perf stat中的公式

时间:2017-09-27 13:50:27

标签: performance profiling performance-testing perf measurement

我想知道perf stat中使用的公式来计算原始数据中的数字。

perf stat -e task-clock,cycles,instructions,cache-references,cache-misses ./myapp

    1080267.226401      task-clock (msec)         #   19.062 CPUs utilized          
 1,592,123,216,789      cycles                    #    1.474 GHz                      (50.00%)
   871,190,006,655      instructions              #    0.55  insn per cycle           (75.00%)
     3,697,548,810      cache-references          #    3.423 M/sec                    (75.00%)
       459,457,321      cache-misses              #   12.426 % of all cache refs      (75.00%)

在这种情况下,如何从缓存引用计算M / sec?

1 个答案:

答案 0 :(得分:2)

公式似乎没有在builtin-stat.cdefault event sets for perf stat are defined)中实施,但可能会在and averaged中计算(perf_stat__print_shadow_stats()带stddev)(以及一些统计信息)被收集到perf_stat__update_shadow_stats())中的数组中:

http://elixir.free-electrons.com/linux/v4.13.4/source/tools/perf/util/stat-shadow.c#L626

计算HW_INSTRUCTIONS时: "每个时钟的指令" = HW_INSTRUCTIONS / HW_CPU_CYCLES; "每个指令的停顿周期" = HW_STALLED_CYCLES_FRONTEND / HW_INSTRUCTIONS

if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
    total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
    if (total) {
        ratio = avg / total;
        print_metric(ctxp, NULL, "%7.2f ",
                "insn per cycle", ratio);
    } else {
        print_metric(ctxp, NULL, NULL, "insn per cycle", 0);
    }

分支未命中来自print_branch_misses为HW_BRANCH_MISSES / HW_BRANCH_INSTRUCTIONS

perf_stat__print_shadow_stats()中有几个缓存未命中率计算,就像HW_CACHE_MISSES / HW_CACHE_REFERENCES和一些更详细的(perf stat -d模式)。

将百分比are computed停止为HW_STALLED_CYCLES_FRONTEND / HW_CPU_CYCLES和HW_STALLED_CYCLES_BACKEND / HW_CPU_CYCLES

GHz计算为HW_CPU_CYCLES / runtime_nsecs_stats,其中runtime_nsecs_stats自2010年起在LKML中从任何软件事件task-clockcpu-clock(SW_TASK_CLOCK& SW_CPU_CLOCK,We still know no exact difference between them two更新和2014年在SO)

if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK) ||
    perf_evsel__match(counter, SOFTWARE, SW_CPU_CLOCK))
    update_stats(&runtime_nsecs_stats[cpu], count[0]);

还有several formulas for transactionsperf stat -T模式)。

"CPU utilized" is from task-clockcpu-clock / walltime_nsecs_stats,其中walltime由the perf stat itself (in userspace使用墙上的时钟计算(天文时间):

static inline unsigned long long rdclock(void)
{
    struct timespec ts;

    clock_gettime(CLOCK_MONOTONIC, &ts);
    return ts.tv_sec * 1000000000ULL + ts.tv_nsec;
}

...

static int __run_perf_stat(int argc, const char **argv)
{    
...
    /*
     * Enable counters and exec the command:
     */
    t0 = rdclock();
    clock_gettime(CLOCK_MONOTONIC, &ref_time);
    if (forks) {
        ....
    }
    t1 = rdclock();

    update_stats(&walltime_nsecs_stats, t1 - t0);

自上而下方法(some estimationsTuning Applications Using a Top-down Microarchitecture Analysis MethodSoftware Optimizations Become Simple with Top-Down Analysis .. Name Skylake, IDF2015中的#22还有Gregg's Methodology List。2016年由Andi Kleen https://lwn.net/Articles/688335/描述"将自上而下的指标添加到perf stat"(perf stat --topdown -I 1000 cmd模式)。

最后,如果目前的印刷活动没有确切的公式,那么通用"%c / sec" (K / sec或M / sec)度量标准:http://elixir.free-electrons.com/linux/v4.13.4/source/tools/perf/util/stat-shadow.c#L845任何除以运行时nsec(任务时钟或cpu-clock事件,如果它们出现在perf stat事件集中)

} else if (runtime_nsecs_stats[cpu].n != 0) {
    char unit = 'M';
    char unit_buf[10];

    total = avg_stats(&runtime_nsecs_stats[cpu]);

    if (total)
        ratio = 1000.0 * avg / total;
    if (ratio < 0.001) {
        ratio *= 1000;
        unit = 'K';
    }
    snprintf(unit_buf, sizeof(unit_buf), "%c/sec", unit);
    print_metric(ctxp, NULL, "%8.3f", unit_buf, ratio);
}