- Preemption Context Switches测量操作系统任务调度线程处理器上执行的次数,以及切换到较高-priority螺纹,数。
- Synchronization context switches度量的是因为显式调用线程同步API而发生线程切换的次数。如给多线程共享的变量加锁,多线程共同去改动。有些线程要堵塞在lock。直至占用锁的线程释放lock。这个度量反映的是线程间竞争的程度。
以下的实验来自VTune。旨在探究Preemption Context Switches的来源。
实验一:多线程无锁保护
speedup-example-no-mutex.cpp
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <assert.h>
#define N 4
#define M 30000
int nwait = 0;
volatile long long sum;
long loops = 6e3;
void set_affinity(int core_id) {
cpu_set_t cpuset;
CPU_ZERO(&cpuset);
CPU_SET(core_id, &cpuset);
assert(pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset) == 0);
}
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
int main(int argc, char *argv[]) {
set_affinity(23);
pthread_t th[N];
int ret;
for(unsigned i=0; i<N; ++i) {
ret = pthread_create(&th[i], NULL, thread_func, (void*)i);
assert(!ret && "pthread_create() failed!");
}
for(unsigned i=0; i<N; ++i)
pthread_join(th[i], NULL);
exit(0);
}
VTune现象:
Preemption Context Switches由两部分组成:clone和Unknown stack frame(s)。
- 后者的Preemption稳定在5:在这个程序中,共同拥有5个线程在执行,VTune显示每一个线程各占1,所以后者的Preemption才稳定在5上。为了验证,我们让N等于8,结果是每一个线程各占1。Unknown stack frame(s)处的Preemption稳定在9。
- clone处的Preemption不是一个确定的数。有可能是6、7、8等。
通过上图能够发现clone处的Preemption都分布在四个子线程中。以下再来一组:
通过比較上面三幅图。我们发现四个子线程所占的Preemption数并不总是均等。
为了验证,我们让N等于8,结果例如以下:
果然clone处的Preemption并非由子线程均分。只是随着线程数添加,clone处Preemption的添加幅度要大于Unknown stack frame(s)处。
通过上面的现象,我们尝试做出结论:
由于没有锁,所以线程间是独立的,我们单独分析一个线程中Preemption Context Switches的来源就可以(事实上这样的如果是有问题的,由于我们上面提到随着线程数添加,Preemption并没有线性添加,如果各线程间相互独立。理应是线性添加的,只是我们先从简单情况入手)。我们尝试逐步降低子线程运行任务的办法:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++)
sum += i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
nwait++;
}
}
无clone处的Preemption Context Switches
通过上面我们就断定当子线程计算任务变轻时。clone处的Preemption会变少,这是武断的。由于例如以下:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
}
}
这个子线程的计算任务要比上面三个中的第一个要轻。但它的Preemption数却要多,所以我初步猜想是第二层for循环的个数决定了clone处的Preemption数,于是做下面验证:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
}
}
确实是随着for循环的增多,clone处的Preemption在增多,但以此下结论还是不妥,合理的验证还应有下面工作:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++) {
sum += i;
sum += i;
sum += i;
sum += i;
}
}
}
奇怪明明这个子线程的工作量和上面验证中的第二个一样,并且它仅仅有一个for。但clone处的Preemption却很多其它,于是继续做验证:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++) {
sum += i;
sum += i;
sum += i;
sum += i;
sum += i;
sum += i;
sum += i;
}
}
}
终于结论:
也就是说随着第二层for的个数添加,clone处Preemption在添加。假设第二层仅仅有一个for,那么随着这个for中的子句(上面的实验仅仅能说明本例中出现的子句sum+=i有这样的情况)的增多,clone处的Preemption在添加。
分析:
假设说这是结论,那为什么?子线程在执行时,频繁被更高优先级的进程给抢占,可能是时间,执行时间,当子线程执行时间长时,系统中更高优先级的进程抢占它的情况很多其它。果然,我们又一次执行上述那些验证程序。发现——clone处Preemption多的程序,它的执行时间越长。
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
for (long i = 0; i < loops; i++) {
sum += i;
sum += i;
sum += i;
sum += i;
}
}
}
至于为什么,也许是由于编译器的优化。这里我们要专注于我们一開始的问题:Preemption Context Switches从何而来。
从运行时间而来。
当然这仅仅是针对多线程间无锁情况,以下给它加上锁。看看是否有哪个因素也会影响到Preemption Context Switches。
实验二:多线程加锁
speedup-example-mutex-only.cpp
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <assert.h>
#define N 4
#define M 30000
int nwait = 0;
volatile long long sum;
long loops = 6e3;
pthread_mutex_t mutex;
void set_affinity(int core_id) {
cpu_set_t cpuset;
CPU_ZERO(&cpuset);
CPU_SET(core_id, &cpuset);
assert(pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset) == 0);
}
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
int main(int argc, char *argv[]) {
set_affinity(23);
pthread_t th[N];
int ret;
for(unsigned i=0; i<N; ++i) {
ret = pthread_create(&th[i], NULL, thread_func, (void*)i);
assert(!ret && "pthread_create() failed!");
}
for(unsigned i=0; i<N; ++i)
pthread_join(th[i], NULL);
exit(0);
}
VTune现象:
接下来我们改变线程数。即N等于8:(我们期望Unknown处的Preemption添加类似线性,而clone处的添加幅度大。即与多线程无锁的情况类似)
Unkown stack frame(s)的对Preemption Context Switches的贡献率任然不如clone。且在同等数目线程下,加锁情况下的clone要比不加锁的制造很多其它的Preemption Context Switches。假设用我们上面的“时间理论”来解释——加锁的执行时间明显比不加锁要多,也能解释,只是这并不充分,让我们执行下面验证:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
phtread_mutex_unlock(&mutex);
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++) {
sum += i;
sum += i;
sum += i;
sum += i;
}
phtread_mutex_unlock(&mutex);
}
}
我们发现,基本上加锁情况与无锁情况一致。只是我们还需做下面验证:
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
- void* thread_func(void *arg) { set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++) {
sum += i*i*i*i*i*i;
sum += i*i*i*i*i*i;
sum += i*i*i*i*i*i;
sum += i*i*i*i*i*i;
}
}
}
果然。在一定误差可容忍下,for循环是不区别加锁for和不加锁for,它们取得的效果基本一样——随着第二层for的数目添加,clone处的Preemption在添加;只是这里,单个for中添加子句的效果和添加for数目的效果基本一样。这与无锁是不同的。并且,有一个比較重要的区别:
在无锁的情况下,
A
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
和
B
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
}
}
clone处Preemption的数目基本一致,但在加锁的情况下:
C
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
和
D
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
}
}
clone处Preemption的数目不一样。前者要明显多于后者。可是假设我们将后者改为:
E
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
for (long i = 0; i < loops; i++)
sum += i;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i;
}
}
则VTune分析有:
这就和C效果基本一样了。
而解释C、D、E三者之间的差异,也许也能够用我们的“时间理论”。运行三者:
C
D
E
尽管D的执行时比C和E稍小。但我们不能直接将无锁情况下的时间理论应用到加锁情况。
在说明原因之前。先看还有一个程序:
F
void* thread_func(void *arg) {
set_affinity((int)(long)arg);
for (int j = 0; j < M; j++) {
phtread_mutex_lock(&mutex);
nwait++;
phtread_mutex_unlock(&mutex);
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
for (long i = 0; i < loops; i++)
sum += i*i*i*i*i*i;
}
}
和D在clone处拥有基本一样的Preemption数。但二者的执行时间却大不一样。
F
所以“执行时间不一样。clone处的Preemption数不一样”。在这里就不适用了。
看来无锁和加锁还是有个重要区别的。我们都知道在无锁情况下,全部子线程并行执行。VTune中有例如以下调度:
我们通过大量的观察发现,对于每一个线程。每相隔1s就会有一次Preemption Context Switches,所以无锁情况下。随着执行时间的添加。clone处的Preemption数会增多。
事实上“时间理论”也适用于加锁情况,那为什么会出现上面C、D、E的情况,以及D和F的情况?我们也从调度图入手:
C
D
F
事实上加锁和无锁的“时间理论”的差别在于:加锁情况中的C和D(基本串行化)。并非每个线程中每隔1s就有一个Preemption。而加锁情况中的F(拥有并行化),每个线程中每隔1s会有一个Preemption。
这样对于C和D。因为C的执行时较D长。当中包括的Preemption比D多;而F尽管执行时比D短,但每一个线程中的Preemption汇总就会和D一样多。
终于我们得出结论:
Preemption Context Switches的来源是——
对于拥有并行化的程序。执行时间越长,Preemption Context Switches越多;对于加锁导致串行化的程序,执行时间越长,Preemption Context Switches越多;对于加锁仍保留并行化的程序。执行时间越长,Preemption Context Switches越多。