GPU性能分析工具，进程实时CPU/内存消耗解析工具下载网址：
https://www.imgtec.com/developers/powervr-sdk-tools/installers/
本工具可以用来分析以进程PID为单的资源使用，可以实时抓取CPU、内存、GPU资源瞬时使用状态。
使用方法如下：
GPU Profiling
Pre-requisite
Installation of the server into the embedded system
Connect the client to the server
This page explains how profile a PowerVR GPU with tools provided by Imagination Technology
Pre-requisite
Install the PowerVR SDK for your OS from the Imagination website: https://www.imgtec.com/developers/powervr-sdk-tools/installers/
You shall install the PVRTune, it's the profiler tools
For Linux users, the installation path is /opt/Imagination/PowerVR_Graphics
Installation of the server into the embedded system
The first step is the installation of the PVRTune Server, called , into the board:
For linux, the path of the server is PVRPerfServerDeveloper
For a pulsar target, use scp, assume the pulsar address is 10.0.1.1:
scp -p /opt/Imagination/PowerVR_Graphics/PowerVR_Tools/PVRTuneDeveloper/PVRPerfServer/Linux_armv8_64
/PVRPerfServerDeveloper [email protected]:/home/root
#or use ADB
adb push /opt/Imagination/PowerVR_Graphics/PowerVR_Tools/PVRTuneDeveloper/PVRPerfServer/Linux_armv8_64
/PVRPerfServerDeveloper
For an Android target, the server is also a CLI program, so you should push the exec file via adb. An APK is also provided, called PVRHub.apk
#for CLI program
adb push /opt/Imagination/PowerVR_Graphics/PowerVR_Tools/PVRTuneDeveloper/PVRPerfServer/Android/CLI/arm64-v8a
/PVRPerfServerDeveloper /storage/self/primary/Download
#for APK
adb install /opt/Imagination/PowerVR_Graphics/PowerVR_Tools/PVRHub/Android/PVRHub.apk
Connect to the board via ssh or adb, and launch the PVRPerfServerDeveloper program
The program wait a connection from the client
Connect the client to the server
Launch the PVRTuneDeveloperGUI program, which contains the client to the PVRPerfServerDeveloper:
cd /opt/Imagination/PowerVR_Graphics/PowerVR_Tools/PVRTuneDeveloper/GUI/Linux_x86_64
./PVRTuneDeveloperGUI
Here, we can select the address of the server, write the address and connect to the server
Now you get the access to the graphics profiling and the CPU viewed by the OS, with graph, percentage of use....
You have
You have also access to the GPU consumption for each process