Benchmark de soma de iteração de arraylist vs stream

Summation.java

package jeffque.bench;

import org.openjdk.jmh.annotations.*;
import org.openjdk.jmh.infra.Blackhole;

import java.util.List;
import java.util.Random;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;

@Warmup(iterations = 2)
@Measurement(iterations = 3, time = 2, timeUnit = TimeUnit.SECONDS)
@Timeout(time = 500, timeUnit = TimeUnit.SECONDS)
@BenchmarkMode(Mode.Throughput)
public class Summation {

    @State(Scope.Benchmark)
    public static class Input {
//        @Param({ "10", "100", "1000", "100000", "10000000", "50000000" })
        @Param({ "100000", "50000000" })
        public int size;

        List<Integer> lista;

        @Setup(Level.Trial)
        public void criaEstado() {
            final var r = new Random(15);
            lista = IntStream.range(0, size)
                    .mapToObj(id -> r.nextInt())
                    .toList();
        }
    }

    @Benchmark
    public long withForIndex(Input i) {
        long result = 0;
        for (int idx = 0; idx < i.lista.size(); idx++) {
            result += i.lista.get(idx);
        }
        return result;
    }

    @Benchmark
    public long withForeach(Input i) {
        long result = 0;
        for (int l: i.lista) {
            result += l;
        }
        return result;
    }

    @Benchmark
    public long withStream(Input i) {
        return i.lista.stream().mapToInt(Integer::intValue).sum();
    }

    @Benchmark
    public void noopForeach(Input i, Blackhole b) {
        for (int l: i.lista) {
            b.consume(l);
        }
    }

    @Benchmark
    public int withForeachInt(Input i) {
        int result = 0;
        for (int l: i.lista) {
            result += l;
        }
        return result;
    }

    // baseado em https://www.baeldung.com/java-streams-vs-loops
    @Benchmark
    public int withStreamAsWrapper(Input i) {
        return i.lista.stream().reduce(0, Integer::sum);
    }
}

Resultados que eu obtive localmente:

Benchmark                 (size)   Mode  Cnt      Score     Error  Units
Summation.noopForeach     100000  thrpt   15  24917.405 ±  75.933  ops/s
Summation.noopForeach   50000000  thrpt   15     38.930 ±   0.397  ops/s
Summation.withForIndex    100000  thrpt   15  15457.297 ± 414.089  ops/s
Summation.withForIndex  50000000  thrpt   15     28.070 ±   0.421  ops/s
Summation.withForeach     100000  thrpt   15  15472.856 ± 196.674  ops/s
Summation.withForeach   50000000  thrpt   15     27.336 ±   1.675  ops/s
Summation.withStream      100000  thrpt   15  19552.178 ± 566.742  ops/s
Summation.withStream    50000000  thrpt   15     35.858 ±   0.419  ops/s

Resultados que obtive localmente após adicionar a função withStreamAsWrapper (para simular o resultado de https://www.baeldung.com/java-streams-vs-loops) e usando int em withForeachInt como tipo acumulador:

Benchmark                        (size)   Mode  Cnt      Score     Error  Units
Summation.noopForeach            100000  thrpt   15  25074.165 ± 145.263  ops/s
Summation.noopForeach          50000000  thrpt   15     39.605 ±   0.493  ops/s
Summation.withForIndex           100000  thrpt   15  15779.702 ±  23.341  ops/s
Summation.withForIndex         50000000  thrpt   15     28.676 ±   0.109  ops/s
Summation.withForeach            100000  thrpt   15  15765.148 ±  21.495  ops/s
Summation.withForeach          50000000  thrpt   15     28.494 ±   0.305  ops/s
Summation.withForeachInt         100000  thrpt   15  24297.451 ± 125.790  ops/s
Summation.withForeachInt       50000000  thrpt   15     38.663 ±   0.238  ops/s
Summation.withStream             100000  thrpt   15  19967.495 ± 113.473  ops/s
Summation.withStream           50000000  thrpt   15     36.249 ±   0.916  ops/s
Summation.withStreamAsWrapper    100000  thrpt   15   3804.950 ±  63.529  ops/s
Summation.withStreamAsWrapper  50000000  thrpt   15      5.544 ±   1.966  ops/s

Dados sobre ambiente de test:

JMH version: 1.37
VM version: JDK 21.0.3, OpenJDK 64-Bit Server VM, 21.0.3+9-LTS

Rodado em um MacBook Pro, macOs 15.6.1, 15 GB de RAM, processador Apple M1 Pro