Open Source Software in Silicon Manufacturing
Matthew Hiltner
https://linux.conf.au/schedule/30139/...
Intel uses feedback from manufacturing test to continuously improve product yields and drive cost savings. Open source software plays an important role in this process, as its flexibility and transparency allow rapid innovation in manufacturing test development.
This presentation explores the use of open source software as part of system-level test for silicon components. By using open and well-understood customer-like workloads, Intel is able to effectively identify manufacturing defects. More importantly, key workloads that expose abnormal failure rates are integrated into more directed test environments, such as a test of a chip before it is even packaged.
The presentation will then focus on the cost impact of late defect detection, and how open source software specifically enables earlier detection. Techniques such as execution recording and CPU state insertion by PSMI, RTOS-like execution of directed workloads within NMISR will be described.
The presentation ends by describing previous (and ongoing) challenges working in a closed-source environment, and how key learnings have dramatically increased awareness and appreciation of open source software within a decidedly hardware-centric field.