338, 0.345, 0.352), stdev = 0.005
CI (99.9%): [0.324, 0.365] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.sequential
# Parameters: (n = 100, q = 100)
# Run progress: 58.33% complete, ETA 00:00:41
# Fork: 1 of 1
# Warmup Iteration 1: 13.752 us/op
# Warmup Iteration 2: 13.339 us/op
# Warmup Iteration 3: 13.433 us/op
Iteration 1: 13.190 us/op
Iteration 2: 12.791 us/op
Iteration 3: 12.906 us/op
Iteration 4: 12.862 us/op
Iteration 5: 13.182 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.sequential":
12.986 ±(99.9%) 0.719 us/op [Average]
(min, avg, max) = (12.791, 12.986, 13.190), stdev = 0.187
CI (99.9%): [12.267, 13.705] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.sequential
# Parameters: (n = 10000, q = 1)
# Run progress: 66.67% complete, ETA 00:00:33
# Fork: 1 of 1
# Warmup Iteration 1: 33.847 us/op
# Warmup Iteration 2: 34.601 us/op
# Warmup Iteration 3: 32.702 us/op
Iteration 1: 31.444 us/op
Iteration 2: 31.912 us/op
Iteration 3: 31.441 us/op
Iteration 4: 36.858 us/op
Iteration 5: 33.124 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.sequential":
32.956 ±(99.9%) 8.808 us/op [Average]
(min, avg, max) = (31.441, 32.956, 36.858), stdev = 2.287
CI (99.9%): [24.147, 41.764] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.sequential
# Parameters: (n = 10000, q = 100)
# Run progress: 75.00% complete, ETA 00:00:24
# Fork: 1 of 1
# Warmup Iteration 1: 1323.332 us/op
# Warmup Iteration 2: 1291.482 us/op
# Warmup Iteration 3: 1267.897 us/op
Iteration 1: 1298.120 us/op
Iteration 2: 1318.923 us/op
Iteration 3: 1306.951 us/op
Iteration 4: 1287.234 us/op
Iteration 5: 1314.010 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.sequential":
1305.048 ±(99.9%) 48.775 us/op [Average]
(min, avg, max) = (1287.234, 1305.048, 1318.923), stdev = 12.667
CI (99.9%): [1256.273, 1353.823] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.sequential
# Parameters: (n = 1000000, q = 1)
# Run progress: 83.33% complete, ETA 00:00:16
# Fork: 1 of 1
# Warmup Iteration 1: 3185.234 us/op
# Warmup Iteration 2: 3270.377 us/op
# Warmup Iteration 3: 3403.997 us/op
Iteration 1: 3126.314 us/op
Iteration 2: 3046.140 us/op
Iteration 3: 3579.626 us/op
Iteration 4: 3076.687 us/op
Iteration 5: 3034.810 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.sequential":
3172.715 ±(99.9%) 886.464 us/op [Average]
(min, avg, max) = (3034.810, 3172.715, 3579.626), stdev = 230.212
CI (99.9%): [2286.252, 4059.179] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.NqBenchmark.sequential
# Parameters: (n = 1000000, q = 100)
# Run progress: 91.67% complete, ETA 00:00:08
# Fork: 1 of 1
# Warmup Iteration 1: 125134.566 us/op
# Warmup Iteration 2: 121151.145 us/op
# Warmup Iteration 3: 126327.662 us/op
Iteration 1: 127210.986 us/op
Iteration 2: 127633.563 us/op
Iteration 3: 126364.109 us/op
Iteration 4: 131880.637 us/op
Iteration 5: 124313.179 us/op
Result "com.ankurm.parallelstreams.NqBenchmark.sequential":
127480.495 ±(99.9%) 10673.912 us/op [Average]
(min, avg, max) = (124313.179, 127480.495, 131880.637), stdev = 2771.981
CI (99.9%): [116806.582, 138154.407] (assumes normal distribution)
# Run complete. Total time: 00:01:39
REMEMBER: The numbers below are just data. To gain reusable insights, you need to follow up on
why the numbers are the way they are. Use profilers (see -prof, -lprof), design factorial
experiments, perform baseline and negative tests that provide experimental control, make sure
the benchmarking environment is safe on JVM/OS/HW level, ask for reviews from the domain experts.
Do not assume the numbers tell you what you want them to tell.
NOTE: Current JVM experimentally supports Compiler Blackholes, and they are in use. Please exercise
extra caution when trusting the results, look into the generated code to check the benchmark still
works, and factor in a small probability of new VM bugs. Additionally, while comparisons between
different JVMs are already problematic, the performance difference caused by different Blackhole
modes can be very significant. Please make sure you use the consistent Blackhole mode for comparisons.
Benchmark (n) (q) Mode Cnt Score Error Units
NqBenchmark.parallel 100 1 avgt 5 3.227 ± 0.227 us/op
NqBenchmark.parallel 100 100 avgt 5 17.024 ± 5.796 us/op
NqBenchmark.parallel 10000 1 avgt 5 45.297 ± 1.633 us/op
NqBenchmark.parallel 10000 100 avgt 5 688.996 ± 41.521 us/op
NqBenchmark.parallel 1000000 1 avgt 5 1952.430 ± 781.640 us/op
NqBenchmark.parallel 1000000 100 avgt 5 66180.327 ± 5356.241 us/op
NqBenchmark.sequential 100 1 avgt 5 0.345 ± 0.020 us/op
NqBenchmark.sequential 100 100 avgt 5 12.986 ± 0.719 us/op
NqBenchmark.sequential 10000 1 avgt 5 32.956 ± 8.808 us/op
NqBenchmark.sequential 10000 100 avgt 5 1305.048 ± 48.775 us/op
NqBenchmark.sequential 1000000 1 avgt 5 3172.715 ± 886.464 us/op
NqBenchmark.sequential 1000000 100 avgt 5 127480.495 ± 10673.912 us/op
@@ -0,0 +1,263 @@
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.arrayList_par
# Run progress: 0.00% complete, ETA 00:01:04
# Fork: 1 of 1
# Warmup Iteration 1: 875.141 us/op
# Warmup Iteration 2: 1000.046 us/op
# Warmup Iteration 3: 887.776 us/op
Iteration 1: 892.010 us/op
Iteration 2: 897.553 us/op
Iteration 3: 799.533 us/op
Iteration 4: 810.756 us/op
Iteration 5: 923.902 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.arrayList_par":
864.751 ±(99.9%) 215.142 us/op [Average]
(min, avg, max) = (799.533, 864.751, 923.902), stdev = 55.872
CI (99.9%): [649.608, 1079.893] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.arrayList_seq
# Run progress: 12.50% complete, ETA 00:00:57
# Fork: 1 of 1
# Warmup Iteration 1: 1707.637 us/op
# Warmup Iteration 2: 1671.232 us/op
# Warmup Iteration 3: 1639.488 us/op
Iteration 1: 1646.433 us/op
Iteration 2: 1587.087 us/op
Iteration 3: 1617.898 us/op
Iteration 4: 1563.128 us/op
Iteration 5: 1565.962 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.arrayList_seq":
1596.102 ±(99.9%) 137.297 us/op [Average]
(min, avg, max) = (1563.128, 1596.102, 1646.433), stdev = 35.656
CI (99.9%): [1458.805, 1733.399] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.intArray_par
# Run progress: 25.00% complete, ETA 00:00:49
# Fork: 1 of 1
# Warmup Iteration 1: 844.636 us/op
# Warmup Iteration 2: 812.772 us/op
# Warmup Iteration 3: 799.070 us/op
Iteration 1: 934.718 us/op
Iteration 2: 807.934 us/op
Iteration 3: 833.626 us/op
Iteration 4: 805.721 us/op
Iteration 5: 808.579 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.intArray_par":
838.116 ±(99.9%) 212.527 us/op [Average]
(min, avg, max) = (805.721, 838.116, 934.718), stdev = 55.192
CI (99.9%): [625.589, 1050.642] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.intArray_seq
# Run progress: 37.50% complete, ETA 00:00:41
# Fork: 1 of 1
# Warmup Iteration 1: 1557.406 us/op
# Warmup Iteration 2: 1627.566 us/op
# Warmup Iteration 3: 1530.733 us/op
Iteration 1: 1672.832 us/op
Iteration 2: 1485.387 us/op
Iteration 3: 1511.766 us/op
Iteration 4: 1480.313 us/op
Iteration 5: 1506.690 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.intArray_seq":
1531.397 ±(99.9%) 308.809 us/op [Average]
(min, avg, max) = (1480.313, 1531.397, 1672.832), stdev = 80.197
CI (99.9%): [1222.588, 1840.206] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.iterate_par
# Run progress: 50.00% complete, ETA 00:00:32
# Fork: 1 of 1
# Warmup Iteration 1: 3539.271 us/op
# Warmup Iteration 2: 3281.146 us/op
# Warmup Iteration 3: 2699.139 us/op
Iteration 1: 2826.627 us/op
Iteration 2: 2769.716 us/op
Iteration 3: 2790.944 us/op
Iteration 4: 2817.520 us/op
Iteration 5: 2756.321 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.iterate_par":
2792.226 ±(99.9%) 115.852 us/op [Average]
(min, avg, max) = (2756.321, 2792.226, 2826.627), stdev = 30.086
CI (99.9%): [2676.374, 2908.078] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.iterate_seq
# Run progress: 62.50% complete, ETA 00:00:24
# Fork: 1 of 1
# Warmup Iteration 1: 2122.584 us/op
# Warmup Iteration 2: 2051.970 us/op
# Warmup Iteration 3: 1984.262 us/op
Iteration 1: 1958.438 us/op
Iteration 2: 1903.397 us/op
Iteration 3: 1902.650 us/op
Iteration 4: 2086.336 us/op
Iteration 5: 1962.841 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.iterate_seq":
1962.732 ±(99.9%) 288.329 us/op [Average]
(min, avg, max) = (1902.650, 1962.732, 2086.336), stdev = 74.878
CI (99.9%): [1674.404, 2251.061] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.linkedList_par
# Run progress: 75.00% complete, ETA 00:00:16
# Fork: 1 of 1
# Warmup Iteration 1: 1723.307 us/op
# Warmup Iteration 2: 1523.676 us/op
# Warmup Iteration 3: 1797.511 us/op
Iteration 1: 1736.234 us/op
Iteration 2: 1713.686 us/op
Iteration 3: 1867.840 us/op
Iteration 4: 1627.651 us/op
Iteration 5: 1840.748 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.linkedList_par":
1757.232 ±(99.9%) 376.977 us/op [Average]
(min, avg, max) = (1627.651, 1757.232, 1867.840), stdev = 97.900
CI (99.9%): [1380.255, 2134.208] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.SourceBenchmark.linkedList_seq
# Run progress: 87.50% complete, ETA 00:00:08
# Fork: 1 of 1
# Warmup Iteration 1: 1624.775 us/op
# Warmup Iteration 2: 1606.300 us/op
# Warmup Iteration 3: 1722.162 us/op
Iteration 1: 1703.266 us/op
Iteration 2: 1687.384 us/op
Iteration 3: 1581.179 us/op
Iteration 4: 1695.401 us/op
Iteration 5: 1660.109 us/op
Result "com.ankurm.parallelstreams.SourceBenchmark.linkedList_seq":
1665.468 ±(99.9%) 191.938 us/op [Average]
(min, avg, max) = (1581.179, 1665.468, 1703.266), stdev = 49.846
CI (99.9%): [1473.530, 1857.405] (assumes normal distribution)
# Run complete. Total time: 00:01:05
REMEMBER: The numbers below are just data. To gain reusable insights, you need to follow up on
why the numbers are the way they are. Use profilers (see -prof, -lprof), design factorial
experiments, perform baseline and negative tests that provide experimental control, make sure
the benchmarking environment is safe on JVM/OS/HW level, ask for reviews from the domain experts.
Do not assume the numbers tell you what you want them to tell.
NOTE: Current JVM experimentally supports Compiler Blackholes, and they are in use. Please exercise
extra caution when trusting the results, look into the generated code to check the benchmark still
works, and factor in a small probability of new VM bugs. Additionally, while comparisons between
different JVMs are already problematic, the performance difference caused by different Blackhole
modes can be very significant. Please make sure you use the consistent Blackhole mode for comparisons.
Benchmark Mode Cnt Score Error Units
SourceBenchmark.arrayList_par avgt 5 864.751 ± 215.142 us/op
SourceBenchmark.arrayList_seq avgt 5 1596.102 ± 137.297 us/op
SourceBenchmark.intArray_par avgt 5 838.116 ± 212.527 us/op
SourceBenchmark.intArray_seq avgt 5 1531.397 ± 308.809 us/op
SourceBenchmark.iterate_par avgt 5 2792.226 ± 115.852 us/op
SourceBenchmark.iterate_seq avgt 5 1962.732 ± 288.329 us/op
SourceBenchmark.linkedList_par avgt 5 1757.232 ± 376.977 us/op
SourceBenchmark.linkedList_seq avgt 5 1665.468 ± 191.938 us/op
@@ -0,0 +1,283 @@
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.BoxingBenchmark.boxed_par
# Run progress: 0.00% complete, ETA 00:00:32
# Fork: 1 of 1
# Warmup Iteration 1: 974.986 us/op
# Warmup Iteration 2: 848.792 us/op
# Warmup Iteration 3: 796.211 us/op
Iteration 1: 870.351 us/op
gc.alloc.rate: 1.213 MB/sec
gc.alloc.rate.norm: 1107.770 B/op
gc.count: ≈ 0 counts
Iteration 2: 824.111 us/op
gc.alloc.rate: 1.281 MB/sec
gc.alloc.rate.norm: 1108.642 B/op
gc.count: ≈ 0 counts
Iteration 3: 845.852 us/op
gc.alloc.rate: 1.243 MB/sec
gc.alloc.rate.norm: 1107.741 B/op
gc.count: ≈ 0 counts
Iteration 4: 861.177 us/op
gc.alloc.rate: 1.227 MB/sec
gc.alloc.rate.norm: 1109.962 B/op
gc.count: ≈ 0 counts
Iteration 5: 886.850 us/op
gc.alloc.rate: 1.190 MB/sec
gc.alloc.rate.norm: 1106.539 B/op
gc.count: ≈ 0 counts
Result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_par":
857.668 ±(99.9%) 92.144 us/op [Average]
(min, avg, max) = (824.111, 857.668, 886.850), stdev = 23.930
CI (99.9%): [765.524, 949.812] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_par:gc.alloc.rate":
1.231 ±(99.9%) 0.132 MB/sec [Average]
(min, avg, max) = (1.190, 1.231, 1.281), stdev = 0.034
CI (99.9%): [1.099, 1.362] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_par:gc.alloc.rate.norm":
1108.131 ±(99.9%) 4.882 B/op [Average]
(min, avg, max) = (1106.539, 1108.131, 1109.962), stdev = 1.268
CI (99.9%): [1103.249, 1113.013] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_par:gc.count":
≈ 0 counts
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.BoxingBenchmark.boxed_seq
# Run progress: 25.00% complete, ETA 00:00:24
# Fork: 1 of 1
# Warmup Iteration 1: 1431.858 us/op
# Warmup Iteration 2: 1497.100 us/op
# Warmup Iteration 3: 1432.571 us/op
Iteration 1: 1389.004 us/op
gc.alloc.rate: 0.232 MB/sec
gc.alloc.rate.norm: 337.709 B/op
gc.count: ≈ 0 counts
Iteration 2: 1421.737 us/op
gc.alloc.rate: 0.227 MB/sec
gc.alloc.rate.norm: 337.943 B/op
gc.count: ≈ 0 counts
Iteration 3: 1439.071 us/op
gc.alloc.rate: 0.224 MB/sec
gc.alloc.rate.norm: 338.029 B/op
gc.count: ≈ 0 counts
Iteration 4: 1446.336 us/op
gc.alloc.rate: 0.223 MB/sec
gc.alloc.rate.norm: 338.116 B/op
gc.count: ≈ 0 counts
Iteration 5: 1482.209 us/op
gc.alloc.rate: 0.218 MB/sec
gc.alloc.rate.norm: 338.584 B/op
gc.count: ≈ 0 counts
Result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_seq":
1435.671 ±(99.9%) 131.474 us/op [Average]
(min, avg, max) = (1389.004, 1435.671, 1482.209), stdev = 34.143
CI (99.9%): [1304.197, 1567.146] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_seq:gc.alloc.rate":
0.224 ±(99.9%) 0.020 MB/sec [Average]
(min, avg, max) = (0.218, 0.224, 0.232), stdev = 0.005
CI (99.9%): [0.205, 0.244] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_seq:gc.alloc.rate.norm":
338.076 ±(99.9%) 1.239 B/op [Average]
(min, avg, max) = (337.709, 338.076, 338.584), stdev = 0.322
CI (99.9%): [336.837, 339.315] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.boxed_seq:gc.count":
≈ 0 counts
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.BoxingBenchmark.primitive_par
# Run progress: 50.00% complete, ETA 00:00:16
# Fork: 1 of 1
# Warmup Iteration 1: 205.937 us/op
# Warmup Iteration 2: 170.623 us/op
# Warmup Iteration 3: 162.141 us/op
Iteration 1: 161.161 us/op
gc.alloc.rate: 5.462 MB/sec
gc.alloc.rate.norm: 923.506 B/op
gc.count: ≈ 0 counts
Iteration 2: 164.425 us/op
gc.alloc.rate: 5.369 MB/sec
gc.alloc.rate.norm: 925.919 B/op
gc.count: ≈ 0 counts
Iteration 3: 166.830 us/op
gc.alloc.rate: 5.271 MB/sec
gc.alloc.rate.norm: 922.734 B/op
gc.count: ≈ 0 counts
Iteration 4: 155.618 us/op
gc.alloc.rate: 5.651 MB/sec
gc.alloc.rate.norm: 922.257 B/op
gc.count: ≈ 0 counts
Iteration 5: 153.176 us/op
gc.alloc.rate: 5.726 MB/sec
gc.alloc.rate.norm: 919.993 B/op
gc.count: ≈ 0 counts
Result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_par":
160.242 ±(99.9%) 22.209 us/op [Average]
(min, avg, max) = (153.176, 160.242, 166.830), stdev = 5.768
CI (99.9%): [138.033, 182.451] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_par:gc.alloc.rate":
5.496 ±(99.9%) 0.733 MB/sec [Average]
(min, avg, max) = (5.271, 5.496, 5.726), stdev = 0.190
CI (99.9%): [4.763, 6.229] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_par:gc.alloc.rate.norm":
922.882 ±(99.9%) 8.253 B/op [Average]
(min, avg, max) = (919.993, 922.882, 925.919), stdev = 2.143
CI (99.9%): [914.629, 931.135] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_par:gc.count":
≈ 0 counts
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.BoxingBenchmark.primitive_seq
# Run progress: 75.00% complete, ETA 00:00:08
# Fork: 1 of 1
# Warmup Iteration 1: 273.688 us/op
# Warmup Iteration 2: 265.275 us/op
# Warmup Iteration 3: 264.934 us/op
Iteration 1: 274.675 us/op
gc.alloc.rate: 0.645 MB/sec
gc.alloc.rate.norm: 185.921 B/op
gc.count: ≈ 0 counts
Iteration 2: 271.925 us/op
gc.alloc.rate: 0.651 MB/sec
gc.alloc.rate.norm: 185.902 B/op
gc.count: ≈ 0 counts
Iteration 3: 269.727 us/op
gc.alloc.rate: 0.655 MB/sec
gc.alloc.rate.norm: 185.887 B/op
gc.count: ≈ 0 counts
Iteration 4: 266.509 us/op
gc.alloc.rate: 0.665 MB/sec
gc.alloc.rate.norm: 185.865 B/op
gc.count: ≈ 0 counts
Iteration 5: 265.184 us/op
gc.alloc.rate: 0.668 MB/sec
gc.alloc.rate.norm: 185.847 B/op
gc.count: ≈ 0 counts
Result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_seq":
269.604 ±(99.9%) 14.943 us/op [Average]
(min, avg, max) = (265.184, 269.604, 274.675), stdev = 3.881
CI (99.9%): [254.660, 284.547] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_seq:gc.alloc.rate":
0.657 ±(99.9%) 0.036 MB/sec [Average]
(min, avg, max) = (0.645, 0.657, 0.668), stdev = 0.009
CI (99.9%): [0.621, 0.693] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_seq:gc.alloc.rate.norm":
185.884 ±(99.9%) 0.113 B/op [Average]
(min, avg, max) = (185.847, 185.884, 185.921), stdev = 0.029
CI (99.9%): [185.771, 185.997] (assumes normal distribution)
Secondary result "com.ankurm.parallelstreams.BoxingBenchmark.primitive_seq:gc.count":
≈ 0 counts
# Run complete. Total time: 00:00:33
REMEMBER: The numbers below are just data. To gain reusable insights, you need to follow up on
why the numbers are the way they are. Use profilers (see -prof, -lprof), design factorial
experiments, perform baseline and negative tests that provide experimental control, make sure
the benchmarking environment is safe on JVM/OS/HW level, ask for reviews from the domain experts.
Do not assume the numbers tell you what you want them to tell.
NOTE: Current JVM experimentally supports Compiler Blackholes, and they are in use. Please exercise
extra caution when trusting the results, look into the generated code to check the benchmark still
works, and factor in a small probability of new VM bugs. Additionally, while comparisons between
different JVMs are already problematic, the performance difference caused by different Blackhole
modes can be very significant. Please make sure you use the consistent Blackhole mode for comparisons.
Benchmark Mode Cnt Score Error Units
BoxingBenchmark.boxed_par avgt 5 857.668 ± 92.144 us/op
BoxingBenchmark.boxed_par:gc.alloc.rate avgt 5 1.231 ± 0.132 MB/sec
BoxingBenchmark.boxed_par:gc.alloc.rate.norm avgt 5 1108.131 ± 4.882 B/op
BoxingBenchmark.boxed_par:gc.count avgt 5 ≈ 0 counts
BoxingBenchmark.boxed_seq avgt 5 1435.671 ± 131.474 us/op
BoxingBenchmark.boxed_seq:gc.alloc.rate avgt 5 0.224 ± 0.020 MB/sec
BoxingBenchmark.boxed_seq:gc.alloc.rate.norm avgt 5 338.076 ± 1.239 B/op
BoxingBenchmark.boxed_seq:gc.count avgt 5 ≈ 0 counts
BoxingBenchmark.primitive_par avgt 5 160.242 ± 22.209 us/op
BoxingBenchmark.primitive_par:gc.alloc.rate avgt 5 5.496 ± 0.733 MB/sec
BoxingBenchmark.primitive_par:gc.alloc.rate.norm avgt 5 922.882 ± 8.253 B/op
BoxingBenchmark.primitive_par:gc.count avgt 5 ≈ 0 counts
BoxingBenchmark.primitive_seq avgt 5 269.604 ± 14.943 us/op
BoxingBenchmark.primitive_seq:gc.alloc.rate avgt 5 0.657 ± 0.036 MB/sec
BoxingBenchmark.primitive_seq:gc.alloc.rate.norm avgt 5 185.884 ± 0.113 B/op
BoxingBenchmark.primitive_seq:gc.count avgt 5 ≈ 0 counts
@@ -0,0 +1,201 @@
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.OrderedOpsBenchmark.distinct_par_ordered
# Run progress: 0.00% complete, ETA 00:00:48
# Fork: 1 of 1
# Warmup Iteration 1: 8704.466 us/op
# Warmup Iteration 2: 9269.009 us/op
# Warmup Iteration 3: 9734.713 us/op
Iteration 1: 9733.449 us/op
Iteration 2: 9639.780 us/op
Iteration 3: 8957.786 us/op
Iteration 4: 11247.905 us/op
Iteration 5: 9896.392 us/op
Result "com.ankurm.parallelstreams.OrderedOpsBenchmark.distinct_par_ordered":
9895.062 ±(99.9%) 3221.657 us/op [Average]
(min, avg, max) = (8957.786, 9895.062, 11247.905), stdev = 836.654
CI (99.9%): [6673.405, 13116.719] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.OrderedOpsBenchmark.distinct_par_unordered
# Run progress: 16.67% complete, ETA 00:00:41
# Fork: 1 of 1
# Warmup Iteration 1: 13742.908 us/op
# Warmup Iteration 2: 7226.162 us/op
# Warmup Iteration 3: 5339.945 us/op
Iteration 1: 6457.032 us/op
Iteration 2: 7957.638 us/op
Iteration 3: 5756.201 us/op
Iteration 4: 6078.999 us/op
Iteration 5: 5919.050 us/op
Result "com.ankurm.parallelstreams.OrderedOpsBenchmark.distinct_par_unordered":
6433.784 ±(99.9%) 3429.350 us/op [Average]
(min, avg, max) = (5756.201, 6433.784, 7957.638), stdev = 890.591
CI (99.9%): [3004.434, 9863.134] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.OrderedOpsBenchmark.distinct_seq
# Run progress: 33.33% complete, ETA 00:00:33
# Fork: 1 of 1
# Warmup Iteration 1: 4882.585 us/op
# Warmup Iteration 2: 4988.734 us/op
# Warmup Iteration 3: 4556.600 us/op
Iteration 1: 5324.242 us/op
Iteration 2: 5486.219 us/op
Iteration 3: 4496.437 us/op
Iteration 4: 4425.314 us/op
Iteration 5: 4358.564 us/op
Result "com.ankurm.parallelstreams.OrderedOpsBenchmark.distinct_seq":
4818.155 ±(99.9%) 2083.875 us/op [Average]
(min, avg, max) = (4358.564, 4818.155, 5486.219), stdev = 541.176
CI (99.9%): [2734.281, 6902.030] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.OrderedOpsBenchmark.limit_par_ordered
# Run progress: 50.00% complete, ETA 00:00:24
# Fork: 1 of 1
# Warmup Iteration 1: 258.706 us/op
# Warmup Iteration 2: 228.867 us/op
# Warmup Iteration 3: 218.050 us/op
Iteration 1: 212.607 us/op
Iteration 2: 215.389 us/op
Iteration 3: 226.594 us/op
Iteration 4: 210.350 us/op
Iteration 5: 210.371 us/op
Result "com.ankurm.parallelstreams.OrderedOpsBenchmark.limit_par_ordered":
215.062 ±(99.9%) 26.069 us/op [Average]
(min, avg, max) = (210.350, 215.062, 226.594), stdev = 6.770
CI (99.9%): [188.993, 241.131] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.OrderedOpsBenchmark.limit_par_unordered
# Run progress: 66.67% complete, ETA 00:00:16
# Fork: 1 of 1
# Warmup Iteration 1: 80.165 us/op
# Warmup Iteration 2: 57.729 us/op
# Warmup Iteration 3: 57.777 us/op
Iteration 1: 55.429 us/op
Iteration 2: 54.038 us/op
Iteration 3: 59.503 us/op
Iteration 4: 57.574 us/op
Iteration 5: 56.395 us/op
Result "com.ankurm.parallelstreams.OrderedOpsBenchmark.limit_par_unordered":
56.588 ±(99.9%) 8.020 us/op [Average]
(min, avg, max) = (54.038, 56.588, 59.503), stdev = 2.083
CI (99.9%): [48.568, 64.608] (assumes normal distribution)
# JMH version: 1.37
# VM version: JDK 25.0.5, OpenJDK 64-Bit Server VM, 25.0.5+8-LTS
# VM invoker: /tmp/tools/jdk-25.0.5+8/bin/java
# VM options: -Djavax.net.ssl.trustStore=/etc/ssl/certs/java/cacerts -Djavax.net.ssl.trustStoreType=JKS -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=42575 -Dhttp.nonProxyHosts=localhost|127.0.0.1|::1|127.*|0.*|::|169.254.*|api.anthropic.com|api-staging.anthropic.com|api-pr-preview.anthropic.com|mcp-proxy.anthropic.com|mcp-proxy-staging.anthropic.com|registry.npmjs.org|jsr.io|npm.jsr.io|pypi.org|files.pythonhosted.org|index.crates.io|proxy.golang.org|host.docker.internal|10.*|172.16.*|172.17.*|172.18.*|172.19.*|172.20.*|172.21.*|172.22.*|172.23.*|172.24.*|172.25.*|172.26.*|172.27.*|172.28.*|172.29.*|172.30.*|172.31.*|192.168.*|100.64.0.0/10|*.svc.cluster.local|*.svc.cluster.local -Djdk.http.auth.tunneling.disabledSchemes= -Djdk.http.auth.proxying.disabledSchemes=
# Blackhole mode: compiler (auto-detected, use -Djmh.blackhole.autoDetect=false to disable)
# Warmup: 3 iterations, 1 s each
# Measurement: 5 iterations, 1 s each
# Timeout: 10 min per iteration
# Threads: 1 thread, will synchronize iterations
# Benchmark mode: Average time, time/op
# Benchmark: com.ankurm.parallelstreams.OrderedOpsBenchmark.limit_seq
# Run progress: 83.33% complete, ETA 00:00:08
# Fork: 1 of 1
# Warmup Iteration 1: 63.583 us/op
# Warmup Iteration 2: 68.744 us/op
# Warmup Iteration 3: 61.862 us/op
Iteration 1: 60.503 us/op
Iteration 2: 37.577 us/op
Iteration 3: 38.935 us/op
Iteration 4: 37.643 us/op
Iteration 5: 36.974 us/op
Result "com.ankurm.parallelstreams.OrderedOpsBenchmark.limit_seq":
42.326 ±(99.9%) 39.223 us/op [Average]
(min, avg, max) = (36.974, 42.326, 60.503), stdev = 10.186
CI (99.9%): [3.103, 81.549] (assumes normal distribution)
# Run complete. Total time: 00:00:49
REMEMBER: The numbers below are just data. To gain reusable insights, you need to follow up on
why the numbers are the way they are. Use profilers (see -prof, -lprof), design factorial
experiments, perform baseline and negative tests that provide experimental control, make sure
the benchmarking environment is safe on JVM/OS/HW level, ask for reviews from the domain experts.
Do not assume the numbers tell you what you want them to tell.
NOTE: Current JVM experimentally supports Compiler Blackholes, and they are in use. Please exercise
extra caution when trusting the results, look into the generated code to check the benchmark still
works, and factor in a small probability of new VM bugs. Additionally, while comparisons between
different JVMs are already problematic, the performance difference caused by different Blackhole
modes can be very significant. Please make sure you use the consistent Blackhole mode for comparisons.
Benchmark Mode Cnt Score Error Units
OrderedOpsBenchmark.distinct_par_ordered avgt 5 9895.062 ± 3221.657 us/op
OrderedOpsBenchmark.distinct_par_unordered avgt 5 6433.784 ± 3429.350 us/op
OrderedOpsBenchmark.distinct_seq avgt 5 4818.155 ± 2083.875 us/op
OrderedOpsBenchmark.limit_par_ordered avgt 5 215.062 ± 26.069 us/op
OrderedOpsBenchmark.limit_par_unordered avgt 5 56.588 ± 8.020 us/op
OrderedOpsBenchmark.limit_seq avgt 5 42.326 ± 39.223 us/op
+4
View File
@@ -0,0 +1,4 @@
[INFO] Running com.ankurm.parallelstreams.ParallelStreamsTest
[INFO] Tests run: 5, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.115 s -- in com.ankurm.parallelstreams.ParallelStreamsTest
[INFO] Tests run: 5, Failures: 0, Errors: 0, Skipped: 0
[INFO] BUILD SUCCESS
+87
View File
@@ -0,0 +1,87 @@
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<parent>
<groupId>com.ankurm</groupId>
<artifactId>java-core-examples</artifactId>
<version>1.0</version>
</parent>
<artifactId>parallel-streams</artifactId>
<name>parallel-streams</name>
<description>When parallel streams are slower: JMH benchmarks for the N*Q model, source splitting, boxing and ordered operations, plus runnable demos of common-pool starvation and parallel-stream correctness bugs.</description>
<properties>
<jmh.version>1.37</jmh.version>
</properties>
<dependencies>
<dependency>
<groupId>org.openjdk.jmh</groupId>
<artifactId>jmh-core</artifactId>
<version>${jmh.version}</version>
</dependency>
<dependency>
<groupId>org.openjdk.jmh</groupId>
<artifactId>jmh-generator-annprocess</artifactId>
<version>${jmh.version}</version>
</dependency>
<dependency>
<groupId>org.junit.jupiter</groupId>
<artifactId>junit-jupiter</artifactId>
<version>5.11.0</version>
<scope>test</scope>
</dependency>
</dependencies>
<build>
<finalName>benchmarks</finalName>
<plugins>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-compiler-plugin</artifactId>
<version>3.13.0</version>
<configuration>
<release>25</release>
<annotationProcessorPaths>
<path>
<groupId>org.openjdk.jmh</groupId>
<artifactId>jmh-generator-annprocess</artifactId>
<version>${jmh.version}</version>
</path>
</annotationProcessorPaths>
</configuration>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-surefire-plugin</artifactId>
<version>3.2.5</version>
<configuration>
<!-- Pin the common pool so the tests behave the same on a 2-vCPU laptop and a 64-core server. -->
<argLine>-Djava.util.concurrent.ForkJoinPool.common.parallelism=3</argLine>
</configuration>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-shade-plugin</artifactId>
<version>3.5.1</version>
<executions>
<execution>
<phase>package</phase>
<goals><goal>shade</goal></goals>
<configuration>
<transformers>
<transformer implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer">
<mainClass>org.openjdk.jmh.Main</mainClass>
</transformer>
</transformers>
</configuration>
</execution>
</executions>
</plugin>
</plugins>
</build>
</project>
+23
View File
@@ -0,0 +1,23 @@
#!/usr/bin/env bash
# Regenerates every file in output/. About 25 minutes on 2 vCPUs, almost all of it JMH.
set -euo pipefail
cd "$(dirname "$0")/.."
mkdir -p output
mvn -q -B package -DskipTests
JMH="java -jar target/benchmarks.jar -f 1 -wi 3 -w 1s -i 5 -r 1s"
echo "availableProcessors = $(java -cp target/classes com.ankurm.parallelstreams.CpuInfo 2>/dev/null)" > output/00-environment.txt
java -version 2>> output/00-environment.txt || true
java -Djava.util.concurrent.ForkJoinPool.common.parallelism=3 -cp target/classes \
com.ankurm.parallelstreams.CorrectnessBugsDemo > output/01-correctness-bugs.txt
java -Djava.util.concurrent.ForkJoinPool.common.parallelism=3 -cp target/classes \
com.ankurm.parallelstreams.CommonPoolStarvationDemo > output/02-common-pool-starvation.txt
$JMH NqBenchmark > output/03-nq-benchmark.txt
$JMH SourceBenchmark > output/04-source-benchmark.txt
$JMH BoxingBenchmark -prof gc > output/05-boxing-benchmark.txt
$JMH OrderedOpsBenchmark > output/06-ordered-ops-benchmark.txt
mvn -B test | grep -E 'Tests run|BUILD|ParallelStreamsTest' > output/07-tests.txt
# Strip JVM environment noise (proxy banner, Unsafe deprecation warnings) from the committed files.
sed -i '/^Picked up JAVA_TOOL_OPTIONS/d;/^WARNING:/d' output/*.txt
+6
View File
@@ -0,0 +1,6 @@
#!/usr/bin/env bash
# Run one demo class ad hoc, e.g.: ./scripts/run.sh CorrectnessBugsDemo
set -euo pipefail
cd "$(dirname "$0")/.."
mvn -q -B package -DskipTests
java -Djava.util.concurrent.ForkJoinPool.common.parallelism=3 -cp target/classes "com.ankurm.parallelstreams.$1"
@@ -0,0 +1,25 @@
package com.ankurm.parallelstreams;
import org.openjdk.jmh.annotations.*;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
/**
* A cheap operation (adding numbers) over 1,000,000 elements, with and without boxing. The
* question is how much the boxed pipeline costs per element and whether parallel rescues it.
* Run with {@code -prof gc} to see the allocation per call.
*/
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
@State(Scope.Benchmark)
public class BoxingBenchmark {
static final int N = 1_000_000;
@Benchmark public long primitive_seq() { return IntStream.range(0, N).asLongStream().sum(); }
@Benchmark public long primitive_par() { return IntStream.range(0, N).parallel().asLongStream().sum(); }
@Benchmark public long boxed_seq() { return IntStream.range(0, N).boxed().mapToLong(Integer::longValue).sum(); }
@Benchmark public long boxed_par() { return IntStream.range(0, N).boxed().parallel().mapToLong(Integer::longValue).sum(); }
}
@@ -0,0 +1,79 @@
package com.ankurm.parallelstreams;
import java.util.concurrent.ForkJoinPool;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
/**
* Every parallel stream in the JVM shares ONE pool: ForkJoinPool.commonPool(). Block its threads
* and every other parallel stream (and CompletableFuture.supplyAsync without an executor) slows
* down with them.
*
* <p>Run with {@code -Djava.util.concurrent.ForkJoinPool.common.parallelism=3} so the numbers are
* the same on any machine. The tasks sleep instead of burning CPU, which stands in for a blocking
* call (HTTP, JDBC) and keeps the result independent of how many cores the host has.
*/
public class CommonPoolStarvationDemo {
static void sleep(long ms) {
try { Thread.sleep(ms); } catch (InterruptedException e) { Thread.currentThread().interrupt(); }
}
/** The "victim": 16 short tasks of 50 ms. Ideal with 4 threads (3 workers + caller): about 200 ms. */
static long victimMillis() {
long t0 = System.nanoTime();
IntStream.range(0, 16).parallel().forEach(i -> sleep(50));
return (System.nanoTime() - t0) / 1_000_000;
}
/** The "noisy neighbour": 8 tasks that each block for 2 seconds. */
static Thread startNeighbour(boolean ownPool) {
Thread t = new Thread(() -> {
if (ownPool) {
try (ForkJoinPool pool = new ForkJoinPool(3)) {
pool.submit(() -> IntStream.range(0, 8).parallel().forEach(i -> sleep(2_000))).join();
}
} else {
IntStream.range(0, 8).parallel().forEach(i -> sleep(2_000));
}
}, "neighbour");
t.start();
return t;
}
public static void main(String[] args) throws Exception {
System.out.println("availableProcessors = " + Runtime.getRuntime().availableProcessors());
System.out.println("commonPool parallelism = " + ForkJoinPool.getCommonPoolParallelism());
victimMillis(); // warm up the pool
System.out.println();
System.out.println("1. Victim alone");
System.out.println(" victim took " + victimMillis() + " ms");
System.out.println();
System.out.println("2. Victim while a neighbour BLOCKS the common pool (8 tasks x 2000 ms)");
Thread n = startNeighbour(false);
sleep(200);
System.out.println(" victim took " + victimMillis() + " ms");
n.join();
System.out.println();
System.out.println("3. Same neighbour, but it runs in its OWN ForkJoinPool(3)");
n = startNeighbour(true);
sleep(200);
System.out.println(" victim took " + victimMillis() + " ms");
n.join();
System.out.println();
System.out.println("4. 20 blocking tasks x 200 ms: parallel stream vs virtual threads");
long t0 = System.nanoTime();
IntStream.range(0, 20).parallel().forEach(i -> sleep(200));
System.out.println(" parallel stream : " + (System.nanoTime() - t0) / 1_000_000 + " ms");
t0 = System.nanoTime();
try (var ex = Executors.newVirtualThreadPerTaskExecutor()) {
for (int i = 0; i < 20; i++) ex.submit(() -> sleep(200));
}
System.out.println(" virtual threads : " + (System.nanoTime() - t0) / 1_000_000 + " ms");
}
}
@@ -0,0 +1,73 @@
package com.ankurm.parallelstreams;
import java.util.ArrayList;
import java.util.List;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
/** Three ways a parallel stream returns the wrong answer without any exception. */
public class CorrectnessBugsDemo {
/** Bug 1: forEach into a non-thread-safe list. Returns how many elements ended up in it. */
public static int sharedListSize(int n) {
List<Integer> out = new ArrayList<>();
try {
IntStream.range(0, n).parallel().forEach(out::add);
} catch (RuntimeException e) {
return -1; // ArrayIndexOutOfBoundsException from a racing resize
}
return out.size();
}
/** The fix: let the stream do the collecting. Always n. */
public static int collectedListSize(int n) {
return IntStream.range(0, n).parallel().boxed().collect(Collectors.toList()).size();
}
/** Bug 2: a "reduce" whose identity is not an identity. Sequential adds the 10 once. */
public static int sumWithBadIdentity(boolean parallel) {
var s = IntStream.rangeClosed(1, 100);
return (parallel ? s.parallel() : s).reduce(10, Integer::sum);
}
/** Bug 3: a non-associative accumulator. Subtraction gives different answers per split. */
public static int subtract(boolean parallel) {
var s = IntStream.rangeClosed(1, 1000);
return (parallel ? s.parallel() : s).reduce(0, (a, b) -> a - b);
}
/** Bug 3b: a stateful lambda. A counter shared by parallel workers is a data race. */
public static int racyCounter(int n) {
int[] counter = {0};
IntStream.range(0, n).parallel().forEach(i -> counter[0]++);
return counter[0];
}
public static int atomicCounter(int n) {
AtomicInteger c = new AtomicInteger();
IntStream.range(0, n).parallel().forEach(i -> c.incrementAndGet());
return c.get();
}
public static void main(String[] args) {
int n = 100_000;
System.out.println("1. forEach(list::add), n = " + n + ", ten runs (-1 = exception):");
StringBuilder sb = new StringBuilder(" sizes: ");
for (int i = 0; i < 10; i++) sb.append(sharedListSize(n)).append(' ');
System.out.println(sb);
System.out.println(" collect(toList()) size: " + collectedListSize(n));
System.out.println();
System.out.println("2. reduce(10, Integer::sum) over 1..100");
System.out.println(" sequential: " + sumWithBadIdentity(false));
System.out.println(" parallel : " + sumWithBadIdentity(true));
System.out.println();
System.out.println("3. reduce(0, (a, b) -> a - b) over 1..1000");
System.out.println(" sequential: " + subtract(false));
System.out.println(" parallel : " + subtract(true));
System.out.println();
System.out.println("4. counter[0]++ in forEach, n = " + n);
System.out.println(" racy : " + racyCounter(n));
System.out.println(" atomic : " + atomicCounter(n));
}
}
@@ -0,0 +1,10 @@
package com.ankurm.parallelstreams;
/** Prints the CPU count and the default common-pool parallelism, for output/00-environment.txt. */
public class CpuInfo {
public static void main(String[] args) {
System.out.println(Runtime.getRuntime().availableProcessors()
+ " (default commonPool parallelism = "
+ java.util.concurrent.ForkJoinPool.getCommonPoolParallelism() + ")");
}
}
@@ -0,0 +1,42 @@
package com.ankurm.parallelstreams;
import org.openjdk.jmh.annotations.*;
import org.openjdk.jmh.infra.Blackhole;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
/**
* The "N x Q" model: N elements, Q units of CPU work per element.
*
* <p>Parallel only wins when N * Q is large enough to pay for splitting, task hand-off and merging.
* {@code Blackhole.consumeCPU(q)} burns a calibrated amount of CPU that the JIT cannot remove, so Q
* is a dial we control. Run with {@code -prof gc} or plain; the interesting output is the
* sequential/parallel pair at each (n, q).
*/
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
@State(Scope.Benchmark)
public class NqBenchmark {
@Param({"100", "10000", "1000000"})
public int n;
@Param({"1", "100"})
public long q;
@Benchmark
public long sequential() {
return IntStream.range(0, n).map(i -> work(i, q)).sum();
}
@Benchmark
public long parallel() {
return IntStream.range(0, n).parallel().map(i -> work(i, q)).sum();
}
static int work(int i, long q) {
Blackhole.consumeCPU(q);
return i & 1;
}
}
@@ -0,0 +1,37 @@
package com.ankurm.parallelstreams;
import org.openjdk.jmh.annotations.*;
import java.util.List;
import java.util.Random;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
/**
* Stateful and order-sensitive operations (limit, distinct) preserve encounter order. In a
* parallel pipeline that costs buffering and coordination. Calling {@code unordered()} tells the
* framework you do not care which elements win, and the cost mostly disappears.
*/
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
@State(Scope.Benchmark)
public class OrderedOpsBenchmark {
static final int N = 200_000;
List<Integer> data;
@Setup
public void setup() {
Random r = new Random(42);
data = IntStream.range(0, N).map(i -> r.nextInt(50_000)).boxed().toList();
}
@Benchmark public long limit_seq() { return data.stream().filter(x -> x % 3 == 0).limit(5_000).count(); }
@Benchmark public long limit_par_ordered() { return data.parallelStream().filter(x -> x % 3 == 0).limit(5_000).count(); }
@Benchmark public long limit_par_unordered() { return data.parallelStream().unordered().filter(x -> x % 3 == 0).limit(5_000).count(); }
@Benchmark public long distinct_seq() { return data.stream().distinct().count(); }
@Benchmark public long distinct_par_ordered() { return data.parallelStream().distinct().count(); }
@Benchmark public long distinct_par_unordered() { return data.parallelStream().unordered().distinct().count(); }
}
@@ -0,0 +1,55 @@
package com.ankurm.parallelstreams;
import org.openjdk.jmh.annotations.*;
import org.openjdk.jmh.infra.Blackhole;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.LinkedList;
import java.util.List;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
import java.util.stream.Stream;
/**
* Same work (n = 100,000 elements, a little CPU each), six different sources. What changes is how
* cheaply the source can be split in half: an array or ArrayList splits by index arithmetic;
* a LinkedList has to be walked; Stream.iterate cannot be split at all, so the stream framework
* copies batches of elements into arrays first.
*/
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
@State(Scope.Benchmark)
public class SourceBenchmark {
static final int N = 100_000;
static final long Q = 20;
int[] array;
List<Integer> arrayList;
List<Integer> linkedList;
@Setup
public void setup() {
array = IntStream.range(0, N).toArray();
arrayList = new ArrayList<>(Arrays.stream(array).boxed().toList());
linkedList = new LinkedList<>(arrayList);
}
static int work(int i) {
Blackhole.consumeCPU(Q);
return i & 1;
}
@Benchmark public long intArray_seq() { return Arrays.stream(array).map(SourceBenchmark::work).sum(); }
@Benchmark public long intArray_par() { return Arrays.stream(array).parallel().map(SourceBenchmark::work).sum(); }
@Benchmark public long arrayList_seq() { return arrayList.stream().mapToInt(SourceBenchmark::work).sum(); }
@Benchmark public long arrayList_par() { return arrayList.parallelStream().mapToInt(SourceBenchmark::work).sum(); }
@Benchmark public long linkedList_seq() { return linkedList.stream().mapToInt(SourceBenchmark::work).sum(); }
@Benchmark public long linkedList_par() { return linkedList.parallelStream().mapToInt(SourceBenchmark::work).sum(); }
@Benchmark public long iterate_seq() { return Stream.iterate(0, i -> i + 1).limit(N).mapToInt(SourceBenchmark::work).sum(); }
@Benchmark public long iterate_par() { return Stream.iterate(0, i -> i + 1).limit(N).parallel().mapToInt(SourceBenchmark::work).sum(); }
}
@@ -0,0 +1,38 @@
package com.ankurm.parallelstreams;
import org.junit.jupiter.api.Test;
import java.util.concurrent.ForkJoinPool;
import static org.junit.jupiter.api.Assertions.*;
/** Pins the deterministic claims only. No timing assertions: those live in output/. */
class ParallelStreamsTest {
@Test
void commonPoolIsPinnedBySurefireArgLine() {
assertEquals(3, ForkJoinPool.getCommonPoolParallelism());
}
@Test
void collectingIsAlwaysCorrect() {
for (int i = 0; i < 20; i++) assertEquals(100_000, CorrectnessBugsDemo.collectedListSize(100_000));
}
@Test
void badIdentityChangesParallelResultButNotSequential() {
assertEquals(5060, CorrectnessBugsDemo.sumWithBadIdentity(false)); // 5050 + one 10
assertNotEquals(5060, CorrectnessBugsDemo.sumWithBadIdentity(true)); // one 10 per leaf
}
@Test
void nonAssociativeSubtractionDiffersBetweenModes() {
assertEquals(-500_500, CorrectnessBugsDemo.subtract(false));
assertNotEquals(-500_500, CorrectnessBugsDemo.subtract(true));
}
@Test
void atomicCounterIsExact() {
assertEquals(100_000, CorrectnessBugsDemo.atomicCounter(100_000));
}
}
+1
View File
@@ -41,6 +41,7 @@
<module>copying</module>
<module>datetimeformatter</module>
<module>bufferedreader</module>
<module>parallel-streams</module>
</modules>
<properties>