Tutorial on G-QED Pre-Silicon Verification: Part 2 Tuesday, April 23, 12-2 pm EDT
You are cordially invited to the second part of the Tutorial on G-QED Pre-Silicon Verification. In the first part, we learned about G-QED (Generalized Quick Error Detection), a highly thorough and highly efficient pre-silicon verification technique, and how designers can utilize G-QED in their design flows. The tutorial also provided the interested participants the opportunity to verify an open-source memory controller design using G-QED in a commercial verification environment. In the second session, we demonstrate in depth how to verify the given design using G-QED and answer questions about this experiment. At the end of this session, the code for G-QED checks for a set of open-source designs will be shared with the participants.
Why G-QED? Extreme heterogeneity in digital systems due to the wide variety of hardware accelerators and processors has made verification challenging. While this heterogeneity limits verification setup reuse, verification effort remains a hurdle to meeting the tight design-to-deployment timelines in current design flows. Over 80% of design projects grapple with critical bug escapes and fall behind schedule, even after extensive verification. G-QED addresses this pressing issue and drastically improves productivity and coverage of verification: As opposed to traditional verification approaches, G-QED does not require understanding internal design details, significantly reducing verification effort. Within the capabilities of existing Bounded Model Checking (BMC) tools, G-QED is provably sound and complete, i.e., detects all logic bugs without any false fails. In an industrial case study on production-ready AI engines (previously verified extensively for over a year), G-QED detected 9 new critical bugs, in addition to all bugs detected by the industrial verification flow, in just 3 person weeks -- an 18-fold productivity boost.
The second part of the tutorial of the will take place on Tuesday April 23, 12-2pm ET.
If you have not registered for Part 1 but would like to attend Part 2, please register at the following link: https://forms.gle/6FJDDWRQCgFsHj2v5 (new participants only).