2025-12-03 TSC Meeting
Date
Dec 3, 2025
Disclosures
Participants
Agenda
Slides: link
Slides: link
Draft paper: link
Slides: link
Overflow Meeting Topics (next week)
Presentations
Votes
Notes & Action Items
Meeting recording and transcript: link
Meeting Summary
Sail Yearly Update
The discussion focused on the annual update of the RISC-V Sail project, presented by Jeff Scheel. The community has made significant progress, with binary releases now available every 3-6 months and a CI system in place. They completed 17 new extensions and reduced their backlog from 57 to 27 PRs. Configuration work for RVA23 certification has been completed, and the mandatory gaps have been reduced to 1.1 PY and 13 PRs. The team is now focusing on privilege extensions like hypervisor and debug. Jeff mentioned that the balance between paid resources and volunteers has helped accelerate contributions. The conversation ended with a discussion on making RISC-V Sail model code more consumable, potentially through an online documentation system.
The group discussed the need to track and address inconsistencies between different implementations of specifications, particularly between Sail and ACTs. Jeff agreed to investigate the process for handling clarifications in ISA specifications and ensure Sail and ACTs stay in sync. The team also discussed creating a centralized system for tracking clarifications and updates to the ISA manual, with a suggestion to use GitHub issues labeled appropriately.
Continuation: Golden Reference Model Survey
The discussion focused on reviewing and refining a survey designed to gather input from members on the Golden Reference Model's priorities and future development. Derek presented version 4 of the survey, which included updated questions addressing controversial topics from previous versions. Greg and Manu proposed using a ranking system to determine priorities, with Greg suggesting a flexible point allocation method. The group discussed the survey's distribution by the ISA Infrastructure HC and its potential review by the Golden Model TG and TSC. Ken inquired about the Golden Model SIG's internal priorities, which Rafael clarified is not a SIG but a TG led by staff.
The team discussed a survey about Golden Reference Models (GRMs) and their features. Ken requested to avoid initialisms (acronymns). They reviewed questions about the importance of various GRM features, including instruction set simulators, documentation generation, formal analysis, and artifact generation for downstream projects. Derek explained that they want to generate human-readable documentation from GRM sources to reduce the need to keep multiple specifications in sync. Krste warned against trying to design a system through surveys, suggesting instead that they should present concrete proposals for different tool organizations and let people provide feedback about them.
The group discussed the approach to evaluating tool proposals, with Krste emphasizing the need for concrete, complete proposals rather than surveys, as surveys alone cannot reach concrete conclusions. They agreed that different people should put together different proposals capturing all important features, which would then be reviewed and potentially iterated upon. The discussion also touched on the need for a complete tool flow rather than focusing solely on the Golden Reference Model (SAIL), with Krste noting that different proposals could include SAIL as one component among others.
Functional Safety White Paper
Daniel presented a presentation about the functional safety white paper that provides an overview of safety mechanisms and processes, identifies existing RISC-V specifications that can be used for safety solutions, and offers recommendations for implementation. Greg and Philipp asked questions about whether the white paper could serve as a checklist or catalog of requirements for new specifications, but Daniel explained that safety domains are too diverse to provide a single comprehensive set of requirements.
Daniel presented a white paper on functional safety in RISC-V, which is split into two documents: a public white paper detailing current functional safety coverage and a separate document highlighting future recommendations for the TSC. The white paper covers topics like caches, performance counters, quality of service, redundancy, and error reporting, with existing RISC-V solutions discussed for each area. Daniel noted that redundancy and error detection/correction are not standardized in RISC-V but are not identified as gaps, while quality of service is addressed at the system level.
The discussion focused on the white paper discussion about error correction codes and redundancy mechanisms in RISC-V architecture. Daniel explained that while other architectures provide solutions for dual-core lockstep and TMR at the platform level, RISC-V does not have specific ISA-level guidance. The group discussed next steps, including continuing review and feedback on the paper before the holidays, with formal TSC approval to follow. Jeff mentioned that after the gap analysis is complete, they will work with Daniel to determine what actions are needed and how to present the findings to task group chairs.
Detailed Minutes
Sail Yearly Update
This is the 3rd yearly update of Sail progress.
Slides from previous presentations are in the backup of the presentation. The format of certain slides have been reused for bridging.
The 2025 progress was impressive from a community contribution perspective.
Additional RVI Staffing (contractor new for 2025) has provided a good balance between member contributions and community needs
Progress on RVA23 has nearly completed all Unpriv instructions (a few late configuration items were found recently and added, see purple items). Focus turning to Priv.
A few large items like Debug and Hypervisor remain but are starting to be included in small chunks.
Action Item:
Readiness for RVI20 by Sail (Jeff, Prashanth)Process for handling ISA Clarifications as they apply to Sail (Jeff, Prashanth)ISA clarifications agreed to by ARC need to ensure that impacts to Sail and ACT are understood and addressed by issue raiser. (Greg, Krste, ARC)
Continuation: Golden Reference Model Survey
Version 4 of survey doc is available and has been reviewed
Reminder that this a proposal to get information on how the Golden Models are being used.
Discussion:
Question 4:
Has been slimmed down from previous version (version 3).
There’s not cross-ranking across all things so how do we prevent everything from being a 1 or a 5.
Does the ISA Infra HC have expected priorities for these items?
Perhaps we want priorities from the CSC?
Suggest we remove the TLAs.
Question 5:
GRM used as a term because of the fact that they can generate simulators.
What flavors of documentation is intended?
Answer: Ops codes, formal description, etc. Want toDesign by survey is difficult is accomplish. The alternative approach is to put a couple proposals and solicit feedback. The design work is inevitable in this process.
The survey is attempting to prioritize the “whats”. An alternate technique is to declare the “whats” and review them. The advantage of the latter is to help people understand the environment.
Question 6: Simply focused on use
Each simulator/model has a role and we need to ensure synergy and appropriate usage.
Next steps:
Call for proposals need to be articulated for the flow. (Derek and Manu)
Note: there has been some work in this space between Derek and Andrea.Review of flow and proposals, starting at email but back into TSC meeting as-needed. (Derek)
Functional Safety White Paper
Presentation included here: link
Discussion: