linux下面使用gprof,分析多线程

我们知道可以使用,gprof工具来分析系统性能，但是在linux下面，多线程则不适用,多线程下，只能采集主线程性能数据。.原因是gprof采用ITIMER_PROF信号，在多线程内，只有主线程才能响应该信号。那么采用什么方法，才能够分析所有
线程呢？  关键是能够让各个线程响应ITIMER_PROF信号，下面是通过桩子函数来进行的，就是重写pthread_create函数。



//////////////////// gprof-helper.c////////////////////////////
#define _GNU_SOURCE
#include <sys/time.h>
#include <stdio.h>
#include <stdlib.h>
#include <dlfcn.h>
#include <pthread.h>

static void * wrapper_routine(void *);

/* Original pthread function */
static int (*pthread_create_orig)(pthread_t *__restrict,
                                 __const pthread_attr_t *__restrict,
                                 void *(*)(void *),
                                 void *__restrict) = NULL;

/* Library initialization function */
void wooinit(void) __attribute__((constructor));

void wooinit(void)
{
   pthread_create_orig = dlsym(RTLD_NEXT, "pthread_create");
   fprintf(stderr, "pthreads: using profiling hooks for gprof/n");
   if(pthread_create_orig == NULL)
   {
       char *error = dlerror();
       if(error == NULL)
       {
           error = "pthread_create is NULL";
       }
       fprintf(stderr, "%s/n", error);
       exit(EXIT_FAILURE);
   }
}

/* Our data structure passed to the wrapper */
typedef struct wrapper_s
{
   void * (*start_routine)(void *);
   void * arg;

   pthread_mutex_t lock;
   pthread_cond_t  wait;

   struct itimerval itimer;

} wrapper_t;

/* The wrapper function in charge for setting the itimer value */
static void * wrapper_routine(void * data)
{
   /* Put user data in thread-local variables */
   void * (*start_routine)(void *) = ((wrapper_t*)data)->start_routine;
   void * arg = ((wrapper_t*)data)->arg;

   /* Set the profile timer value */
   setitimer(ITIMER_PROF, &((wrapper_t*)data)->itimer, NULL);

   /* Tell the calling thread that we don't need its data anymore */
   pthread_mutex_lock(&((wrapper_t*)data)->lock);
   pthread_cond_signal(&((wrapper_t*)data)->wait);
   pthread_mutex_unlock(&((wrapper_t*)data)->lock);

   /* Call the real function */
   return start_routine(arg);
}

/* Our wrapper function for the real pthread_create() */
int pthread_create(pthread_t *__restrict thread,
                  __const pthread_attr_t *__restrict attr,
                  void * (*start_routine)(void *),
                  void *__restrict arg)
{
   wrapper_t wrapper_data;
   int i_return;

   /* Initialize the wrapper structure */
   wrapper_data.start_routine = start_routine;
   wrapper_data.arg = arg;
   getitimer(ITIMER_PROF, &wrapper_data.itimer);
   pthread_cond_init(&wrapper_data.wait, NULL);
   pthread_mutex_init(&wrapper_data.lock, NULL);
   pthread_mutex_lock(&wrapper_data.lock);

   /* The real pthread_create call */
   i_return = pthread_create_orig(thread,
                                  attr,
                                  &wrapper_routine,
                                  &wrapper_data);

   /* If the thread was successfully spawned, wait for the data
    * to be released */
   if(i_return == 0)
   {
       pthread_cond_wait(&wrapper_data.wait, &wrapper_data.lock);
   }

   pthread_mutex_unlock(&wrapper_data.lock);
   pthread_mutex_destroy(&wrapper_data.lock);
   pthread_cond_destroy(&wrapper_data.wait);

   return i_return;
}

///////////////////
然后编译成动态库 gcc -shared -fPIC gprof-helper.c -o gprof-helper.so -lpthread -ldl

使用例子：
/////////////////////a.c/////////////////////////////
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <pthread.h>
#include <string.h>
void fun1();
void fun2();
void* fun(void * argv);
int main()
{
int i =0;
int id;
pthread_t    thread[100];
for(i =0 ;i< 100; i++)
{

id = pthread_create(&thread[i], NULL, fun, NULL);

printf("thread =%d/n",i);
}
printf("dsfsd/n");
return 0;
}
void*  fun(void * argv)
{

fun1();
fun2();
return NULL;
}

void fun1()
{
int i = 0;

while(i<100)
{
i++;
printf("fun1/n");
}
}

void fun2()
{
int i = 0;
int b;
while(i<50)
{
i++;
printf("fun2/n");
//b+=i;
}
}
///////////////

gcc -pg a.c  gprof-helper.so

运行程序:
./a.out

分析gmon.out:
gprof -b a.out gmon.out