Introduction:
A tool of testing and analyzing the performance of full mobile platform (iOS/Android). Quickly locating and analyzing the performance issues， improving the performance and quality of apps and games. No need of ROOT/Jailbreak on the phone. No need of extra settings in mobile hardware, games and apps. Extremely simplified, a plug-and-play testing tool.
PerfDog supports all applications on the mobile platform (games, apps, browsers, mini programs, mini-games, H5, background system processes, etc.) and Android Emulator, Cloud Real Machine performance tests. Support APP multi-process testing, such as Android multi-child process and iOS extension process APP Extension.
PerfDog's Windows & Mac OS X desktop app support users testing on iOS and Android devices. PerfDog can test multiple mobile devices at the same time on a single PC.
Important notes Q&A: https://bbs.perfdog.qq.com/detail-132.html
Guide for Special Phones: https://bbs.perfdog.qq.com/detail-133.html
Installation and running:
Step1: Log in to https://perfdog.qq.com/ and select the desktop app you want to download based on your PC platform
Windows platform: Unzip the download package and double-click to open the PerfDog.exe desktop app.
Mac platform: Run the dmg desktop app directly.
Remarks: PerfDog uses a free installation software mode, if you want to update, you can download the latest version on the official website.
Step2: Connect the phone with USB, automatically detect and add the phone to the list of applications.
iOS: Plug and play, no need for users to do anything. If PerfDog can't detect the connection or can't test, please ensure that the latest iTunes can connect to the phone.
Remarks: If you cannot connect, please refer to: https://bbs.perfdog.qq.com/detail-132.html
Android：There are two modes, non-installation mode and installation mode. (Note: USB debugging mode needs to be turned on).
a. Non-installation mode:
(Plug and Play) for mobile phones is very simple to use without any settings and installation, but there is no real-time performance data on the screen of mobile phones.
b. Installation mode:
PerfDog APK needs to be installed automatically on the mobile phone, and real-time performance data can be displayed on the screen of the mobile phone. (Please open debugging mode, allow USB installation and PerfDog suspension window management rights). Start PC version PerfDog.exe, then automatically install PerfDog APK on the mobile phone, specific installation similar to the installation of third-party APP. Note: Because many mobile phone installations require account and password, it is possible that automatic installation fails. If the automatic installation fails, the installation file PerfDog APK will be released into the current folder, you can finish the installation of PerfDog. APK manually. In this mode, during the test process, due to the lack of system resources, the Kill PerfDog display app (the mobile phone interface performance data display) may be displayed, but it does not affect the normal test function.
Remarks: If you cannot connect, please refer to: https://bbs.perfdog.qq.com/detail-132.html
Step3：Testing mode.
l USB mode test:
USB connection, select USB icon in device list for USB mode test
l WIFI mode test (power testing):
Information such as power can only be tested in WIFI mode. After the USB connection, select the WIFI icon device in the device list to test the WIFI mode. After WIFI detects the connection is successful, unplug the USB cable. (Note: The PC and the phone under test need to be connected to the same WIFI. After the WIFI detection connection is successful, unplug the USB cable of the phone under test (USB mode test power has no meaning and will charge). Restrictions make iOS platform unavailable under company WIFI)
Step4：Choose the mobile game & App.
PC as follows:
Android platform, installation mode, the upper left corner of the mobile phone screen has real-time performance data display (Android mobile phone, please open PerfDog suspension window management permission, otherwise the mobile phone will not display performance parameters).
Software Performance Information Collection:
iOS platform (aligned with the parameters of Apple's official Xcode tool. Note: The iOS system has strict limits on process performance parameters. If the limit is exceeded, the system will be killed.)
l ScreenShot (only supports USB mode)
l FPS (The average number of refresh of game interface or application interface in 1 second, commonly known as Frames Per Second/FPS)
1) Avg (FPS): The average number of FPS (FPS average over a period of time)
2) Var (FPS): FPS variance (FPS variance over a period of time)
3) Drop (FPS):The number of frame drop(The average number of adjacent two FPS points drop more than 8 frames per hour)
l Jank (The number of lag in 1 second, iOS9.1 and below systems are temporarily not supported. Similar to Android's Jank and iOS's FramePacing smoothness statistics principle. FPS is high doesn't mean smooth. For example: The FPS is 50, rendering 1 frame in the first 200ms, rendering 49 frames in the remaining 800ms, although the FPS is 50, it doesn't feel smooth. At the same time, FPS is low, which does not represent a lag. For example, if there is no lag, the average FPS is 15 frames. Therefore, the average FPS has no direct relationship with lag.)
Reference：https://bbs.perfdog.qq.com/article-detail.html?id=38
PerfDog calculation method:
1. At the same time satisfying two conditions, it is considered to be a Jank.
a) Current frame time consumption > Twice the average time of the previous three frames.
b) Current frame time > Two frames of movie frame time (1000ms/24*2=84ms).
2. At the same time, satisfying two conditions is considered to be a serious BigJank.
a) Current frame time consumption > Twice the average time of the previous three frames.
b) Current frame time > Three frames of movie frame time (1000ms/24*3=125ms).
1) BigJank: The number of serious lag in 1 second
2) Jank (/10min): The average number of Jank per 10 minutes
3) BigJank (/10min): The average number of BigJank per 10 minutes
l FTime ( Up and down frame display time interval, that is frame time-consuming, iOS9.1 and below systems are temporarily not supported.)
1) Avg(FTime): The average number of FTime
2) Delta (FTime): Delta time-consuming (The average number of time difference between two frames > 100ms per hour)
l CPU Usage (TotalCPU represents the CPU usage of the entire machine, AppCPU represents the CPU usage of the process, and the statistical results are consistent with Xcode. PerfDog usage = Xcode usage / cores. Remarks: iOS Platform CPU Usage is Not Normalized CPU Usage, please reference: https://bbs.perfdog.qq.com/detail-147.html)
l Memory (a statistical Footprint. Note: OOM is related to Footprint and has nothing to do with system or model. It is only related to RAM, such as 1G memory machine. Footprint over 650MB, triggering OOM. Due to the limitation of iOS platform, it is temporarily impossible to obtain the memory of iOS10 and below systems. The function will be added in subsequent versions. For performance testing, it is recommended to upgrade iOS system versions
l XCode Memory (XCode Debug Gauges statistics means XCode Memory). Due to the limitation of iOS platform, it is impossible to obtain Xcode Memory of iOS10 and below systems. The function will be added in subsequent versions. For performance testing, it is recommended to upgrade iOS system version
l Real Memory (Xcode Instrument statistics means Real Memory, which actually occupies physical memory. Note: Physical memory system strategy is related, and it will not pay attention when measuring memory indicators, but it will help analyze and locate overall performance issues. For example: the footprint has not decreased, It means that the application has not released the memory, but the real memory has decreased, indicating that the system has compressed the memory. Since compression consumes CPU resources, it will also reduce the FPS accordingly)
l Virtual Memory
l Wakeups (number of thread wakeups. Note: more than 150 processes are likely to be killed by the system. A sleep / wake cycle on each thread per second, Exceinging limit of 150 wakeups per second over 300 seconds, especially iOS 13.2 suffocating the background Process events, it is recommended to focus on)
l CSwitch (context switching test). Note: Single-core processes over 14,000 will be killed by the system
l GPU Utilization
1) Render: Renderer utilization (Pixel coloring processing stage, if the ratio is high, it means that there is a bottleneck in the PS stage, the shader is too complicated or the texture size, sampling complexity, etc.)
2) Tiler: Tiler utilization (Vertex coloring processing stage, if the ratio is high, it means that there is a bottleneck in the VS stage, too many vertices, etc.)
3) Device: Device utilization (Overall GPU utilization)
l Network (Recv/Send, test target process traffic, consistent with Xcode results)
l BTemp (Battery Temperature)
l Battery Power (Real-time Current, Voltage, and Power Consumption of the entire machine) (Note: Obtained once every 20s, the most accurate statistical method at present. The results are consistent with Battery life results, and all iOS models are supported.)
Note: Sum(Battery) is power consumption.
l Energy Usage (Xcode Energy Impact. Monitor the energy consumption of the application (including CPU, GPU, NetWork, Location, Display (iPhone X only), Overhead). (Note: Consistent with Xcode Energy Impact results. Test in wired mode .Total Energy <270 is Low, 270 < Total Energy < 1000 is High, Total Energy> 1000 is Very High).
Reference：https://help.apple.com/xcode/mac/11.0/index.html?localePath=en.lproj#/devf7f7c5fcd
l Log collection (Log collection is not supported in WIFI mode)
Android Platform
ScreenShot (only supports USB mode)
FPS (The average number of refresh of game interface or application interface in 1 second, commonly known as Frames Per Second/FPS)
1) Avg (FPS): The average number of FPS (FPS average over a period of time)
2)Var (FPS): FPS variance (FPS variance over a period of time)
3)Drop (FPS):The number of frame drop(The average number of adjacent two FPS points drop more than 8 frames per hour)
Remarks: The relationship between Android-Window and FPS
Please refer to: https://bbs.perfdog.qq.com/detail-202.html
l Jank (The number of lag in 1 second, explain as explained on the iOS platform)
1) BigJank: The number of serious lag in 1 second
2) Jank (/10min): The average number of Jank per 10 minutes
3) BigJank (/10min): The average number of BigJank per 10 minutes
l FTime (Up and down frame display time interval, that is frame time-consuming )
1) Avg(FTime): The average number of FTime
2) Delta (FTime): Delta time-consuming (The average number of time difference between two frames > 100ms per hour)
l CPU Usage (Not Normalized CPU Usage that use traditional calculations, TotalCPU represents the Not normalized CPU usage of the entire machine, AppCPU represents the Not normalized CPU usage of the process, and the statistical results are consistent with Android Studio Profiler. please reference: https://bbs.perfdog.qq.com/detail-147.html )
l CPU Usage (Normalized) (Normalized CPU Usage, TotalCPU represents the normalized CPU usage of the entire machine, AppCPU represents the normalized CPU usage of the process. Remarks: It is recommended to use Normalized CPU Usage for more accurate performance. please reference: https://bbs.perfdog.qq.com/detail-147.html )
l CPU Clock (Not normalized frequency and Not normalized usage of each CPU core)
l CPU Clock (Normalized) (Normalized frequency and Normalized usage of each CPU core)
l Memory (PSS Memory, statistical results are consistent with Android Java API and Meminfo. Note: Some Samsung machine systems have modified Meminfo underlying statistical methods, resulting in inconsistent statistical results between Meminfo and Java API, the new Samsung machine has been fixed)
l Swap Memory (Swap Memory, some devices support the Swap function. After the Swap function is enabled, the system will compress the PSS memory. When the swap increases, the PSS will decrease accordingly. Due to the compression, CPU resources will be consumed and the FPS will decrease accordingly)
l Virtual Memory(VSS)
l Memory Detail(NativePSS、GFX、GL、Unknown)
l GPU Usage (currently only supported Qualcomm chipset phones). Note: Top Android GPU test models, please refer to: https://bbs.perfdog.qq.com/detail-195.html
l GPU Frequency (currently only supported Qualcomm chipset phones)
l Network (Recv/Send, test target process traffic)
l CTemp (CPU temperature)
l Battery Power (Current, Voltage, Power) (Note: The error with the instrument test is less than 3%)
Note: Sum(Battery) is power consumption.
Log collection (Log collection is not supported in WIFI mode)
Web Performance Information Management:
l Account information management
l Performance information management, chart display, editing, version comparison
l Performance data statistics, analysis and multi-dimensional comparison
l Performance test task management
Functional introduction:
1. Page of performance parameter control:
Select performance parameters that need to be collected and control performance parameters display pages.
2. Record preservation:
At the end of the testing, data can be processed in two ways:
1) Synchronizing data with Cloud-based dashboard.
2) Export Excel files locally.
3. Record playback (easy review and analysis):
Open the Excel file to play back.
The playback result is shown in the figure:
4. Annotation and calibration:
Double-click the left mouse button, add comments, double-click again, then cancel comments. Click left mouse, add calibration, click again, re-calibration.)：
5. Setting performance parameter statistical analysis threshold, editing statistical parameter analysis and box selection record preservation:
You can select / drag queries through the mouse box within the time cycle, statistical data, etc. Also the selected data can be saved. ‘Setting’ on the left side of the tool can set the corresponding statistical analysis threshold of performance parameters, etc.
6. Scene Label tag:
Label the performance data by the label button. Double-click the left mouse button on the color area can modify the label name of the corresponding area.
7. Screen Recording:
Note: Screen capture opening affects performance.
8. Multi-process testing:
On the iOS platform, the APP multi-process is divided into APP Extension and System XPC Server.
For example: Penguin E-sports Live software uses the APP Extension process (extension process name LABroadcastUpload). Of course, the system XPC Server service process may also be used. For example, a web browser will use webkit.
Android platform, generally large-scale APP, such as games sometimes multi-process collaborative operation (WeChat mini games, WeiShi and other apps and AOV and other multi-child games), you can select the target sub-process for targeted testing. The default is the main process. As the AOV.
WeChat games, mini program testing, etc., as follows: WeChat mini game - romantic rose garden:
Note: The subprocess name is highlighted, indicating that the current subprocess is at the top level.
9. FPS advanced functions (Android platform):
Satisfy advanced users to test window frame rate more accurately (games, mini games, mini programs, live apps, video apps, web and other applications are applicable).
Please refer to: https://bbs.perfdog.qq.com/?m=app&c=detail&a=index&id=202
10. Log function:
Collect system log information.
11. Stop function:
You can stop collecting information by selecting Stop Profilling without pulling out your cell phone.
12. Customize the data file server address and build own Web-Cloud:
The upload address of the data file server can be modified. Post upload http protocol format file to own server address
Reference Demo: https://bbs.perfdog.qq.com/detail-218.html
Web cloud platform features:
1. Personal performance information“MyData”.
Specific performance information display.
2. Test task management.
Create tasks that invite test members to join, or join test tasks created by others.
3. Performance information sharing induction.
4. Test task performance information “Task Data”.
5. Data report private sharing
Visit the others share link to view the data report:
https://perfdog.qq.com/case_detail/367665 Password: xs2BPa
6. Case report multi-dimensional comparison (details, models, versions, proportions, intervals, trends, drag and zoom, etc.).
Comparison of ratio, interval, trend, drag and zoom (mouse left and right performance curve)
扫一扫
2062
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
