Skip to contentSkip to conversation

BPF WG ISA Document: Last Call

DISCUSSION David Vernet 2 participants 2 messages Mar 7, 2024 – Apr 5, 2024 BPF

lore
Share & tools
lore
#0 lore raw

Hello everyone,

We're very excited to meet in less than two weeks at IETF 119!

We've made great progress since IETF 118, and we're thrilled to announce that the Instruction Set Architecture (ISA) document is now in working group last call. Please review the document at [0], and share any remaining feedback that you believe needs to be addressed before we send the document to the ADs. As a reminder, the official working group document is in [0], but any suggestions for updates should be sent as patches to the instruction-set.rst document in the Linux kernel tree [1].

[0]: https://datatracker.ietf.org/doc/draft-ietf-bpf-isa/
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/Documentation/bpf/standardization/instruction-set.rst

Because we're so close to IETF 119, the document will remain in last call for at least two weeks following the conclusion of the conference.

Finally, as a friendly reminder, we're still soliciting agenda items for IETF
119. As mentioned in [2], please email myself and Suresh at [email protected] with any items you'd like to discuss. Note that there are still plenty of remaining milestones in the working group charter [3]. Any topic related to the charter or BPF at large are within scope for discussion.

[2]: https://mailarchive.ietf.org/arch/msg/bpf/6pJCj9ujMA5ffb0yKv6DrIOzXOo/
[3]: https://datatracker.ietf.org/wg/bpf/about/

Regards,
David and Suresh

#1 lore raw
Quoted 1 line · David Vernet

We've made great progress since IETF 118, and we're thrilled to announce

that

Quoted 1 line

the Instruction Set Architecture (ISA) document is now in working group

last

Quoted 1 line

call. Please review the document at [0], and share any remaining feedback

that

Quoted 1 line

you believe needs to be addressed before we send the document to the ADs.

[...]

Quoted 1 line

Because we're so close to IETF 119, the document will remain in last call

for at

Quoted 1 line

least two weeks following the conclusion of the conference.

FYI to all, per IETF 119 minutes the WGLC closes today.

I just did another read-through and found some more editorial nits which I plan to send a patch for, but wanted to get this email out before WGLC ends.

Section 1.1: "_s16_ is a types whose..."
s/types/type/

Section 1.4: "tools as such compilers that..."
s/as such/such as/

Section 2.1: the src_reg field correctly says "the source register number (0-10),
except where otherwise specified", but dst_reg and offset are missing the "except where otherwise specified" phrases on their definitions, since some new instructions (SDIV, SMOD, MOVSX) reuse offset for other purposes, and the thread around future callx would use dst_reg for another purpose so having
similar caveats on all three fields seems warranted.

Section 3.1: most places in the doc use "8-bit", "16-bit", "32-bit", and "64-bit"
as adjectives. A couple places in this section are missing the hyphen so are inconsistent.

Section 3.2: most places in the doc put field names in single quotes like 'imm'
when referencing them in normal text. A few places in this section, and one in
section 3.1, are missing the single quotes so are inconsistent.

Section 4.2: "Where size is one of:"
s/size/'<size>'/
for consistency with Section 4.1.

Section 5.1: "a IANA sub-registry"
s/a/an/

Section 5.1.2: "to supercede the existing conformance group" s/supercede/supersede/

Dave