DFT Orchestrator

Overview

This agent orchestrates the entire Design for Testability (DFT) flow, managing a complex sequence of stages: architecture setup $\rightarrow$ scan insertion $\rightarrow$ Automatic Test Pattern Generation (ATPG) $\rightarrow$ Built-In Self-Test (BIST) $\rightarrow$ JTAG setup $\rightarrow$ Final Sign-off. It adheres to strict loop-back rules and sign-off criteria to ensure comprehensive test coverage.

Capabilities

• Stage Sequencing: Executes the defined flow: dft_architecture $\rightarrow$ scan_insertion $\rightarrow$ atpg $\rightarrow$ bist_insertion $\rightarrow$ jtag_setup $\rightarrow$ dft_signoff.
• Tool Management: Supports both open-source (Yosys, OpenROAD) and proprietary tools (Synopsys TetraMAX, Cadence Modus). It prioritizes the MCP server execution method.
• Error Handling & Retries: Implements specific loop-back rules for failures (e.g., retrying scan_insertion on DRC errors).
• State Tracking: Tracks fault coverage across ATPG iterations and maintains a history of all executed stages in memory.
• Documentation: Reads initial knowledge from memory/dft/knowledge.md and logs detailed experience records to memory/dft/experiences.jsonl upon termination.

Example Use Cases

• Full Flow Execution: Initiate the process by calling the orchestrator when a new RTL design requires full DFT verification, ensuring all stages complete successfully.
• Troubleshooting: If ATPG fails due to low SAF coverage, the agent will automatically loop back and suggest re-running scan_insertion with adjustments.
• Sign-off Verification: Use it to confirm that the final netlist meets sign-off criteria (zero DRC errors, $\ge 99.0%$ SAF coverage, BIST pass).