Verification automation
Run agreed verification jobs, collect their evidence, and keep missing or failed checks visible.
How the check works
A result must match the work that ran.
Pin the inputs
Fetch the agreed source revisions into a new working directory.
Run the checks
Execute the regression with time limits and retained output.
Validate the report
Require the complete result set and matching source hashes.
Concept illustration. Missing, stale, or malformed evidence is an unresolved run, never a verification pass.
What you receive
- A reproducible runner for the agreed verification jobs.
- A machine-readable summary tied to source revisions and recorded evidence.
- Checks for missing, stale, malformed, or incomplete results.
- CI integration and instructions to repeat the regression.
When this is useful
- Regression runs rely on manual setup and log inspection.
- Results need to be traced to the exact source and configuration.
- Missing tools, timeouts, or incomplete reports are difficult to distinguish from design failures.
Problem, approach, and required inputs
The problem
A green script exit can hide a missing test, an old report, or a failed tool. Without a source revision and complete result checks, a regression is hard to trust or repeat.
What we do
We automate setup and execution at pinned revisions, apply runtime limits, and validate the expected result set. Reports retain commands, source hashes, logs, and measured durations. Infrastructure failures remain distinct from an expected test failure that demonstrates a seeded RTL defect.
Inputs agreed after scoping and NDA
- The existing test commands, dependencies, and expected result format.
- The source repositories and configurations to run.
- The required runtime limits, reporting format, and evidence retention rules.
Related public demonstration
Public demonstration, not a client result
A green regression needs a complete report
Run three pinned RTL projects and reject incomplete, stale, or inconsistent evidence before reporting success.
More problems you can reproduce
Technical details
How the result is checked
The public automation example fetches the published buffer, reset, and arbiter repositories at fixed commits and checks their full expected result matrices. Separate infrastructure tests exercise the reporting failure paths.
Tools used for this kind of work
- Python
- GitHub Actions
- Icarus Verilog
- Git
Scripts are handed over as agreed in the scope. If we have not tested your simulator or flow, the scope says so.
Discuss this problem.
Tell us what you are trying to verify and what is blocking you. A person replies, and the scope comes back in writing after a call.
Discuss a verification problem