Shakespeare plays Reactive Scrabble with Queues

Use JMH to benchmark several queue libraries and their ability to coordinate work on a number of threads.

Queues

Conversant Disruptor: a fast blocking queue
Java 8 streams
JcTools: psy-lob-saw's concurrent data structures
Kilim: fibers, continuations and message passing for java
Quasar: fibers and message passing for java and kotlin
RxJava: reactive extensions for the JVM

Note: the intent of these implementations is that buffering should be limited to simulate a live connection in which real resources are consumed and show the advantages of a reactive approach.

for Java 8 streams, it's not clear to what degree the iterable is buffered
I'm not experienced enough with RxJava to know how much buffering occurs
the other libraries provide obvious buffer sizes (which I'm trusting)

Running

sudo cpupower frequency-set -u 2.8ghz
java -version 2>&1 | head -n1 | grep -q "version\W*1.8" && git checkout quasar7
cp=$(mvn -q dependency:build-classpath -Dmdep.outputFile=/dev/fd/1)
quasar="-javaagent:$(echo $cp | grep -o '[^:]*' | grep quasar-core)"
mvn package
java $quasar -jar target/benchmarks.jar
java $quasar -cp target/classes:$cp direct.ShakespearePlaysScrabbleWithQueues

Notes:

depending on your machine, you may experience thermal throttling. Adjust the first line accordingly.
quasar 0.8 supports java 9 and later, and is incompatible with java 8. this is the default. for java 8, the above checks out the quasar7 tag

Flags

For the direct implementations, some jvm -D flags are accepted.

-Dsuffix=xxx: for words matching this suffix, hash the word and modify the score. try "-Dsuffix=ks"
-Dnh=1000: number of sha-256 hashes to perform. default is 1000
-Dfast: only store the 3 best scores at any time
-Dnp=4: number of cpus to assume. default is number of available cpus
-Dsize=10: the number of bits of buffer to use. default is 10

these flags can be useful for understanding how the implementations perform.

Results

License and History

Apache 2.0 licence

This project is a merged fork of two projects

This project adds several multi-threaded implementations using different queues for communicating between the threads.