CG-07-06.md

Agenda for the July 6th video call of WebAssembly's Community Group

• Host: Google Hangouts
• Dates: Thursday July 6, 2017
• Times: 9:00am–10:00am Pacific Time
• Location: link sent to registered participants
• Contact:
  • Name: JF Bastien
  • Email: [email protected]

Registration

Sign-up

Logistics

The meeting will be a Google Hangouts call.

Agenda items

1. Opening, welcome and roll call
   1. Opening of the meeting
   2. Introduction of attendees
2. Find volunteers for note taking (chair to volunteer)
3. Adoption of the agenda
4. Proposals and discussions
   1. Follow-up from the May meeting:
      • Non-trapping float-to-int conversions open on the design repository
      • Threads
        • Update Overview.md to match CG meeting decisions:
          • Remove float atomics
          • Use 0xfe as prefix byte
          • Add sign-extending conversion operators
          • Remove sign-extending atomic-operators (encoding proposal 3)
        • Reference ECMAScript Agent/AgentCluster
        • Rationale examples for WebAssembly.Global
        • WebAssembly.Memory clarified, and potentially frozen buffer
        • Sign-extend: fix naming and add i64.extend32_s
      • SIMD
        • Added a separate document for the binary encoding of SIMD.
        • Update per CG meeting consensus:
          • Changed the float-to-int conversions to saturating semantics instead of trapping.
          • Removed some of the i64x2 operations.
          • Removed all shuffles and swizzles except for v8x16.shuffle.
        • Removed the boolean types. The CG meeting did not reach a consensus on the direction to take, so we'll go with the simpler version for now.
   2. Adopt Online consensus
   3. Upcoming Kirkland meeting: discuss potential agenda items
   • Specification
   • Multiple return values
   • Tail call
   • GC
   1. Working Group formation
5. Closure

Schedule constraints

None.

Dates and locations of future meetings

Dates	Location	Host
2017-07-18 to 2017-07-20	Kirkland, WA	Google
2017-11-06 to 2017-11-07	Burlingame, CA	TPAC

Meeting notes

Roll call

• Alan Cezar Araujo da Silva
• Andreas Rossberg
• Arun Purushan
• Ben Smith
• Ben Titzer
• Bobby Powers
• Brad Nelson
• Clemens Hammacher
• Dan Gohman
• David Voyles
• Deepti Gandluri
• Derek Schuff
• Dimitar Bunav
• Domenic Denicola
• Ed Coyne
• Ido Neeman
• JF Bastien
• Jakob Geavelle
• Jakob Olesen
• Jason Yu
• Keith Miller
• Lars hansen
• Luke Wagner
• Mark Miller
• Michael Ferris
• Michael Holman
• Peter Jensen
• Sercan Sulun
• Sergey Rubanov
• Vincent Belliard
• Yan Z
• fanmingfei

Agenda adoption

Schuff sconds meeting adoption.

Follow-up from May meeting

float-to-int conversion

Gohman presenting.

• PR open.
• Not merged yet, waiting for Google review.
• Titzer asks about process. Bastien explains how CG to WG process is intended to work. Process still being clarified, mentions Nelson's PR (discussed later).
• Rossberg mentions the spec draft is already in good shape, he'd be ready to try merging this proposal to the spec. Asks that we do the full process with a fork of the design repo.
• Wagner proposes that we could just do a PR against the design repo.
• Titzer thinks we should do the full process.
• Nelson said the proposed process used to have this, but he removed it.
• Rossberg says having a fork allows getting the tests and specs right, and uploading a draft spec properly on github.io.
• Denicola do like TC39? Don't need design process for bug fix, but you shouldn't edit spec if doing a substantial change.
• Nelson spec is fairly readable, maybe we should consider deprecating design repo.
• Gohman volunteers to try out this process.

Threads

Smith presenting.

• Match CG choices as detailed in agenda above (list not repeated for notes).
• Rossberg found a difference in extend, different from CG discussion.
• Miller asks what it means to extend an i64. Smith: it refers to sign extension, the low 32 bits only.
• Another question about wait operator which takes a float timeout. Could take i64 instead.
• Schuff: some work on the toolchain, in LLVM and binaryen, not fully ready yet.
• Smith: Scheidecker has implemented the older proposal in wavm.
• Bastien: no complete implementation in browsers yet.
• Nelson hopes to have implementation in V8 by July, may be missing 64-bit opcodes but will have prototype.

SIMD

Olesen presenting.

• Match CG choices as detailed in agenda above (list not repeated for notes).
• Not sure about benchmarking.
• Nelson: hoping to do a presentation about general numbers on portable SIMD, representing a subset of interesting applications. Should be a quick update, not complete.

Online consensus

• Bastien explains the proposal.
• Was adopted on GitHub, want to make sure wider group gets a say.
• Titzer: ship it!
• Bastien asks for unanimous consensus to adopt? Yes.

Upcoming meeting

Bastien presenting the current list of topics (see agenda above).

• Bastien asks for more detailed agenda items, including proposal details and straw polls.
• Miller emphasizes that GC really changes the character of WebAssembly.
• Rossberg agrees that all language communities are really interested in WebAssembly. Was surprised by interest in places such as PLDI. For some the lack of GC is a showstopper.
• Titzer thinks we should really call it "managed data", it's not really about "garbage". Brings different conotations, not all positive.
• Rossberg: there are two directions, either as a primitive or in some way provide necessary hooks so you can implement it yourself. Titzer and him have strong opinions about this and think the latter have huge issues, but we need to discuss this in-person.
• Miller: the surprising thing to him was it changes from Von-Neumann to heap allocated opaque things. It's not a flat address space anymore.
• Bastien: technically WebAssembly is currently Harvard.
• Rossberg: yes, this changes things from "assembly" to higher level, similar raising of abstraction level to exceptions.
• Nelson: yes, exceptions aren't on the agenda, we want to try out C++ exceptions more before presenting to in-person group. Work happening in repository.
• Titzer: we're not sure yet how languages would use the managed object proposal. The discussion should be about the context and gathering people who would target it.
• Rossberg: yes we're far away from a final thing yet.
• Limin: GC proposal will attract people who are interested about just GC. Would be good to allow folks to attend just that discussion.
• Nelson: we could do a day for this.
• Bastien: will send a PR for Wednesday all day GC discussion.
• Wagner asks about cadence of meetings? Every two months seems a lot.
• Bastien: was thinking about doing every 3 months for now.
• Wagner: 3 or 4 a year is OK.
• Bastien: still look for a venue in November. Maybe Munich next year.

Working group formation

Nelson presenting.

• There's a draft charter out for review for the W3C. Out for vote with advisory committee. Ending in July 12th. Will ping WebAssembly folks whose company is on the advisory group. Need 17 votes, only have 7 right now. Hope to have an announcement shortly after. Hope by Kirkland.
• Bastien charter proposal on mailing list.
• Nelson: charter is super standard W3C, nothing special.

Meeting adjourned.