TSC Strategic Priorities Meeting
Date
Jun 11, 2025 Jul 9, 2025 Aug 6, 2025
Disclosures
Participants
Agenda
Presentations
Title | Presenter | File | |
|---|---|---|---|
| 1 |
|
|
|
| 2 |
|
|
|
Votes
None
Notes & Action Items
Audio links:
Detailed Notes
Next TSC Meeting: Platform SIG Needs Discussion (Andrea, Greg) (10 Minutes)
Background for meeting:
The goal is alignment within RISC-V
Participation is important
Resources are limited, thus let’s focus on items we can achieve within the next year
Collaboration makes us all more efficient and improves ROI
Premier Member Gaps Discussion
ISA
Virtualization for Real-time Processors noted as important to Automotive – BOSC, SiFive, Infineon, Synopsys, and others…
Enterprise/Datacenter/… gaps (#3)
Needs to be split in multiple pieces
Will focus on ISA items more than non-ISA
Vector item (#5) has new item of 56x56 carry-save-multi-add.
Speculation barriers being worked
Scalar (#7)
Most being worked starting with SIG, but also includes proposed Long Instruction TG
Sail (#8) wants Sail to move faster.
Long Instruction (#9) supported by Akeana as well.
Atomic Enqueue (#10) newly raised to mailing lists. Support being posted to list would be beneficial.
Vector Extension (#11)
crcfolder and vgmulxor instructions apply to storage. No need for Storage SIG at this time but will explore. Note: Carryless multiply helps here too.
fp4 not yet being looked at in FP SIG and Vector SIG
SFU item are motivated by AI activation functions, DSP, and other specialized applications
Non-ISA
Opensource IOMMU implementation (#1) available at University of Minho (Portugal): link | source
Discussion around reference implementations occurred: RISC-V remains focused on being a standards body
I/D Cache coherency (#2) started under J-ext TG (moved to SIG). FT extension being explored as starting point. More work expected. (Should move to ISA, 3-6)
HPE being worked (#4)
User mode timer and interrupts (#5) has been raised through HP IO SIG.
Documentation (#7) being shared with TSC, CSC, Doc SIG.
A lot of work to do around normative rule declaration. CSC, UDB SIG, Doc SIG building a proposal.
CSC and TSC will need to collaborate on the list to achieve successful results.
Side-channel (#10) also includes work on Speculation Barriers (new TG)
Profiles
Next RVA Profile (#1)
Next major 2+ years away (Profiles SIG)
Expect 1-2 minor releases pending
Security Profile (#2) discussion
Being built as “Security Model” specification
Might need a platform to build extensions + Security Model
RVA23+ challenges… (#3)
OSs tend to do WARL discovery making VM migration difficult
Debug/Trace and Interrupt Profiles (#6)
This is a request for inclusion of items in this area
Debug and interrupts are more of a Platform level consideration
These are key areas for Certification work as well
Platforms
#2 not solvable by RISC-V
#3 needs more refinement to clarify which areas of Automotive
Appears #4 not solvable by RISC-V as written, but may need more deep look
Discussion: are there areas of standardization that might not be purely RISC-V centric but still beneficial, e.g. AXI, AMBA.
Would recommend SOC Infrastructure HC lead this look
Certification Authority (#8) for software signing
Certificate to be provided to Secure Boot.
Andrea Gallo will help drive this considering the strategic and operational fronts.
#9 Platform Systems
Future new Industrial SIG Andrea Gallo and Philipp Tomsich and Manu Gulati (Deactivated)
Software
#1 AI extensions
Compilers already in Acceptance Criteria (Jira subtasks) so will apply to AI extensions
For AI/ML, need to ensure we cover up to frameworks/tools such as MLIR
Interlock with RISE, including TSC
#2
#7 Unified DB
Generated items need a process for review. Separating syntax and semantics as 2 parts may help. Finalize work on encodings (syntax).
Numerous items in the current process (opcodes db, Sail) are in use but clearly more efficiencies here.
UDB is working on how to help resolve duplications and provide greater efficiencies. But this will be complex to cover everything.
#8 appears to be in support of existing work
Strategic Member Gaps
ISA
#1, #2 have been covered by Premier gaps
#3 Custom Co-processors – defer until Krste and Greg are available
Custom Composable Extensions may help here too.
Intel work in this space for RISC-V: https://github.com/IntelLabs/riscv-vector
Profiles
#1 Automotive profiles
Already in progress in Automotive SIG and RVM TG.
#2 server virtualization
Covered by Server SOC work, RVA23, Server Platform
#3 Edge AI
Merge into other AI topics from Premier.
May need to to increase focus of AI beyond accelerators to Edge solutions and AIIoT. These solutions are mainly other architectures.
Platforms
#1 - covered
#2 → Server Platform spec
#3 Need more information
Is this what’s referenced, “Approximate Computing: Concepts, Architectures,
Challenges, Applications, and Future Directions“? https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10693435
#4 and #5 are underway or available
Software
#1, #2 are downstream issues that are software ecosystem basis
Engagement in existing groups would help close images.
May to investigate where/which container images are built in the community v. enterprise Distro images
“Base images” may be key part
#3 of ISVs being slow
Not a technology/standards issue
Engagement by members will be required to solve
#4 Debuggers
May be a requirement for VTune. Unclear.
Enabling base tools. Becomes an ISV issue (like #3) when software becomes proprietary.
Ecosystem will continue to grow with hardware availability and solution growth
#5 Thermal and Power Management
RISC-V is standardizing around interfaces in BRS and ACPI
SBI implementing support
Awaiting HW support
#6 Mixed precision computing compiler support
Clarity required because ISA doesn’t support mixed precision
#7 OS support upstream
Need community education of partners to drive into community
#8 Boot standards
BRS for Servers but not really defined for embedded, but bespoke solution available in BRS
#9 AI
Batch with AI topics
#10 Vector optimizations
Merge with ISV item
Synergy with RISE
#11 Common Primitives for Matrix
Being driven by Toolchain SIG
Community Organization and Individual Gaps
ISA
#1 Neuromorphic extensions: has been raised before, but work never proceeded. If multiple companies who are interested, they can work with AI/ML SIG.
#2 Next gen crypto: PQC TG underway
#3 Power efficiency: RPMI interfaces recently ratified. May simply be micro-architecture but need more details.
#4 Standardized Crypto extension: covered by Crypto SIG
#5 Avoid fragmentation of extensions: role of ARC
#6 Vector and DSP: We don’t recognize Kinchit so need to understand this.
#7 Built in encryption commands: need specifics.
#8 Enhanced Bit-manip: Scalar Efficiency TG looking at this
#9 HW security integration: need specifics.
#10 Security/Crypto support: need specifics. HAC, PQC are occuring.
#11 Reconfig GPU design: unclear what is meant.
#12 Chains of parallel ops: AME, IME, VME being worked.
#13 General ISA security support: Too vague
#14 Vector, Matrix, Virtualization extensions: Appears to be covered.
Non-ISA
#1, #2: unclear
#3 Standard accelerator interfaces: CX TG
#4 Trusted Execution Environment: Supervisor domains, RISC-V Worlds
#5 Power: Again RPMI
#6: FPGA hardware gaps: unclear, maybe covered by CX TG
#7: Looking for SBI for micro-controllers.
Profiles
#1 Automotive: Automotive SIG and RVM Profile cover
#2 Compliance: CSC working
#3: Vague. We do mentorships, education, local events.
#4: Sounds like TEE.
#5, #6: need to understand who Kinchit.
#7: RVM.
#8: Profiles process documentation should address. Further speculation is difficult.
Platforms
#1: has to come from Members
Action items:
Adjournment
Motion to Adjourn Made By:
Seconded By:
Time Adjourned: