Running a TF-Lite Model on RENODE (Litex Vexx RISC-V SoC)

In today’s embedded world, having a reliable emulation platform can drastically reduce development time and facilitate system-level testing. Renode is one such platform, and today, we’ll delve into running a TensorFlow Lite application on the LiteX Vexx RISC-V SoC under the Renode emulation environment.

Prerequisites

1. Linux environment (for our steps).
2. Knowledge about Zephyr, an open-source Real-Time Operating System (RTOS).

STEP 1: Setting Up Renode

Head to the Renode Github Repository.

STEP 2: Downloading Necessary Files

From the releases section, download the following files:

1. renode-1.12.0–1-x86_64.pkg.tar.xz8.13 MB
2. renode-1.12.0–1.f23.x86_64.rpm12.5 MB
3. renode-1.12.0.linux-portable.tar.gz46.1 MB
4. renode_1.12.0.dmg14.3 MB
5. renode_1.12.0.msi15.7 MB
6. renode_1.12.0.zip15.2 MB
7. renode_1.12.0_amd64.deb12.6 MB
8. Source code (zip)
9. Source code (tar.gz)

STEP 3: Unpacking The Files

mkdir renode_portable
tar xf renode-*.linux-portable.tar.gz -C renode_portable --strip-components=1

STEP 4: Adding to System Path

To conveniently call Renode from any terminal:

cd renode_portable
export PATH="`pwd`:$PATH"

STEP 5: Launching Renode

Open a terminal and simply type:

renode

This will launch the Renode monitor (CLI) where the emulation environment can be controlled.

STEP 6: Loading a Script

Scripts are crucial in Renode. To load one, use:

include @/path/to/script.resc

For instance:

include @/home/mahnoor/renode/renode_portable/scripts/single-node/litex.resc

STEP 7: Installing Zephyr

Zephyr can be installed following the guidelines provided here.

STEP 8: Building the Zephyr Shell Demo Application

cd ~/zephyrproject/zephyr
source zephyr-env.sh
west build -p auto -b litex_vexriscv samples/subsys/shell/shell_module/

STEP 9: Relaunching Renode

Again, type:

renode

STEP 10: Running the Zephyr Binary on Renode

In the Renode monitor, run:

(monitor) $zephyr=@/path/to/zephyrproject/zephyr/build/zephyr/zephyr.elf
(monitor) start @scripts/single-node/litex_vexriscv_zephyr.resc

E: if you will get an error from the above command; goto: https://www.mono-project.com/download/stable/

It looks like mono can’t find the libdl.so file. Perhaps it’s the .2 postfix in the filename.

Then Run this command

sudo ln -s /usr/lib/x86_64-linux-gnu/libdl.so.2 /usr/lib/dl

E: if you get an error :

(monitor) $zephyr?=@/path/to/zephyrproject/zephyr/build/zephyr/zephyr.elf                (zephyr binary)
(monitor) start @scripts/single-node/litex_vexriscv_zephyr.resc                         (soc/board )

STEP 11: Running an Application on Renode

The power of Renode extends beyond the basic Zephyr shell. You can also run specialized applications, like TensorFlow Lite, on the emulated LiteX Vexx RISC-V SoC. Let’s dive into the steps:

1. Setting Up The Application: Navigate to the ‘hello world’ folder within Zephyr’s sample applications.
2. Building the Binary: Open your command prompt or terminal and execute:

west build -b {board_name}

1. As an example, if you’re using the litex_vexriscv board:

west build -b litex_vexriscv

1. This will generate the application’s binary. Note that litex_vexriscv is just one of the many boards supported by Zephyr. You can replace it with any board name supported by Zephyr.
2. Loading the Binary in Renode: Launch Renode, and in the monitor, input:

(monitor) $zephyr?=@/path/to/zephyrproject/zephyr/samples/modules/tflite-micro/build/zephyr/zephyr.elf

1. Make sure to replace the path with the exact location of your Zephyr binary.
2. Starting the Emulation: Now, to initiate the emulation, enter:

(monitor) start @scripts/single-node/litex_vexriscv_zephyr.resc

1. Observing the Output: Once the command is executed, Renode will bring up the UART window, showcasing the application’s output.

Congratulations!

Your TensorFlow Lite Application is now humming smoothly on the LiteX Vexx RISC-V SoC in Renode’s emulated environment.

For those deeply interested in embedded systems, the approach outlined in this blog offers a promising way to run and test applications in a risk-free environment before transferring them to real hardware. Happy coding!