Background

Challenges in Traditional Physical Design Implementation

Manual methods and fragmented toolchains cause:

• Disconnected Flow: Synthesis, placement, routing, and signoff tools lack integration, leading to manual errors.
• Limited Flexibility: Rigid scripts hinder tuning for timing, power, or area optimization.
• Error-Prone Tasks: Manual script handling increases mistakes and debugging time.
• Inconsistent Results: Non-standard methods cause variability across teams and runs.
• Slow Debugging: No automated checks; tracking timing, DRC, or LVS errors is manual and slow.
• Poor Adaptability: Script-heavy flows struggle with varying tech nodes or project scales.
• Limited Visibility: Lack of centralized control impairs tracking and collaboration.

Our Solution:

An integrated, automated physical design flow for reliable, flexible, and scalable RTL-to-GDSII implementation.

Methodology

Proposals to Overcome Challenges:

• Unified Toolchain: Streamline the entire RTL-to-GDSII process through an integrated ecosystem, eliminating manual data transfers.
• Configurable Flexibility: Enable easy tuning of tool parameters and flow settings to meet diverse design objectives.
• Automated Workflow: Reduce manual intervention by automating repetitive tasks, improving accuracy and efficiency.
• Standardized Processes: Ensure consistent results across runs and teams through well-defined methodologies.
• Built-In Error Detection: Incorporate diagnostics for early identification of issues, significantly reducing debug time.
• Scalable Architecture: Support projects of varying complexity and technology nodes without major flow changes.
• Centralized Control & Monitoring: Integrate version control and real-time dashboards for enhanced visibility, traceability, and project management.

Customization and Configurability:

• Fully Configurable Settings: Modify timing, power, or routing settings for optimal results.
• Custom Flow Options: Tailored flow behavior to meet specific design goals.
• Project Flexibility: Adapts to diverse designs without major script changes

Advantages of the Custom Flow:

• Faster Execution: Automation and tool integration speed up project timelines.
• Reliable Results: Standard methods ensure consistent, high-quality outcomes.
• High Adaptability: Modular design handles varying project sizes and technology nodes.
• User-Friendly: Clear guides make the flow easy for all engineers to use.
• Optimized Designs: Flexible settings improve performance, power, and area efficiency.
• Improved Oversight: Git-based version control and a dashboard provide clear visibility into design progress and status.
• PDK Flexibility: Integrates any PDK with minimal modifications, supporting diverse process nodes.

Result & Analysis

Adoption Time:

• Custom flow reduced adoption time from 12 weeks to 4 weeks.
• Fresh engineers adopted the flow more quickly and confidently.
• Enabled faster onboarding and early productivity.

PPA Improvement of a Design:

• It took less time with our flow to improve the PPA than the conventional flow.
• Fewer engineers are needed to improve the PPA with the help of this advanced flow

Augmenting OpenROAD with Domain-Specific Insights through a Tool-Agnostic Automated Design Flow

Using insights from our custom RTL-to-GDSII flow, we participated in a competition based on OpenROAD, where we enhanced OpenROAD’s default scripts to optimize synthesis and implementation, thereby improving design efficiency and performance, and ultimately won first prize in the performance category.

OpenRoad Competition: Design Requirements

Design	riscv32i and ibex
Process Node	7nm
Instance Count	riscv32i: 18k ibex: 11k
Number of Macros (riscv32i)	4
Methodology	Synthesis to Implementation
Target Frequency	Maximum Achievable for Both Designs
Tools	Yosys and OpenROAD
Requirements	DRC cleaned design
Prize Money	$800

OpenRoad Competition: Participants

• Professionals
• Academics
• Students
• PHd Students
• Industry Experts
• Hobbyists

OpenRoad Competition: Improvement Strategy

Implications

Enhancements:

To elevate our custom RTL-to-GDSII flow, we are focusing on:

• AI Integration: Leveraging AI for tool tuning and issue prediction to accelerate design convergence.
• Broader Tool Support: Enabling compatibility with multiple EDA vendors for increased flexibility.
• Advanced Dashboards: Introducing real-time analytics and predictive insights for improved tracking.
• Centralized Design Databases: Building unified repositories to access reports and summaries at any stage without reruns.

These enhancements aim to boost the flow’s efficiency, scalability, and reliability, supporting evolving VLSI design demands and advancing our engineering expertise.