schedule performance between goroutine and pthread

gistfile1.txt

king@debian:~$ time ./a.out 

real	0m10.261s
user	0m3.040s
sys	0m17.312s

king@debian:~$ time go run pthread.go

real	0m4.017s
user	0m5.444s
sys	0m1.052s

schedule.c

#include <pthread.h>
#include <stdio.h>

void* thread_func(void* arg)
{
	int i;
	for (i = 0; i < 20000; i++) {
		pthread_yield();
	}
	return NULL;
}

int main()
{
	const int max_threads = 1024;
	pthread_t pth[max_threads];
	int i;
	for (i=0;i<max_threads;i++) {
		pthread_create(&pth[i], NULL, thread_func, NULL);
	}
	for (i=0;i<max_threads;i++) {
		pthread_join(pth[i], NULL);
	}
}

schedule.go

package main

import "runtime"

func thread_func(done chan struct{}) {
	for i := 0; i < 20000; i++ {
		runtime.Gosched()
	}
	done <- struct{}{}
}

func main() {
	runtime.GOMAXPROCS(runtime.NumCPU())
	max_threads := 1024
	done := make(chan struct{})
	for i := 0; i < max_threads; i++ {
		go thread_func(done)
	}
	for i := 0; i < max_threads; i++ {
		<-done
	}
}