Skip to content

WGSL 2022 04 12 Minutes

Jump to bottom
Kelsey Gilbert edited this page Apr 13, 2022 · 1 revision

WGSL 2022-04-12 Minutes

🪑 Chair: Kelsey Gilbert

⌨️ Scribes: DN

🗺 Location: meet.google.com

⌚ Time: Tuesday 11am-noon Americas/Los_Angeles

Specification: https://webgpu.dev/wgsl

Meeting Issues: Marked Issues

Open Issues: WGSL Issues

**Todos doc: **WGSN TODOs

Previous: 2022-04-05 - WGSL - Agenda / Minutes (APAC)

Note: These are the minutes taken in real-time. The official minutes can be found on the WebGPU wiki.

If you didn't receive a meet.google.com invitation and plan on participating, please send dneto a Google Apps enabled address and he'll add you.

📋 Attendance

WIP, the list of all the people invited to the meeting. In bold, the people that have been seen in the meeting:

  • Apple
    • Myles C. Maxfield
    • Robin Morisset
    • Daniel Glastonbury
  • Google
    • Alan Baker
    • Antonio Maiorano
    • Ben Clayton
    • Brandon Jones
    • Corentin Wallez
    • Dan Sinclair
    • David Neto
    • Ekaterina Ignasheva
    • Kai Ninomiya
    • James Darpinian
    • James Price
    • Rahul Garg
    • Ryan Harrison
    • Sarah Mashayekhi
    • Jaebaek Seo
  • Intel
    • Hao Li
    • Jiajia Qin
    • Jiawei Shao
    • Narifumi Iwamoto
    • Shaobo Yan
    • Yang Gu
    • Yunchao He
    • Zhaoming Jiang
  • Microsoft
    • Damyan Pepper
    • Greg Roth
    • Michael Dougherty
    • Rafael Cintron
    • Tex Riddell
  • Mozilla
    • Dzmitry Malyshau
    • Jim Blandy
    • Kelsey Gilbert
  • Connecting Matrix
    • Muhammad Abeer
  • Kings Distributed Systems
    • Daniel Desjardins
    • Hamada Gasmallah
    • Wes Garland
  • UC Santa Cruz
    • Tyler Sorensen
    • Reese Levine
  • Dominic Cerisano
  • Eduardo H.P. Souza
  • Joshua Groves
  • Kris & Paul Leathers
  • Lukasz Pasek
  • Matijs Toonen
  • Mehmet Oguz Derin
  • Michael Shannon
  • Pelle Johnsen
  • Timo de Kort
  • Tyler Larson

📢 Announcements

Daylight savings shifts are complete for now!

Office Hour

⏳ Timeboxes

FP16 Extension against spec #2696

  • Offline:
    • David and Alan reviewed and approve
  • Alan: API signed off already. Good to land.
  • Merge

Use concise forms for entry point stage attributes. #2652

  • Alan: Looks good.
  • Merge

WGSL: Add dot4U8Packed and dot4I8Packed as new built-in functions #2738

  • Alan: Agreed post-v1 on this. So we haven’t reviewed. Things to be worked out: do we want to force all implementations to emulate. And whether to saturate. Suggest set this aside for now
  • Reconsider post-mvp
  • MM: Goes into the extensions directory?
  • DN: It’s already in mergeable form, can leave it in PR state.

Make matrix notation more intuitive #2745

  • KG: Changing notation matCxR.
  • DN: Yes, just editorial. And the edit found typos where we had swapped the M and N originally.
  • Merge when editor satisfied typos fixed.

WGSL: Explain parameterized type rules more explicitly. #2682

  • JB: Needs updates due to review. Fine to make those changes.

[wgsl] Minor formatting and language changes. #2726

  • DN: This is editorial.
  • DS: Had one fix to apply from David. Applied it.
  • Merge when conflicts done.

Change integer texture parameters to unsigned #2722

  • KG: textureDimensions yield unsigned.
  • AB: Didn’t go as far as we had proposed at the F2F. This is the half-stage: if the value can’t be negative then change it to unsigned. Should be a reasonable first step.
  • MM: Think we should go farther. Using int’s is idiomatic in other shading languages. And would be unfortunate to require casts.
  • DN: Are we going to add the signed forms again?
  • Yes
  • DN: So let’s not remove them now.
  • AB: Ok, so add the variants. The whole cross product of cases.
  • KG: consensus to have both overloads.
  • KG: Suggest have all-signed forms, and all-unsigned forms. Rather than full cross product. Wait for feedback if people miss it.
  • AB: Ok, will update the PR that way.

⚖️ Discussions

Add creation-time expressions #2668

  • Previously:
    • KG: I promise to review this week.
  • KG**: **Think we should land this. Progress
  • AB to resolve conflicts, then David to land.

WGSL Internationalization Adjustment for Code Points #2734

  • MOD: Changes “character” to “code point” and explicitly lists the code point numbers.
  • MOD: Feedback from Myles: use “ignorable” set. But that conflicts with XID sets and is not guaranteed stable.
  • MOD: Based the change on “pattern whitespace” set. Think it should be mergeable.
  • KG: Is there an official category for that?
  • MOD: Yes, can mention it in the spec. But in this current one enumerated each one. Can mention the
  • KG: I like having the category there as an intuition and indication of justification.
  • MM: May have feedback. Doesn’t mention other bidi characters. Only lists one quarter of the bidi characters.
  • MOD: Similarly, there are characters here that are not in pattern_whitespace. Unicode had some considerations.
  • KG: So decision is do we trust other people’s categories.
  • MM: So if you did have a bidi code point there, would it be part of an id?
  • MOD: no, because not part of XID. Would yield invalid program.
  • MM: So part of the bidi characters are blank space, and others are not (only usable in comments) because Unicode guides us this way.
  • MOD: Right.
  • MM: Ok, agree with KG generally that that spec should reference to character categories where possible. Generally, stay away from listing individual characters.
  • KG: Ok, so with addition of naming the referenced pattern category, we can land this.

WGSL parsing assumes context-aware tokenization #2717

  • Previously:
    • Leave this for a week. Call to action: review your recursive descent parser to see
  • DN: Landed the grammar refactoring separately, as https://github.com/gpuweb/gpuweb/pull/2737
  • DN: Still working on programmatically extracting the lookahaeds from the grammar.
  • MM: Our concerns last week was number of tokens to lookahead to be bounded.
  • JB: Have not reviewed parser structure.
  • Check back next week.

📆 Next Meeting Agenda

  • Next week: Tuesday, April 19, 11am-noon (America/Los_Angeles)
  • #2741 Pointers in uniformity analysis
  • https://github.com/gpuweb/gpuweb/issues/2632 Reserve fewer words.
    • DN: Already removed ‘box’ and words that are types in other languages
    • DN: Close this?
Clone this wiki locally