[Webkit-unassigned] [Bug 243810] New: Watchdog can't work on Mac with m1 chip because wrong thread cpu time.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=243810

            Bug ID: 243810
           Summary: Watchdog can't work on Mac with m1 chip because wrong
                    thread cpu time.
           Product: WebKit
           Version: Other
          Hardware: Unspecified
                OS: macOS 12
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: JavaScriptCore
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: yuweishan at 126.com

Watchdog.cpp use CPUTime::forCurrentThread, which use clock_gettime(CLOCK_THREAD_CPUTIME_ID, &ts), to get the cpu clock of thread. But on Mac with m1, the clock_gettime(CLOCK_THREAD_CPUTIME_ID) is always incredibly small.


Here is the test code:

```
double CPUTime()
{
    struct timespec ts;
    int ret = clock_gettime(CLOCK_THREAD_CPUTIME_ID, &ts);
    return ts.tv_sec + ts.tv_nsec/1000.0/1000.0/1000.0;
}

double RealTime()
{
    struct timespec ts;
    int ret = clock_gettime(CLOCK_MONOTONIC, &ts);
    return ts.tv_sec + ts.tv_nsec/1000.0/1000.0/1000.0;
}

void longTask()
{
    double begin = RealTime();
    double limit = 1.0;
    static double g_result = 0;
    while (1) {
        for (int i=0; i<1000000; i++) {
            g_result = sin(g_result+5);
        }
        double now = RealTime();
        if (now - begin >= limit) break;
    }
    return;
}

void testClocks()
{
    double t1, t0, cpuTimeCost, realTimeCost;

    t0 = CPUTime();
    longTask();
    t1 = CPUTime();
    cpuTimeCost = t1 - t0;

    double t10 = RealTime();
    longTask();
    double t11 = RealTime();
    realTimeCost = t11 - t10;
    NSLog(@"\nthread cpu: %lfs\nreal: %lfs", cpuTimeCost, realTimeCost);
    return;
}
```

It will output: 

```
thread cpu: 0.024395s
real: 1.006627s
```

Only the M1 chip has this problem. On Mac with Intel, these two clocks are almost the same.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20220811/021f5e3a/attachment.htm>