-
Notifications
You must be signed in to change notification settings - Fork 24
Expand file tree
/
Copy pathindex.html
More file actions
615 lines (587 loc) · 45.9 KB
/
index.html
File metadata and controls
615 lines (587 loc) · 45.9 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title data-i18n="title">perf-prof - Linux 实时性能分析框架</title>
<link rel="icon" href="data:image/svg+xml,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 64 64'><polygon points='32,4 55.3,17.5 55.3,46.5 32,60 8.7,46.5 8.7,17.5' fill='%23E8A735'/><rect x='22' y='18' width='20' height='4' rx='2' fill='%23fff'/><rect x='17' y='24.5' width='10' height='4' rx='2' fill='%23fff'/><rect x='30' y='24.5' width='17' height='4' rx='2' fill='%23fff'/><rect x='14' y='31' width='36' height='4' rx='2' fill='%23fff'/><rect x='17' y='37.5' width='17' height='4' rx='2' fill='%23fff'/><rect x='37' y='37.5' width='10' height='4' rx='2' fill='%23fff'/><rect x='22' y='44' width='20' height='4' rx='2' fill='%23fff'/></svg>">
<link rel="stylesheet" href="style.css">
</head>
<body>
<!-- Language toggle -->
<div class="lang-toggle" id="langToggle">
<button class="lang-btn" id="langEn" onclick="setLang('en')">EN</button>
<button class="lang-btn" id="langZh" onclick="setLang('zh')">中文</button>
</div>
<!-- ====== Hero ====== -->
<div class="hero">
<div class="hero-inner">
<div class="logo">
<svg viewBox="0 0 64 64" width="48" height="48"><polygon points="32,4 55.3,17.5 55.3,46.5 32,60 8.7,46.5 8.7,17.5" fill="#E8A735"/><rect x="22" y="18" width="20" height="4" rx="2" fill="#fff"/><rect x="17" y="24.5" width="10" height="4" rx="2" fill="#fff"/><rect x="30" y="24.5" width="17" height="4" rx="2" fill="#fff"/><rect x="14" y="31" width="36" height="4" rx="2" fill="#fff"/><rect x="17" y="37.5" width="17" height="4" rx="2" fill="#fff"/><rect x="37" y="37.5" width="10" height="4" rx="2" fill="#fff"/><rect x="22" y="44" width="20" height="4" rx="2" fill="#fff"/></svg>
<span>perf-prof</span>
</div>
<h1 data-i18n-html="hero_h1">基于 perf_event 的 Linux 系统级分析工具<br>权衡性能与灵活性，兼容旧内核，可长期运行</h1>
<div class="hero-badges">
<span class="badge"><span class="dot c"></span><span data-i18n="badge_c">C 语言</span></span>
<span class="badge"><span class="dot k"></span><span>Linux 3.10+</span></span>
<span class="badge"><span class="dot p"></span><span data-i18n="badge_p">30+ 分析器</span></span>
</div>
<div class="hero-actions">
<a class="btn btn-primary" href="https://github.com/OpenCloudOS/perf-prof" target="_blank" rel="noopener">
<svg width="18" height="18" viewBox="0 0 98 96" fill="currentColor"><path fill-rule="evenodd" clip-rule="evenodd" d="M48.854 0C21.839 0 0 22 0 49.217c0 21.756 13.993 40.172 33.405 46.69 2.427.49 3.316-1.059 3.316-2.362l-.08-9.127c-13.59 2.934-16.42-5.867-16.42-5.867-2.184-5.704-5.42-7.17-5.42-7.17-4.448-3.015.324-3.015.324-3.015 4.934.326 7.523 5.052 7.523 5.052 4.367 7.496 11.404 5.378 14.235 4.074.404-3.178 1.699-5.378 3.074-6.6-10.839-1.141-22.243-5.378-22.243-24.283 0-5.378 1.94-9.778 5.014-13.2-.485-1.222-2.184-6.275.486-13.038 0 0 4.125-1.304 13.426 5.052a46.97 46.97 0 0 1 12.214-1.63c4.125 0 8.33.571 12.213 1.63 9.302-6.356 13.427-5.052 13.427-5.052 2.67 6.763.97 11.816.485 13.038 3.155 3.422 5.015 7.822 5.015 13.2 0 18.905-11.404 23.06-22.324 24.283 1.78 1.548 3.316 4.481 3.316 9.126l-.08 13.526c0 1.304.89 2.853 3.316 2.364 19.412-6.52 33.405-24.935 33.405-46.691C97.707 22 75.788 0 48.854 0z"/></svg>
GitHub
</a>
<a class="btn btn-outline" href="#quick-start" data-i18n="btn_start">快速开始</a>
<a class="btn btn-outline" href="#profilers" data-i18n="btn_profilers">分析器一览</a>
</div>
</div>
</div>
<!-- ====== Architecture ====== -->
<section>
<div class="wrap">
<h2 class="section-title" data-i18n="arch_title">架构概览</h2>
<p class="section-sub" data-i18n="arch_sub">内核态采样 → 三层过滤器 → 环形缓冲区 → 时序排序 → 分析器实时处理</p>
<div class="framework-img">
<img src="images/perf-prof_framework.png" alt="perf-prof framework" loading="lazy">
</div>
</div>
</section>
<!-- ====== Design Philosophy ====== -->
<section>
<div class="wrap">
<h2 class="section-title" data-i18n="phil_title">设计理念：围绕 Tracepoint 构建分析能力</h2>
<p class="section-sub" data-i18n="phil_sub">内核 maintainer 在关键流程上预埋了 tracepoint，理解这些点就是理解内核</p>
<div style="display:grid;gap:20px">
<div class="card">
<h3 data-i18n-html="tp_h3">🎯 Tracepoint：内核的观测基础设施</h3>
<p data-i18n-html="tp_p1">内核 maintainer 在<strong>调度、内存、文件系统、IO、网络</strong>等子系统的关键节点预埋了 tracepoint，每个点标记一个关键的状态转换或决策点。</p>
<p data-i18n-html="tp_p2"><strong>解决性能问题的本质，就是跟踪和分析 tracepoint 之间的关系</strong> —— 事件的先后顺序、延迟、配对、聚合。</p>
</div>
<div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(320px,1fr));gap:16px">
<div class="card">
<h3 style="color:var(--red)" data-i18n-html="raw_h3">⚠ 原始事件的困境</h3>
<p data-i18n-html="raw_p">perf、ftrace 可以采集原始 tracepoint 事件，但一秒内就能产生数十万条。<strong>海量原始数据既超出 AI 上下文窗口，也远超人的阅读能力</strong>，靠逐条分析无法定位问题。</p>
</div>
<div class="card card-accent">
<h3 style="color:var(--lk)" data-i18n-html="sol_h3">✓ perf-prof 的方案</h3>
<p data-i18n="sol_p">在用户态实时处理事件关系，将海量事件压缩为分析结论：</p>
<ul>
<li data-i18n-html="sol_li1">• <strong>multi-trace</strong> — tracepoint 之间的延迟与因果</li>
<li data-i18n-html="sol_li2">• <strong>top / sql</strong> — 按维度聚合，输出 Top-N</li>
<li data-i18n-html="sol_li3">• <strong>task-state</strong> — 调度事件流 → 状态耗时分布</li>
<li data-i18n-html="sol_li4">• <strong>kmemleak</strong> — 配对 alloc/free，只输出泄漏</li>
</ul>
</div>
</div>
<div class="card">
<h3 data-i18n-html="ai_h3">🤖 AI + perf-prof 协作</h3>
<div class="flow">
<div class="flow-step"><strong data-i18n="ai_s1">AI 理解问题</strong><span data-i18n="ai_s1d">选择 tracepoint</span></div>
<div class="flow-arrow">→</div>
<div class="flow-step"><strong data-i18n="ai_s2">perf-prof 处理</strong><span data-i18n="ai_s2d">过滤、关联、聚合</span></div>
<div class="flow-arrow">→</div>
<div class="flow-step"><strong data-i18n="ai_s3">精简结论</strong><span data-i18n="ai_s3d">延迟分布 / Top-N / 泄漏栈</span></div>
<div class="flow-arrow">→</div>
<div class="flow-step accent"><strong data-i18n="ai_s4">AI 解读定位</strong><span data-i18n="ai_s4d">上下文可控，结论可达</span></div>
</div>
<p style="margin-top:14px" data-i18n-html="ai_p">perf-prof 将数十万条原始事件压缩为数行结论，<strong>比 AI 直接消化原始事件流高效几个数量级。</strong></p>
</div>
</div>
</div>
</section>
<!-- ====== Features ====== -->
<section>
<div class="wrap">
<h2 class="section-title" data-i18n="feat_title">核心特性</h2>
<p class="section-sub" data-i18n="feat_sub">事件在内存中实时处理后直接丢弃，可长期运行，安全可靠</p>
<div class="features">
<div class="feat">
<div class="feat-header"><span class="feat-icon">⚡</span><h3 data-i18n="f1_h">内存中实时处理</h3></div>
<p data-i18n="f1_p">采样事件不写文件，处理后直接丢弃。无磁盘瓶颈，适合长期运行。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🔌</span><h3 data-i18n="f2_h">广泛兼容性</h3></div>
<p data-i18n="f2_p">支持 Linux 3.10+（仅需 perf_event），远低于 BCC/bpftrace 的 4.1+ 要求。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🔍</span><h3 data-i18n="f3_h">延迟根因分析</h3></div>
<p data-i18n="f3_p">multi-trace 事件配对与延迟分析，还原中间事件细节。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🛠</span><h3 data-i18n="f4_h">三层过滤 + eBPF</h3></div>
<p data-i18n="f4_p">eBPF / trace event / PMU 三层内核态过滤。支持 eBPF 过滤器和 BPF skeleton 分析器。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">📈</span><h3 data-i18n="f5_h">30+ 内建分析器</h3></div>
<p data-i18n="f5_p">覆盖 CPU、内存、调度、IO、虚拟化等场景，紧急问题直接拼命令。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🐍</span><h3 data-i18n="f6_h">Python 脚本扩展</h3></div>
<p data-i18n="f6_p">事件自动转为 PerfEvent 对象，只需了解字段即可分析，无需编写内核代码。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🔥</span><h3 data-i18n="f7_h">火焰图 & 热图</h3></div>
<p data-i18n="f7_p">内建火焰图折叠栈和延迟热图，一条命令完成可视化。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🌐</span><h3 data-i18n="f8_h">跨主机联合分析</h3></div>
<p data-i18n="f8_p">通过 virtio-ports / TCP 传播事件，Guest ↔ Host 跨系统联合分析。</p>
</div>
<div class="feat">
<div class="feat-header"><span class="feat-icon">🤖</span><h3 data-i18n="f9_h">AI 技能集成</h3></div>
<p data-i18n-html="f9_p">提供 <a href="https://github.com/OpenCloudOS/perf-prof/tree/main/skills/perf-prof">perf-prof skill</a>，AI 自动选择分析器、拼命令、解读结果。</p>
</div>
</div>
</div>
</section>
<!-- ====== Comparison ====== -->
<section>
<div class="wrap">
<h2 class="section-title" data-i18n="cmp_title">工具定位</h2>
<p class="section-sub" data-i18n="cmp_sub">在性能、灵活性、跟踪时效、兼容性之间权衡</p>
<div class="card" style="margin-bottom:28px">
<h3 data-i18n="cmp_prio">Linux 跟踪工具优先级</h3>
<div class="flow" style="margin-bottom:16px">
<div class="flow-step" style="border-radius:8px 0 0 8px"><strong style="color:var(--grn)">bcc / bpftrace / bpftime</strong><span data-i18n="prio_1">性能影响最小，内核态聚合</span></div>
<div class="flow-arrow">›</div>
<div class="flow-step accent"><strong style="color:var(--lk)">perf-prof</strong><span data-i18n="prio_2">兼容旧内核，延迟根因分析</span></div>
<div class="flow-arrow">›</div>
<div class="flow-step"><strong style="color:var(--ylw)">perf / ftrace</strong><span data-i18n="prio_3">短时采样，离线分析</span></div>
<div class="flow-arrow">›</div>
<div class="flow-step" style="border-radius:0 8px 8px 0"><strong style="color:var(--tx3)" data-i18n="prio_4n">其他</strong><span data-i18n="prio_4">特定场景</span></div>
</div>
<p style="font-size:14px;color:var(--tx3);margin:0" data-i18n-html="prio_lang">类比编程语言：<span style="color:var(--grn)">C/C++/Rust</span> › <span style="color:var(--lk)">JavaScript/TypeScript</span> › <span style="color:var(--ylw)">Python</span> › <span style="color:var(--tx3)">Shell</span></p>
</div>
<div class="cmp">
<div class="cmp-item">
<h4>bcc / bpftrace</h4>
<div class="sub" data-i18n="cmp1_sub">类比 C/C++/Rust</div>
<ul>
<li data-i18n="cmp1_l1">性能影响最小，内核态聚合</li>
<li data-i18n="cmp1_l2">高频事件首选</li>
<li data-i18n="cmp1_l3">需 4.1+ 内核</li>
<li data-i18n="cmp1_l4">需掌握 eBPF C 代码细节</li>
</ul>
</div>
<div class="cmp-item highlight">
<h4>perf-prof</h4>
<div class="sub" data-i18n="cmp2_sub">类比 JavaScript/TypeScript</div>
<ul>
<li data-i18n="cmp2_l1">兼容 3.10 旧内核</li>
<li data-i18n="cmp2_l2">延迟根因分析（multi-trace）</li>
<li data-i18n="cmp2_l3">紧急问题快速拼命令</li>
<li data-i18n="cmp2_l4">eBPF 过滤器 + BPF 分析器</li>
<li data-i18n="cmp2_l5">Python 分析器降低门槛</li>
</ul>
</div>
<div class="cmp-item">
<h4>perf / ftrace</h4>
<div class="sub" data-i18n="cmp3_sub">类比 Python</div>
<ul>
<li data-i18n="cmp3_l1">短时采样后离线分析</li>
<li data-i18n="cmp3_l2">无长期运行需求时使用</li>
<li data-i18n="cmp3_l3">事件写文件，磁盘开销</li>
</ul>
</div>
</div>
<div style="margin-top:24px;display:grid;grid-template-columns:repeat(auto-fit,minmax(300px,1fr));gap:14px">
<div class="card">
<h3 style="color:var(--grn);font-size:15px" data-i18n-html="pick1_h">❯ 优先选 bcc/bpftrace</h3>
<p style="margin:0" data-i18n="pick1_p">性能影响最小化，高频事件内核态聚合，新内核（4.1+）环境</p>
</div>
<div class="card card-accent">
<h3 style="color:var(--lk);font-size:15px" data-i18n-html="pick2_h">❯ 选择 perf-prof</h3>
<p style="margin:0" data-i18n="pick2_p">旧内核（3.10+）且需要性能保障；延迟根因分析；紧急问题快速拼命令</p>
</div>
<div class="card">
<h3 style="color:var(--ylw);font-size:15px" data-i18n-html="pick3_h">❯ 可以用 perf/ftrace</h3>
<p style="margin:0" data-i18n="pick3_p">无长期采样需求，短时采样后离线分析即可</p>
</div>
</div>
</div>
</section>
<!-- ====== Python Analyzer ====== -->
<section>
<div class="wrap">
<h2 class="section-title" data-i18n="py_title">Python 分析器</h2>
<p class="section-sub" data-i18n="py_sub">降低跟踪分析的门槛，让更多人能分析性能问题</p>
<div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(300px,1fr));gap:16px">
<div class="card">
<h3 style="color:var(--red);font-size:15px" data-i18n="py_bcc_h">bcc 的门槛</h3>
<ul>
<li data-i18n-html="py_bcc1">✗ 需掌握 eBPF C 代码（不同 prog 类型参数不同）</li>
<li data-i18n-html="py_bcc2">✗ 需了解不同 prog 可调用的 bpf helper</li>
<li data-i18n-html="py_bcc3">✗ 需理解 perf_event、tracepoint、kprobe、poll</li>
</ul>
</div>
<div class="card">
<h3 style="color:var(--ylw);font-size:15px" data-i18n="py_bt_h">bpftrace 的门槛</h3>
<ul>
<li data-i18n-html="py_bt1">✗ 需学习一套全新的跟踪语言</li>
<li data-i18n-html="py_bt2">✗ 大量新函数需要记忆</li>
<li data-i18n-html="py_bt3">✗ 限制性较大，复杂逻辑难以表达</li>
</ul>
</div>
<div class="card card-accent">
<h3 style="color:var(--lk);font-size:15px">perf-prof python</h3>
<ul>
<li data-i18n-html="py_pp1">✓ 采样事件自动转换为 PerfEvent 对象</li>
<li data-i18n-html="py_pp2">✓ 不需要关注如何采样和内核逻辑</li>
<li data-i18n-html="py_pp3">✓ 只需了解事件成员即可开始分析</li>
<li data-i18n-html="py_pp4">✓ 使用 <code>help</code> 生成模板后修改</li>
</ul>
</div>
</div>
</div>
</section>
<!-- ====== Profilers ====== -->
<section id="profilers">
<div class="wrap">
<h2 class="section-title" data-i18n="prof_title">分析器一览</h2>
<p class="section-sub" data-i18n="prof_sub">每个分析器都是独立模块，针对特定场景优化</p>
<div class="tbl-wrap">
<table>
<thead><tr><th data-i18n="th_prof">分析器</th><th data-i18n="th_func">功能</th><th data-i18n="th_area">领域</th></tr></thead>
<tbody>
<tr><td><a href="profilers/profile.md">profile</a></td><td data-i18n="pd_profile">CPU 采样：热点函数定位、火焰图，支持内核/用户态</td><td><span class="tag tag-cpu">CPU</span></td></tr>
<tr><td><a href="profilers/trace.md">trace</a></td><td data-i18n="pd_trace">通用事件跟踪：打印事件与调用栈，支持 kprobe / uprobe</td><td><span class="tag tag-trace" data-i18n="tag_trace">跟踪</span></td></tr>
<tr><td><a href="profilers/multi-trace.md">multi-trace</a></td><td data-i18n-html="pd_mt">多事件关系：延迟分析、事件配对、<code>--detail</code> 还原中间细节</td><td><span class="tag tag-sched" data-i18n="tag_latency">延迟</span></td></tr>
<tr><td><a href="profilers/top.md">top</a></td><td data-i18n="pd_top">键值聚合统计：按进程/线程/自定义键排序</td><td><span class="tag tag-trace" data-i18n="tag_stat">统计</span></td></tr>
<tr><td><a href="profilers/sql.md">sql</a></td><td data-i18n="pd_sql">SQL 聚合查询：基于 SQLite 的灵活事件分析</td><td><span class="tag tag-trace" data-i18n="tag_stat2">统计</span></td></tr>
<tr><td><a href="profilers/task-state.md">task-state</a></td><td data-i18n="pd_ts">进程状态监控：R/S/D/T 耗时分布与根因分析</td><td><span class="tag tag-sched" data-i18n="tag_sched">调度</span></td></tr>
<tr><td><a href="profilers/oncpu.md">oncpu</a></td><td data-i18n="pd_oncpu">CPU 运行监控：实时显示各 CPU 运行进程及时间</td><td><span class="tag tag-cpu">CPU</span></td></tr>
<tr><td><a href="profilers/blktrace.md">blktrace</a></td><td data-i18n="pd_blk">块设备 IO：全生命周期延迟、毛刺检测</td><td><span class="tag tag-io">IO</span></td></tr>
<tr><td><a href="profilers/kvm-exit.md">kvm-exit</a></td><td data-i18n="pd_kvm">KVM VM-Exit/Entry 延迟分析与退出原因统计</td><td><span class="tag tag-vm" data-i18n="tag_vm">虚拟化</span></td></tr>
<tr><td><a href="profilers/kmemleak.md">kmemleak</a></td><td data-i18n="pd_kml">内存泄漏检测：内核/用户态分配追踪</td><td><span class="tag tag-mem" data-i18n="tag_mem">内存</span></td></tr>
<tr><td><a href="profilers/python.md">python</a></td><td data-i18n="pd_py">Python 脚本：自定义脚本处理 perf 事件</td><td><span class="tag tag-trace" data-i18n="tag_script">脚本</span></td></tr>
<tr><td><a href="profilers/multi-trace.md#syscalls---系统调用耗时分析">syscalls</a></td><td data-i18n="pd_sys">系统调用延迟：完整生命周期分析</td><td><span class="tag tag-sched" data-i18n="tag_latency2">延迟</span></td></tr>
<tr><td><a href="profilers/multi-trace.md#rundelay---调度延迟分析">rundelay</a></td><td data-i18n="pd_rd">调度延迟：唤醒到实际运行的延迟</td><td><span class="tag tag-sched" data-i18n="tag_sched2">调度</span></td></tr>
<tr><td>kmemprof</td><td data-i18n="pd_kmp">内存分配统计：分配器热点与大小分布</td><td><span class="tag tag-mem" data-i18n="tag_mem2">内存</span></td></tr>
<tr><td>bpf:kvm_exit</td><td data-i18n="pd_bpfkvm">基于 eBPF 的 KVM 退出事件生成</td><td><span class="tag tag-vm" data-i18n="tag_vm2">虚拟化</span></td></tr>
<tr><td>watchdog</td><td data-i18n="pd_wd">硬锁/软锁检测：预测 lockup 并输出内核栈</td><td><span class="tag tag-cpu" data-i18n="tag_debug">调试</span></td></tr>
<tr><td>irq-off</td><td data-i18n="pd_irq">中断关闭检测：定位长时间关中断代码段</td><td><span class="tag tag-cpu" data-i18n="tag_debug2">调试</span></td></tr>
<tr><td>num-dist</td><td data-i18n="pd_nd">数值分布：任意字段的直方图与百分位</td><td><span class="tag tag-trace" data-i18n="tag_stat3">统计</span></td></tr>
<tr><td>hwstat / llcstat / tlbstat</td><td data-i18n="pd_hw">硬件计数器：IPC、LLC 命中率、TLB 命中率</td><td><span class="tag tag-cpu" data-i18n="tag_hw">硬件</span></td></tr>
<tr><td>breakpoint</td><td data-i18n="pd_bp">硬件断点：监控指定地址的读写</td><td><span class="tag tag-cpu" data-i18n="tag_debug3">调试</span></td></tr>
</tbody>
</table>
</div>
</div>
</section>
<!-- ====== Problem Guide ====== -->
<section>
<div class="wrap">
<h2 class="section-title" data-i18n="guide_title">按问题类型查找</h2>
<p class="section-sub" data-i18n="guide_sub">快速定位合适的分析器</p>
<div class="guide">
<div class="guide-item">
<h4 data-i18n-html="g1_h">🔴 CPU 使用率高</h4>
<p data-i18n="g1_p">采样热点函数，生成火焰图</p>
<code>profile</code> <code>oncpu</code> <code>cpu-util</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g2_h">🟠 调度延迟</h4>
<p data-i18n="g2_p">唤醒 → 运行延迟、进程状态分析</p>
<code>multi-trace</code> <code>rundelay</code> <code>task-state</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g3_h">🟡 IO 性能</h4>
<p data-i18n="g3_p">块设备 IO 全链路延迟、毛刺检测</p>
<code>blktrace</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g4_h">🟣 内存泄漏</h4>
<p data-i18n="g4_p">分配/释放配对追踪，泄漏字节统计</p>
<code>kmemleak</code> <code>kmemprof</code> <code>page-faults</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g5_h">🔵 虚拟化开销</h4>
<p data-i18n="g5_p">VM-Exit 延迟与退出原因统计</p>
<code>kvm-exit</code> <code>bpf:kvm_exit</code> <code>kvmmmu</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g6_h">⬛ 进程 D/S 状态</h4>
<p data-i18n="g6_p">状态耗时分布与堆栈定位</p>
<code>task-state</code> <code>syscalls</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g7_h">🔸 系统调用慢</h4>
<p data-i18n="g7_p">系统调用延迟全链路分析</p>
<code>syscalls</code> <code>multi-trace</code>
</div>
<div class="guide-item">
<h4 data-i18n-html="g8_h">⬜ 事件聚合统计</h4>
<p data-i18n="g8_p">按维度聚合，Top-N 排序</p>
<code>top</code> <code>sql</code> <code>num-dist</code> <code>stat</code>
</div>
</div>
</div>
</section>
<!-- ====== Quick Start ====== -->
<section id="quick-start">
<div class="wrap">
<h2 class="section-title" data-i18n="qs_title">快速开始</h2>
<p class="section-sub" data-i18n="qs_sub">安装构建与常见分析场景示例</p>
<div class="workflows">
<div class="wf">
<h3 data-i18n-html="qs1_h">① 构建安装</h3>
<pre><code># Install dependencies
yum install -y xz-devel elfutils-libelf-devel libunwind-devel python3-devel
# Clone and build
git clone https://github.com/OpenCloudOS/perf-prof.git
cd perf-prof && make</code></pre>
</div>
<div class="wf">
<h3 data-i18n-html="qs2_h">② CPU 热点分析</h3>
<pre><code># Sample kernel CPU hotspots
perf-prof profile -F 997 -g --exclude-user --than 30
# Generate flame graph
perf-prof profile -F 997 -g --flame-graph cpu.folded
flamegraph.pl cpu.folded.folded > cpu.svg</code></pre>
</div>
<div class="wf">
<h3 data-i18n-html="qs3_h">③ 延迟分析</h3>
<pre><code># Scheduling latency > 4ms, with intermediate details
perf-prof multi-trace \
-e sched:sched_wakeup,sched:sched_wakeup_new \
-e 'sched:sched_switch//key=next_pid/' \
-k pid --order --than 4ms --detail</code></pre>
</div>
<div class="wf">
<h3 data-i18n-html="qs4_h">④ 内存泄漏检测</h3>
<pre><code># Track kmalloc/kfree pairs
perf-prof kmemleak \
--alloc "kmem:kmalloc//ptr=ptr/size=bytes_alloc/stack/" \
--free "kmem:kfree//ptr=ptr/" \
--order -m 128 -g</code></pre>
</div>
<div class="wf">
<h3 data-i18n-html="qs5_h">⑤ IO 跟踪</h3>
<pre><code># IO latency > 10ms
perf-prof blktrace -d /dev/sda -i 1000 --than 10ms
# IO latency heatmap
perf-prof blktrace -d /dev/sda --heatmap io_lat</code></pre>
</div>
<div class="wf">
<h3 data-i18n-html="qs6_h">⑥ 进程状态分析</h3>
<pre><code># D state > 100ms, with call stacks
perf-prof task-state -D --than 100ms -g
# S state analysis
perf-prof task-state -S -p <pid> --than 50ms -g</code></pre>
</div>
</div>
</div>
</section>
<!-- ====== Footer ====== -->
<footer>
<div class="wrap">
<p style="margin-bottom:8px" data-i18n-html="footer_t"><strong>perf-prof</strong> — Linux 实时性能分析框架</p>
<p>
<a href="https://github.com/OpenCloudOS/perf-prof" target="_blank" rel="noopener">GitHub</a>
•
<a href="https://man7.org/linux/man-pages/man2/perf_event_open.2.html" target="_blank" rel="noopener" data-i18n="footer_perf">perf_event 文档</a>
•
<a href="https://www.kernel.org/doc/html/latest/trace/events.html" target="_blank" rel="noopener" data-i18n="footer_tp">Tracepoint 文档</a>
</p>
</div>
</footer>
<script>
var T = {
// title
title: { zh: 'perf-prof - Linux 实时性能分析框架', en: 'perf-prof - Linux Real-time Performance Analysis Framework' },
// hero
hero_h1: { zh: '基于 perf_event 的 Linux 系统级分析工具<br>权衡性能与灵活性，兼容旧内核，可长期运行', en: 'Linux System-level Analysis Tool Based on perf_event<br>Balancing performance and flexibility, compatible with legacy kernels, designed for long-running' },
badge_c: { zh: 'C 语言', en: 'C Language' },
badge_p: { zh: '30+ 分析器', en: '30+ Profilers' },
btn_start: { zh: '快速开始', en: 'Quick Start' },
btn_profilers: { zh: '分析器一览', en: 'Profilers' },
// architecture
arch_title: { zh: '架构概览', en: 'Architecture Overview' },
arch_sub: { zh: '内核态采样 → 三层过滤器 → 环形缓冲区 → 时序排序 → 分析器实时处理', en: 'Kernel sampling → 3-layer filters → Ring buffer → Time-ordered → Real-time profiler processing' },
// design philosophy
phil_title: { zh: '设计理念：围绕 Tracepoint 构建分析能力', en: 'Design Philosophy: Building Analysis Around Tracepoints' },
phil_sub: { zh: '内核 maintainer 在关键流程上预埋了 tracepoint，理解这些点就是理解内核', en: 'Kernel maintainers embed tracepoints at critical paths — understanding these points is understanding the kernel' },
tp_h3: { zh: '🎯 Tracepoint：内核的观测基础设施', en: '🎯 Tracepoint: The Kernel\'s Observability Infrastructure' },
tp_p1: { zh: '内核 maintainer 在<strong>调度、内存、文件系统、IO、网络</strong>等子系统的关键节点预埋了 tracepoint，每个点标记一个关键的状态转换或决策点。', en: 'Kernel maintainers embed tracepoints at critical nodes in <strong>scheduling, memory, filesystem, IO, networking</strong> subsystems, each marking a key state transition or decision point.' },
tp_p2: { zh: '<strong>解决性能问题的本质，就是跟踪和分析 tracepoint 之间的关系</strong> —— 事件的先后顺序、延迟、配对、聚合。', en: '<strong>Solving performance problems is essentially about tracing and analyzing relationships between tracepoints</strong> — ordering, latency, pairing, and aggregation.' },
raw_h3: { zh: '⚠ 原始事件的困境', en: '⚠ The Raw Event Dilemma' },
raw_p: { zh: 'perf、ftrace 可以采集原始 tracepoint 事件，但一秒内就能产生数十万条。<strong>海量原始数据既超出 AI 上下文窗口，也远超人的阅读能力</strong>，靠逐条分析无法定位问题。', en: 'perf and ftrace can collect raw tracepoint events, but hundreds of thousands can be generated per second. <strong>The massive raw data overwhelms both AI context windows and human reading capacity</strong>, making line-by-line analysis impractical.' },
sol_h3: { zh: '✓ perf-prof 的方案', en: '✓ The perf-prof Solution' },
sol_p: { zh: '在用户态实时处理事件关系，将海量事件压缩为分析结论：', en: 'Process event relationships in user space in real time, compressing massive events into analytical conclusions:' },
sol_li1: { zh: '• <strong>multi-trace</strong> — tracepoint 之间的延迟与因果', en: '• <strong>multi-trace</strong> — Latency and causality between tracepoints' },
sol_li2: { zh: '• <strong>top / sql</strong> — 按维度聚合，输出 Top-N', en: '• <strong>top / sql</strong> — Aggregate by dimensions, output Top-N' },
sol_li3: { zh: '• <strong>task-state</strong> — 调度事件流 → 状态耗时分布', en: '• <strong>task-state</strong> — Scheduling events → State duration distribution' },
sol_li4: { zh: '• <strong>kmemleak</strong> — 配对 alloc/free，只输出泄漏', en: '• <strong>kmemleak</strong> — Pair alloc/free, report only leaks' },
ai_h3: { zh: '🤖 AI + perf-prof 协作', en: '🤖 AI + perf-prof Collaboration' },
ai_s1: { zh: 'AI 理解问题', en: 'AI Understands' },
ai_s1d: { zh: '选择 tracepoint', en: 'Select tracepoints' },
ai_s2: { zh: 'perf-prof 处理', en: 'perf-prof Processes' },
ai_s2d: { zh: '过滤、关联、聚合', en: 'Filter, correlate, aggregate' },
ai_s3: { zh: '精简结论', en: 'Condensed Results' },
ai_s3d: { zh: '延迟分布 / Top-N / 泄漏栈', en: 'Latency dist / Top-N / Leak stacks' },
ai_s4: { zh: 'AI 解读定位', en: 'AI Interprets' },
ai_s4d: { zh: '上下文可控，结论可达', en: 'Context manageable, conclusions actionable' },
ai_p: { zh: 'perf-prof 将数十万条原始事件压缩为数行结论，<strong>比 AI 直接消化原始事件流高效几个数量级。</strong>', en: 'perf-prof compresses hundreds of thousands of raw events into a few lines of conclusions, <strong>orders of magnitude more efficient than AI consuming raw event streams.</strong>' },
// features
feat_title: { zh: '核心特性', en: 'Core Features' },
feat_sub: { zh: '事件在内存中实时处理后直接丢弃，可长期运行，安全可靠', en: 'Events processed in memory in real time then discarded — long-running, safe, and reliable' },
f1_h: { zh: '内存中实时处理', en: 'Real-time In-memory Processing' },
f1_p: { zh: '采样事件不写文件，处理后直接丢弃。无磁盘瓶颈，适合长期运行。', en: 'Sampled events are never written to disk — processed and discarded. No disk bottleneck, suitable for long-running.' },
f2_h: { zh: '广泛兼容性', en: 'Broad Compatibility' },
f2_p: { zh: '支持 Linux 3.10+（仅需 perf_event），远低于 BCC/bpftrace 的 4.1+ 要求。', en: 'Supports Linux 3.10+ (only requires perf_event), well below the 4.1+ requirement of BCC/bpftrace.' },
f3_h: { zh: '延迟根因分析', en: 'Latency Root Cause Analysis' },
f3_p: { zh: 'multi-trace 事件配对与延迟分析，还原中间事件细节。', en: 'multi-trace event pairing and latency analysis, reconstructing intermediate event details.' },
f4_h: { zh: '三层过滤 + eBPF', en: '3-Layer Filtering + eBPF' },
f4_p: { zh: 'eBPF / trace event / PMU 三层内核态过滤。支持 eBPF 过滤器和 BPF skeleton 分析器。', en: 'eBPF / trace event / PMU three-layer kernel-space filtering. Supports eBPF filters and BPF skeleton profilers.' },
f5_h: { zh: '30+ 内建分析器', en: '30+ Built-in Profilers' },
f5_p: { zh: '覆盖 CPU、内存、调度、IO、虚拟化等场景，紧急问题直接拼命令。', en: 'Covering CPU, memory, scheduling, IO, virtualization scenarios. Quickly compose commands for urgent issues.' },
f6_h: { zh: 'Python 脚本扩展', en: 'Python Scripting' },
f6_p: { zh: '事件自动转为 PerfEvent 对象，只需了解字段即可分析，无需编写内核代码。', en: 'Events auto-converted to PerfEvent objects — just understand the fields to start analyzing, no kernel code needed.' },
f7_h: { zh: '火焰图 & 热图', en: 'Flame Graphs & Heatmaps' },
f7_p: { zh: '内建火焰图折叠栈和延迟热图，一条命令完成可视化。', en: 'Built-in flame graph folded stacks and latency heatmaps — one command for visualization.' },
f8_h: { zh: '跨主机联合分析', en: 'Cross-host Joint Analysis' },
f8_p: { zh: '通过 virtio-ports / TCP 传播事件，Guest ↔ Host 跨系统联合分析。', en: 'Propagate events via virtio-ports / TCP for Guest ↔ Host cross-system joint analysis.' },
f9_h: { zh: 'AI 技能集成', en: 'AI Skill Integration' },
f9_p: { zh: '提供 <a href="https://github.com/OpenCloudOS/perf-prof/tree/main/skills/perf-prof">perf-prof skill</a>，AI 自动选择分析器、拼命令、解读结果。', en: 'Provides <a href="https://github.com/OpenCloudOS/perf-prof/tree/main/skills/perf-prof">perf-prof skill</a> — AI automatically selects profilers, composes commands, and interprets results.' },
// comparison
cmp_title: { zh: '工具定位', en: 'Tool Positioning' },
cmp_sub: { zh: '在性能、灵活性、跟踪时效、兼容性之间权衡', en: 'Balancing performance, flexibility, tracing timeliness, and compatibility' },
cmp_prio: { zh: 'Linux 跟踪工具优先级', en: 'Linux Tracing Tool Priority' },
prio_1: { zh: '性能影响最小，内核态聚合', en: 'Minimal overhead, kernel-space aggregation' },
prio_2: { zh: '兼容旧内核，延迟根因分析', en: 'Legacy kernel compatible, latency root cause' },
prio_3: { zh: '短时采样，离线分析', en: 'Short sampling, offline analysis' },
prio_4n: { zh: '其他', en: 'Others' },
prio_4: { zh: '特定场景', en: 'Specific scenarios' },
prio_lang: { zh: '类比编程语言：<span style="color:var(--grn)">C/C++/Rust</span> › <span style="color:var(--lk)">JavaScript/TypeScript</span> › <span style="color:var(--ylw)">Python</span> › <span style="color:var(--tx3)">Shell</span>', en: 'Language analogy: <span style="color:var(--grn)">C/C++/Rust</span> › <span style="color:var(--lk)">JavaScript/TypeScript</span> › <span style="color:var(--ylw)">Python</span> › <span style="color:var(--tx3)">Shell</span>' },
cmp1_sub: { zh: '类比 C/C++/Rust', en: 'Analogy: C/C++/Rust' },
cmp1_l1: { zh: '性能影响最小，内核态聚合', en: 'Minimal overhead, kernel-space aggregation' },
cmp1_l2: { zh: '高频事件首选', en: 'Best for high-frequency events' },
cmp1_l3: { zh: '需 4.1+ 内核', en: 'Requires kernel 4.1+' },
cmp1_l4: { zh: '需掌握 eBPF C 代码细节', en: 'Requires eBPF C code expertise' },
cmp2_sub: { zh: '类比 JavaScript/TypeScript', en: 'Analogy: JavaScript/TypeScript' },
cmp2_l1: { zh: '兼容 3.10 旧内核', en: 'Compatible with 3.10 legacy kernels' },
cmp2_l2: { zh: '延迟根因分析（multi-trace）', en: 'Latency root cause analysis (multi-trace)' },
cmp2_l3: { zh: '紧急问题快速拼命令', en: 'Quick command composition for urgent issues' },
cmp2_l4: { zh: 'eBPF 过滤器 + BPF 分析器', en: 'eBPF filters + BPF profilers' },
cmp2_l5: { zh: 'Python 分析器降低门槛', en: 'Python profiler lowers the barrier' },
cmp3_sub: { zh: '类比 Python', en: 'Analogy: Python' },
cmp3_l1: { zh: '短时采样后离线分析', en: 'Short sampling then offline analysis' },
cmp3_l2: { zh: '无长期运行需求时使用', en: 'Use when no long-running requirement' },
cmp3_l3: { zh: '事件写文件，磁盘开销', en: 'Events written to files, disk overhead' },
pick1_h: { zh: '❯ 优先选 bcc/bpftrace', en: '❯ Prefer bcc/bpftrace' },
pick1_p: { zh: '性能影响最小化，高频事件内核态聚合，新内核（4.1+）环境', en: 'Minimize overhead, kernel-space aggregation for high-frequency events, kernel 4.1+ environment' },
pick2_h: { zh: '❯ 选择 perf-prof', en: '❯ Choose perf-prof' },
pick2_p: { zh: '旧内核（3.10+）且需要性能保障；延迟根因分析；紧急问题快速拼命令', en: 'Legacy kernel (3.10+) with performance requirements; latency root cause analysis; quick command for urgent issues' },
pick3_h: { zh: '❯ 可以用 perf/ftrace', en: '❯ Use perf/ftrace' },
pick3_p: { zh: '无长期采样需求，短时采样后离线分析即可', en: 'No long-running requirement, short sampling with offline analysis is sufficient' },
// python analyzer
py_title: { zh: 'Python 分析器', en: 'Python Profiler' },
py_sub: { zh: '降低跟踪分析的门槛，让更多人能分析性能问题', en: 'Lowering the barrier to tracing analysis, enabling more people to analyze performance issues' },
py_bcc_h: { zh: 'bcc 的门槛', en: 'bcc Barrier' },
py_bcc1: { zh: '✗ 需掌握 eBPF C 代码（不同 prog 类型参数不同）', en: '✗ Must master eBPF C code (different prog types have different params)' },
py_bcc2: { zh: '✗ 需了解不同 prog 可调用的 bpf helper', en: '✗ Must understand bpf helpers available to each prog type' },
py_bcc3: { zh: '✗ 需理解 perf_event、tracepoint、kprobe、poll', en: '✗ Must understand perf_event, tracepoint, kprobe, poll' },
py_bt_h: { zh: 'bpftrace 的门槛', en: 'bpftrace Barrier' },
py_bt1: { zh: '✗ 需学习一套全新的跟踪语言', en: '✗ Must learn a completely new tracing language' },
py_bt2: { zh: '✗ 大量新函数需要记忆', en: '✗ Many new functions to memorize' },
py_bt3: { zh: '✗ 限制性较大，复杂逻辑难以表达', en: '✗ Limited expressiveness, complex logic is hard to implement' },
py_pp1: { zh: '✓ 采样事件自动转换为 PerfEvent 对象', en: '✓ Sampled events auto-converted to PerfEvent objects' },
py_pp2: { zh: '✓ 不需要关注如何采样和内核逻辑', en: '✓ No need to worry about sampling or kernel internals' },
py_pp3: { zh: '✓ 只需了解事件成员即可开始分析', en: '✓ Just understand event fields to start analyzing' },
py_pp4: { zh: '✓ 使用 <code>help</code> 生成模板后修改', en: '✓ Use <code>help</code> to generate templates, then customize' },
// profiler table
prof_title: { zh: '分析器一览', en: 'Profiler Overview' },
prof_sub: { zh: '每个分析器都是独立模块，针对特定场景优化', en: 'Each profiler is an independent module optimized for specific scenarios' },
th_prof: { zh: '分析器', en: 'Profiler' },
th_func: { zh: '功能', en: 'Function' },
th_area: { zh: '领域', en: 'Domain' },
pd_profile: { zh: 'CPU 采样：热点函数定位、火焰图，支持内核/用户态', en: 'CPU sampling: hotspot identification, flame graphs, kernel/user space' },
pd_trace: { zh: '通用事件跟踪：打印事件与调用栈，支持 kprobe / uprobe', en: 'General event tracing: print events & call stacks, supports kprobe / uprobe' },
pd_mt: { zh: '多事件关系：延迟分析、事件配对、<code>--detail</code> 还原中间细节', en: 'Multi-event correlation: latency analysis, event pairing, <code>--detail</code> for intermediate details' },
pd_top: { zh: '键值聚合统计：按进程/线程/自定义键排序', en: 'Key-value aggregation: sort by process/thread/custom key' },
pd_sql: { zh: 'SQL 聚合查询：基于 SQLite 的灵活事件分析', en: 'SQL aggregation: flexible event analysis based on SQLite' },
pd_ts: { zh: '进程状态监控：R/S/D/T 耗时分布与根因分析', en: 'Process state monitoring: R/S/D/T duration distribution & root cause' },
pd_oncpu: { zh: 'CPU 运行监控：实时显示各 CPU 运行进程及时间', en: 'CPU run monitoring: real-time display of running processes per CPU' },
pd_blk: { zh: '块设备 IO：全生命周期延迟、毛刺检测', en: 'Block device IO: full lifecycle latency, spike detection' },
pd_kvm: { zh: 'KVM VM-Exit/Entry 延迟分析与退出原因统计', en: 'KVM VM-Exit/Entry latency analysis and exit reason statistics' },
pd_kml: { zh: '内存泄漏检测：内核/用户态分配追踪', en: 'Memory leak detection: kernel/user space allocation tracking' },
pd_py: { zh: 'Python 脚本：自定义脚本处理 perf 事件', en: 'Python scripting: custom scripts to process perf events' },
pd_sys: { zh: '系统调用延迟：完整生命周期分析', en: 'Syscall latency: full lifecycle analysis' },
pd_rd: { zh: '调度延迟：唤醒到实际运行的延迟', en: 'Scheduling delay: wakeup to actual run latency' },
pd_kmp: { zh: '内存分配统计：分配器热点与大小分布', en: 'Memory allocation stats: allocator hotspots & size distribution' },
pd_bpfkvm: { zh: '基于 eBPF 的 KVM 退出事件生成', en: 'eBPF-based KVM exit event generation' },
pd_wd: { zh: '硬锁/软锁检测：预测 lockup 并输出内核栈', en: 'Hard/soft lockup detection: predict lockups and dump kernel stacks' },
pd_irq: { zh: '中断关闭检测：定位长时间关中断代码段', en: 'IRQ-off detection: locate long interrupt-disabled code sections' },
pd_nd: { zh: '数值分布：任意字段的直方图与百分位', en: 'Numeric distribution: histograms & percentiles for any field' },
pd_hw: { zh: '硬件计数器：IPC、LLC 命中率、TLB 命中率', en: 'Hardware counters: IPC, LLC hit rate, TLB hit rate' },
pd_bp: { zh: '硬件断点：监控指定地址的读写', en: 'Hardware breakpoint: monitor reads/writes at specified addresses' },
tag_trace: { zh: '跟踪', en: 'Trace' },
tag_latency: { zh: '延迟', en: 'Latency' },
tag_latency2: { zh: '延迟', en: 'Latency' },
tag_stat: { zh: '统计', en: 'Stats' },
tag_stat2: { zh: '统计', en: 'Stats' },
tag_stat3: { zh: '统计', en: 'Stats' },
tag_sched: { zh: '调度', en: 'Sched' },
tag_sched2: { zh: '调度', en: 'Sched' },
tag_vm: { zh: '虚拟化', en: 'VM' },
tag_vm2: { zh: '虚拟化', en: 'VM' },
tag_mem: { zh: '内存', en: 'Memory' },
tag_mem2: { zh: '内存', en: 'Memory' },
tag_script: { zh: '脚本', en: 'Script' },
tag_debug: { zh: '调试', en: 'Debug' },
tag_debug2: { zh: '调试', en: 'Debug' },
tag_debug3: { zh: '调试', en: 'Debug' },
tag_hw: { zh: '硬件', en: 'HW' },
// problem guide
guide_title: { zh: '按问题类型查找', en: 'Find by Problem Type' },
guide_sub: { zh: '快速定位合适的分析器', en: 'Quickly locate the right profiler' },
g1_h: { zh: '🔴 CPU 使用率高', en: '🔴 High CPU Usage' },
g1_p: { zh: '采样热点函数，生成火焰图', en: 'Sample hotspot functions, generate flame graphs' },
g2_h: { zh: '🟠 调度延迟', en: '🟠 Scheduling Latency' },
g2_p: { zh: '唤醒 → 运行延迟、进程状态分析', en: 'Wakeup → run latency, process state analysis' },
g3_h: { zh: '🟡 IO 性能', en: '🟡 IO Performance' },
g3_p: { zh: '块设备 IO 全链路延迟、毛刺检测', en: 'Block device IO full-path latency, spike detection' },
g4_h: { zh: '🟣 内存泄漏', en: '🟣 Memory Leak' },
g4_p: { zh: '分配/释放配对追踪，泄漏字节统计', en: 'Alloc/free pair tracking, leaked bytes statistics' },
g5_h: { zh: '🔵 虚拟化开销', en: '🔵 Virtualization Overhead' },
g5_p: { zh: 'VM-Exit 延迟与退出原因统计', en: 'VM-Exit latency and exit reason statistics' },
g6_h: { zh: '⬛ 进程 D/S 状态', en: '⬛ Process D/S State' },
g6_p: { zh: '状态耗时分布与堆栈定位', en: 'State duration distribution and stack trace location' },
g7_h: { zh: '🔸 系统调用慢', en: '🔸 Slow Syscalls' },
g7_p: { zh: '系统调用延迟全链路分析', en: 'Syscall latency full-path analysis' },
g8_h: { zh: '⬜ 事件聚合统计', en: '⬜ Event Aggregation' },
g8_p: { zh: '按维度聚合，Top-N 排序', en: 'Aggregate by dimension, Top-N sorting' },
// quick start
qs_title: { zh: '快速开始', en: 'Quick Start' },
qs_sub: { zh: '安装构建与常见分析场景示例', en: 'Installation and common analysis scenario examples' },
qs1_h: { zh: '① 构建安装', en: '① Build & Install' },
qs2_h: { zh: '② CPU 热点分析', en: '② CPU Hotspot Analysis' },
qs3_h: { zh: '③ 延迟分析', en: '③ Latency Analysis' },
qs4_h: { zh: '④ 内存泄漏检测', en: '④ Memory Leak Detection' },
qs5_h: { zh: '⑤ IO 跟踪', en: '⑤ IO Tracing' },
qs6_h: { zh: '⑥ 进程状态分析', en: '⑥ Process State Analysis' },
// footer
footer_t: { zh: '<strong>perf-prof</strong> — Linux 实时性能分析框架', en: '<strong>perf-prof</strong> — Linux Real-time Performance Analysis Framework' },
footer_perf: { zh: 'perf_event 文档', en: 'perf_event Docs' },
footer_tp: { zh: 'Tracepoint 文档', en: 'Tracepoint Docs' }
};
var curLang = localStorage.getItem('pp_lang') || ((navigator.language || '').startsWith('zh') ? 'zh' : 'en');
function setLang(lang) {
curLang = lang;
localStorage.setItem('pp_lang', lang);
document.documentElement.lang = lang === 'zh' ? 'zh-CN' : 'en';
document.getElementById('langEn').className = 'lang-btn' + (lang === 'en' ? ' active' : '');
document.getElementById('langZh').className = 'lang-btn' + (lang === 'zh' ? ' active' : '');
document.querySelectorAll('[data-i18n]').forEach(function(el) {
var k = el.getAttribute('data-i18n');
if (T[k]) el.textContent = T[k][lang];
});
document.querySelectorAll('[data-i18n-html]').forEach(function(el) {
var k = el.getAttribute('data-i18n-html');
if (T[k]) el.innerHTML = T[k][lang];
});
}
setLang(curLang);
</script>
</body>
</html>