ISA: do individual instructions still need their own IANA status?
Previously (draft -00) we said that each instruction would have a status of
Permanent, Provisional, or Historical in the IANA registry.
However, we now have conformance groups about to be merged into the ISA doc,
and at IETF 118 we discussed having each conformance group have a status of
Permanent, Provisional, or Historical. That is, it makes sense for the status to
be at the granularity of conformance group since one should implement all
instructions in a conformance group together.
As a result I now believe that each individual instruction no longer needs its own
status since it can be derived from the status of the conformance group(s) it
belongs to. So in the IANA Considerations section, I plan to remove "status"
from the list of fields in the instruction sub-registry and ONLY have "status"
in the list of fields for the conformance group sub-registry).
Let me know if anyone has a good reason to keep it per-instruction.
Dave
Signature · 3 lines
Bpf mailing list [email protected] https://www.ietf.org/mailman/listinfo/bpf
Quoted 15 lines · On Fri, Feb 02, 2024 at 08:36:02PM -0800…
Previously (draft -00) we said that each instruction would have a status of Permanent, Provisional, or Historical in the IANA registry.
However, we now have conformance groups about to be merged into the ISA doc, and at IETF 118 we discussed having each conformance group have a status of Permanent, Provisional, or Historical. That is, it makes sense for the status to be at the granularity of conformance group since one should implement all instructions in a conformance group together.
As a result I now believe that each individual instruction no longer needs its own status since it can be derived from the status of the conformance group(s) it belongs to. So in the IANA Considerations section, I plan to remove "status"
Quoted 5 lines
from the list of fields in the instruction sub-registry and ONLY have "status" in the list of fields for the conformance group sub-registry).
Let me know if anyone has a good reason to keep it per-instruction.
No objection from me. AFAICT this matches what RISC-V does, which again seems reasonable to emulate.
Thanks,
David