1.
本书在第6章中有过介绍,gettimeofday函数最多精确到us,并且大致精确(并不完全精确),需要多迭代几次减少误差,循环次数太多也会导致结束时间小于开始时间(即回滚)的现象,需要权衡。
2.
代码如下:
#include <stdio.h>
#include <sys/time.h>
#include <stdlib.h>#define PGSIZE 4096//return us
int time_diff(struct timeval* t1, struct timeval* t2) {return (t2->tv_sec - t1->tv_sec) * 1000000 + (t2->tv_usec - t1->tv_usec);
}int main(int argc, char* argv[]) {if (argc != 3) {fprintf(stderr, "too many or too few arguments\n");exit(1);}int i;int pageCount = atoi(argv[1]);int num = atoi(argv[2]);struct timeval curr, now;int size = (pageCount * PGSIZE) / sizeof(int);int arr[size];if (gettimeofday(&curr, NULL) == -1) {fprintf(stderr, "gettimeofday error\n");exit(1);}for (i = 0; i < num; ++i) {arr[(rand() % pageCount) * (PGSIZE / sizeof(int))] = i;}if (gettimeofday(&now, NULL) != 0) {fprintf(stderr, "gettimeofday error\n");exit(1);}printf("time elaps %d us\n", time_diff(&curr, &now));return 0;
}
运行结果如下:
100000次访问消耗1230us,平均每次访问时间为12.3ns
3.
#!/usr/bin/env python
import os
i = 1
while i < 2000 :print '\npage num ' + str(i)val = os.system('./tlb ' + str(i) + ' ' + str(10000))i = i * 2
本处有点问题,可参考python——报错解决:/usr/bin/env: ‘python\r’: No such file or directory
运行结果如下所示:
4.
略
5.
禁止编译器优化,有两个方案,一个是在编译的时候增加选项-O0
6.
如果进程运行移到了另一个cpu,则很容易导致tlb需要重新缓存,命中率大大降低,需要加以限制
7.
修改代码tlb.c,添加初始化逻辑:
测试结果如下:
和题2测试结果相比,确实时间减少了不少