[manoelcampos]

Related to cloudsimplus/cloudsimplus#586 (PR cloudsimplus/cloudsimplus#593)

Summary

Adds ListenersNotificationBenchmark (new org.cloudsimplus.listeners package) to check if storing listeners in plain arrays, instead of the immutable lists created by the Listeners class, makes notification faster.

A for-each loop over a List calls iterator(), which creates an object, unless the JIT removes it with escape analysis. A for-each loop over an array creates no object. The benchmark should be run with -prof gc to see if the iterator is created (gc.alloc.rate.norm).

• The listeners just increment a counter and all get the same EventInfo, so only the iteration is measured.
• single*: notifies one list (or array) with 0 to 3 listeners. Each param value runs in its own fork, so the loop only sees one List class.
• mixed*: notifies lists with 0 to 3 listeners in each operation, like a simulation with many VMs that have different numbers of listeners. This way, the loop sees both List12 and ListN, as in a real simulation.

Results

java -jar target/cloudsimplus-benchmarks-9.0.0-SNAPSHOT-uber.jar ListenersNotificationBenchmark -f 1 -wi 3 -w 1s -i 5 -r 1s -prof gc (JDK 25)

Benchmark	Listeners	List (ns/op)	Array (ns/op)	List alloc (B/op)	Array alloc (B/op)
single	0	0.86	0.65	0	0
single	1	2.30	2.30	0	0
single	2	4.51	4.60	0	0
single	3	6.76	6.78	0	0
mixed (sizes 0–3)	—	23.2	13.9	128	0

When the loop sees only one list class, escape analysis removes the iterator, and lists and arrays perform the same.
When it sees both classes (the realistic case), an iterator is created for each list (32 B × 4 lists = 128 B/op), and the list version is ~67% slower.

🤖 Generated with Claude Code