From a7a85f04c49818e0ebad636ecfa3d00f360ef057 Mon Sep 17 00:00:00 2001 From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com> Date: Wed, 3 Sep 2025 00:57:22 +0000 Subject: [PATCH 1/2] Initial plan From 766099d70fb0ab18f86124ba7c293a69786b7ad3 Mon Sep 17 00:00:00 2001 From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com> Date: Wed, 3 Sep 2025 01:03:12 +0000 Subject: [PATCH 2/2] Add comprehensive LLM4ChipDesign course syllabus with 14-week curriculum Co-authored-by: FCHXWH823 <176723341+FCHXWH823@users.noreply.github.com> --- LLM4ChipDesign_Course_Syllabus.html | 561 ++++++++++++++++++++++++++++ LLM4ChipDesign_Course_Syllabus.md | 282 ++++++++++++++ README.md | 8 + 3 files changed, 851 insertions(+) create mode 100644 LLM4ChipDesign_Course_Syllabus.html create mode 100644 LLM4ChipDesign_Course_Syllabus.md diff --git a/LLM4ChipDesign_Course_Syllabus.html b/LLM4ChipDesign_Course_Syllabus.html new file mode 100644 index 0000000..a9cb7e5 --- /dev/null +++ b/LLM4ChipDesign_Course_Syllabus.html @@ -0,0 +1,561 @@ + + + + + + LLM4ChipDesign Course Syllabus + + + +
+

LLM4ChipDesign Course Syllabus

+

Generative AI for Chip Design - 14 Week Curriculum

+
+ +
+

Course Information

+ +
+ +
+

Course Objectives

+

By the end of this course, students will:

+ +
+ +
+

Module Schedule Overview

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
WeeksModuleTopicKey Technologies
1-2Module 1AutoChip - Automated Verilog GenerationBasic LLM prompting, Error feedback loops
3-4Module 2VeriThoughts - Reasoning-Based GenerationFormal verification, Reasoning frameworks
5-6Module 3ROME - Hierarchical PromptingHierarchical design, Complex systems
7-8Module 4LLM-Aided Testbench GenerationVerification, Coverage analysis
9-10Module 5Hybrid-NL2SVA - Assertion GenerationSystemVerilog assertions, Security
11-12Module 6C2HLSC - Software to Hardware BridgeHigh-level synthesis, C-to-hardware
13-14Module 7PrefixLLM - Specialized Circuit DesignPrefix circuits, Design optimization
+
+ +
+

Detailed Module Breakdown

+ +
+

Module 1: AutoChip - Automated Verilog Generation (Weeks 1-2)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Understand the basics of LLM-assisted Verilog generation
    • +
  • Learn AutoChip's iterative feedback mechanism
    • +
  • Explore error-driven design refinement
    • +
+
+
+ Homework Assignment: +
    +
  • Design a simple combinational circuit (e.g., 4-bit adder) using AutoChip
    • +
  • Analyze and document the iterative refinement process
    • +
  • Compare LLM-generated code with manual implementation
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • Review homework solutions and common issues
    • +
  • Hands-on workshop with AutoChip tutorial
    • +
  • Debugging techniques for LLM-generated Verilog
    • +
  • Best practices for prompt design
    • +
+
+
+ +
+

Module 2: VeriThoughts - Reasoning-Based Verilog Generation (Weeks 3-4)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Understand reasoning-based approaches to hardware description
    • +
  • Learn formal verification integration with LLM generation
    • +
  • Explore specialized small-scale models for Verilog
    • +
+
+
+ Homework Assignment: +
    +
  • Implement a finite state machine using VeriThoughts methodology
    • +
  • Apply formal verification to validate correctness
    • +
  • Compare reasoning-based vs. direct generation approaches
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • Solution walkthrough for FSM implementation
    • +
  • Formal verification workshop
    • +
  • Performance comparison analysis
    • +
  • Discussion on correctness guarantees
    • +
+
+
+ +
+

Module 3: ROME - Hierarchical Prompting for Complex Designs (Weeks 5-6)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Master hierarchical design decomposition strategies
    • +
  • Understand multi-level prompting techniques
    • +
  • Learn to handle complex hardware modules
    • +
+
+
+ Homework Assignment: +
    +
  • Design a hierarchical processor component (e.g., ALU with multiple sub-modules)
    • +
  • Implement using both flat and hierarchical prompting
    • +
  • Analyze cost and time benefits of hierarchical approach
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • Hierarchical design solution review
    • +
  • Advanced ROME techniques workshop
    • +
  • Cost-benefit analysis discussion
    • +
  • Complex design case studies
    • +
+
+
+ +
+

Module 4: LLM-Aided Testbench Generation and Verification (Weeks 7-8)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Understand automated testbench generation principles
    • +
  • Learn coverage-driven verification with LLMs
    • +
  • Master EDA tool integration techniques
    • +
+
+
+ Homework Assignment: +
    +
  • Generate comprehensive testbenches for a finite state machine
    • +
  • Achieve target coverage metrics
    • +
  • Implement bug detection and reporting workflow
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • Testbench solution analysis
    • +
  • Coverage optimization techniques
    • +
  • EDA tool integration workshop
    • +
  • Real-world verification scenarios
    • +
+
+
+ +
+

Module 5: Hybrid-NL2SVA - Natural Language to SystemVerilog Assertions (Weeks 9-10)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Master assertion-based verification concepts
    • +
  • Learn natural language to SystemVerilog translation
    • +
  • Understand security-centric assertion development
    • +
+
+
+ Homework Assignment: +
    +
  • Convert natural language security requirements to SVA
    • +
  • Implement assertion suite for a cryptographic module
    • +
  • Validate assertions using formal verification
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • SVA implementation review
    • +
  • Security assertion patterns workshop
    • +
  • Formal verification of assertions
    • +
  • Industry best practices discussion
    • +
+
+
+ +
+

Module 6: C2HLSC - Software to Hardware Design Bridge (Weeks 11-12)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Understand High-Level Synthesis (HLS) principles
    • +
  • Learn C-to-hardware translation techniques
    • +
  • Master LLM-guided code transformation
    • +
+
+
+ Homework Assignment: +
    +
  • Transform a C algorithm (e.g., sorting, encryption) to HLS-compatible format
    • +
  • Optimize for area and timing constraints
    • +
  • Compare different optimization strategies
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • HLS optimization solution review
    • +
  • Performance analysis workshop
    • +
  • Area vs. timing trade-offs discussion
    • +
  • Advanced HLS techniques
    • +
+
+
+ +
+

Module 7: PrefixLLM - Specialized Circuit Design (Weeks 13-14)

+ +
+

Week 1: Fundamentals and Introduction

+
+ Learning Objectives: +
    +
  • Understand prefix circuit design principles
    • +
  • Learn structured representation for circuit synthesis
    • +
  • Master design space exploration with LLMs
    • +
+
+
+ Homework Assignment: +
    +
  • Design optimized prefix circuits for different bit widths
    • +
  • Implement design space exploration framework
    • +
  • Compare multiple optimization objectives
    • +
+
+
+ +
+

Week 2: Problem Solving and Practice

+
    +
  • Optimization results analysis
    • +
  • Design space exploration workshop
    • +
  • Multi-objective optimization discussion
    • +
  • Course wrap-up and future directions
    • +
+
+
+
+ +
+

Assessment Methods

+

Continuous Assessment (70%)

+ +

Final Project (30%)

+ +
+ +
+

Required Tools and Resources

+
+
+

Software Tools

+
    +
  • Verilator or equivalent Verilog simulator
    • +
  • Python environment with LLM libraries
    • +
  • Access to LLM APIs (OpenAI, Anthropic, etc.)
    • +
  • EDA tools for synthesis and verification
    • +
  • Git for version control
    • +
+
+
+

Hardware Requirements

+
    +
  • Computer with sufficient computational resources
    • +
  • GPU access recommended for training custom models
    • +
+
+
+
+ +
+

Course Resources

+

Tutorial Notebooks

+
+ AutoChip Tutorial + VeriThoughts Tutorial + ROME Demo + Testbench Generation + Hybrid-NL2SVA + C2HLSC Tutorial + PrefixLLM +
+ +

Reference Papers

+
+ AutoChip Paper + VeriThoughts Paper + ROME Paper + LLM Testbench Paper + Hybrid-NL2SVA Paper + C2HLSC Paper + PrefixLLM Paper +
+
+ +
+

Prerequisites

+ +
+ +
+

Contact Information

+ +

This syllabus is subject to modification as the course progresses. Students will be notified of any changes in advance.

+
+ + \ No newline at end of file diff --git a/LLM4ChipDesign_Course_Syllabus.md b/LLM4ChipDesign_Course_Syllabus.md new file mode 100644 index 0000000..3f4c3a5 --- /dev/null +++ b/LLM4ChipDesign_Course_Syllabus.md @@ -0,0 +1,282 @@ +# LLM4ChipDesign Course Syllabus +## Generative AI for Chip Design - 14 Week Curriculum + +--- + +### Course Information +- **Course Title:** Generative AI for Chip Design +- **Duration:** 14 Weeks (1 Semester) +- **Format:** 7 Modules, 2 Weeks per Module +- **Teaching Structure:** + - Week 1 of each module: Fundamental concepts and homework assignment + - Week 2 of each module: Homework solutions and hands-on practice + +--- + +### Course Objectives +By the end of this course, students will: +- Understand how Large Language Models (LLMs) can be applied to hardware design automation +- Master various LLM-based tools and techniques for Verilog generation, verification, and optimization +- Develop practical skills in prompt engineering for hardware design tasks +- Gain hands-on experience with state-of-the-art AI-assisted chip design workflows +- Learn to integrate LLM tools into existing hardware design methodologies + +--- + +### Prerequisites +- Basic knowledge of digital logic design +- Familiarity with Verilog/SystemVerilog +- Understanding of hardware design verification concepts +- Programming experience (Python preferred) +- Basic machine learning concepts (helpful but not required) + +--- + +## Module Schedule + +### **Module 1: AutoChip - Automated Verilog Generation (Weeks 1-2)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Understand the basics of LLM-assisted Verilog generation + - Learn AutoChip's iterative feedback mechanism + - Explore error-driven design refinement +- **Topics Covered:** + - Introduction to generative AI in hardware design + - AutoChip framework overview + - Prompt engineering for Verilog generation + - Compilation and simulation feedback loops +- **Homework Assignment:** + - Design a simple combinational circuit (e.g., 4-bit adder) using AutoChip + - Analyze and document the iterative refinement process + - Compare LLM-generated code with manual implementation + +**Week 2: Problem Solving and Practice** +- **Activities:** + - Review homework solutions and common issues + - Hands-on workshop with AutoChip tutorial + - Debugging techniques for LLM-generated Verilog + - Best practices for prompt design + +--- + +### **Module 2: VeriThoughts - Reasoning-Based Verilog Generation (Weeks 3-4)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Understand reasoning-based approaches to hardware description + - Learn formal verification integration with LLM generation + - Explore specialized small-scale models for Verilog +- **Topics Covered:** + - Reasoning frameworks in hardware design + - Formal verification methods for generated code + - VeriThoughts dataset and benchmarking + - Quality metrics for generated hardware descriptions +- **Homework Assignment:** + - Implement a finite state machine using VeriThoughts methodology + - Apply formal verification to validate correctness + - Compare reasoning-based vs. direct generation approaches + +**Week 2: Problem Solving and Practice** +- **Activities:** + - Solution walkthrough for FSM implementation + - Formal verification workshop + - Performance comparison analysis + - Discussion on correctness guarantees + +--- + +### **Module 3: ROME - Hierarchical Prompting for Complex Designs (Weeks 5-6)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Master hierarchical design decomposition strategies + - Understand multi-level prompting techniques + - Learn to handle complex hardware modules +- **Topics Covered:** + - Hierarchical design principles + - Bottom-up vs. top-down generation approaches + - ROME framework and scripted prompting + - Scaling challenges and solutions +- **Homework Assignment:** + - Design a hierarchical processor component (e.g., ALU with multiple sub-modules) + - Implement using both flat and hierarchical prompting + - Analyze cost and time benefits of hierarchical approach + +**Week 2: Problem Solving and Practice** +- **Activities:** + - Hierarchical design solution review + - Advanced ROME techniques workshop + - Cost-benefit analysis discussion + - Complex design case studies + +--- + +### **Module 4: LLM-Aided Testbench Generation and Verification (Weeks 7-8)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Understand automated testbench generation principles + - Learn coverage-driven verification with LLMs + - Master EDA tool integration techniques +- **Topics Covered:** + - Testbench generation methodologies + - Coverage metrics and analysis + - Iterative feedback from EDA tools + - Bug detection and classification +- **Homework Assignment:** + - Generate comprehensive testbenches for a finite state machine + - Achieve target coverage metrics + - Implement bug detection and reporting workflow + +**Week 2: Problem Solving and Practice** +- **Activities:** + - Testbench solution analysis + - Coverage optimization techniques + - EDA tool integration workshop + - Real-world verification scenarios + +--- + +### **Module 5: Hybrid-NL2SVA - Natural Language to SystemVerilog Assertions (Weeks 9-10)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Master assertion-based verification concepts + - Learn natural language to SystemVerilog translation + - Understand security-centric assertion development +- **Topics Covered:** + - SystemVerilog assertion (SVA) fundamentals + - Natural language processing for hardware properties + - Security assertion patterns + - RAG (Retrieval-Augmented Generation) techniques +- **Homework Assignment:** + - Convert natural language security requirements to SVA + - Implement assertion suite for a cryptographic module + - Validate assertions using formal verification + +**Week 2: Problem Solving and Practice** +- **Activities:** + - SVA implementation review + - Security assertion patterns workshop + - Formal verification of assertions + - Industry best practices discussion + +--- + +### **Module 6: C2HLSC - Software to Hardware Design Bridge (Weeks 11-12)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Understand High-Level Synthesis (HLS) principles + - Learn C-to-hardware translation techniques + - Master LLM-guided code transformation +- **Topics Covered:** + - HLS design flow and challenges + - C code refactoring for hardware synthesis + - Streaming data and hardware-specific signals + - Hierarchical design decomposition +- **Homework Assignment:** + - Transform a C algorithm (e.g., sorting, encryption) to HLS-compatible format + - Optimize for area and timing constraints + - Compare different optimization strategies + +**Week 2: Problem Solving and Practice** +- **Activities:** + - HLS optimization solution review + - Performance analysis workshop + - Area vs. timing trade-offs discussion + - Advanced HLS techniques + +--- + +### **Module 7: PrefixLLM - Specialized Circuit Design (Weeks 13-14)** + +**Week 1: Fundamentals and Introduction** +- **Learning Objectives:** + - Understand prefix circuit design principles + - Learn structured representation for circuit synthesis + - Master design space exploration with LLMs +- **Topics Covered:** + - Prefix adder architectures + - Structured Prefix Circuit Representation (SPCR) + - Automated design space exploration + - Area and delay optimization techniques +- **Homework Assignment:** + - Design optimized prefix circuits for different bit widths + - Implement design space exploration framework + - Compare multiple optimization objectives + +**Week 2: Problem Solving and Practice** +- **Activities:** + - Optimization results analysis + - Design space exploration workshop + - Multi-objective optimization discussion + - Course wrap-up and future directions + +--- + +### Assessment Methods + +**Continuous Assessment (70%)** +- Weekly homework assignments (40%) +- Module quizzes (20%) +- Participation and engagement (10%) + +**Final Project (30%)** +- Comprehensive project integrating multiple modules +- Original research or advanced implementation +- Presentation and documentation + +--- + +### Required Tools and Resources + +**Software Tools:** +- Verilator or equivalent Verilog simulator +- Python environment with LLM libraries +- Access to LLM APIs (OpenAI, Anthropic, etc.) +- EDA tools for synthesis and verification +- Git for version control + +**Hardware:** +- Computer with sufficient computational resources +- GPU access recommended for training custom models + +**Reading Materials:** +- Research papers provided for each module +- Online documentation and tutorials +- Supplementary materials in colab-scripts directory + +--- + +### Additional Resources + +**Jupyter Notebooks:** +- `AutoChip_Tutorial.ipynb` +- `VeriThoughts_Tutorial.ipynb` +- `ROME_demo.ipynb` +- `LLM_Aided_Testbench_Generation_for_FSM.ipynb` +- `Hybrid_NL2SVA.ipynb` +- `C2HLSC_Tutorial.ipynb` +- `PrefixLLM.ipynb` + +**Reference Papers:** +- AutoChip: https://arxiv.org/abs/2311.04887 +- VeriThoughts: https://arxiv.org/abs/2505.20302 +- ROME: https://arxiv.org/abs/2407.18276 +- LLM Testbench: https://arxiv.org/html/2406.17132v1 +- Hybrid-NL2SVA: https://arxiv.org/pdf/2506.21569 +- C2HLSC: https://arxiv.org/abs/2412.00214 +- PrefixLLM: https://arxiv.org/abs/2412.02594 + +--- + +### Contact Information +- **Instructor:** [To be filled] +- **Office Hours:** [To be scheduled] +- **Course Repository:** https://github.com/FCHXWH823/LLM4ChipDesign + +--- + +*This syllabus is subject to modification as the course progresses. Students will be notified of any changes in advance.* \ No newline at end of file diff --git a/README.md b/README.md index 5a0c872..d38ef3b 100644 --- a/README.md +++ b/README.md @@ -1,5 +1,13 @@ # Generative AI for Chip Design +## 📚 Course Syllabus +**NEW:** [**LLM4ChipDesign Course Syllabus**](LLM4ChipDesign_Course_Syllabus.md) - A comprehensive 14-week curriculum covering all modules with structured learning objectives, homework assignments, and assessment methods. + +- 📄 **Markdown Version:** [LLM4ChipDesign_Course_Syllabus.md](LLM4ChipDesign_Course_Syllabus.md) +- 🌐 **HTML Version:** [LLM4ChipDesign_Course_Syllabus.html](LLM4ChipDesign_Course_Syllabus.html) + +--- + ## Table of Contents - [AutoChip to Generate Functional Verilog](#autochip-to-generate-functional-verilog) - [VeriThoughts: Enabling Automated Verilog Code Generation using Reasoning and Formal Verification](#verithoughts-enabling-automated-verilog-code-generation-using-reasoning-and-formal-verification)