UVM Methodology Skill

SkillDev tools

Deep expertise in Universal Verification Methodology (IEEE 1800.2) for FPGA verification

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the UVM Methodology Skill skill

What this skill tells your AI

The instructions your AI receives, as published by a5c-ai/babysitter in library/specializations/fpga-programming/skills/uvm-methodology/SKILL.md and read by ahel’s review.

Overview

Expert skill for Universal Verification Methodology (UVM) development following IEEE 1800.2 standards for comprehensive FPGA verification.

Capabilities

  • Generate UVM agent architecture (driver, monitor, sequencer)
  • Create UVM environments and scoreboards
  • Implement uvm_sequence and virtual sequences
  • Configure UVM factory and config_db
  • Implement functional coverage with covergroups
  • Design UVM register models (RAL)
  • Apply UVM phasing and objections correctly
  • Debug UVM testbenches effectively

Target Processes

  • uvm-testbench.js
  • constrained-random-verification.js
  • testbench-development.js

Usage Guidelines

Agent Architecture

  • Driver: Converts sequence items to pin-level activity
  • Monitor: Observes DUT interface and creates transactions
  • Sequencer: Routes sequence items to driver
  • Agent: Contains driver, monitor, sequencer; configurable active/passive

Environment Structure

  • Top-level environment contains agents and scoreboard
  • Scoreboard performs reference model comparison
  • Config objects distribute configuration
  • Virtual sequencer coordinates multiple agents

Sequence Development

  • Extend from uvm_sequence#(item_type)
  • Use start_item() / finish_item() paradigm
  • Create layered sequences for complex scenarios
  • Use virtual sequences for multi-agent coordination

Coverage Strategy

  • Embed covergroups in monitors
  • Sample on transaction completion
  • Cross functional coverage points
  • Track coverage closure progress

Best Practices

  • Use factory for all component creation
  • Configure via config_db, not constructors
  • Raise/drop objections properly
  • Use UVM reporting macros consistently

Dependencies

  • UVM 1.2 or UVM IEEE 1800.2 library
  • SystemVerilog expertise
  • Verification methodology knowledge

Signals

GitHub stars
2k
Forks
112
Last commit
Sep 2026
Advanced
Item type
skill
Key
uvm-methodology
Source
github.com/a5c-ai/babysitter