571 lines
19 KiB
Markdown
571 lines
19 KiB
Markdown
# Android 设备卡顿诊断报告（系统层+应用层综合优化版）
|
||
|
||
**诊断时间：2026-04-18**
|
||
**应用代码路径：F:\code\git_project\xcmg\xcmg**
|
||
**系统代码路径：S:\znzc\034_xcmg\x9hp_xcmg\androd10_x9hp_ptg4.3p3\android10**
|
||
|
||
---
|
||
|
||
## 一、诊断数据汇总
|
||
|
||
### 设备信息
|
||
|
||
| 项目 | 值 |
|
||
|------|-----|
|
||
| 设备型号 | x9 for arm64 (车载设备) |
|
||
| Android 版本 | 10 (API 29) |
|
||
| CPU | 8核 arm64, 1.6GHz |
|
||
| 总内存 | 4,383 MB |
|
||
| Swap | 1,753 MB |
|
||
|
||
### 系统状态诊断
|
||
|
||
| 指标 | 数值 | 状态 | 说明 |
|
||
|------|------|------|------|
|
||
| **Load Average** | **12.7** | **🔴 严重过载** | 8核CPU，正常应<8，当前超载1.6倍 |
|
||
| 可用内存 | 584 MB (13%) | 🔴 紧张 | 低于安全阈值20% |
|
||
| IRQ/SIRQ | 13% | 🔴 过高 | 中断处理负担重 |
|
||
| surfaceflinger CPU | 18% | 🟡 偏高 | 图形合成压力大 |
|
||
|
||
### 高资源占用进程
|
||
|
||
| 进程 | CPU | 内存 | 问题 |
|
||
|------|-----|------|------|
|
||
| **com.yaxon.avm** | **89%** | 255MB | AVM全景监控CPU异常 |
|
||
| **com.txzing.voiceassistant** | **76%** | 563MB | 语音助手CPU+内存双高 |
|
||
| surfaceflinger | 18% | 51MB | 渲染合成压力 |
|
||
|
||
---
|
||
|
||
## 二、问题根因分析
|
||
|
||
### 🔴 根因链
|
||
|
||
```
|
||
┌─────────────────────────────────────────────────────────────────┐
|
||
│ 卡顿根本原因链 │
|
||
├─────────────────────────────────────────────────────────────────┤
|
||
│ │
|
||
│ [应用层问题] │
|
||
│ ├─ AVM无限激活重试 + Native密集调用 → 89% CPU │
|
||
│ ├─ TXZ语音助手WebSocket失败无限重试 → 76% CPU + 563MB │
|
||
│ └─ YX_Voice无重试限制 → 持续触发连接 │
|
||
│ │
|
||
│ ║ │
|
||
│ ║ 累计占用165%+ CPU │
|
||
│ ▼ │
|
||
│ [系统层压力] │
|
||
│ ├─ Load Average 12.7 → 任务队列过长 │
|
||
│ ├─ 内存可用13% → LMKD频繁杀进程 │
|
||
│ ├─ surfaceflinger无法及时合成 → 18% CPU │
|
||
│ └─ IRQ/SIRQ 13% → 硬件中断处理慢 │
|
||
│ │
|
||
│ ║ │
|
||
│ ▼ │
|
||
│ [渲染管线阻塞] │
|
||
│ ├─ Launcher主线程等待CPU → 88.89%卡顿帧 │
|
||
│ ├─ SystemUI主线程阻塞 → 302帧跳帧 │
|
||
│ └─ VSync丢失 → 渲染节奏紊乱 │
|
||
│ │
|
||
└─────────────────────────────────────────────────────────────────┘
|
||
```
|
||
|
||
---
|
||
|
||
## 三、系统层代码分析与优化
|
||
|
||
### 3.1 ProcessList.java - OOM配置分析
|
||
|
||
**文件位置：** `frameworks/base/services/core/java/com/android/server/am/ProcessList.java`
|
||
|
||
```java
|
||
// 第285-298行：OOM adj级别定义
|
||
private final int[] mOomAdj = new int[] {
|
||
FOREGROUND_APP_ADJ, VISIBLE_APP_ADJ, PERCEPTIBLE_APP_ADJ,
|
||
PERCEPTIBLE_LOW_APP_ADJ, CACHED_APP_MIN_ADJ, CACHED_APP_LMK_FIRST_ADJ
|
||
};
|
||
|
||
// 第291-294行：低内存设备的minfree配置（KB）
|
||
private final int[] mOomMinFreeLow = new int[] {
|
||
12288, 18432, 24576, 36864, 43008, 49152 // 12MB, 18MB, 24MB, 36MB, 42MB, 48MB
|
||
};
|
||
|
||
// 第297-299行：高内存设备的minfree配置（KB）
|
||
private final int[] mOomMinFreeHigh = new int[] {
|
||
73728, 92160, 110592, 129024, 147456, 184320 // 72MB, 90MB, 108MB, 126MB, 144MB, 180MB
|
||
};
|
||
```
|
||
|
||
**问题分析：**
|
||
- 当前设备4.3GB内存，使用的是动态计算的minfree值
|
||
- 可用内存仅584MB，触发频繁的进程回收
|
||
|
||
**优化方案：**
|
||
|
||
```java
|
||
// ===== 优化：调整OOM minfree阈值 =====
|
||
// 修改 ProcessList.java 第291-299行
|
||
|
||
// 针对4GB内存车载设备，适当提高阈值避免频繁杀进程
|
||
// 原值基于1GB设备设计，对于4GB设备过于激进
|
||
private final int[] mOomMinFreeHigh = new int[] {
|
||
122880, 153600, 184320, // 120MB, 150MB, 180MB (前台/可见/感知)
|
||
245760, 307200, 368640 // 240MB, 300MB, 360MB (缓存级别)
|
||
};
|
||
// 提高约33%，减少频繁杀进程导致的卡顿
|
||
```
|
||
|
||
---
|
||
|
||
### 3.2 ActivityManagerConstants.java - 服务管理优化
|
||
|
||
**文件位置：** `frameworks/base/services/core/java/com/android/server/am/ActivityManagerConstants.java`
|
||
|
||
```java
|
||
// 第81行：最大缓存进程数
|
||
private static final int DEFAULT_MAX_CACHED_PROCESSES = 8;
|
||
|
||
// 第103行：服务最大不活跃时间
|
||
private static final long DEFAULT_MAX_SERVICE_INACTIVITY = 30*60*1000; // 30分钟
|
||
|
||
// 第107行：绑定服务崩溃最大重试
|
||
private static final int DEFAULT_BOUND_SERVICE_CRASH_MAX_RETRY = 16;
|
||
```
|
||
|
||
**问题分析：**
|
||
- MAX_CACHED_PROCESSES=8 对于车载系统偏少
|
||
- 服务重试次数过多可能导致资源浪费
|
||
|
||
**优化方案：**
|
||
|
||
```java
|
||
// ===== 优化1：增加缓存进程数 =====
|
||
// 车载系统有更多常驻服务，适当增加缓存容量
|
||
private static final int DEFAULT_MAX_CACHED_PROCESSES = 12; // 原8改为12
|
||
|
||
// ===== 优化2：减少服务崩溃重试次数 =====
|
||
// 避免崩溃服务反复重启占用资源
|
||
private static final int DEFAULT_BOUND_SERVICE_CRASH_MAX_RETRY = 5; // 原16改为5
|
||
```
|
||
|
||
---
|
||
|
||
### 3.3 lmkd.c - 低内存杀手配置优化
|
||
|
||
**文件位置：** `system/core/lmkd/lmkd.c`
|
||
|
||
```c
|
||
// 第182-186行：PSI压力阈值配置
|
||
static struct psi_threshold psi_thresholds[VMPRESS_LEVEL_COUNT] = {
|
||
{ PSI_SOME, 70 }, // 70ms out of 1sec for partial stall
|
||
{ PSI_SOME, 100 }, // 100ms out of 1sec for partial stall
|
||
{ PSI_FULL, 70 }, // 70ms out of 1sec for complete stall
|
||
};
|
||
```
|
||
|
||
**问题分析：**
|
||
- PSI阈值偏激进，触发过早导致频繁杀进程
|
||
|
||
**优化方案：**
|
||
|
||
```c
|
||
// ===== 优化：调整PSI阈值减少过早触发 =====
|
||
// 修改 lmkd.c 第182-186行
|
||
static struct psi_threshold psi_thresholds[VMPRESS_LEVEL_COUNT] = {
|
||
{ PSI_SOME, 100 }, // 100ms（原70ms）- 放宽容忍度
|
||
{ PSI_SOME, 150 }, // 150ms（原100ms）
|
||
{ PSI_FULL, 100 }, // 100ms（原70ms）
|
||
};
|
||
// 减少内存压力误触发，避免频繁杀进程
|
||
```
|
||
|
||
---
|
||
|
||
### 3.4 SurfaceFlinger.cpp - 渲染管线优化
|
||
|
||
**文件位置：** `frameworks/native/services/surfaceflinger/SurfaceFlinger.cpp`
|
||
|
||
当前渲染合成占用18% CPU，问题在于：
|
||
1. CPU过载导致SurfaceFlinger无法及时获得调度
|
||
2. 多图层合成计算量大
|
||
|
||
**优化方案（Scheduler层面）：**
|
||
|
||
**文件位置：** `frameworks/native/services/surfaceflinger/Scheduler/Scheduler.cpp`
|
||
|
||
```cpp
|
||
// ===== 优化：提高渲染线程优先级 =====
|
||
// 在 Scheduler.cpp 中添加优先级提升
|
||
|
||
// 第134-150行附近，创建EventThread时提高优先级
|
||
std::unique_ptr<EventThread> eventThread =
|
||
makeEventThread(connectionName, mPrimaryDispSync.get(), phaseOffsetNs,
|
||
offsetThresholdForNextVsync, std::move(interceptCallback));
|
||
|
||
// 提高渲染相关线程的调度优先级
|
||
// 在EventThread创建后设置：
|
||
if (eventThread) {
|
||
// 设置更高的调度优先级，确保渲染及时
|
||
eventThread->setSchedPriority(SCHED_PRIORITY_REALTIME);
|
||
}
|
||
```
|
||
|
||
---
|
||
|
||
### 3.5 系统属性配置优化
|
||
|
||
**建议添加的系统属性（build.prop或通过adb设置）：**
|
||
|
||
```bash
|
||
# ===== 内存管理优化 =====
|
||
# 提高LMKD阈值，减少过早触发
|
||
ro.lmk.psi_partial_stall_ms=150
|
||
ro.lmk.psi_complete_stall_ms=100
|
||
ro.lmk.low=100
|
||
ro.lmk.medium=200
|
||
ro.lmk.critical=500
|
||
|
||
# ===== 进程管理优化 =====
|
||
# 增加缓存进程数
|
||
ro.sys.fw.bg_apps_limit=12
|
||
ro.sys.fw.max_cached_processes=12
|
||
|
||
# ===== 渲染优化 =====
|
||
# 提高渲染线程优先级
|
||
debug.sf.set_idle_timer_ms=100
|
||
debug.sf.enable_gl_backpressure=1
|
||
|
||
# ===== CPU调度优化 =====
|
||
# 降低后台进程对CPU的影响
|
||
ro.sys.fw.bg_cpu_percent=10
|
||
```
|
||
|
||
---
|
||
|
||
## 四、应用层代码优化（结合之前分析）
|
||
|
||
### 4.1 YX_Avm模块优化
|
||
|
||
**文件位置：** `F:\code\git_project\xcmg\xcmg\YX_Avm\src\main\java\com\android\v4l2camera\V4L2CameraManager.java`
|
||
|
||
```java
|
||
// ===== 优化1：添加激活超时机制 =====
|
||
// 修改 avmActivatedIfNeed() 方法（第166-181行）
|
||
|
||
private static final long ACTIVATION_TIMEOUT_MS = 30000; // 30秒超时
|
||
private final ExecutorService mActivationExecutor = Executors.newSingleThreadExecutor();
|
||
|
||
public void avmActivatedIfNeed() {
|
||
if (isAvmActivated) {
|
||
Log.i(TAG, "avmActivatedIfNeed isAvmActivated -> true");
|
||
return;
|
||
}
|
||
if (mActivatedThread != null && mActivatedThread.isAlive()) {
|
||
Log.i(TAG, "avmActivatedIfNeed mActivatedThread is alive");
|
||
return;
|
||
}
|
||
|
||
mActivatedThread = new Thread(() -> {
|
||
Future<Boolean> future = mActivationExecutor.submit(() ->
|
||
mV4L2CameraManager.activatedAvmModule());
|
||
|
||
try {
|
||
future.get(ACTIVATION_TIMEOUT_MS, TimeUnit.MILLISECONDS);
|
||
} catch (TimeoutException e) {
|
||
Log.w(TAG, "AVM activation timeout, cancelling");
|
||
future.cancel(true);
|
||
// 延迟重试，避免立即重试占用CPU
|
||
mHandler.postDelayed(() -> avmActivatedIfNeed(), 60000);
|
||
} catch (Exception e) {
|
||
Log.e(TAG, "AVM activation error", e);
|
||
}
|
||
});
|
||
mActivatedThread.setName("T_AvmActivated");
|
||
mActivatedThread.start();
|
||
}
|
||
|
||
// ===== 优化2：图像处理移至子线程 =====
|
||
// 修改 saveYuvImageToFrontPng() 等方法（第695-820行）
|
||
|
||
private final ExecutorService mImageProcessExecutor = Executors.newFixedThreadPool(2);
|
||
|
||
private void saveYuvImageToFrontPng(byte[] yuvData, int width, int height) {
|
||
if (calibDirection != AvmConstant.CalibCameraView.AVM_CALIBRATION_FRONT) {
|
||
return;
|
||
}
|
||
|
||
// 异步处理，避免阻塞回调线程
|
||
mImageProcessExecutor.execute(() -> {
|
||
String filePath = AvmConstant.SDCARD_CALIBRATION_PATH;
|
||
File file = new File(filePath);
|
||
if (!file.exists()) file.mkdirs();
|
||
|
||
String temFile = file.getAbsolutePath() + File.separator + "front.jpg";
|
||
try {
|
||
YuvImage image = new YuvImage(yuvData, ImageFormat.NV21, width, height, null);
|
||
ByteArrayOutputStream baos = new ByteArrayOutputStream();
|
||
image.compressToJpeg(new Rect(0, 0, width, height), 100, baos);
|
||
|
||
Bitmap bmp = BitmapFactory.decodeByteArray(baos.toByteArray(), 0, baos.toByteArray().length);
|
||
FileOutputStream fos = new FileOutputStream(temFile);
|
||
bmp.compress(Bitmap.CompressFormat.JPEG, 90, fos); // 降低质量减少CPU
|
||
fos.close();
|
||
baos.close();
|
||
|
||
if (calibraResultCallBack != null) {
|
||
mHandler.post(() -> calibraResultCallBack.onFrontCapture(temFile));
|
||
}
|
||
} catch (Exception e) {
|
||
Log.e(TAG, "saveYuvImageToFrontPng error", e);
|
||
}
|
||
});
|
||
}
|
||
|
||
// ===== 优化3：添加ContentObserver防抖 =====
|
||
// 修改 AvmService.java 第160-217行
|
||
|
||
private ContentObserver mContentObserver = new ContentObserver(null) {
|
||
private long mLastCallbackTime = 0;
|
||
private static final long MIN_CALLBACK_INTERVAL = 200; // 200ms最小间隔
|
||
|
||
@Override
|
||
public void onChange(boolean selfChange, Uri uri) {
|
||
super.onChange(selfChange, uri);
|
||
if (uri == null) return;
|
||
|
||
// 防抖处理
|
||
long currentTime = System.currentTimeMillis();
|
||
if (currentTime - mLastCallbackTime < MIN_CALLBACK_INTERVAL) {
|
||
Log.d(TAG, "onChange throttled: " + uri.toString());
|
||
return;
|
||
}
|
||
mLastCallbackTime = currentTime;
|
||
|
||
// 原有的处理逻辑...
|
||
}
|
||
};
|
||
```
|
||
|
||
---
|
||
|
||
### 4.2 YX_Voice模块优化
|
||
|
||
**文件位置：** `F:\code\git_project\xcmg\xcmg\YX_Voice\src\main\java\com\yaxon\voice\service\VoiceService.java`
|
||
|
||
```java
|
||
// ===== 优化：添加连接重试限制 =====
|
||
// 修改 VoiceService.java
|
||
|
||
private int mConnectRetryCount = 0;
|
||
private static final int MAX_CONNECT_RETRY = 3;
|
||
private static final long[] RETRY_DELAY_MS = {5000, 15000, 60000}; // 5s, 15s, 60s
|
||
|
||
private void initTxz() {
|
||
try {
|
||
// 检测网络可用性
|
||
if (!isNetworkAvailable()) {
|
||
Log.w(TAG, "Network not available, delay connect");
|
||
scheduleReconnect(10000);
|
||
return;
|
||
}
|
||
|
||
// ... 原有的初始化代码
|
||
} catch (Throwable e) {
|
||
Log.e(TAG, "initTxz error", e);
|
||
scheduleReconnect(RETRY_DELAY_MS[Math.min(mConnectRetryCount, RETRY_DELAY_MS.length - 1)]);
|
||
}
|
||
}
|
||
|
||
@Override
|
||
public void onDisconnected() {
|
||
Log.d(TAG,"TXZ onDisconnected");
|
||
VoiceManager.getInstance().setVoiceBind(false);
|
||
|
||
if (mConnectRetryCount < MAX_CONNECT_RETRY) {
|
||
long delay = RETRY_DELAY_MS[mConnectRetryCount];
|
||
Log.d(TAG, "Schedule reconnect in " + delay + "ms, retry=" + (mConnectRetryCount + 1));
|
||
scheduleReconnect(delay);
|
||
mConnectRetryCount++;
|
||
} else {
|
||
Log.w(TAG, "Max retry count reached (" + MAX_CONNECT_RETRY + "), stop reconnecting");
|
||
// 网络恢复后再重试
|
||
registerNetworkRecoveryListener();
|
||
}
|
||
}
|
||
|
||
@Override
|
||
public void onConnected() {
|
||
Log.d(TAG,"TXZ onConnected");
|
||
mConnectRetryCount = 0; // 连接成功重置计数
|
||
// ... 原有代码
|
||
}
|
||
|
||
private void scheduleReconnect(long delayMs) {
|
||
mHandler.removeCallbacks(mReconnectRunnable);
|
||
mHandler.postDelayed(mReconnectRunnable, delayMs);
|
||
}
|
||
|
||
private final Runnable mReconnectRunnable = () -> {
|
||
if (isNetworkAvailable()) {
|
||
mConnectRetryCount = 0;
|
||
initTxz();
|
||
}
|
||
};
|
||
|
||
private boolean isNetworkAvailable() {
|
||
ConnectivityManager cm = (ConnectivityManager) getSystemService(CONNECTIVITY_SERVICE);
|
||
NetworkInfo activeNetwork = cm.getActiveNetworkInfo();
|
||
return activeNetwork != null && activeNetwork.isConnectedOrConnecting();
|
||
}
|
||
```
|
||
|
||
---
|
||
|
||
### 4.3 TXZ语音助手SDK优化需求
|
||
|
||
**需联系TXZ技术支持（同舟行科技）提出以下需求：**
|
||
|
||
| 序号 | 问题描述 | 优化要求 | 优先级 |
|
||
|------|----------|----------|--------|
|
||
| 1 | WebSocket连接失败后无限重试 | 添加重试次数限制（最多3次） | **P0** |
|
||
| 2 | 重试间隔固定约3秒 | 添加指数退避（5s→15s→60s） | **P0** |
|
||
| 3 | 网络不可达时仍重试 | 检测网络状态，不可达时暂停 | **P0** |
|
||
| 4 | 内存占用563MB过高 | 优化内存使用，减少缓存 | **P1** |
|
||
| 5 | CPU占用76%异常 | 检查SDK内部循环/定时器 | **P0** |
|
||
|
||
---
|
||
|
||
## 五、综合优化方案实施清单
|
||
|
||
### 优先级排序
|
||
|
||
| 优先级 | 层级 | 模块 | 优化内容 | 预期效果 |
|
||
|--------|------|------|----------|----------|
|
||
| **P0** | 应用 | TXZ SDK | 重试限制、指数退避 | 释放76% CPU |
|
||
| **P0** | 应用 | YX_Avm | 激活超时、图像异步 | 降低CPU峰值 |
|
||
| **P0** | 应用 | YX_Voice | 重试限制、网络检测 | 降低重试频率 |
|
||
| **P1** | 系统 | lmkd.c | PSI阈值调整 | 减少频繁杀进程 |
|
||
| **P1** | 系统 | ProcessList | minfree调整 | 增加缓存容量 |
|
||
| **P1** | 系统 | ActivityManagerConstants | 缓存进程数 | 减少进程重建 |
|
||
| **P2** | 系统 | Scheduler | 渲染优先级 | 提升UI响应 |
|
||
| **P2** | 系统 | build.prop | 系统属性 | 全局优化 |
|
||
|
||
---
|
||
|
||
### 实施步骤
|
||
|
||
#### 步骤1：立即生效（无需重新编译系统）
|
||
|
||
```bash
|
||
# 通过adb设置临时属性验证效果
|
||
adb shell "setprop ro.lmk.psi_partial_stall_ms 150"
|
||
adb shell "setprop ro.lmk.psi_complete_stall_ms 100"
|
||
adb shell "setprop ro.sys.fw.bg_apps_limit 12"
|
||
|
||
# 强制停止问题应用
|
||
adb shell "am force-stop com.yaxon.avm"
|
||
adb shell "am force-stop com.txzing.voiceassistant"
|
||
```
|
||
|
||
#### 步骤2：应用层修改（需要重新编译应用）
|
||
|
||
修改以下文件并重新编译：
|
||
1. `YX_Avm/.../V4L2CameraManager.java`
|
||
2. `YX_Avm/.../AvmService.java`
|
||
3. `YX_Voice/.../VoiceService.java`
|
||
|
||
#### 步骤3：系统层修改（需要重新编译系统）
|
||
|
||
修改以下文件并重新编译系统镜像：
|
||
1. `system/core/lmkd/lmkd.c` - PSI阈值
|
||
2. `frameworks/base/services/core/java/com/android/server/am/ProcessList.java` - minfree
|
||
3. `frameworks/base/services/core/java/com/android/server/am/ActivityManagerConstants.java` - 缓存配置
|
||
4. `device/.../build.prop` - 系统属性
|
||
|
||
---
|
||
|
||
## 六、代码修改文件清单
|
||
|
||
### 应用层修改
|
||
|
||
| 文件路径 | 修改内容 | 状态 |
|
||
|----------|----------|------|
|
||
| `YX_Avm/src/.../V4L2CameraManager.java` | 激活超时、图像异步处理 | 待修改 |
|
||
| `YX_Avm/src/.../AvmService.java` | ContentObserver防抖 | 待修改 |
|
||
| `YX_Voice/src/.../VoiceService.java` | 连接重试限制、网络检测 | 待修改 |
|
||
|
||
### 系统层修改
|
||
|
||
| 文件路径 | 修改内容 | 状态 |
|
||
|----------|----------|------|
|
||
| `system/core/lmkd/lmkd.c` | PSI阈值调整 | 待修改 |
|
||
| `frameworks/base/services/.../ProcessList.java` | minfree阈值调整 | 待修改 |
|
||
| `frameworks/base/services/.../ActivityManagerConstants.java` | 缓存配置调整 | 待修改 |
|
||
| `frameworks/native/services/surfaceflinger/Scheduler/Scheduler.cpp` | 渲染优先级 | 待修改 |
|
||
| `device/.../system.prop` 或 `build.prop` | 系统属性配置 | 待修改 |
|
||
|
||
---
|
||
|
||
## 七、验证与监控
|
||
|
||
### 优化效果验证命令
|
||
|
||
```bash
|
||
# 检查CPU负载
|
||
adb shell "top -n 1 | head -15"
|
||
|
||
# 检查内存状态
|
||
adb shell "cat /proc/meminfo | grep -E 'MemTotal|MemAvailable'"
|
||
|
||
# 检查进程杀次数
|
||
adb shell "cat /proc/vmstat | grep -i kill"
|
||
|
||
# 检查渲染性能
|
||
adb shell "dumpsys gfxinfo com.yaxon.launcher"
|
||
|
||
# 检查LMKD状态
|
||
adb shell "logcat -d | grep lmkd | tail -20"
|
||
```
|
||
|
||
### 持续监控脚本
|
||
|
||
```bash
|
||
# 创建监控脚本 monitor.sh
|
||
#!/bin/bash
|
||
while true; do
|
||
echo "=== $(date) ==="
|
||
echo "CPU Load:"
|
||
adb shell "top -n 1 | head -5"
|
||
echo "Memory:"
|
||
adb shell "cat /proc/meminfo | grep -E 'MemTotal|MemFree|MemAvailable'"
|
||
echo "Janky frames:"
|
||
adb shell "dumpsys gfxinfo com.yaxon.launcher | grep 'Janky frames'"
|
||
sleep 10
|
||
done
|
||
```
|
||
|
||
---
|
||
|
||
## 八、结论
|
||
|
||
### 核心问题
|
||
|
||
| 层级 | 问题 | 占用 | 根因 |
|
||
|------|------|------|------|
|
||
| 应用层 | AVM | 89% CPU | 激活无超时、图像处理阻塞 |
|
||
| 应用层 | TXZ语音助手 | 76% CPU | WebSocket失败无限重试 |
|
||
| 系统层 | 内存紧张 | 13%可用 | PSI阈值激进、频繁杀进程 |
|
||
| 系统层 | 渲染阻塞 | 18% CPU | surfaceflinger调度不及时 |
|
||
|
||
### 解决方向
|
||
|
||
1. **应用层优先**：修复AVM和TXZ的CPU异常占用（预期释放165%+ CPU）
|
||
2. **系统层配合**：调整内存管理策略，减少进程回收导致的二次卡顿
|
||
3. **渲染优化**：提高SurfaceFlinger调度优先级，保障UI流畅
|
||
|
||
---
|
||
|
||
**报告生成完毕**
|
||
|
||
诊断工具：Android Lag Analyzer v1.0 + 代码静态分析
|
||
生成时间：2026-04-18 |