CKTIS / Java
Performance and measurement in Java
Performance and measurement in Java: Java performance depends first on algorithmic complexity and data structures, then on allocation, locality, synchronization, GC, and JIT optimization.
Version: Java SE 21 (JLS and Java SE API); examples verified on the repository Java runtime
Syntax
performanceJava performance depends first on algorithmic complexity and data structures, then on allocation, locality, synchronization, GC, and JIT optimization.
Profile a representative workload first, choose collections for the dominant operations, and measure elapsed time as a difference of System.nanoTime values rather than wall-clock time.
A single cold measurement is distorted by JIT compilation, dead-code elimination, and GC; a hand-written microbenchmark may measure code unlike production.
The origin of nanoTime is arbitrary and only differences are meaningful; parallelism can slow a small task, and a better asymptotic algorithm wins only beyond a relevant input size.
For target 31 the counters yield linear=32 versus binary=6, while elapsed-nonnegative=true demonstrates a System.nanoTime difference without machine-dependent timing output.
Examples
public class Main {
static int linearComparisons(int[] values, int target) {
int comparisons = 0;
for (int value : values) {
comparisons++;
if (value == target) break;
}
return comparisons;
}
static int binaryComparisons(int[] values, int target) {
int low = 0, high = values.length - 1, comparisons = 0;
while (low <= high) {
int middle = (low + high) >>> 1;
comparisons++;
if (values[middle] == target) break;
if (values[middle] < target) low = middle + 1;
else high = middle - 1;
}
return comparisons;
}
public static void main(String[] args) {
int[] values = new int[32];
for (int i = 0; i < values.length; i++) values[i] = i;
System.out.println("linear=" + linearComparisons(values, 31));
System.out.println("binary=" + binaryComparisons(values, 31));
long started = System.nanoTime();
long elapsed = System.nanoTime() - started;
System.out.println("elapsed-nonnegative=" + (elapsed >= 0));
}
}
Result
linear=32 binary=6 elapsed-nonnegative=true
For target 31 the counters yield linear=32 versus binary=6, while elapsed-nonnegative=true demonstrates a System.nanoTime difference without machine-dependent timing output.