I am trying to create a little benchmark (in Groovy) that shows high thread contention on a couple of synchronized methods. High contention should show up when monitoring voluntary context switches, and in Linux this can be achieved thanks to "pidstat".
The program is the following:
class Res {
private int n;
synchronized public void inc() {
n++;
def foo = []
for (int i = 0; i < 1000; ++i) foo << "hello"
}
synchronized public int getN() {
return n;
}
}
while (true) {
Res res = new Res()
int N = 100000
for (int i = 0; i < N; ++i) {
new Thread({
res.inc()
if (res.getN() == N) {
println "ok"
}
}).start()
}
while (res.getN() < N) {
}
println "========================="
}
but the command
pidstat -w -I -p 26848 5
is printing 0 on the voluntary context switches column. The program creates 100000 thread that concurrently access a synchronized method. I can't believe that with such workload, no context switching is happening.
What's wrong with my benchmark?
