BeagleBoard/GSoC/2023 Projects

From eLinux.org
Jump to: navigation, search


Links

Weekly reports

  1. All weekly reports will be sent to the mailing list (as that is our primary support venue outside of live chat) on a single thread (to avoid e-mail thrash).
  2. They must be sent on Monday to allow for mentors to respond ahead of Wednesdays IRC meeting where all blockers will be discussed live.
  3. They must include the following sections:
    • Accomplishments
    • Resolutions to blockers
    • On-going blockers
    • Plans for the next week

Projects

Project Student Mentors Videos Reports Code Documentation
Replace GBridge Ayush Singh (Ayush1325) Jason Kridner (jkridner), Vaishnav Achath (Vaishnav) Intro weekly cc1352-firmware, beagleplay-greybus-driver TBD Cross-compilation Task
Zephyr on R5/M4F (K3) Prashanth S (Slpp95prashanth) Nishanth Menon, Vaishnav Achath (Vaishnav), Kumar Abhishek, Dhruva Gole, Deepak Khatri (lorforlinux) Intro weekly Gitlab TBD Cross Compilation Task
OpenGLES acceleration for DL Pratham Deshmukh (Pratham Deshmukh) Shreyas Atre, Kumar Abhishek, Deepak Khatri (lorforlinux) Intro weekly Github TBD Cross Compilation Task
Building an LLVM Backend for PRU Khushi Balia (Khushi Balia) Vedant Paranjape, Shreyas Atre, Kumar Abhishek Intro weekly GitHub TBD Cross Compilation Task

Replace GBridge

The main goals of this project is as follows:

  1. Introduce a Platform Driver that facilitates communication between AM6254 and CC1352 in BeaglePlay over UART.
  2. Move SVC and APBridge roles from GBridge into CC1352 firmware.
  3. Eliminate GBridge and other components that are no longer necessary.
  4. Add a tutorial to the documentation to use the new setup.

Milestones

  1. Introductory Video
  2. Introductory Blog Post
  3. Run old GBridge setup on BeaglePlay and BeagleConnect Freedom
  4. Write a Hello World Linux Driver for BeaglePlay.
  5. Write a Hello World Zephyr Module for CC1352.
  6. Implement sending data from AM62 over UART on Linux Driver.
  7. Implement receiving data from AM62 over UART on Zephyr Module.
  8. Implement receiving data from CC1352 over UART on Linux Driver.
  9. Implement sending data from CC1352 over UART on Zephyr Module.
  10. Establish socket connection with beagleconnect freedom node.
  11. Implement HDLC communication between Linux driver and beagle play CC1352.
  12. Implement sending/receiving Greybus-specific events and messages over UART.
  13. Submit the initial Linux driver to the mailing list.
  14. Implement SVC Role in the Zephyr Module.
  15. Implement APBridge Role in Zephyr Module.
  16. Test the complete new BeagleConnect set up with all the parts.
  17. Add documentation about the new setup.
  18. Get Linux Driver merged upstream.
  19. Final Youtube Video.

Zephyr on R5/M4F (K3)]

The goals of the project:

  1. Add Zephyr RTOS support to run on Cortex R5 processor cores loaded from A72 core running Linux through remoteproc.
  2. Add few peripheral support (Interrupts, Gpio, UART, Timers) on TDA4VM SoC.

Milestones

  1. Introductory Video.
  2. Create essential board directory, defconfig and device tree bindings for board.
  3. Add pinctrl support
  4. UART controller support for TDA4VM in Zephyr.
  5. Understand VIM interrupt programming model with AM263x SDK as reference.
  6. VIM Interrupt Controller support to Cortex R5F.
  7. UART in Interrupt mode.
  8. Understanding Timers programming model in TDA4VM.
  9. Timers and System Tick Support in Zephyr.
  10. Understanding Gpio programming model in TDA4VM.
  11. Gpio controller support in Zephyr.
  12. upstreaming the developed features to Zephyr.
  13. Final Project Video.

OpenGLES Acceleration for DL

The goals of the project:

  1. To accelerate the performance of the DL models.
  2. Identify and optimise the deep learning layers within the darknet CNN framework.
  3. Writing optimised shader program that can perform necessary computations in parallel.
  4. Evaluate the performance of the accelerated darknet framework by comparing it with the original CPU based implementation.

Milestones

  1. Introductory Video.
  2. Implement darknet on the host(laptop)
  3. Benchmark the performance of the pre-existing model on the laptop.
  4. Cross-compile darknet for 64-bit architecture (Beaglebone AI-64).
  5. Go through the TIDL implementation
  6. Benchmarking the performnace of the framework on BB-AI 64.
  7. Adding some simple layers to the framework.
  8. Adding support for these layers and get the performance benefits.
  9. Implement Opengles Support to utilise OpenGLES for specific layer computation.
  10. Cross-compile the modified darknet.
  11. Test the GPU-accelerated darknet and benchmark its performance.
  12. Analysing the benchmark result and optimising it.
  13. Final Project Video

Building an LLVM Backend for PRU

The goals of the project:

  1. To add an LLVM support for PRU.
  2. Upstream the backend for PRU.

Milestones

  1. Introductory Video.
  2. Get thorough with the PRU architecture, instructions.
  3. Implement pru-gcc examples on the board, understand assembly code.
  4. Implement cpu0 backend.
  5. Add the register set and register classes.
  6. Add the Instruction set.
  7. Describe the Target Machine for PRU.
  8. Describe the Target Registration.
  9. Construct the SelectionDAG.
  10. Add the Instruction Selector.
  11. Add the Assembly printer.
  12. Add the Machine code Emitter (JIT support).
  13. Testing and Debugging on various boards.
  14. Upstream the backend.
  15. Final Project Video.