/*
* Context switch microbenchmark.
*
* Copyright (C) 2015 Anton Blanchard <anton@au.ibm.com>, IBM
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#define _GNU_SOURCE
#include <sched.h>
#include <string.h>
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <getopt.h>
#include <signal.h>
#include <assert.h>
#include <pthread.h>
#include <limits.h>
#include <sys/time.h>
#include <sys/syscall.h>
#include <sys/types.h>
#include <sys/shm.h>
#include <linux/futex.h>
#ifdef __powerpc__
#include <altivec.h>
#endif
#include "utils.h"
static unsigned int timeout = 30;
static int touch_vdso;
struct timeval tv;
static int touch_fp = 1;
double fp;
static int touch_vector = 1;
vector int a, b, c;
#ifdef __powerpc__
static int touch_altivec = 1;
/*
* Note: LTO (Link Time Optimisation) doesn't play well with this function
* attribute. Be very careful enabling LTO for this test.
*/
static void __attribute__((__target__("no-vsx"))) altivec_touch_fn(void)
{
c = a + b;
}
#endif
static void touch(void)
{
if (touch_vdso)
gettimeofday(&tv, NULL);
if (touch_fp)
fp += 0.1;
#ifdef __powerpc__
if (touch_altivec)
altivec_touch_fn();
#endif
if (touch_vector)
c = a + b;
asm volatile("# %0 %1 %2": : "r"(&tv), "r"(&fp), "r"(&c));
}
static void start_thread_on(void *(*fn)(void *), void *arg, unsigned long cpu)
{
pthread_t tid;
cpu_set_t cpuset;
pthread_attr_t attr;
CPU_ZERO(&cpuset);
CPU_SET(cpu, &cpuset);
pthread_attr_init(&attr);
if (pthread_attr_setaffinity_np(&attr, sizeof(cpu_set_t), &cpuset)) {
perror("pthread_attr_setaffinity_np");
exit(1);
}
if (pthread_create(&tid, &attr, fn, arg)) {
perror("pthread_create");
exit(1);
}
}
static void start_process_on(void *(*fn)(void *), void *arg, unsigned long cpu)
{
int pid;
cpu_set_t cpuset;
pid = fork();
if (pid == -1) {
perror("fork");
exit(1);
}
if (pid)
return;
CPU_ZERO(&cpuset);
CPU_SET(cpu, &cpuset);
if (sched_setaffinity(0, sizeof(cpuset), &cpuset)) {
perror("sched_setaffinity");
exit(1);
}
fn(arg);
exit(0);
}
static unsigned long iterations;
static unsigned long iterations_prev;
static void sigalrm_handler(int junk)
{
unsigned long i = iterations;
printf("%ld\n", i - iterations_prev);
iterations_prev = i;
if (--timeout == 0)
kill(0, SIGUSR1);
alarm(1);
}
static void sigusr1_handler(int junk)
{
exit(0);
}
struct actions {
void (*setup)(int, int);
void *(*thread1)(void *);
void *(*thread2)(void *);
};
#define READ 0
#define WRITE 1
static int pipe_fd1[2];
static int pipe_fd2[2];
static void pipe_setup(int cpu1, int cpu2)
{
if (pipe(pipe_fd1) || pipe(pipe_fd2))
exit(1);
}
static void *pipe_thread1(void *arg)
{
signal(SIGALRM, sigalrm_handler);
alarm(1);
while (1) {
assert(read(pipe_fd1[READ], &c, 1) == 1);
touch();
assert(write(pipe_fd2[WRITE], &c, 1) == 1);
touch();
iterations += 2;
}
return NULL;
}
static void *pipe_thread2(void *arg)
{
while (1) {
assert(write(pipe_fd1[WRITE], &c, 1) == 1);
touch();
assert(read(pipe_fd2[READ], &c, 1) == 1);
touch();
}
return NULL;
}
static struct actions pipe_actions = {
.setup = pipe_setup,
.thread1 = pipe_thread1,
.thread2 = pipe_thread2,
};
static void yield_setup(int cpu1, int cpu2)
{
if (cpu1 != cpu2) {
fprintf(stderr, "Both threads must be on the same CPU for yield test\n");
exit(1);
}
}
static void *yield_thread1(void *arg)
{
signal(SIGALRM, sigalrm_handler);
alarm(1);
while (1) {
sched_yield();
touch();
iterations += 2;
}
return NULL;
}
static void *yield_thread2(void *arg)
{
while (1) {
sched_yield();
touch();
}
return NULL;
}
static struct actions yield_actions = {
.setup = yield_setup,
.thread1 = yield_thread1,
.thread2 = yield_thread2,
};
static long sys_futex(void *addr1, int op, int val1, struct timespec *timeout,
void *addr2, int val3)
{
return syscall(SYS_futex, addr1, op, val1, timeout, addr2, val3);
}
static unsigned long cmpxchg(unsigned long *p, unsigned long expected,
unsigned long desired)
{
unsigned long exp = expected;
__atomic_compare_exchange_n(p, &exp, desired, 0,
__ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST);
return exp;
}
static unsigned long xchg(unsigned long *p, unsigned long val)
{
return __atomic_exchange_n(p, val, __ATOMIC_SEQ_CST);
}
static int mutex_lock(unsigned long *m)
{
int c;
c = cmpxchg(m, 0, 1);
if (!c)
return 0;
if (c == 1)
c = xchg(m, 2);
while (c) {
sys_futex(m, FUTEX_WAIT, 2, NULL, NULL, 0);
c = xchg(m, 2);
}
return 0;
}
static int mutex_unlock(unsigned long *m)
{
if (*m == 2)
*m = 0;
else if (xchg(m, 0) == 1)
return 0;
sys_futex(m, FUTEX_WAKE, 1, NULL, NULL, 0);
return 0;
}
static unsigned long *m1, *m2;
static void futex_setup(int cpu1, int cpu2)
{
int shmid;
void *shmaddr;
shmid = shmget(IPC_PRIVATE, getpagesize(), SHM_R | SHM_W);
if (shmid < 0) {
perror("shmget");
exit(1);
}
shmaddr = shmat(shmid, NULL, 0);
if (shmaddr == (char *)-1) {
perror("shmat");
shmctl(shmid, IPC_RMID, NULL);
exit(1);
}
shmctl(shmid, IPC_RMID, NULL);
m1 = shmaddr;
m2 = shmaddr + sizeof(*m1);
*m1 = 0;
*m2 = 0;
mutex_lock(m1);
mutex_lock(m2);
}
static void *futex_thread1(void *arg)
{
signal(SIGALRM, sigalrm_handler);
alarm(1);
while (1) {
mutex_lock(m2);
mutex_unlock(m1);
iterations += 2;
}
return NULL;
}
static void *futex_thread2(void *arg)
{
while (1) {
mutex_unlock(m2);
mutex_lock(m1);
}
return NULL;
}
static struct actions futex_actions = {
.setup = futex_setup,
.thread1 = futex_thread1,
.thread2 = futex_thread2,
};
static int processes;
static struct option options[] = {
{ "test", required_argument, 0, 't' },
{ "process", no_argument, &processes, 1 },
{ "timeout", required_argument, 0, 's' },
{ "vdso", no_argument, &touch_vdso, 1 },
{ "no-fp", no_argument, &touch_fp, 0 },
#ifdef __powerpc__
{ "no-altivec", no_argument, &touch_altivec, 0 },
#endif
{ "no-vector", no_argument, &touch_vector, 0 },
{ 0, },
};
static void usage(void)
{
fprintf(stderr, "Usage: context_switch2 <options> CPU1 CPU2\n\n");
fprintf(stderr, "\t\t--test=X\tpipe, futex or yield (default)\n");
fprintf(stderr, "\t\t--process\tUse processes (default threads)\n");
fprintf(stderr, "\t\t--timeout=X\tDuration in seconds to run (default 30)\n");
fprintf(stderr, "\t\t--vdso\t\ttouch VDSO\n");
fprintf(stderr, "\t\t--no-fp\t\tDon't touch FP\n");
#ifdef __powerpc__
fprintf(stderr, "\t\t--no-altivec\tDon't touch altivec\n");
#endif
fprintf(stderr, "\t\t--no-vector\tDon't touch vector\n");
}
int main(int argc, char *argv[])
{
signed char c;
struct actions *actions = &yield_actions;
int cpu1;
int cpu2;
static void (*start_fn)(void *(*fn)(void *), void *arg, unsigned long cpu);
while (1) {
int option_index = 0;
c = getopt_long(argc, argv, "", options, &option_index);
if (c == -1)
break;
switch (c) {
case 0:
if (options[option_index].flag != 0)
break;
usage();
exit(1);
break;
case 't':
if (!strcmp(optarg, "pipe")) {
actions = &pipe_actions;
} else if (!strcmp(optarg, "yield")) {
actions = &yield_actions;
} else if (!strcmp(optarg, "futex")) {
actions = &futex_actions;
} else {
usage();
exit(1);
}
break;
case 's':
timeout = atoi(optarg);
break;
default:
usage();
exit(1);
}
}
if (processes)
start_fn = start_process_on;
else
start_fn = start_thread_on;
if (((argc - optind) != 2)) {
cpu1 = cpu2 = pick_online_cpu();
} else {
cpu1 = atoi(argv[optind++]);
cpu2 = atoi(argv[optind++]);
}
printf("Using %s with ", processes ? "processes" : "threads");
if (actions == &pipe_actions)
printf("pipe");
else if (actions == &yield_actions)
printf("yield");
else
printf("futex");
printf(" on cpus %d/%d touching FP:%s altivec:%s vector:%s vdso:%s\n",
cpu1, cpu2, touch_fp ? "yes" : "no", touch_altivec ? "yes" : "no",
touch_vector ? "yes" : "no", touch_vdso ? "yes" : "no");
/* Create a new process group so we can signal everyone for exit */
setpgid(getpid(), getpid());
signal(SIGUSR1, sigusr1_handler);
actions->setup(cpu1, cpu2);
start_fn(actions->thread1, NULL, cpu1);
start_fn(actions->thread2, NULL, cpu2);
while (1)
sleep(3600);
return 0;
}
