简体繁体 English

用Clocks（）测量CPU时间

[英]Measuring CPU Time with clocks()

原文 2015-04-07 12:41:12 9 2 c/ linux/ time/ linux-kernel/ libc

Libc provides the clock function for measuring the CPU Time of a Linux process. Libc提供了时钟功能，用于测量Linux进程的CPU时间。 I wonder if this approach is still reliable/meaningful in modern computers? 我想知道这种方法在现代计算机中是否仍然可靠/有意义？ Why "CLOCKS_PER_SEC" is a constant? 为什么“ CLOCKS_PER_SEC”是一个常量？ Why is 1e6 instructions per second assumed for every machine? 为什么每台机器假设每秒1e6条指令？ Moreover, modern processors even scale the clock frequency. 此外，现代处理器甚至可以扩展时钟频率。

2 个解决方案

No, it's not relevant anymore; 不，不再相关了； it was poorly designed when it was added, and no longer fits its intended use case. 添加时它的设计很差，不再适合其预期的用例。 Many of the details are frozen to keep backwards compatibility. 许多细节被冻结以保持向后兼容性。

Use POSIX.1-2001 clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &ts) instead. 请改用POSIX.1-2001 clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &ts) 。 You can theoretically reach nanosecond precision using it. 理论上，您可以使用它达到纳秒级精度。

Let me share my understanding. 让我分享我的理解。

CLOCKS_PER_SEC have no direct relation with CPU Clock. CLOCKS_PER_SEC与CPU时钟没有直接关系。 Imagine that you have abstract timer which configured with frequency 1000000 ticks per second. 想象一下，您有一个抽象计时器，它的配置频率为每秒1000000个滴答。 Actually this frequency is very low and could be achieved literally everywhere by dividing MAIN_CLK. 实际上，该频率非常低，实际上可以通过除以MAIN_CLK来实现。 If your OS wants to support POSIX syscall clock, then it must implement this clock() syscall which returns quantity of this low freq timer ticks. 如果您的操作系统要支持POSIX系统调用时钟，则它必须实现此clock（）系统调用，该时钟返回此低频率计时器滴答的数量。 That is why you can measure only with 1ms granularity with clock(). 因此，使用clock（）只能以1ms的粒度进行测量。