systrace/perfetto中需要actrace打tag相关方法-车载车机framework系统开发实战

背景:

经常在看systrace、perfetto相关trace时候,其实我们主要就是看各种方法的调用tag,如下图所示
在这里插入图片描述正因为有了系统中各个地方埋下的这些tag,才让我们可以根据这些方法tag分析出整个系统的运行情况。但是大家有没有想过,请问这些tag是怎么打上去的?如果我们自己要打印自己方法的tag应该怎么搞?

使用atrace相关类方法介绍

这里还需要看对应的trace源码是最全面的
路径:system/core/libcutils/include/cutils/trace.h
源码:

/*
 * Copyright (C) 2012 The Android Open Source Project
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#ifndef _LIBS_CUTILS_TRACE_H
#define _LIBS_CUTILS_TRACE_H

#include <inttypes.h>
#include <stdatomic.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <sys/cdefs.h>
#include <sys/types.h>
#include <unistd.h>
#include <cutils/compiler.h>

__BEGIN_DECLS

/**
 * The ATRACE_TAG macro can be defined before including this header to trace
 * using one of the tags defined below.  It must be defined to one of the
 * following ATRACE_TAG_* macros.  The trace tag is used to filter tracing in
 * userland to avoid some of the runtime cost of tracing when it is not desired.
 *
 * Defining ATRACE_TAG to be ATRACE_TAG_ALWAYS will result in the tracing always
 * being enabled - this should ONLY be done for debug code, as userland tracing
 * has a performance cost even when the trace is not being recorded.  Defining
 * ATRACE_TAG to be ATRACE_TAG_NEVER or leaving ATRACE_TAG undefined will result
 * in the tracing always being disabled.
 *
 * ATRACE_TAG_HAL should be bitwise ORed with the relevant tags for tracing
 * within a hardware module.  For example a camera hardware module would set:
 * #define ATRACE_TAG  (ATRACE_TAG_CAMERA | ATRACE_TAG_HAL)
 *
 * Keep these in sync with frameworks/base/core/java/android/os/Trace.java.
 */
#define ATRACE_TAG_NEVER            0       // This tag is never enabled.
#define ATRACE_TAG_ALWAYS           (1<<0)  // This tag is always enabled.
#define ATRACE_TAG_GRAPHICS         (1<<1)
#define ATRACE_TAG_INPUT            (1<<2)
#define ATRACE_TAG_VIEW             (1<<3)
#define ATRACE_TAG_WEBVIEW          (1<<4)
#define ATRACE_TAG_WINDOW_MANAGER   (1<<5)
#define ATRACE_TAG_ACTIVITY_MANAGER (1<<6)
#define ATRACE_TAG_SYNC_MANAGER     (1<<7)
#define ATRACE_TAG_AUDIO            (1<<8)
#define ATRACE_TAG_VIDEO            (1<<9)
#define ATRACE_TAG_CAMERA           (1<<10)
#define ATRACE_TAG_HAL              (1<<11)
#define ATRACE_TAG_APP              (1<<12)
#define ATRACE_TAG_RESOURCES        (1<<13)
#define ATRACE_TAG_DALVIK           (1<<14)
#define ATRACE_TAG_RS               (1<<15)
#define ATRACE_TAG_BIONIC           (1<<16)
#define ATRACE_TAG_POWER            (1<<17)
#define ATRACE_TAG_PACKAGE_MANAGER  (1<<18)
#define ATRACE_TAG_SYSTEM_SERVER    (1<<19)
#define ATRACE_TAG_DATABASE         (1<<20)
#define ATRACE_TAG_NETWORK          (1<<21)
#define ATRACE_TAG_ADB              (1<<22)
#define ATRACE_TAG_VIBRATOR         (1<<23)
#define ATRACE_TAG_AIDL             (1<<24)
#define ATRACE_TAG_NNAPI            (1<<25)
#define ATRACE_TAG_RRO              (1<<26)
#define ATRACE_TAG_THERMAL          (1 << 27)
#define ATRACE_TAG_LAST             ATRACE_TAG_THERMAL

// Reserved for initialization.
#define ATRACE_TAG_NOT_READY        (1ULL<<63)

#define ATRACE_TAG_VALID_MASK ((ATRACE_TAG_LAST - 1) | ATRACE_TAG_LAST)

#ifndef ATRACE_TAG
#define ATRACE_TAG ATRACE_TAG_NEVER
#elif ATRACE_TAG > ATRACE_TAG_VALID_MASK
#error ATRACE_TAG must be defined to be one of the tags defined in cutils/trace.h
#endif

/**
 * Opens the trace file for writing and reads the property for initial tags.
 * The atrace.tags.enableflags property sets the tags to trace.
 * This function should not be explicitly called, the first call to any normal
 * trace function will cause it to be run safely.
 */
void atrace_setup();

/**
 * If tracing is ready, set atrace_enabled_tags to the system property
 * debug.atrace.tags.enableflags. Can be used as a sysprop change callback.
 */
void atrace_update_tags();

/**
 * Set whether tracing is enabled for the current process.  This is used to
 * prevent tracing within the Zygote process.
 */
void atrace_set_tracing_enabled(bool enabled);

/**
 * This is always set to false. This forces code that uses an old version
 * of this header to always call into atrace_setup, in which we call
 * atrace_init unconditionally.
 */
extern atomic_bool atrace_is_ready;

/**
 * Set of ATRACE_TAG flags to trace for, initialized to ATRACE_TAG_NOT_READY.
 * A value of zero indicates setup has failed.
 * Any other nonzero value indicates setup has succeeded, and tracing is on.
 */
extern uint64_t atrace_enabled_tags;

/**
 * Handle to the kernel's trace buffer, initialized to -1.
 * Any other value indicates setup has succeeded, and is a valid fd for tracing.
 */
extern int atrace_marker_fd;

/**
 * atrace_init readies the process for tracing by opening the trace_marker file.
 * Calling any trace function causes this to be run, so calling it is optional.
 * This can be explicitly run to avoid setup delay on first trace function.
 */
#define ATRACE_INIT() atrace_init()
#define ATRACE_GET_ENABLED_TAGS() atrace_get_enabled_tags()

void atrace_init();
uint64_t atrace_get_enabled_tags();

/**
 * Test if a given tag is currently enabled.
 * Returns nonzero if the tag is enabled, otherwise zero.
 * It can be used as a guard condition around more expensive trace calculations.
 */
#define ATRACE_ENABLED() atrace_is_tag_enabled(ATRACE_TAG)
static inline uint64_t atrace_is_tag_enabled(uint64_t tag)
{
    return atrace_get_enabled_tags() & tag;
}

/**
 * Trace the beginning of a context.  name is used to identify the context.
 * This is often used to time function execution.
 */
#define ATRACE_BEGIN(name) atrace_begin(ATRACE_TAG, name)
static inline void atrace_begin(uint64_t tag, const char* name)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_begin_body(const char*);
        atrace_begin_body(name);
    }
}

/**
 * Trace the end of a context.
 * This should match up (and occur after) a corresponding ATRACE_BEGIN.
 */
#define ATRACE_END() atrace_end(ATRACE_TAG)
static inline void atrace_end(uint64_t tag)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_end_body();
        atrace_end_body();
    }
}

/**
 * Trace the beginning of an asynchronous event. Unlike ATRACE_BEGIN/ATRACE_END
 * contexts, asynchronous events do not need to be nested. The name describes
 * the event, and the cookie provides a unique identifier for distinguishing
 * simultaneous events. The name and cookie used to begin an event must be
 * used to end it.
 */
#define ATRACE_ASYNC_BEGIN(name, cookie) \
    atrace_async_begin(ATRACE_TAG, name, cookie)
static inline void atrace_async_begin(uint64_t tag, const char* name,
        int32_t cookie)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_async_begin_body(const char*, int32_t);
        atrace_async_begin_body(name, cookie);
    }
}

/**
 * Trace the end of an asynchronous event.
 * This should have a corresponding ATRACE_ASYNC_BEGIN.
 */
#define ATRACE_ASYNC_END(name, cookie) atrace_async_end(ATRACE_TAG, name, cookie)
static inline void atrace_async_end(uint64_t tag, const char* name, int32_t cookie)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_async_end_body(const char*, int32_t);
        atrace_async_end_body(name, cookie);
    }
}

/**
 * Trace the beginning of an asynchronous event. In addition to the name and a
 * cookie as in ATRACE_ASYNC_BEGIN/ATRACE_ASYNC_END, a track name argument is
 * provided, which is the name of the row where this async event should be
 * recorded. The track name, name, and cookie used to begin an event must be
 * used to end it.
 */
#define ATRACE_ASYNC_FOR_TRACK_BEGIN(track_name, name, cookie) \
    atrace_async_for_track_begin(ATRACE_TAG, track_name, name, cookie)
static inline void atrace_async_for_track_begin(uint64_t tag, const char* track_name,
                                                const char* name, int32_t cookie) {
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_async_for_track_begin_body(const char*, const char*, int32_t);
        atrace_async_for_track_begin_body(track_name, name, cookie);
    }
}

/**
 * Trace the end of an asynchronous event.
 * This should correspond to a previous ATRACE_ASYNC_FOR_TRACK_BEGIN.
 */
#define ATRACE_ASYNC_FOR_TRACK_END(track_name, name, cookie) \
    atrace_async_for_track_end(ATRACE_TAG, track_name, name, cookie)
static inline void atrace_async_for_track_end(uint64_t tag, const char* track_name,
                                              const char* name, int32_t cookie) {
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_async_for_track_end_body(const char*, const char*, int32_t);
        atrace_async_for_track_end_body(track_name, name, cookie);
    }
}

/**
 * Trace an instantaneous context. name is used to identify the context.
 *
 * An "instant" is an event with no defined duration. Visually is displayed like a single marker
 * in the timeline (rather than a span, in the case of begin/end events).
 *
 * By default, instant events are added into a dedicated track that has the same name of the event.
 * Use atrace_instant_for_track to put different instant events into the same timeline track/row.
 */
#define ATRACE_INSTANT(name) atrace_instant(ATRACE_TAG, name)
static inline void atrace_instant(uint64_t tag, const char* name) {
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_instant_body(const char*);
        atrace_instant_body(name);
    }
}

/**
 * Trace an instantaneous context. name is used to identify the context.
 * track_name is the name of the row where the event should be recorded.
 *
 * An "instant" is an event with no defined duration. Visually is displayed like a single marker
 * in the timeline (rather than a span, in the case of begin/end events).
 */
#define ATRACE_INSTANT_FOR_TRACK(trackName, name) \
    atrace_instant_for_track(ATRACE_TAG, trackName, name)
static inline void atrace_instant_for_track(uint64_t tag, const char* track_name,
                                            const char* name) {
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_instant_for_track_body(const char*, const char*);
        atrace_instant_for_track_body(track_name, name);
    }
}

/**
 * Traces an integer counter value.  name is used to identify the counter.
 * This can be used to track how a value changes over time.
 */
#define ATRACE_INT(name, value) atrace_int(ATRACE_TAG, name, value)
static inline void atrace_int(uint64_t tag, const char* name, int32_t value)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_int_body(const char*, int32_t);
        atrace_int_body(name, value);
    }
}

/**
 * Traces a 64-bit integer counter value.  name is used to identify the
 * counter. This can be used to track how a value changes over time.
 */
#define ATRACE_INT64(name, value) atrace_int64(ATRACE_TAG, name, value)
static inline void atrace_int64(uint64_t tag, const char* name, int64_t value)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_int64_body(const char*, int64_t);
        atrace_int64_body(name, value);
    }
}

__END_DECLS

#endif // _LIBS_CUTILS_TRACE_H

日常使用trace打TAG其实本质上都是基于这个类的,在这个的基础上会加一些额外包装
比如常见的几个方法:
1、ATRACE_CALL();
这个方法,代表对一个function的开头和结尾进行tag,它的源码如下
#define ATRACE_CALL() ATRACE_NAME(FUNCTION)
即本质调用的是ATRACE_NAME

调用如下:
在这里插入图片描述

ATRACE_NAME方法

这个方法的本质定义如下

//这里的ATRACE_NAME其实本质是个临时对象构造和析构进行调用的atrace_begin和atrace_end,这两个其实就是trace.h原始方法
#define ATRACE_NAME(name) ::android::ScopedTrace PASTE(___tracer, __LINE__)(ATRACE_TAG, name)
namespace android {
//临时对象,调用该时候该对象刚好构造,一般方法域结束就代表作用无效了,因为会进行对象的析构
class ScopedTrace {
public:
    inline ScopedTrace(uint64_t tag, const char* name) : mTag(tag) {
        atrace_begin(mTag, name);//构造时候进行begin
            }

    inline ~ScopedTrace() {
        atrace_end(mTag);//析构时候代表trace结束end
    }

private:
    uint64_t mTag;
};

调用如下:
在这里插入图片描述
效果如下:
在这里插入图片描述
其他常见还有

ATRACE_INT

具体源码:

/**
 * Traces an integer counter value.  name is used to identify the counter.
 * This can be used to track how a value changes over time.
 */
#define ATRACE_INT(name, value) atrace_int(ATRACE_TAG, name, value)
static inline void atrace_int(uint64_t tag, const char* name, int32_t value)
{
    if (CC_UNLIKELY(atrace_is_tag_enabled(tag))) {
        void atrace_int_body(const char*, int32_t);
        atrace_int_body(name, value);
    }
}

代码调用:
在这里插入图片描述
效果如下:
![在这里插入图片描述](https://img-blog.csdnimg.cn/6f4e7a7f62c24145af81e537611ba155.png#pic_center

trace相关实战使用:

1、引入相关的头文件及相关库
#include <utils/Trace.h>
编译时候还得考率相关utils的lib是否引入
如这里以截图代码为例就有引入libutils
frameworks/base/cmds/screencap/Android.bp

cc_binary {
    name: "screencap",

    srcs: ["screencap.cpp"],

    shared_libs: [
        "libcutils",
        "libutils",
        "libbinder",
        "libjnigraphics",
        "libui",
        "libgui",
    ],

    cflags: [
        "-Wall",
        "-Werror",
        "-Wunused",
        "-Wunreachable-code",
    ],
}

2、定义好相关的TAG

 #undef ATRACE_TAG
#define ATRACE_TAG ATRACE_TAG_GRAPHICS

这里就选了一个ATRACE_TAG_GRAPHICS即graphics这个组

3、调用相关actrace方法
ATRACE_CALL
ATRACE_NAME
一般针对整个方法或者单独作用域,ATRACE_CALL和ATRACE_NAME本质没有区别,只是把名字变成了function的name
在这里插入图片描述具体效果如下:
在这里插入图片描述优点:只需要一个调用既可以实现tag打印,自动跟随作用域进行tag的结束
缺点:需要自己熟练掌握好作用域,没有可以手动控制的tag结束的点

4、相关灵活控制开始与结束的trace方法
直接使用atrace_begin和atrace_end方法,使用如下:

GraphicBuffer::GraphicBuffer()
    : BASE(), mOwner(ownData), mBufferMapper(GraphicBufferMapper::get()),
      mInitCheck(NO_ERROR), mId(getUniqueId()), mGenerationNumber(0)
{
    ATRACE_NAME("GraphicBuffer::GraphicBuffer()1");
    atrace_begin(ATRACE_TAG, "GraphicBuffer 1");
    width  =
    height =
    stride =
    format =
    usage_deprecated = 0;
    atrace_end(ATRACE_TAG);
    usage  = 0;
    layerCount = 0;
    handle = nullptr;
}

具体效果如下:
在这里插入图片描述
同时也可以使用对应宏定义:
ATRACE_BEGIN(name);
ATRACE_END();
来控制
在这里插入图片描述
执行后效果如下:
在这里插入图片描述

ATRACE_FORMAT

有时候想要Trace上的显示的标签是动态可以变化的比如打印一个数字,或者是某个变量的值
在这里插入图片描述比如上面这种,你看到很多数字其实都是每次执行都是不一样的,那么这个时候打印trace名字时候,就会要求有动态变化的需求,需要把变量值也打印出来,这个时候就需要使用ATRACE_FORMAT这个宏,当然你可以自己每次把变量组合到一个string变量,然后ATRACE_NAME打印这个string既可以,但是这样比较麻烦
使用方式:
ATRACE_FORMAT和上面的ATRACE_NAME没啥区别
不过include文件夹有区别需要引入TraceUtils
#include <gui/TraceUtils.h>
调用方式和打log没啥差别:

  ATRACE_FORMAT("%s %" PRId64 " vsyncIn %.2fms%s", __func__, vsyncId, vsyncIn,
                  mExpectedPresentTime == expectedVsyncTime ? "" : " (adjusted)");

相关源码如下:
frameworks/native/include/gui/TraceUtils.h

#pragma once

#include <stdarg.h>

#include <cutils/trace.h>
#include <utils/Trace.h>

#define ATRACE_FORMAT(fmt, ...)           \
    TraceUtils::TraceEnder __traceEnder = \
            (TraceUtils::atraceFormatBegin(fmt, ##__VA_ARGS__), TraceUtils::TraceEnder())

#define ATRACE_FORMAT_BEGIN(fmt, ...) TraceUtils::atraceFormatBegin(fmt, ##__VA_ARGS__)

namespace android {

class TraceUtils {
public:
    class TraceEnder {
    public:
        ~TraceEnder() { ATRACE_END(); }
    };

    static void atraceFormatBegin(const char* fmt, ...) {
        if (CC_LIKELY(!ATRACE_ENABLED())) return;

        const int BUFFER_SIZE = 256;
        va_list ap;
        char buf[BUFFER_SIZE];

        va_start(ap, fmt);
        vsnprintf(buf, BUFFER_SIZE, fmt, ap);
        va_end(ap);

        ATRACE_BEGIN(buf);
    }

}; // class TraceUtils

} /* namespace android */

java端的trace打印

app层面时候也有trace打印方式

  Trace.traceBegin(Trace.TRACE_TAG_ALWAYS,"aaaaa");//开始trace 字符为aaaaa
  Trace.traceEnd(Trace.TRACE_TAG_ALWAYS);//结束trace

源码如下:
frameworks/base/core/java/android/os/Trace.java


    /**
     * Writes a trace message to indicate that a given section of code has
     * begun. Must be followed by a call to {@link #traceEnd} using the same
     * tag.
     *
     * @param traceTag The trace tag.
     * @param methodName The method name to appear in the trace.
     *
     * @hide
     */
    @UnsupportedAppUsage
    @SystemApi(client = MODULE_LIBRARIES)
    public static void traceBegin(long traceTag, @NonNull String methodName) {
        if (isTagEnabled(traceTag)) {
            nativeTraceBegin(traceTag, methodName);
        }
    }

    /**
     * Writes a trace message to indicate that the current method has ended.
     * Must be called exactly once for each call to {@link #traceBegin} using the same tag.
     *
     * @param traceTag The trace tag.
     *
     * @hide
     */
    @UnsupportedAppUsage
    @SystemApi(client = MODULE_LIBRARIES)
    public static void traceEnd(long traceTag) {
        if (isTagEnabled(traceTag)) {
            nativeTraceEnd(traceTag);
        }
    }

具体使用方法:

  protected void onStart() {
        Trace.traceBegin(Trace.TRACE_TAG_ALWAYS,"Settings Home onStart()");
        ((SettingsApplication) getApplication()).setHomeActivity(this);
        super.onStart();
        doAidlHalCall();
        Trace.traceEnd(Trace.TRACE_TAG_ALWAYS);
    }

然后再抓对应trace
在这里插入图片描述

是不是把我们Settings Home onStart()在trace中给打印出来了

更多framework干货请关注“千里马学框架”

### Perfetto UI Introduction and Usage Perfetto is an open-source performance tracing platform designed to provide detailed insights into system operations on Android devices as well as other systems that support this tool. The Perfetto user interface (UI), accessible via a web-based application, serves as a powerful visualization tool for analyzing trace data collected by the Perfetto tracer. The Perfetto UI allows users to load traces either directly from connected devices or through uploading pre-recorded trace files. Once loaded, various aspects of device performance can be explored including CPU scheduling, memory allocation, disk I/O activity such as those obtained using `storaged` command with `-u` option[^1], network traffic, and more. For interacting effectively within the Perfetto UI: - **Navigating Traces:** Users may zoom in/out along timelines; pan across different sections of interest. - **Inspecting Events:** Clickable events display additional metadata about specific occurrences during execution periods. - **Filtering Data:** Filters help narrow down displayed information based upon event types, processes involved, etc., making it easier to focus analysis efforts where needed most. An example Python script demonstrating how one might interact programmatically with Perfetto's capabilities could look like below but note actual implementation would depend heavily on intended use case specifics beyond simple interaction patterns shown here: ```python import subprocess def start_trace(): """Starts a perfetto trace session.""" cmd = ["perfetto", "-c", "config.txt"] process = subprocess.Popen(cmd) def stop_and_pull_trace(file_path): """Stops current perfetto trace and pulls generated file.""" pull_cmd = f"adb shell ls /data/misc/perfetto-traces/*.trace > {file_path}" subprocess.run(pull_cmd, shell=True) start_trace() # ... perform actions you want to profile ... stop_and_pull_trace("/path/to/local/file.trace") ``` --related questions-- 1. How does one configure custom metrics collection when setting up Perfetto? 2. What are some common pitfalls encountered while interpreting results inside the Perfetto UI? 3. Can multiple instances of Perfetto tracers run simultaneously without interfering each other? 4. Is there any way to automate periodic capturing of performance snapshots over extended durations using Perfetto scripts?
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

千里马学框架

帮助你了,就请我喝杯咖啡

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付:¥1
获取中
扫码支付

您的余额不足,请更换扫码支付或充值

打赏作者

实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值