Loops vs. Streams Benchmark

A simple micro-benchmark to compare the performance of external iteration methods, such as the for-loop, to streams. The test generates n random numbers which are used for the iterations. Each random number is doubled, and the maximum value is determined.

Built with Maven and JMH. Inspired by "Mastering Lambdas: Java Programming in a Multicore World" by Maurice Naftalin (Oracle Press, 2015).

Executing the Benchmark

Tested with: Maven 3.8.1 and Oracle JDK 16

• Step 1: $ git clone git://github.com/rharri/streams-benchmark.git
• Step 2: $ cd streams-benchmark
• Step 3: $ mvn clean verify
• Step 4: $ java -jar target/benchmarks.jar -t 1

Note: If you encounter heap space errors, try using java -XX:MaxRAMPercentage=50.0 -jar target/benchmarks.jar -t 1 to allow the JVM to allocate more memory. The -XX:MaxRAMPercentage= option accepts a double value representing the desired percentage of total memory to be made available to the JVM.

Test Machine

	MacBook Pro
Operating System	macOS Big Sur
CPU	2.5 GHz Quad-Core Intel Core i7
RAM	16 GB 1600 MHz DDR3
JVM	Java HotSpot(TM) 64-Bit Server VM (build 16+36-2231)

Results

The streamSize represents the number of random numbers generated, and thus number of elements to be iterated by each method. In this case: 1 Million, 10 Million, and 100 Million.

Benchmark	(streamSize)	Mode	Cnt	Score	Error	Units
_getMaxNumberWithLoops	1000000	avgt	6	2.078	± 0.021	ms/op
_getMaxNumberWithLoops	10000000	avgt	6	23.501	± 0.525	ms/op
_getMaxNumberWithLoops	100000000	avgt	6	303.414	± 2.349	ms/op
_getMaxNumberWithParallelStream	1000000	avgt	6	0.887	± 0.006	ms/op
_getMaxNumberWithParallelStream	10000000	avgt	6	8.909	± 0.277	ms/op
_getMaxNumberWithParallelStream	100000000	avgt	6	124.410	± 1.162	ms/op
_getMaxNumberWithStream	1000000	avgt	6	1.482	± 0.004	ms/op
_getMaxNumberWithStream	10000000	avgt	6	15.631	± 0.162	ms/op
_getMaxNumberWithStream	100000000	avgt	6	193.641	± 9.973	ms/op

avgt = Average Time