Intel

The problem

Validating a hardware video decoder before silicon exists means you can't just play videos at it. You need bitstreams crafted to exercise specific corners of the design.

What I built

On Intel's graphics chipset division, I worked in a small team on pre-silicon validation of a hardware-based VC-1 (formerly WMV-9) video decoder. The novel piece: I built a video encoder in object-oriented Perl that gave testers fine-grained control over generated bitstreams. Testers could target exact decoder states instead of hoping real-world content happened to hit them.

It was my first taste of a lesson that has repeated ever since: the highest leverage engineering is often the tooling that makes everyone else's testing sharper.