分析trace 知识点分析的很全面

Total time breakdown：依据关键tag拆分，比如binderapplication/activitystart/doframe
找出时间差异大tag的cpu status ：Running/Runnable/Sleeping/Uninterruptible sleep
Running:SW-diff-check 或 cpu能力(大小核或频率）
Runnable:schedule 或 cpuloading重(列出top 5 thread)
Sleeping:wakeup event
Memory:kswapd 排名top 3
BlockIO:IO performance

Activity Launcher Flow Chart

Activity Resume Flow Chart

1、Running time

• 关注tag launch flow不同->google版本升级/UI不同

• 列出diff时间段对比/测试机各个function level差别
  

• APK版本是否一致

• 测试机和对比机软件大版本是否一致

• DriverOnly是否输对比机

• Check table，检查diff

2、Runnable

• Top thread做优化，如果是apk，可以卸载再测试
  

3、Sleeping

唤醒源查找
第一步：Bindapplication sleeping时间长-> wait for binder reply

第二步：Check wakeup event

Binder reply sleeping -> wakeup from 2566 (gservice 后台app升级)

4、Uninterruptible sleep - memory

• Uninterruptible sleep caused by do_page_fault à low memory
  

• Memory parameter check point
  

• Lowmemory check point
  

• Memory tuning
  

• 进程优先级与oom adj关系
  

5、Uninterruptble sleep -blockio

• Uninterruptible sleep caused by waiton_page_bit_killable à IO
  
• Io parameter check point
  
• Perlyzer Block TAG：trace format