Question

所以我运行这个命令：

$ redis-cli --intrinsic-latency 100 
... some lines ...

11386032 total runs (avg latency: 8.7827 microseconds / 87826.91 nanoseconds per run).
Worst run took 5064x longer than the average latency.

此报告中的问题是87826.91纳秒不等于至8.7827微秒。
正确答案为8782.69纳秒

关于版本：

$ redis-cli -v      
redis-cli 3.0.5

$ redis-server -v
Redis server v=3.0.5 sha=00000000:0 malloc=jemalloc-3.6.0 bits=64 build=9e32aff68ca15a3f

Answer 1

在redis-cli.c中有以下代码：

static void intrinsicLatencyMode(void) {
    .......
    double avg_us = (double)run_time/runs;
    double avg_ns = avg_us * 10e3;
    if (force_cancel_loop || end > test_end) {
        printf("\n%lld total runs "
            "(avg latency: "
            "%.4f microseconds / %.2f nanoseconds per run).\n",
            runs, avg_us, avg_ns);
        printf("Worst run took %.0fx longer than the average latency.\n",
            max_latency / avg_us);
        exit(0);
    }

问题在于将微秒转换为纳秒的行：

double avg_ns = avg_us * 10e3;

代替10e3，代码应该使用1e3：

>gdb -q
(gdb) print 10e3
$1 = 10000
(gdb) print 1e3
$2 = 1000
(gdb)

redis-cli中的纳秒次数不正确 - 内在延迟

1 个答案: