一、核心公式(工程上真正使用的)
进程 CPU 使用率必须是“两个时间点之间的差值”,公式如下:
CPU% = (Δ(utime + stime) / Δtotal_cpu_time) × NCPU × 100%
或更常见等价形式:
CPU% = Δproc_time / Δwall_time × 100%
取决于你用哪种参照系。
下面解释清楚。
二、utime / stime 是什么单位?
单位不是秒,而是:
clock ticks
转换成秒:
seconds = ticks / sysconf(_SC_CLK_TCK)
查看:
getconf CLK_TCK
通常:
100
表示:
1 tick = 10 ms
三、方法 1:最直观工程计算法(推荐)
步骤:
Step 1:读取进程 CPU 时间
cat /proc/<pid>/stat
第 14、15 字段:
utime stime
例如:
第一次:
utime = 1200
stime = 300
total = 1500
1 秒后:
utime = 1300
stime = 400
total = 1700
增量:
Δproc = 200 ticks
Step 2:换算为秒
如果 CLK_TCK = 100:
200 ticks = 2 seconds CPU time
Step 3:除以实际时间
假设采样间隔 1 秒:
CPU% = 2 / 1 × 100% = 200%
表示:
使用了 2 个 CPU 核心满载
四、为什么可以超过 100%?
因为进程可以多线程并行:
1 核满载 = 100%
2 核满载 = 200%
8 核满载 = 800%
top 的显示就是这样。
五、方法 2:用 /proc/stat 作为基准(更精确)
读取:
cat /proc/stat
第一行:
cpu user nice system idle iowait irq softirq steal …
计算:
total_cpu_time = 所有字段之和
然后:
CPU% = Δproc_time / Δtotal_cpu_time × NCPU × 100%
这是 top 的实现方式。
六、完整 Python 示例(工程级)
import time
CLK_TCK = 100
def read_proc_time(pid):
with open(f"/proc/{pid}/stat") as f:
fields = f.read().split()
utime = int(fields[13])
stime = int(fields[14])
return utime + stime
pid = 1234
t1 = read_proc_time(pid)
time.sleep(1)
t2 = read_proc_time(pid)
delta = t2 – t1
cpu_percent = delta / CLK_TCK * 100
print(cpu_percent, "%")
七、utime vs stime 的工程意义
它们回答不同问题:
| utime | 用户态执行时间 | 算法 / 计算 |
| stime | 内核态执行时间 | syscall / IO / kernel |
八、工程诊断意义(非常重要)
情况 1:utime 高
utime >> stime
说明:
CPU 花在用户代码
原因:
- 算法复杂
- 死循环
- 计算密集
情况 2:stime 高
stime >> utime
说明:
CPU 花在内核
原因:
- write()
- read()
- 网络
- 文件系统
- 锁竞争
情况 3:CPU% 低,但进程卡
检查:
D state
说明:
在等 IO
不是 CPU 问题。
九、top 的 %CPU 本质就是这个公式
top 每隔一段时间:
读取 proc stat
计算差值
显示 %
不是瞬时值。
十、一张完整因果链图
#mermaid-svg-IeFCxSuCTm4tThe6{font-family:\”trebuchet ms\”,verdana,arial,sans-serif;font-size:16px;fill:#333;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#mermaid-svg-IeFCxSuCTm4tThe6 .error-icon{fill:#552222;}#mermaid-svg-IeFCxSuCTm4tThe6 .error-text{fill:#552222;stroke:#552222;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-thickness-normal{stroke-width:1px;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-thickness-thick{stroke-width:3.5px;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-pattern-solid{stroke-dasharray:0;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-svg-IeFCxSuCTm4tThe6 .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-svg-IeFCxSuCTm4tThe6 .marker{fill:#333333;stroke:#333333;}#mermaid-svg-IeFCxSuCTm4tThe6 .marker.cross{stroke:#333333;}#mermaid-svg-IeFCxSuCTm4tThe6 svg{font-family:\”trebuchet ms\”,verdana,arial,sans-serif;font-size:16px;}#mermaid-svg-IeFCxSuCTm4tThe6 p{margin:0;}#mermaid-svg-IeFCxSuCTm4tThe6 .label{font-family:\”trebuchet ms\”,verdana,arial,sans-serif;color:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 .cluster-label text{fill:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 .cluster-label span{color:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 .cluster-label span p{background-color:transparent;}#mermaid-svg-IeFCxSuCTm4tThe6 .label text,#mermaid-svg-IeFCxSuCTm4tThe6 span{fill:#333;color:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 .node rect,#mermaid-svg-IeFCxSuCTm4tThe6 .node circle,#mermaid-svg-IeFCxSuCTm4tThe6 .node ellipse,#mermaid-svg-IeFCxSuCTm4tThe6 .node polygon,#mermaid-svg-IeFCxSuCTm4tThe6 .node path{fill:#ECECFF;stroke:#9370DB;stroke-width:1px;}#mermaid-svg-IeFCxSuCTm4tThe6 .rough-node .label text,#mermaid-svg-IeFCxSuCTm4tThe6 .node .label text,#mermaid-svg-IeFCxSuCTm4tThe6 .image-shape .label,#mermaid-svg-IeFCxSuCTm4tThe6 .icon-shape .label{text-anchor:middle;}#mermaid-svg-IeFCxSuCTm4tThe6 .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-svg-IeFCxSuCTm4tThe6 .rough-node .label,#mermaid-svg-IeFCxSuCTm4tThe6 .node .label,#mermaid-svg-IeFCxSuCTm4tThe6 .image-shape .label,#mermaid-svg-IeFCxSuCTm4tThe6 .icon-shape .label{text-align:center;}#mermaid-svg-IeFCxSuCTm4tThe6 .node.clickable{cursor:pointer;}#mermaid-svg-IeFCxSuCTm4tThe6 .root .anchor path{fill:#333333!important;stroke-width:0;stroke:#333333;}#mermaid-svg-IeFCxSuCTm4tThe6 .arrowheadPath{fill:#333333;}#mermaid-svg-IeFCxSuCTm4tThe6 .edgePath .path{stroke:#333333;stroke-width:2.0px;}#mermaid-svg-IeFCxSuCTm4tThe6 .flowchart-link{stroke:#333333;fill:none;}#mermaid-svg-IeFCxSuCTm4tThe6 .edgeLabel{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-IeFCxSuCTm4tThe6 .edgeLabel p{background-color:rgba(232,232,232, 0.8);}#mermaid-svg-IeFCxSuCTm4tThe6 .edgeLabel rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-IeFCxSuCTm4tThe6 .labelBkg{background-color:rgba(232, 232, 232, 0.5);}#mermaid-svg-IeFCxSuCTm4tThe6 .cluster rect{fill:#ffffde;stroke:#aaaa33;stroke-width:1px;}#mermaid-svg-IeFCxSuCTm4tThe6 .cluster text{fill:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 .cluster span{color:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:\”trebuchet ms\”,verdana,arial,sans-serif;font-size:12px;background:hsl(80, 100%, 96.2745098039%);border:1px solid #aaaa33;border-radius:2px;pointer-events:none;z-index:100;}#mermaid-svg-IeFCxSuCTm4tThe6 .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#333;}#mermaid-svg-IeFCxSuCTm4tThe6 rect.text{fill:none;stroke-width:0;}#mermaid-svg-IeFCxSuCTm4tThe6 .icon-shape,#mermaid-svg-IeFCxSuCTm4tThe6 .image-shape{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-IeFCxSuCTm4tThe6 .icon-shape p,#mermaid-svg-IeFCxSuCTm4tThe6 .image-shape p{background-color:rgba(232,232,232, 0.8);padding:2px;}#mermaid-svg-IeFCxSuCTm4tThe6 .icon-shape rect,#mermaid-svg-IeFCxSuCTm4tThe6 .image-shape rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-IeFCxSuCTm4tThe6 .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#mermaid-svg-IeFCxSuCTm4tThe6 .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#mermaid-svg-IeFCxSuCTm4tThe6 :root{–mermaid-font-family:\”trebuchet ms\”,verdana,arial,sans-serif;}
进程运行
utime 增加
stime 增加
proc total time
两次采样求差
除以时间间隔
得到 CPU%
十一、一句话工程总结
进程 CPU 使用率,本质是“单位时间内消耗的 CPU 时间”,
utime + stime 提供了这个时间,而 CPU% 是它的变化速率。


