最新国产好看的视频,伊人天堂AV在线,国产Aaaaaa视频,蜜臀视频在线观看一区,人妻av色图,密臀久久久精品影片,青青视频免费观看毛片,久草在线观看视,国产三级精品色情在线

C語言實(shí)現(xiàn)用戶態(tài)線程庫案例

 更新時(shí)間:2017年05月14日 09:10:33   投稿:jingxian  
下面小編就為大家?guī)硪黄狢語言實(shí)現(xiàn)用戶態(tài)線程庫案例。小編覺得挺不錯(cuò)的,現(xiàn)在就分享給大家,也給大家做個(gè)參考。一起跟隨小編過來看看吧

輪子年年有人造,我們也來湊熱鬧,參考協(xié)程實(shí)現(xiàn),大概有以下幾種方法:

1)利用setjmp,longjmp

2)利用ucontext接口函數(shù)

3)匯編

(線程無非就是多了個(gè)搶占功能,由定時(shí)器觸發(fā),而非自愿讓出運(yùn)行權(quán)限)

因?yàn)槲覍懙臅r(shí)候還沒看到其他帖子,如果看到了,鐵定會(huì)用最直觀的ucontext接口寫的(注意,在macOSX中已經(jīng)標(biāo)注為廢除,頭文件得換做sys/ucontext.h),結(jié)果就是我用了匯編來寫,但是盡量不用匯編來寫整個(gè)switch_to調(diào)度函數(shù)(這樣有個(gè)明顯的壞處,那就是用gas/nasm的標(biāo)準(zhǔn)匯編格式寫的函數(shù)在macOSX下不能編譯通過,這個(gè)與系統(tǒng)自帶的編譯工具有關(guān)),而用經(jīng)量少的內(nèi)嵌匯編來寫。switch_to函數(shù)參考的是minix操作系統(tǒng)中任務(wù)切換函數(shù)實(shí)現(xiàn)的,用軟件時(shí)鐘器每隔1s發(fā)信號(hào)以激發(fā)switch_to函數(shù)切換任務(wù)。下面直接貼代碼了,對(duì)外提供了類似pthread的接口(只有兩個(gè),分別是threadCreate和threadJoin)?,F(xiàn)在的代碼還非常的buggy,只能安全地支持在線程函數(shù)里頭純計(jì)算,其他的行為非??赡芤l(fā)bus error和segmentation fault。(要更加嚴(yán)謹(jǐn)?shù)匮芯坑脩魬B(tài)線程庫,請(qǐng)去看gnu pth的實(shí)現(xiàn)代碼)

 thread.h

#pragma once
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <string.h>
#include <signal.h>
#include <assert.h>
#include <time.h>

#define JMP(r)  asm volatile \
        (  "pushl %3\n\t" \
          "popfd\n\t" \
          "movl %2, %%ebp\n\t" \
          "movl %0, %%esp\n\t" \
          "jmp *%1\n\t" \
          : \
          : "m"(r._esp),"m"(r._eip),"m"(r._ebp),"m"(r._eflags) \
          : \
        )

#define SAVE()         asm volatile \
              (  "movl %%eax, %0\n\t" \
                "movl %%ecx, %1\n\t" \
                "movl %%edx, %2\n\t" \
                "movl %%ebx, %3\n\t" \
                  "movl %%esp, %4\n\t" \
                "movl %%ebp, %5\n\t" \
                "movl %%esi, %6\n\t" \
                "movl %%edi, %7\n\t" \
                "pushfd\n\t" \
                "movl (%%esp), %%eax\n\t" \
                "movl %%eax, %8\n\t" \
                "popfd\n\t" \
                : "=m"(_eax),"=m"(_ecx),"=m"(_edx),"=m"(_ebx) \
                ,"=m"(_esp),"=m"(_ebp) \
                , "=m"(_esi),"=m"(_edi),"=m"(_eflags) \
                : \
                : "%eax" \
              )

#define RESTORE(r)     asm volatile \
              (  "movl %0, %%eax\n\t" \
                "movl %1, %%ecx\n\t" \
                "movl %1, %%edx\n\t" \
                "movl %3, %%ebx\n\t" \
                "movl %4, %%esi\n\t" \
                "movl %5, %%edi\n\t" \
                : \
                :"m"(r._eax),"m"(r._ecx),"m"(r._edx),"m"(r._ebx) \
                , "m"(r._esi),"m"(r._edi) \
              )

typedef void Func(int);

/* __timer struct is the real Timer struct we use
 * id is unique to each timer
 * intersec is the inteval seconds to each signal forwarding the this Timer
 * sigactor is the handler for this Timer
 * next is a internal member used for linked list
 */
struct __timer
{
  void *next;
  unsigned int sec;
  unsigned int intersec;
  int id;
  Func *sigactor;
};

/* struct alarm is ugly for the compatibility with early struct.
 * I should have used unnamed member instead of __inner.
 */
typedef struct alarm *Timer;
struct alarm
{
  union{
    struct
    {
      Timer next;
      unsigned int sec;
    };
    struct __timer __inner;
  }; 
};

typedef struct list *Header;

struct list
{
  Timer head;
};

typedef struct __thread_table_regs Regs;
struct __thread_table_regs
{
  int _edi;
  int _esi;
  int _ebp;
  int _esp;
  int _ebx;
  int _edx;
  int _ecx;
  int _eax;
  int _eip;
  int _eflags;
};

typedef struct __ez_thread Thread_t;
struct __ez_thread
{
  Regs regs;
  int tid;
  sigset_t sigmask;
  unsigned int priority;
  int tick;
  int state;
  int errno;
  unsigned int stacktop;
  unsigned int stacksize;
  void *stack;
  void *retval;
  volatile int __reenter;
};

typedef struct __pnode pNode;
struct __pnode
{
  pNode *next;
  pNode *prev;
  Thread_t *data;
};

typedef struct __loopcursor Cursor;
struct __loopcursor
{
  int total;
  pNode *current;
};
typedef struct __stack *Stack_t;
struct __stack
{
  int __pad[4096];
};

void switch_to(int);

extern Header hdr_ptr;
extern Cursor live;
extern Cursor dead;
extern Thread_t pmain;

 thread.c

/* MIT License

Copyright (c) 2017 Yuandong-Chen

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE. */

#include "thread.h"
/************************* Alarm facility *************************/

struct list linkedlist;
Header hdr_ptr = &linkedlist;


Timer mallocTimer(int id, Func *actor,unsigned int sec, unsigned int interval)
{
  Timer ret = (Timer)malloc(sizeof(struct alarm));
  assert(ret);
  ret->__inner.id = id;
  ret->__inner.sigactor = actor;
  ret->__inner.intersec = interval;
  ret->sec = sec;
  return ret;
}

/* find Timer in linked list which id is id.
 * return: return NULL if not found, -1 if it's header link, 
 * otherwise prev which is the previous Timer member to this Timer
 */

Timer findTimerPrev(Header h, int id)
{
  assert(h);
  if(h->head == NULL)
    return NULL;

  Timer t = h->head;
  Timer prev = NULL;

  while(t)
  {
    if(t->__inner.id == id){
      if(prev == NULL)
        return (Timer)-1;
      else
        return prev;
    }
    prev = t;
    t = t->next;
  }

  return NULL;
}

/* delete Timer in linked list.
 * return: nothing, we ensure this t is deleted in the linked list.
 */

void delTimer(Header h, Timer t)
{
  assert(h);
  assert(t);
  Timer prevtodel = findTimerPrev(h, t->__inner.id);
  unsigned int base = 0;

  if(prevtodel)
  {
    if(prevtodel == (Timer)-1){

      unsigned int res = (h->head)->sec;
      if(res != 0)
      {
        base = res;
      }
      else
      {
        kill(getpid(),SIGALRM);
        return;
      }
      h->head = (h->head)->next;
      Timer tmp = (h->head);

      while(tmp){
        tmp->sec += base;
        tmp = tmp->next;
      }
      return;
    }
    else
    {
      
      base = (prevtodel->next)->sec;
      prevtodel->next = (prevtodel->next)->next;
      Timer tmp = (prevtodel->next);
      
      while(tmp){
        tmp->sec += base;
        tmp = tmp->next;
      }
      return;
    }
  }

  return;
}

/* append Timer in appropriate place in linked list.
 * the appropriate place means all timers in linked list are arranged 
 * according their next alarm seconds.
 * The algorithm we use here is that the real left alarm seconds for this Timer 
 * is the sum of all the sec member in Timer in linked list prev to this Timer
 * plus its sec member. For example, we add 3 Timers to the linked list,
 * whose sec are 4, 3, 2 respectively. Then the linked list looks like:
 * 2 (real sec = 2) --> 1 (real sec = 2+1 = 3) --> 1 (real sec = 2+1+1 = 4)
 * The advantage is obviously, we dont need to remember how many seconds passed.
 * We always fetch the header to respond the alarm signal and set next alarm sec 
 * as the next timer in the linked list. (The real situation is a little bit more 
 * complex, for example if upcoming timers' sec equals 0, we need to call their
 * handler right away all together in a certain sequence. If its intersec is not 
 * zero, we need to append it to the linked list again as quick as possible)
 * note: delTimer also address this problem. If we delete any Timer, we need to 
 * recalculate the secs after this timer in the linked list.(simply to add sec to 
 * the next timer and delete this timer node)
 * return: only 0 if success, otherwise the hole process failed.
 */

int appendTimer(Header h, Timer t)
{
  assert(h);
  assert(t);
  delTimer(h, t);

  if(h->head == NULL)
  {
    h->head = t;
    return 0;
  }

  Timer tmp = h->head;
  Timer prev = NULL;
  unsigned int prevbase = 0;
  unsigned int base = 0;

  while(tmp)
  {
    prevbase = base;
    base += tmp->sec;
    if(t->sec < base){
      break;
    }
    else{
      prev = tmp;
      tmp = tmp->next;
    }
      
  }

  if(prev == NULL)
  {
    (h->head)->sec -= t->sec;
    t->next = h->head;
    h->head = t;
    return 0;
  }

  if(tmp == NULL)
    t->sec -=base;
  else
    t->sec -=prevbase;

  prev->next = t;
  t->next = tmp;
  if(tmp)
    tmp->sec -= t->sec;

  return 0;
}

/* pop header timer in linked list.
 * return: its hander
 */

Func* popTimer(Header h)
{
  assert(h);
  if(h->head == NULL)
    return (Func *)-1;
  Func *ret = (h->head)->__inner.sigactor;
  Timer todel = h->head;
  h->head = (h->head)->next;
  // if its intersec greater than 0, we append it right away to the linked list
  if(todel->__inner.intersec > 0)
  {
    todel->sec = todel->__inner.intersec;
    appendTimer(h, todel);
  }
  return ret;
}

void printList(Header h)
{
  assert(h);
  if(h->head == NULL)
    return;

  Timer tmp = h->head;

  while(tmp)
  {
    printf("timer[%d] = %u saved %u\n", tmp->__inner.id, tmp->sec, tmp->__inner.intersec);
    tmp = tmp->next;
  }
}

/* it's the real signal handler responding to every SIGALRM.
 */
void sig_alarm_internal(int signo)
{ 
  void funcWrapper(int signo, Func *func);

  if(hdr_ptr->head == NULL)
    return;

  Func *recv;
  if((recv = popTimer(hdr_ptr)) == (Func *)-1){
    funcWrapper(SIGALRM, recv);
  } 
  else
  {
    // signal ourself if next timer's sec = 0
    if(hdr_ptr->head){
      ((hdr_ptr->head)->sec > 0?alarm((hdr_ptr->head)->sec):kill(getpid(), SIGALRM));
    }
    funcWrapper(SIGALRM, recv);
  }
}

/* Alarm function simulates native alarm function.
 * what if SIGALRM arrives when process is running in Alarm?
 * we just block the signal since there is no slow function in Alarm,
 * sig_alarm_internal will for sure address the signal very soon.
 */

unsigned int Alarm(Header h, Timer mtimer)
{
  sigset_t mask;
  sigset_t old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);
  
  unsigned int res = 0;
  Timer t;

  if((t = findTimerPrev(h, mtimer->__inner.id)) == NULL)
    goto LL;

  t = h->head;
  while(t)
  {
    res += t->sec; // it's not precise, we should use alarm(0) for the first sec.
            // However, its simple enough to implement. 
    if(t->__inner.id == mtimer->__inner.id)
      break;

    t = t->next;
  }
LL:
  if(mtimer->sec == 0)
  {
    delTimer(h, mtimer);
    sigprocmask(SIG_SETMASK, &old, NULL);
    return res;
  }
   
  appendTimer(h, mtimer);
  if(mtimer->__inner.id == (h->head)->__inner.id)
    ((h->head)->sec > 0?alarm((h->head)->sec):kill(getpid(), SIGALRM));
  sigprocmask(SIG_SETMASK, &old, NULL);
  return res;
}

void initTimer()
{
  struct sigaction act;
  act.sa_handler = sig_alarm_internal;
  act.sa_flags = SA_RESTART|SA_NODEFER;
  sigemptyset(&act.sa_mask);
  sigaction(SIGALRM, &act, NULL);
}

void funcWrapper(int signo, Func *func)
{
  sigset_t mask;
  sigset_t old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_UNBLOCK, &mask, &old);
  func(signo);
  sigprocmask(SIG_SETMASK, &old, NULL);
}

/************************* Thread facility *************************/


Cursor live;
Cursor dead;
Thread_t pmain;

void initCursor(Cursor *cur)
{
  cur->total = 0;
  cur->current = NULL;
}

Thread_t *findThread(Cursor *cur, int tid)
{
  sigset_t mask,old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);
  int counter = cur->total;
  if(counter == 0){
    sigprocmask(SIG_SETMASK, &old, NULL);
    return NULL;
  }
    

  int i;
  pNode *tmp = cur->current;
  for (int i = 0; i < counter; ++i)
  {
    if((tmp->data)->tid == tid){
      sigprocmask(SIG_SETMASK, &old, NULL);
      return tmp->data;
    }
    tmp = tmp->next;
  }
  sigprocmask(SIG_SETMASK, &old, NULL);
  return NULL;
}

int appendThread(Cursor *cur, Thread_t *pth)
{
  sigset_t mask,old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);
  if(cur->total == 0)
  {
    //note this never freed for simple implementation
    cur->current = (pNode *)malloc(sizeof(pNode));
    assert(cur->current);
    (cur->current)->data = pth;
    (cur->current)->prev = cur->current;
    (cur->current)->next = cur->current;
    cur->total++;
    sigprocmask(SIG_SETMASK, &old, NULL);
    return 0;
  }
  else
  {
    #define MAXTHREADS 5
    if(cur->total > MAXTHREADS)
    {
      assert((cur->total == MAXTHREADS));
      sigprocmask(SIG_SETMASK, &old, NULL);
      return -1;
    }
    //freed at threadJoin for simple implementation
    pNode *tmp = malloc(sizeof(pNode));
    assert(tmp);
    tmp->data = pth;
    tmp->prev = cur->current;
    tmp->next = (cur->current)->next;
    ((cur->current)->next)->prev = tmp;
    (cur->current)->next = tmp;
    cur->total++;
    sigprocmask(SIG_SETMASK, &old, NULL);
    return 0;
  }
}

pNode *deleteThread(Cursor *cur, int tid)
{
  sigset_t mask,old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);

  int counter = cur->total;
  int i;
  pNode *tmp = cur->current;
  for (int i = 0; i < counter; ++i)
  {
    if((tmp->data)->tid == tid){
      (tmp->prev)->next = tmp->next;
      (tmp->next)->prev = tmp->prev;
      if(tmp == cur->current)
      {
        cur->current = cur->current->next;
      } 
      //free(tmp);
      cur->total--;
      assert(cur->total);
      sigprocmask(SIG_SETMASK, &old, NULL);
      return tmp;
    }
    tmp = tmp->next;
  }
  sigprocmask(SIG_SETMASK, &old, NULL);
  return NULL;
}

void printThread(Thread_t *pth)
{
  printf("pth tid: %d\n", pth->tid);
  printf("pth stack top: %x\n", pth->stacktop);
  printf("pth stack size: %u\n", pth->stacksize);
  printf("pth state: %d\n", pth->state);
  printf("pth errno: %d\n", pth->errno);
  printf("pth retval: %p\n", pth->retval);
  printf("pth sigmask: %u\n", pth->sigmask);
  printf("pth priority: %d\n", pth->priority);
  printf("pth tick: %d\n", pth->tick);
  printf("EFLAGS: %x\t", pth->regs._eflags);
  printf("EIP: %x\t", pth->regs._eip);
  printf("EAX: %x\t", pth->regs._eax);
  printf("ECX: %x\n", pth->regs._ecx);
  printf("EDX: %x\t", pth->regs._edx);
  printf("EBX: %x\t", pth->regs._ebx);
  printf("ESP: %x\t", pth->regs._esp);
  printf("EBP: %x\n", pth->regs._ebp);
  printf("ESI: %x\t", pth->regs._esi);
  printf("EDI: %x\n", pth->regs._edi);

}

void printLoop(Cursor *cur)
{
  int count = 0;
  pNode *tmp = cur->current;
  assert(tmp);
  do{
    printThread(tmp->data);
    tmp = tmp->next;
    count ++; 
  }while(tmp != cur->current);
  printf("real total: %d\n", count);
  printf("total record:%d\n", cur->total);
  assert(count == cur->total);
}

int fetchTID()
{
  static int tid;
  return ++tid;
}

void real_entry(Thread_t *pth, void *(*start_rtn)(void *), void* args)
{
  //printf("in real entry: %p\n", start_rtn);
  
  pth->retval = (*start_rtn)(args);
  //deleteThread(&live, pth->tid);
  /* some clean job here */
  //free(pth->stack);
  //pth->stack = NULL;
  //pth->stacktop = 0;
  //pth->stacksize = 0;
  #define DETACHED 1
  deleteThread(&live, pth->tid);
  appendThread(&dead, pth);

  if(pth->state == DETACHED)
    threadJoin(pth, NULL);

  switch_to(-1);
}

int threadCreat(Thread_t **pth, void *(*start_rtn)(void *), void *arg)
{
  sigset_t mask,old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);
  //freed at threadJoin for simple implementation
  *pth = malloc(sizeof(Thread_t));
  #define PTHREAD_STACK_MIN 4096
  //freed at threadJoin for simple implementation
  (*pth)->stack = malloc(PTHREAD_STACK_MIN);
  assert((*pth)->stack);
  (*pth)->stacktop = (((int)(*pth)->stack + PTHREAD_STACK_MIN)&(0xfffff000));
  (*pth)->stacksize = PTHREAD_STACK_MIN - (((int)(*pth)->stack + PTHREAD_STACK_MIN) - (*pth)->stacktop);
  (*pth)->state = 0; // 0 JOINABLE 1 DETACHED
  (*pth)->priority = 1; //one seconds
  (*pth)->tick = (*pth)->priority;
  (*pth)->tid = fetchTID();
  sigprocmask(0,NULL,&((*pth)->sigmask));
  /* set params */
  void *dest = (*pth)->stacktop - 12;
  memcpy(dest, pth, 4);
  dest += 4;
  memcpy(dest, &start_rtn, 4);
  dest += 4;
  memcpy(dest, &arg, 4);
  (*pth)->regs._eip = &real_entry;
  (*pth)->regs._esp = (*pth)->stacktop - 16;
  (*pth)->regs._edi = 0;
  (*pth)->regs._esi = 0;
  (*pth)->regs._ebp = 0;
  (*pth)->regs._eax = 0;
  (*pth)->regs._ebx = 0;
  (*pth)->regs._ecx = 0;
  (*pth)->regs._edx = 0;
  (*pth)->regs._eflags = 0;
  appendThread(&live, (*pth));
  sigprocmask(SIG_SETMASK, &old, NULL);
  return 0;
}

int threadJoin(Thread_t *pth, void **rval_ptr)
{

  sigset_t mask,old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);
  Thread_t *find1, *find2;
  find1 = findThread(&live, pth->tid);
  find2 = findThread(&dead, pth->tid);
  

  if((find1 == NULL)&&(find2 == NULL)){
    sigprocmask(SIG_SETMASK, &old, NULL);
    return -1;
  }

  if(find2){
    if(rval_ptr != NULL)
      *rval_ptr = find2->retval;

    sigprocmask(SIG_SETMASK, &old, NULL);
    return 0;
  }
  sigprocmask(SIG_SETMASK, &old, NULL);
  while(1)
  {
    if((find2 = findThread(&dead, pth->tid))!= NULL){
      if(rval_ptr!= NULL)
        *rval_ptr = find2->retval;

      pNode *tmp = deleteThread(&dead, pth->tid);
      free(tmp);
      free((Stack_t)find2->stack);
      free(find2);
      return 0;
    }
  }
  return -1;
}

void init()
{
  initTimer();
  initCursor(&live);
  initCursor(&dead);
  appendThread(&live, &pmain);
  Alarm(hdr_ptr,mallocTimer(1, switch_to, 1, 1));
}

void switch_to(int signo)
{
  sigset_t mask,old;
  sigemptyset(&mask);
  sigaddset(&mask, SIGALRM);
  sigprocmask(SIG_BLOCK, &mask, &old);
  Regs regs;
  //printf("");
  if(signo == -1)
  {
    regs = live.current->data->regs;
    sigprocmask(SIG_SETMASK, &old, NULL);
    JMP(regs);
    assert(0);
  }
  
  int _edi;
  int _esi;
  int _ebp;
  int _esp;
  int _ebx;
  int _edx;
  int _ecx;
  int _eax;
  int _eip = &&_REENTERPOINT;
  int _eflags;
  live.current->data->__reenter = 0; 
  /* save current context */
  SAVE();

  /* save context in current thread */
  live.current->data->regs._eflags = _eflags;
  live.current->data->regs._eip = _eip;
  live.current->data->regs._eax = _eax;
  live.current->data->regs._ecx = _ecx;
  live.current->data->regs._edx = _edx;
  live.current->data->regs._ebx = _ebx;
  live.current->data->regs._esp = _esp;
  live.current->data->regs._ebp = _ebp;
  live.current->data->regs._esi = _esi;
  live.current->data->regs._edi = _edi;

  if(!live.current->data->__reenter)
  {
    goto _END;
  }

_REENTERPOINT:
  regs = live.current->data->regs;

  if(live.current->data->__reenter){
    live.current->data->__reenter = 0;
    sigprocmask(SIG_SETMASK, &old, NULL);
    return;
  } 

_END:
  live.current->data->__reenter = 1;
  regs = live.current->next->data->regs;
  live.current = live.current->next;
  sigprocmask(SIG_SETMASK, &old, NULL);
  JMP(regs);
  assert(0);
}

/************************* Test *************************/
/**
 * Note: The implementation is really bugy, right now only support compute in thread.
 * Even standard I/O in the thread will cause I/O bus error or segmentation error because
 * all pthread-reentrant function is not guaranteed in our thread model.
 * (pthread_mutex_t cannot block thread in our model cause we modify eip directly)
 */
void *sum1tod(void *d)
{
  int i, k, j=0;

  for (i = 0; i <= (int)d; ++i)
  {
      /* code */
      j+=i;   
  }
  return ((void *)j);
}

int main(int argc, char const *argv[])
{
  int res = 0;
  int i;
  init();
  Thread_t *tid1, *tid2;
  int *res1, *res2;
  threadCreat(&tid1, sum1tod, 100);
  threadCreat(&tid2, sum1tod, 100);
  for (i = 0; i <= 100; ++i){
    res+=i;
  }

  threadJoin(tid1, &res1);
  threadJoin(tid2, &res2);
  printf("parallel compute: %d = 5050 * 3\n", (int)res1+(int)res2+(int)res);
  return 0;
}

以上這篇C語言實(shí)現(xiàn)用戶態(tài)線程庫案例就是小編分享給大家的全部?jī)?nèi)容了,希望能給大家一個(gè)參考,也希望大家多多支持腳本之家。

相關(guān)文章

  • 詳解C++標(biāo)準(zhǔn)庫中處理正則表達(dá)式的類std::regex

    詳解C++標(biāo)準(zhǔn)庫中處理正則表達(dá)式的類std::regex

    std?是?C++?標(biāo)準(zhǔn)庫的命名空間,包含了大量標(biāo)準(zhǔn)的?C++?類、函數(shù)和對(duì)象,這些類和函數(shù)提供了廣泛的功能,包括輸入輸出、容器、算法、字符串處理等,這篇文章主要介紹了C++標(biāo)準(zhǔn)庫中提供的用于處理正則表達(dá)式的類std::regex,需要的朋友可以參考下
    2024-03-03
  • C++深入刨析muduo中的抽象類Poller

    C++深入刨析muduo中的抽象類Poller

    muduo網(wǎng)絡(luò)庫中Poller類是一個(gè)抽象類,用戶使用PollPoller或者EPollPoller類,下面跟隨小編一起來詳細(xì)了解一下
    2022-04-04
  • C++實(shí)現(xiàn)雙向冒泡排序算法

    C++實(shí)現(xiàn)雙向冒泡排序算法

    這篇文章主要為大家詳細(xì)介紹了C++實(shí)現(xiàn)雙向冒泡排序算法,文中示例代碼介紹的非常詳細(xì),具有一定的參考價(jià)值,感興趣的小伙伴們可以參考一下
    2020-04-04
  • C++類模板與模板類深入詳解

    C++類模板與模板類深入詳解

    這篇文章主要介紹了C++類模板與模板類深入詳解,需要的朋友可以參考下
    2014-07-07
  • C++中Boost庫裁剪與其應(yīng)用詳解

    C++中Boost庫裁剪與其應(yīng)用詳解

    大家都知道STL全稱"標(biāo)準(zhǔn)模板庫(Standard Template Library)",其實(shí)它是一套標(biāo)準(zhǔn),可能有不同的實(shí)現(xiàn),它是 C++ 的"標(biāo)準(zhǔn)庫"。Boost 則是一個(gè) C++ 庫,被稱為"C++ 準(zhǔn)標(biāo)準(zhǔn)庫"。那么這篇文章文章我們就來詳細(xì)的介紹C++Boost庫的裁剪與其應(yīng)用,感興趣的朋友可以一起學(xué)習(xí)。
    2016-10-10
  • C語言實(shí)現(xiàn)掃雷游戲及其優(yōu)化

    C語言實(shí)現(xiàn)掃雷游戲及其優(yōu)化

    這篇文章主要為大家詳細(xì)介紹了C語言實(shí)現(xiàn)掃雷游戲及其優(yōu)化,文中示例代碼介紹的非常詳細(xì),具有一定的參考價(jià)值,感興趣的小伙伴們可以參考一下
    2019-08-08
  • C++11新特性之右值引用與完美轉(zhuǎn)發(fā)詳解

    C++11新特性之右值引用與完美轉(zhuǎn)發(fā)詳解

    C++11標(biāo)準(zhǔn)為C++引入右值引用語法的同時(shí),還解決了一個(gè)短板,即使用簡(jiǎn)單的方式即可在函數(shù)模板中實(shí)現(xiàn)參數(shù)的完美轉(zhuǎn)發(fā)。本文就來講講二者的應(yīng)用,需要的可以參考一下
    2022-09-09
  • 詳細(xì)解析命令行的getopt_long()函數(shù)

    詳細(xì)解析命令行的getopt_long()函數(shù)

    getopt_long支持長(zhǎng)選項(xiàng)的命令行解析,函數(shù)中的參數(shù)argc和argv通常直接從main()的兩個(gè)參數(shù)傳遞而來
    2013-09-09
  • 解析C++編程中的#include和條件編譯

    解析C++編程中的#include和條件編譯

    這篇文章主要介紹了解析C++編程中的#include和條件編譯,是C++入門學(xué)習(xí)中的基礎(chǔ)知識(shí),需要的朋友可以參考下
    2015-09-09
  • 基于Matlab實(shí)現(xiàn)水波倒影特效的制作

    基于Matlab實(shí)現(xiàn)水波倒影特效的制作

    這篇文章主要介紹了如何利用Matlab制作出水波倒影的特效,文中的示例代碼講解詳細(xì),對(duì)我們學(xué)習(xí)Matlab有一定幫助,需要的可以參考一下
    2022-03-03

最新評(píng)論

丘北县| 象山县| 常山县| 廊坊市| 哈密市| 新余市| 壶关县| 那曲县| 通州市| 昭觉县| 达日县| 定州市| 三原县| 荥阳市| 边坝县| 托里县| 龙泉市| 枞阳县| 兰溪市| 兴文县| 金沙县| 邹平县| 大足县| 商河县| 宁乡县| 嘉黎县| 宜丰县| 吉林市| 霍城县| 内江市| 吉安县| 襄汾县| 乡城县| 铜鼓县| 柯坪县| 莎车县| 宝丰县| 佳木斯市| 蕲春县| 黎城县| 上饶县|