#include "rq_bench.hfa"

thread __attribute__((aligned(128))) Yielder {
	unsigned long long count;
};
void ?{}( Yielder & this ) {
	((thread&)this){ "Yielder Thread", bench_cluster };
	this.count = 0;
}

void main( Yielder & this ) {
	park();
	for() {
		yield();
		this.count++;
		if( clock_mode && stop) break;
		if(!clock_mode && this.count >= stop_count) break;
	}

	__atomic_fetch_add(&threads_left, -1, __ATOMIC_SEQ_CST);
}

int main(int argc, char * argv[]) {
	cfa_option opt[] = {
		BENCH_OPT
	};
	BENCH_OPT_PARSE("cforall yield benchmark");

	{
		unsigned long long global_counter = 0;

		Time start, end;
		BenchCluster bc = { nprocs };
		{
			threads_left = nthreads;
			Yielder * threads = alloc(nthreads);
			for(i; nthreads) {
				(threads[i]){};
			}

			printf("Starting\n");

			bool is_tty = isatty(STDOUT_FILENO);
			start = timeHiRes();

			for(i; nthreads) {
				unpark( threads[i] );
			}
			wait(start, is_tty);

			stop = true;
			end = timeHiRes();
			printf("\nDone\n");

			for(i; nthreads) {
				Yielder & y = join( threads[i] );
				global_counter += y.count;
				^(threads[i]){};
			}
			free(threads);
		}

		printf("Duration (ms)        : %'ld\n", (end - start)`dms);
		printf("Number of processors : %'d\n", nprocs);
		printf("Number of threads    : %'d\n", nthreads);
		printf("Total Operations(ops): %'15llu\n", global_counter);
		printf("Ops per second       : %'18.2lf\n", ((double)global_counter) / (end - start)`s);
		printf("ns per ops           : %'18.2lf\n", (end - start)`dns / global_counter);
		printf("Ops per threads      : %'15llu\n", global_counter / nthreads);
		printf("Ops per procs        : %'15llu\n", global_counter / nprocs);
		printf("Ops/sec/procs        : %'18.2lf\n", (((double)global_counter) / nprocs) / (end - start)`s);
		printf("ns per ops/procs     : %'18.2lf\n", (end - start)`dns / (global_counter / nprocs));
		fflush(stdout);
	}
}