../
Meeting with Kevin
Minutes
Brad
- Brad talked to a bunch of lawyers
- https://www.ibm.com/docs/en/systems-hardware/zsystems/9175-ME1?topic=introduction-spyre-accelerator
Yasas
- Boom core vcd mining takes a week
- Mining is done using Texada
- HARM still crashes due to the memory requirement
- It seems to convert everything into a csv and then operate on it rather than use the vcd directly
- Maybe removing signals could help?
Kumaresh
- 10 new templates are being tried out on CVA6 using both HARM and texada
Why did we choose HARM over texada?
- Texada seems faster according to Yasas
- But germinianiBaselineFrameworkQualification2025 suggest otherwise. HARM seems magnitudes faster than Texada, at least in
G(A -> X(B)). - HARM doesn’t support other templates
- Texada is actually written in C++. I misspoke. It’s Goldminer that is written in Python.
No meeting next week.