Meeting 2020-10-08

[repi]

Previous meeting: #39

Meeting agenda: Discuss project scope overall and go/nogo for next phase(s)

(add links to issues and material for everyone to readup on beforehand)

• Overview of reasons for the project to exist, and "levels"/milestones/roadmap: Roadmap, milestones, and functionality levels #47. Good to read beforehand, it's a lot.
• After go/nogo, reaffirm our requirements before open-sourcing, and anything the community team can start on now to prepare for that. (@arirawr)

[repi]

Meeting notes

Participants

• @repi
• @khyperia
• @VZout
• @Jasper-Bekkers
• @Jake-Shadle
• @arirawr
• @Tobski

Minutes

Project overview discussion

• Went through and discussed the "feature levels" and overall status of the project (Roadmap, milestones, and functionality levels #47)

  • Level 1 = basically done 🎉

• New ways of exposing compute will be very exciting!

  • "Both GLSL & HLSL ecosystems are stifled, Rust would be awesome" - anonymous
  • @Tobski wanted this for a while

• External users would primarily get value of it once we reach Level 3 (it is friendly to use), but could open source earlier also to start engaging and hopefully get more collaborators as well

• Group decision: Let's Go, continue to work and bring up next level with shaders in Ark! (Embark's internal engine)

People & resourcing

• What do we need to continue the work effectively and scale up development and take on Level 2 support (Roadmap, milestones, and functionality levels #47)?

• Need help on "language design" and adaption as well

  • Not just compiler backend which @khyperia is focusing on
  • To do things properly you need both GPU shading design and language/compiler design experience, which is very hard to find. But also very interesting/attractive area!
  • @Tobski can help a bit. This will be tricky

• What we need the most:

  • Dev full time on build system ("Build system" #48) and shader library, parallel work to the compiler backend
  • Additional compiler dev, that can make language design decisions. Offload @khyperia

• Discussed potential collaborators or people we can engage with or contract/hire

• New devs joining Embark soon!

  • @termhn - internship focusing on Rust GPU project! 2016-10-26
  • @XAMPPRocky - open source dev / dev community, 2020-10-12
  • @lipl - open source dev / compiler experience from Gleam, 2020-10-19

• Start looking for additional compiler-experienced engineer, full-time or contract (remote is fine)

  • Gather contacts / people of interest, esp. in Rust & GPU communities, and start reaching out to them @repi
  • Sketch out "Compiler Engineer" job ad would be excellent to have posted when we open source the project (Open source repo & project #55)! @khyperia

Going open source

• Discussion: When to release?

  • @khyperia: Would be comfortable open sourcing soon after basic build system ("Build system" #48) works well enough
  • @repi: As long as repo is in good shape and we have a buildable example (Good project onboarding experience #28) and we are committed to the project (which we decided today), we could open up early and soon!
  • @arirawr: Pretty comfortable with it right now, good docs!
  • We have multiple new people joining soon with dev community experience that can help manage and supporti t also.
  • Already have a first small easy to run example, just 1 day of cleanup. So ahead of RLSL in first usage experience

• TODOs discussed (move/merge to Open source repo & project #55)

  • Sketch out more on build system & what it involves, and how much work it is ("Build system" #48) - @khyperia
  • Fix and clean up docs. @arirawr will go through it after, and test running
  • Communicate clearly our phase and level of support + version (0.1?) in the readme for everyone. Set expectations
  • Write more of details and wants to get to level 2

• Decided that we want to make the GitHub repo and the Discord channel public at the same together, easiest. Then we start working in the open fully.

• Decided we want to release as soon as possible, within the next week or two! 🚀 Tracked in Open source repo & project #55

Other topics

• Briefly discussed faster paths to language bindings, potentially using bindless
  • Would simplify things if we could make the decision
  • Would work well on desktop, but mobile is problematic
  • @Jasper-Bekkers: Almost no support for bindless on Android, ~1.7% of users
  • Let's avoid not require it for now, but re-evaluate later on

[Tobski]

"Both GLSL & HLSL ecosystems are stifled, Rust would be awesome" - anonymous

I think that was me as well 😅 Otherwise minutes LGTM 👍