Profiling the NPU and Installing gputop for Neural Processing Unit in NXP i.MX8M Plus

Getting started with AI on NXP i.MX8M Plus
Introduction
Overview i.MX8M Plus hardware specifications Development tools Development kit DART-MX8M-PLUS Development kit Quick getting started Yocto minimal image Flashing SD card First steps on the board
Development
Gatesgarth and Hardknott Yocto versions Downloading Yocto images for NXP based BSP Building an imx-core-image Flashing SD card Integrating multimedia software stack Installing GStreamer Installing OpenCV Integrating Artificial Intelligence software stack Installing TensorFlow Lite for NXP based images Installing R2Inference Installing GstInference Developing software for the board Building the SDK Crosscompiling apps for GStreamer, TensorFlow Lite, and OpenCV Exploring TensorFlow Lite delegates for prototyping
Neural Processing Unit
Introduction Profiling the NPU Installing gputop Setting the board on debug mode Use Case experiments: Smart Parking Introduction to the use case License Plate Detection TinyYOLO version 3 Training a model compatible with GstInference License Plate Recognition Rosetta Getting the model prototyping Bash scripts for CPU usage and time estimation Serial experiments Parallel Experiments Results
GStreamer
Exploring Rosetta element for character recognition Example pipelines Smart Parking Video inferences Image inferences Video streaming
Contact Us

Including the gputop module

This binary is provided by the recipes at sources/meta-imx/meta-sdk, so if you are developing for i.MX8M Plus from the Variscite images based on the NXP BSP, you will have the dependencies for this sampler. In order to include gputop to be a GPU/NPU usage sampler for the i.MX8M Plus, then go to the local.conf file, and add the following line:

CORE_IMAGE_EXTRA_INSTALL += "gputop"

After adding this line, please configure the environment and rebake the image:

cd $HOME/<your image folder>

source setup-environment <your build directory>

bitbake imx-image-core

Then flash the image into the SD card, and boot the board. Access it by serial or SSH, and then execute:

gputop

The output of the command above is like the following:

Galcore version:6.4.3.305572, gpuperfcnt:8e763ddca55b, 1.4
3D:GC7000,Rev:6204 Core: 999 MHz, Shader: 999 MHz 
3D:GC8000,Rev:8002 Core: 999 MHz, Shader: 999 MHz 2D:GC520,Rev:5341 
3D Cores:2,2D Cores:1,VG Cores:0

IMX8_DDR0: axid-read:15.66,axid-write:6.46
IMX8_DDR1: 

     PID   RES(kB)   CONT(kB)   VIRT(kB)  Non-PGD(kB)  Total(kB)              CMD
     760     33171       2147          0            0      35318              car

TOT:         33171       2147          0            0      35318
TOT_CON:         -          -          -            -     228972

As you can see in the previous output, there is a process called car with a PID of 760. This means that this process is using the NPU or the GPU since both of the accelerators use axi for communicate. In this specific case, the process was detecting license plates from an image using the NPU for accelerating the inference.