asicdesign | UCSC OSPO

OpenROAD - An Open-Source, Autonomous RTL-GDSII Flow for Chip Design

Sun, 19 Jan 2025 00:00:00 +0000

The OpenROAD project is a non-profit project, originally funded by DARPA with the aim of creating open-source EDA tools; an Autonomous flow from RTL-GDSII that completes < 24 hrs, to lower cost and boost innovation in IC design. This project is now supported by Precision Innovations.

OpenROAD massively scales and supports EWD (Education and Workforce Development) and supports a broad ecosystem making it a vital tool that supports a rapidly growing Semiconductor Industry.

OpenROAD is the fastest onramp to gain knowledge, skills and create pathways for great career opportunities in chip design. You will develop important software and hardware design skills by contributing to these interesting projects. You will also have the opportunity to work with mentors from the OpenROAD project and other industry experts.

We welcome a diverse community of designers, researchers, enthusiasts, software engineers and entrepreneurs to use and contribute to OpenROAD and make a far-reaching impact in the rapidly growing, global Semiconductor Industry.

Improving Code Quality in OpenROAD

Topics: Coding Best Practices in C++, Code Quality Tooling, Continuous Integration
Skills: C++
Difficulty: Medium
Size: Medium (175 hours)
Mentors: Matt Liberty & Arthur Koucher

OpenROAD is a large and complex program. This project is to improve the code quality through resolving issues flagged by tools like Coverity and clang-tidy. New tools like the clang sanitizers ASAN/TSAN/UBSAN should also be set up and integrated with the Jenkins CI.

GUI Testing in OpenROAD

Topics: Testing, Continuous Integration
Skills: C++, Qt
Difficulty: Medium
Size: Large (350 hours)
Mentors: Matt Liberty & Peter Gadfort

The OpenROAD GUI is a crucial set of functionality for users to see and investigate their design. GUI testing is specialized and rather different from standard unit testing. The GUI therefore needs improvements to its testing to cover both interaction and rendering. The GUI uses the Qt framework. An open-source testing tool like https://github.com/faaxm/spix will be set up and key tests developed. This will provide the framework for all future testing.

Rectilinear Floorplans in OpenROAD

Topics: Electronic Design Automation, Algorithms
Skills: C++, data structures and algorithms
Difficulty: Medium
Size: Large (350 hours)
Mentors: Eder Monteiro & Augusto Berndt

OpenROAD supports block floorplans that are rectangular in shape. Some designs may require more complex shapes to fit. This project extends the tool to support rectilinear polygon shapes as floorplans. This will require upgrading data structures and algorithms in various parts of OpenROAD including floor plan generation, pin placement, and global placement.

LEF Reader and Database Enhancements in OpenROAD

Topics: Electronic Design Automation, Database, Parsing
Skills: Boost Spirit parsers, Database, C++
Difficulty: Medium
Size: Medium (175 hours)
Mentors: Osama Hammad & Ethan Mahintorabi

LEF (Library Exchange Format) is a standard format for describing physical design rules for integrated circuits. OpenROAD has support for many constructs but some newer ones for advanced process nodes are not supported. This project is to support parsing such information and storing in the OpenDB for use by the rest of the tool.

ORAssistant - LLM Data Engineering and Testing

Topics: Large Language Model, Machine Learning, Data Engineering, Model Deployment, Testing, Full-Stack Development
Skills: large language model engineering, database, evaluation, CI/CD, open-source or related software development, full-stack
Difficulty: Medium
Size: Medium (175 hours)
Mentor: Jack Luar & Palaniappan R

This project is aimed at enhancing robustness and accuracy for OR Assistant, the conversational assistant for OpenROAD through comprehensive testing and evaluation. You will work with members of the OpenROAD team and other researchers to enhance the existing dataset to cover a wide range of use cases to deliver accurate responses more efficiently. This project will focus on data engineering and benchmarking and you will collaborate on a project on the LLM model engineering. Tasks include: creating evaluation pipelines, building databases to gather feedback, improving CI/CD, writing documentation, and improving the backend and frontend services as needed (non-exhaustive). You will gain valuable experience and skills in understanding chip design flows and applications. Open to proposals from all levels of ML practitioners.

ORAssistant - LLM Model Engineering

Topics: Large Language Model, Machine Learning, Model Architecture, Model Deployment
Skills: large language model engineering, prompt engineering, fine-tuning
Difficulty: Medium
Size: Medium (175 hours)
Mentor: Jack Luar & Palaniappan R

This project is aimed at enhancing robustness and accuracy for OR Assistant, the conversational assistant for OpenROAD through enhanced model architectures. You will work with members of the OpenROAD team and other researchers to explore alternate architectures beyond the existing RAG-based implementation. This project will focus on improving reliability and accuracy of the existing model architecture. You will collaborate on a tandem project on data engineering for OR assistant. Tasks include: reviewing and understanding the state-of-the-art in retrieval augmented generation, implementing best practices, caching prompts, improving relevance and accuracy metrics, writing documentation and improving the backend and frontend services as needed (non-exhaustive). You will gain valuable experience and skills in understanding chip design flows and applications. Open to proposals from all levels of ML practitioners.

OpenROAD - An Open-Source, Autonomous RTL-GDSII Flow for Chip Design

Mon, 22 Jan 2024 00:00:00 +0000

OpenROAD massively scales and supports EWD (Education and Workforce Development) and supports a broad ecosystem making it a vital tool that supports a rapidly growing Semiconductor Industry.

Create OpenROAD Tutorials and Videos

Topics: Documentation, Tutorials, Videos, VLSI design basics
Skills: Video/audio recording and editing, training and education
Difficulty: Medium
Size: Large (350 hours)
Mentor: Indira Iyer, Vitor Bandeira

Create short videos for training and course curriculum highlighting key features and flows in OpenROAD-flow-scripts.

Improve the OpenROAD AutoTuner Flow and documentation

Topics: OpenROAD-flow-scripts, AutoTuner, Design Exploration
Skills: Knowledge of ML for hyperparameter tuning, Cloud-based computation, Basic VLSI design and tools knowledge, python, C/C++
Difficulty: Medium
Size: Large (350 hours)
Mentor: Vitor Bandeira, Indira Iyer

Test, analyze and enhance the AutoTuner to improve usability, documentation and QoR. The Autotuner is an important tool in the OpenROAD flow - OpenROAD-flow-scripts for Chip design exploration that significantly reduces design time. You will use state-of-the-art ML tools to test the current tool exhaustively for good PPA (performance, power, area) results. You will also update existing documentation to reflect any changes to the tool and flow.

Implement a memory compiler in the OpenROAD Flow

Topics: OpenROAD-flow-scripts, Memory Compiler,
Skills: Basic VLSI design and tools knowledge, python, tcl, C/C++, memory design a plus
Difficulty: Medium
Size: Medium (175 hours)
Mentor: Matt Liberty, Austin Rovinski

Implement a memory compiler as part of the OpenROAD flow to improve the placement and layout efficiency of large, memory-intensive designs. You will start with an existing code base to develop this feature: https://github.com/The-OpenROAD-Project-staging/OpenROAD/tree/dffram This is another option: https://github.com/AUCOHL/DFFRAM Enhance code to support DFFRAM support for the OpenROAD native flow, OpenROAD-flow-scripts.

Integrate a tcl and python linter

Topics: Linting, Workflow
Skills: tcl, python, linting
Difficulty: Easy
Size: Small (90 hours)
Mentor: Vitor Bandeira, Austin Rovinski

Integrate a tcl and python linter for tools in OpenROAD and OpenROAD-flow-scripts to enforce error checking, style and best practices.

LLM assistant for OpenROAD - Create Model Architecture and Prototype

Topics: Large Language Model, Machine Learning, Model Architecture, Model Deployment
Skills: large language model engineering, prompt engineering, fine-tuning
Difficulty: Medium
Size: Medium (175 hours)
Mentor: Indira Iyer, Jack Luar

This project involves the creation of a conversational assistant designed around OpenROAD to answer user queries. You will be working in tandem with members of the OpenROAD team and other researchers to deliver a final deployable prototype. You will focus on the design and implementation of modular LLM architectures. You will be experimenting through different architectures and justifying which approach works the best on our domain-specific data. Open to proposals from all levels of ML practitioners.

LLM assistant for OpenROAD - Data Engineering and testing

Topics: Large Language Model, Machine Learning, Data Engineering, Model Deployment, Testing
Skills: large language model engineering, prompt engineering, fine-tuning
Difficulty: Medium
Size: Medium (175 hours)
Mentor: Indira Iyer, Jack Luar

This project involves the creation of a conversational assistant designed around OpenROAD to answer user queries. You will be working in tandem with members of the OpenROAD team and other researchers to deliver a final deployable prototype. This project will focus on the data engineering portion of the project. This may include: training pipelines specifically tailored for fine-tuning LLM models, data annotation, preprocessing and augmentation. Open to proposals from all levels of ML practitioners.

Create Unit tests for OpenROAD tools

Topics: OpenROAD-flow-scripts, unit testing
Skills: Basic VLSI design and tools knowledge, python, tcl, C/C++, Github
Difficulty: Medium
Size: Medium ( 175 hours)
Mentor: Vitor Bandeira, Indira Iyer

You will build unit tests to test specific features of the OpenROAD tool which will become part of the regression test. Here is an example of a test for UPF support: https://github.com/The-OpenROAD-Project/OpenROAD/blob/master/test/upf/mpd_aes.upf. This is a great way to learn VLSI flow basics and the art of testing them for practical applications.