Skip to content
This repository has been archived by the owner on Dec 22, 2021. It is now read-only.

When should our next Sync meeting be? #402

Closed
tlively opened this issue Dec 11, 2020 · 15 comments
Closed

When should our next Sync meeting be? #402

tlively opened this issue Dec 11, 2020 · 15 comments

Comments

@tlively
Copy link
Member

tlively commented Dec 11, 2020

Friday, December 18 is a day off for Google

Friday, December 25 is Christmas

Friday, January 1 is New Years

Friday, January 8 is free but a long way away.

I propose that we next meet on January 8, but try to make progress online in the meantime. Specifically, we should aim to have benchmarking data for all of the outstanding proposals by the January 8th meeting. How does that sound to everyone?

cc @lars-t-hansen @arunetm @abrown @Maratyszcza @omnisip @dtig @rrwinterton

@omnisip
Copy link

omnisip commented Dec 11, 2020

Do we have an expected timeline for finishing all of the prototypes? I'm going to be home for the holidays, so if I have some spare cycles, I can probably help benchmark other proposals too.

@Maratyszcza
Copy link
Contributor

I'm fine with meeting on Dec 18

@tlively
Copy link
Member Author

tlively commented Dec 11, 2020

There are a lot of moving parts, so I don't have a good overall picture. @ngzhian, do you have a sense of when V8 prototyping will be done for the outstanding instructions? (Or would you say it's done right now?) On the tools side, I aim to get up to parity with V8 within the next week.

Another alternative meeting time (suggested by Deepti) would be the 22nd during the normal CG meeting slot, assuming the CG meeting is canceled, which looks likely. This is my favorite idea so far.

@dtig
Copy link
Member

dtig commented Dec 11, 2020

I'm fine with meeting on the 18th, but we could also do the 14th or the 22nd - the 22nd pushes pretty close to the holidays, but as the CG meeting that day will most likely be cancelled it's possibly a slot that most people may have empty on their calendars.

@dtig
Copy link
Member

dtig commented Dec 11, 2020

Oops, added a comment before the page updated, but what @tlively said. :)

@arunetm
Copy link
Collaborator

arunetm commented Dec 12, 2020

22nd works for me too :)

@lars-t-hansen
Copy link
Contributor

I'm not working on the 22nd so won't be able to attend then (18th is my last day before the holidays). I don't think that's a good reason not to meet; I'm OK with reading the minutes.

@ngzhian
Copy link
Member

ngzhian commented Dec 14, 2020

do you have a sense of when V8 prototyping will be done for the outstanding instructions?

All these instructions have been prototyped for at least 1 arch on V8:

  • load store lane
  • prefetch
  • saturating q-format mul
  • i64x2.widen_{low,high}
  • extending miul
  • popcnt
  • extending pairwise add
  • i64x2.eq
  • i64x2.bitmask
  • signselect

If I'm missing prototype on some opcodes please cc me on the issue so I get some visibility, thanks!

@Maratyszcza
Copy link
Contributor

@ngzhian #383 are missing

@omnisip
Copy link

omnisip commented Dec 14, 2020

And #395

@penzn
Copy link
Contributor

penzn commented Dec 15, 2020

I am off on 22nd as well, though I might dial in if I am going to be around my desk (cannot promise), I am also OK with reading the minutes later.

I would like to eventually discuss #403, but we would definitely need larger group for that.

@tlively
Copy link
Member Author

tlively commented Dec 15, 2020

Ok, lets save #403 for January, then. How about @Maratyszcza and I try to get as much benchmarking as possible by the 22nd for the instructions prototyped in V8 and we can focus that meeting on evaluating whether those newly benchmarked instructions should be included in the proposal. Would the folks who are going to be out on the 22nd (@lars-t-hansen, @penzn) be ok with that?

@lars-t-hansen
Copy link
Contributor

lars-t-hansen commented Dec 16, 2020

Yes, with the exception of the prefetch instruction, which I am not in favor of at present.

@penzn
Copy link
Contributor

penzn commented Dec 16, 2020

Same here, I think we should discuss prefetch with everybody present.

@tlively
Copy link
Member Author

tlively commented Dec 22, 2020

Here are the notes from this morning's meeting. We had two provisional votes that achieved consensus: one to include the load/store lane instructions (#350) and one to include the i64x2 widening instructions (#290). Unless the folks who weren't in the meeting today have objections, we will merge these instructions after our next meeting.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

8 participants