Stream: t-compiler/wg-self-profile

Topic: meeting 2019-03-05


Wesley Wiser (Mar 01 2019 at 10:45, on Zulip):

We're holding our first team meeting here, in this stream on March 5th, 2019 at 7AM EST/13:00 CEST. We're anticipating this meeting to be ~30 minutes.

Some topics for discussion:
- Progress update on the current state of the profiler
- Planned next work items
- How/where to track work?
- Path to possible minimum viable product the rest of the compiler team can use?
- Others?

mw (Mar 04 2019 at 16:58, on Zulip):

I tried to expand this list into an agenda for the meeting

mw (Mar 04 2019 at 16:58, on Zulip):

AGENDA

Current Status

Where to track work?

What does a minimum viable product look like?

Next work items

mw (Mar 04 2019 at 17:03, on Zulip):

things marked with TODO are points that I'd like to discuss during the meeting

mw (Mar 04 2019 at 17:05, on Zulip):

I propose that we go over the items in order and see how far we get in the 30 minutes

Wesley Wiser (Mar 04 2019 at 17:19, on Zulip):

Sounds great to me!

simulacrum (Mar 04 2019 at 17:38, on Zulip):

FWIW I should be able to make it to the meeting if I'm looking at timezones correctly, so I can help give guidance from the perf.r-l.o side of things

Wesley Wiser (Mar 04 2019 at 17:44, on Zulip):

@simulacrum That would be great! The Google Calendar time should be adjusted for your local time I think. If not https://www.timeanddate.com/worldclock/converter.html?iso=20190305T120000&p1=263&p2=37

simulacrum (Mar 04 2019 at 17:44, on Zulip):

yep the calendar was correct -- yeah, I can definitely make it. It... will be 5 am, so not sure how functional I'll be, but I'll be there :)

mw (Mar 04 2019 at 21:56, on Zulip):

meanwhile xkcd has decided on a data format for us: https://xkcd.com/2116/

mw (Mar 05 2019 at 12:00, on Zulip):

:wave:

Wesley Wiser (Mar 05 2019 at 12:01, on Zulip):

:wave:

mw (Mar 05 2019 at 12:01, on Zulip):

alright, everone present it seems

mw (Mar 05 2019 at 12:02, on Zulip):

I hope you're all somewhat awake :)

Wesley Wiser (Mar 05 2019 at 12:02, on Zulip):

Making some coffee now :coffee:

mw (Mar 05 2019 at 12:02, on Zulip):

excellent idea

mw (Mar 05 2019 at 12:02, on Zulip):

shall we go through the proposed agenda?

mw (Mar 05 2019 at 12:03, on Zulip):

or is there anything missing that should be added now?

Wesley Wiser (Mar 05 2019 at 12:03, on Zulip):

Let's go through the agenda

simulacrum (Mar 05 2019 at 12:03, on Zulip):

Agenda sgtm -- do we want to extract it to a paper doc or something?

mw (Mar 05 2019 at 12:03, on Zulip):

yes, I'll do a summary after the meeting

mw (Mar 05 2019 at 12:04, on Zulip):

and add it to the wg repo

Wesley Wiser (Mar 05 2019 at 12:04, on Zulip):

So first up is the current status

mw (Mar 05 2019 at 12:04, on Zulip):

@Wesley Wiser do you agree with the current status ?

Wesley Wiser (Mar 05 2019 at 12:04, on Zulip):

Yes. I believe that it's correct

mw (Mar 05 2019 at 12:05, on Zulip):

@simulacrum any questions regarding that?

Wesley Wiser (Mar 05 2019 at 12:05, on Zulip):

The only additional note would be that I'm in progress replacing the codegen TimeLine stuff

simulacrum (Mar 05 2019 at 12:05, on Zulip):

Do we have some sort of versioning built-in to those files?

Wesley Wiser (Mar 05 2019 at 12:05, on Zulip):

https://github.com/rust-lang/rust/pull/58488

Wesley Wiser (Mar 05 2019 at 12:05, on Zulip):

No, not currently. It's pretty adhoc

mw (Mar 05 2019 at 12:05, on Zulip):

@simulacrum that's a later point of the agenda

mw (Mar 05 2019 at 12:06, on Zulip):

i.e. coming up with a way of versioning/stabilizing something

simulacrum (Mar 05 2019 at 12:06, on Zulip):

:thumbs_up:

simulacrum (Mar 05 2019 at 12:06, on Zulip):

I think I understand the current state then

mw (Mar 05 2019 at 12:07, on Zulip):

OK, I've put a marker to your comment @Wesley Wiser

mw (Mar 05 2019 at 12:07, on Zulip):

so I'll add that to the summary

mw (Mar 05 2019 at 12:07, on Zulip):

next up, where to track work

mw (Mar 05 2019 at 12:08, on Zulip):

the WG repo/template gives us some structure here

mw (Mar 05 2019 at 12:08, on Zulip):

but I think it's missing something like a "current status" section somewhere

simulacrum (Mar 05 2019 at 12:08, on Zulip):

(is there a link to said repo?)

Wesley Wiser (Mar 05 2019 at 12:08, on Zulip):

I think we're probably allowed to take some liberty with the WG template format

davidtwco (Mar 05 2019 at 12:09, on Zulip):

(do feel free to add one to your page if you feel it would help you)

mw (Mar 05 2019 at 12:09, on Zulip):

https://github.com/rust-lang/compiler-team

mw (Mar 05 2019 at 12:09, on Zulip):

yes, I propose that we have a tracking issue for the next big milestone on GH

mw (Mar 05 2019 at 12:09, on Zulip):

and a "current status" section in the WG readme that links to it

Wesley Wiser (Mar 05 2019 at 12:10, on Zulip):

Advertising what's currently being worked on and what could use some help is probably a good way to encourage people to get involved

mw (Mar 05 2019 at 12:10, on Zulip):

yes

Wesley Wiser (Mar 05 2019 at 12:10, on Zulip):

yes, I propose that we have a tracking issue for the next big milestone on GH

Sounds good to me!

simulacrum (Mar 05 2019 at 12:10, on Zulip):

rust-lang/rust GH?

mw (Mar 05 2019 at 12:10, on Zulip):

yes

simulacrum (Mar 05 2019 at 12:11, on Zulip):

we could also create an actual milestone if that'd be helpful but tracking issue is definitely good start

mw (Mar 05 2019 at 12:11, on Zulip):

yeah, idk, there's no big tradition for using GH projects and milestones so far

simulacrum (Mar 05 2019 at 12:12, on Zulip):

let's go with just a tracking issue for now

mw (Mar 05 2019 at 12:12, on Zulip):

does not mean we can't start it :)

Wesley Wiser (Mar 05 2019 at 12:12, on Zulip):

I don't have any experience using GH projects/milestones but I'm not opposed :)

mw (Mar 05 2019 at 12:13, on Zulip):

I'd vote for tracking issue because that's a known quantity

mw (Mar 05 2019 at 12:13, on Zulip):

alright

Wesley Wiser (Mar 05 2019 at 12:13, on Zulip):

Should we just do one tracking issue for the next milestone (whatever that is)? Or one for the overall roadmap?

simulacrum (Mar 05 2019 at 12:14, on Zulip):

I think probably just the one issue and we can update as we go (we might have sub-issues, though)

mw (Mar 05 2019 at 12:14, on Zulip):

I think the overall roadmap is not quite clear enough

Wesley Wiser (Mar 05 2019 at 12:14, on Zulip):

That's how I also feel

simulacrum (Mar 05 2019 at 12:15, on Zulip):

Yep, strong agree -- so not laying it out at all yet seems simpler :)

mw (Mar 05 2019 at 12:15, on Zulip):

OK, I think that's it for this topic

mw (Mar 05 2019 at 12:16, on Zulip):

what the milestone is is the next topic :)

Wesley Wiser (Mar 05 2019 at 12:16, on Zulip):

Let me just copy what @mw said so I don't have to keep scrolling back up:

What does a minimum viable product look like?

Some thoughts by mw:

mw (Mar 05 2019 at 12:17, on Zulip):

so, I think the perf.rlo version would be most useful

mw (Mar 05 2019 at 12:17, on Zulip):

because that would make it really accessible to everyone

Wesley Wiser (Mar 05 2019 at 12:17, on Zulip):

Of the table based data?

mw (Mar 05 2019 at 12:17, on Zulip):

yes

mw (Mar 05 2019 at 12:18, on Zulip):

i.e. per query-kind numbers

simulacrum (Mar 05 2019 at 12:18, on Zulip):

I think so long as the format is pretty stable and/or versioned we can definitely do it

mw (Mar 05 2019 at 12:18, on Zulip):

so the goal would be that if there is a regression, one can open a view on perf.rlo and see which query takes longer

simulacrum (Mar 05 2019 at 12:19, on Zulip):

Ah, and also we need the overhead to be about equivalent to perf

mw (Mar 05 2019 at 12:19, on Zulip):

yes

Wesley Wiser (Mar 05 2019 at 12:19, on Zulip):

I'd really like to see the compiler export format stabilized. Once that happens, we can build a lot of tools on top of that

mw (Mar 05 2019 at 12:19, on Zulip):

so, I think the format that perf.rlo uses does not have to be the same that the compiler exports

Wesley Wiser (Mar 05 2019 at 12:19, on Zulip):

Oh for sure

Wesley Wiser (Mar 05 2019 at 12:20, on Zulip):

I mean the (probably) binary event format

mw (Mar 05 2019 at 12:20, on Zulip):

so that should make it simple to have the format for perf.rlo stable

mw (Mar 05 2019 at 12:20, on Zulip):

if it is generic enough

mw (Mar 05 2019 at 12:21, on Zulip):

ok, let's try to come up with structure for this discussion :)

mw (Mar 05 2019 at 12:21, on Zulip):

first, let's decide what the end goal is

Wesley Wiser (Mar 05 2019 at 12:21, on Zulip):

Do we just want to see that table on perf.rlo? Or do we actually want something more structured so that, for example, performance diffs could be generated at the query level?

mw (Mar 05 2019 at 12:21, on Zulip):

I'd like there to be some diffing

mw (Mar 05 2019 at 12:22, on Zulip):

i.e. perf.rlo should display time spent in each query per commit

mw (Mar 05 2019 at 12:22, on Zulip):

and then show which one is slower

mw (Mar 05 2019 at 12:22, on Zulip):

something like the current comparison view

simulacrum (Mar 05 2019 at 12:22, on Zulip):

yeah, I'm envisioning something like https://perf.rust-lang.org/nll-dashboard.html except per-crate probably

Wesley Wiser (Mar 05 2019 at 12:23, on Zulip):

Gotcha

simulacrum (Mar 05 2019 at 12:23, on Zulip):

(where instead of crates we have queries on the left hand side)

mw (Mar 05 2019 at 12:23, on Zulip):

yes, that's how I imagine it to, @simulacrum

mw (Mar 05 2019 at 12:23, on Zulip):

only I want stuff to be green! :)

mw (Mar 05 2019 at 12:23, on Zulip):

or red

Wesley Wiser (Mar 05 2019 at 12:23, on Zulip):

No just green :)

mw (Mar 05 2019 at 12:24, on Zulip):

but yes, I imagine that from the comparison view one clicks on a specific benchmark to "zoom in"

Wesley Wiser (Mar 05 2019 at 12:24, on Zulip):

So I guess we definitely need some structure for that

Wesley Wiser (Mar 05 2019 at 12:24, on Zulip):

Do we just want total times for each query? Or also some kind of mean, max, etc?

simulacrum (Mar 05 2019 at 12:24, on Zulip):

yep that's how I've wanted to build it for a while

mw (Mar 05 2019 at 12:24, on Zulip):

yes, we have multiple values we collect per query

mw (Mar 05 2019 at 12:25, on Zulip):
mw (Mar 05 2019 at 12:25, on Zulip):
mw (Mar 05 2019 at 12:25, on Zulip):
mw (Mar 05 2019 at 12:25, on Zulip):
mw (Mar 05 2019 at 12:25, on Zulip):

and so on

mw (Mar 05 2019 at 12:25, on Zulip):

adding new queries should be easy to handle

simulacrum (Mar 05 2019 at 12:26, on Zulip):

Ah, do we want to do the profile maps? Like, callgraph stuff

simulacrum (Mar 05 2019 at 12:26, on Zulip):

remembered actual name: flamegraphs

Wesley Wiser (Mar 05 2019 at 12:26, on Zulip):

With the current setup, adding new queries should happen automatically

mw (Mar 05 2019 at 12:26, on Zulip):

not yet, I think

simulacrum (Mar 05 2019 at 12:26, on Zulip):

I think that'll be beyond MVP, yeah

mw (Mar 05 2019 at 12:26, on Zulip):

yeah

Wesley Wiser (Mar 05 2019 at 12:26, on Zulip):

There's also the detailed perf.html stuff

mw (Mar 05 2019 at 12:27, on Zulip):

so, I think, for this view we can stabilize a format, right?

mw (Mar 05 2019 at 12:27, on Zulip):

i.e. there's a table of data

mw (Mar 05 2019 at 12:27, on Zulip):

with one line per query

mw (Mar 05 2019 at 12:27, on Zulip):

and a number of columns

mw (Mar 05 2019 at 12:27, on Zulip):

i.e. absolute time spent etc

simulacrum (Mar 05 2019 at 12:27, on Zulip):

we could do something like CSV, but I'd be fine parsing binary data too...

Wesley Wiser (Mar 05 2019 at 12:27, on Zulip):

What would be most helpful for perf.rlo @simulacrum?

mw (Mar 05 2019 at 12:28, on Zulip):

and then one can compare which column to compare

mw (Mar 05 2019 at 12:28, on Zulip):

yes, we can generate whatever is most convenient

mw (Mar 05 2019 at 12:28, on Zulip):

but do we agree on the general form of that data

simulacrum (Mar 05 2019 at 12:28, on Zulip):

yeah, I think so

simulacrum (Mar 05 2019 at 12:28, on Zulip):

though

mw (Mar 05 2019 at 12:28, on Zulip):

new lines could be added any time (i.e. new kinds of queries)

simulacrum (Mar 05 2019 at 12:28, on Zulip):

one line per query -- are we aggregating data then?

mw (Mar 05 2019 at 12:29, on Zulip):

and new columns too (i.e. we collect something new)

mw (Mar 05 2019 at 12:29, on Zulip):

for this view yes

mw (Mar 05 2019 at 12:29, on Zulip):

but that will be done by a postprocessing tool

simulacrum (Mar 05 2019 at 12:29, on Zulip):

sure, okay

mw (Mar 05 2019 at 12:30, on Zulip):

oh my, looks like we are out of time!

simulacrum (Mar 05 2019 at 12:30, on Zulip):

(FWIW I can stick around, not sure about you two)

mw (Mar 05 2019 at 12:30, on Zulip):

I can stick around too

Wesley Wiser (Mar 05 2019 at 12:31, on Zulip):

So, for the raw format, something like:

query name, total time, total invocations, percent of total time, cache hit percent
typeck_tables_of, 120.32, 7361, 60, 10
Wesley Wiser (Mar 05 2019 at 12:31, on Zulip):

(I can as well)

mw (Mar 05 2019 at 12:31, on Zulip):

ok, let's talk some more then

simulacrum (Mar 05 2019 at 12:31, on Zulip):

yep, that looks good to me

mw (Mar 05 2019 at 12:31, on Zulip):

alright, so that's one goal for the MVP: generate this kind of table

mw (Mar 05 2019 at 12:32, on Zulip):

a second goal: create a perf.rlo view that displays this data

mw (Mar 05 2019 at 12:32, on Zulip):

and make perf.rlo generate links to that view from the comparison page

mw (Mar 05 2019 at 12:33, on Zulip):

perf.rlo would also have to run the postprocessing tool on the data the compiler generates

Wesley Wiser (Mar 05 2019 at 12:33, on Zulip):

We probably also need to make the event dump from the compiler much faster if we're going to run the self-profiler during perf runs

mw (Mar 05 2019 at 12:33, on Zulip):

yes, let's talk about that

mw (Mar 05 2019 at 12:33, on Zulip):

so, right now there's a lot of overhead

mw (Mar 05 2019 at 12:33, on Zulip):

but I think most of it is unnecessary

mw (Mar 05 2019 at 12:34, on Zulip):

but it's not clear how low we can make the overhead

mw (Mar 05 2019 at 12:34, on Zulip):

i.e. if it is viable to pass -Zself-profile to all perf.rlo runs

simulacrum (Mar 05 2019 at 12:34, on Zulip):

so perf collects events and then writes them as binary data into the perf.data file; that overhead seems fairly low

mw (Mar 05 2019 at 12:35, on Zulip):

@simulacrum yes, but I think we are collect a lot more events

Wesley Wiser (Mar 05 2019 at 12:35, on Zulip):

I think going to a binary format is going to help a lot

mw (Mar 05 2019 at 12:35, on Zulip):

yes

mw (Mar 05 2019 at 12:35, on Zulip):

but one thing that might come up:

Wesley Wiser (Mar 05 2019 at 12:35, on Zulip):

mmap'd file might as well. I tried playing with that on my computer but haven't had much success yet

mw (Mar 05 2019 at 12:35, on Zulip):

the overhead might always be too high for regular runs

mw (Mar 05 2019 at 12:36, on Zulip):

@Wesley Wiser and I talked about this at the all hands

mw (Mar 05 2019 at 12:36, on Zulip):

and said that maybe we need to do two sets of runs for perf.rlo

mw (Mar 05 2019 at 12:36, on Zulip):

one with self-profiling and one without

simulacrum (Mar 05 2019 at 12:36, on Zulip):

we can definitely enable it for some crates

mw (Mar 05 2019 at 12:36, on Zulip):

I'm not sure we will have to

simulacrum (Mar 05 2019 at 12:36, on Zulip):

that kind of thing has been supported historically for e.g. NLL so it's not too bad to implement

mw (Mar 05 2019 at 12:37, on Zulip):

but I wanted to mention it on the record that this issue exists

simulacrum (Mar 05 2019 at 12:37, on Zulip):

sounds good

Wesley Wiser (Mar 05 2019 at 12:37, on Zulip):

I don't know if this would be any easier but it might be sufficient to have some kind of "nightly" perf run that runs the self-profiler even if it's too expensive to run all the time

mw (Mar 05 2019 at 12:37, on Zulip):

so, I'd say we try to optimize profiling as much as possible and see if that's enough

mw (Mar 05 2019 at 12:37, on Zulip):

if it isn't we can do something about it

mw (Mar 05 2019 at 12:38, on Zulip):

so, I'd like to avoid creating work for @simulacrum before we know it's necessary

Wesley Wiser (Mar 05 2019 at 12:38, on Zulip):

@mw You talked a bit in the compiler meeting about writing a new crate related to the event dump. Do you want to talk a bit about that?

Wesley Wiser (Mar 05 2019 at 12:38, on Zulip):

so, I'd like to avoid creating work for simulacrum before we know it's necessary

Agreed :)

simulacrum (Mar 05 2019 at 12:39, on Zulip):

I will note that if the dump is aggregated then you can somewhat trivially make this be equivalent to ~zero work for me

mw (Mar 05 2019 at 12:39, on Zulip):

let's see where we are in the discussion first ...

Wesley Wiser (Mar 05 2019 at 12:39, on Zulip):

@simulacrum Can you say a bit more about that?

simulacrum (Mar 05 2019 at 12:40, on Zulip):

I think we could discuss that later, perhaps?

mw (Mar 05 2019 at 12:40, on Zulip):

yes

Wesley Wiser (Mar 05 2019 at 12:40, on Zulip):

Ok

mw (Mar 05 2019 at 12:40, on Zulip):

so, for the MVP we have:

mw (Mar 05 2019 at 12:40, on Zulip):
mw (Mar 05 2019 at 12:40, on Zulip):
mw (Mar 05 2019 at 12:40, on Zulip):
mw (Mar 05 2019 at 12:41, on Zulip):

ok, @Wesley Wiser let's talk about that crate :)

mw (Mar 05 2019 at 12:41, on Zulip):

what do you want to discuss?

Wesley Wiser (Mar 05 2019 at 12:42, on Zulip):

Well, I'm not 100% sure what you were thinking about it. It just seemed related to getting the raw data out of the compiler faster :)

Wesley Wiser (Mar 05 2019 at 12:42, on Zulip):

So that might be part of the plan to the MVP

mw (Mar 05 2019 at 12:42, on Zulip):

yes, it is

mw (Mar 05 2019 at 12:42, on Zulip):

so, I'd like us to produce a binary stream of events

mw (Mar 05 2019 at 12:43, on Zulip):

plus some sidetables that allow reconstructing information about the events

mw (Mar 05 2019 at 12:43, on Zulip):

the details are here: https://github.com/rust-lang/rust/issues/58372

mw (Mar 05 2019 at 12:43, on Zulip):

this is slightly complicated, so I want to pull it into its own crate

mw (Mar 05 2019 at 12:43, on Zulip):

so that the postprocessing tool can just use that

mw (Mar 05 2019 at 12:44, on Zulip):

but, we should discuss first if the general implementation strategy makes sense to you

Wesley Wiser (Mar 05 2019 at 12:44, on Zulip):

Ok gotcha. I wasn't sure if that's what you were referring to

mw (Mar 05 2019 at 12:45, on Zulip):

I think it's clear that we want to do as little processing as possible per profiler.record call

Wesley Wiser (Mar 05 2019 at 12:45, on Zulip):

Implementing that seems like it might be the next most important thing to work on in terms of getting to the MVP

mw (Mar 05 2019 at 12:45, on Zulip):

so sticking a 32 bit event id seems like a good idea

mw (Mar 05 2019 at 12:45, on Zulip):

yes

mw (Mar 05 2019 at 12:46, on Zulip):

so, as far as stabilizing the compiler's export format goes: this crate would be the stable interface

mw (Mar 05 2019 at 12:47, on Zulip):

so we'd come up with a struct definition for the "expanded" event

mw (Mar 05 2019 at 12:47, on Zulip):

and the post processing tool(s) would work with that

mw (Mar 05 2019 at 12:47, on Zulip):

how events are stored on disk and generated by the compiler would be abstracted away

mw (Mar 05 2019 at 12:48, on Zulip):

the expanded event would be all strings and integers, I think

mw (Mar 05 2019 at 12:48, on Zulip):

i.e. it would not (just) have the event id. but already the query name and key

Wesley Wiser (Mar 05 2019 at 12:48, on Zulip):

That seems reasonable to me

mw (Mar 05 2019 at 12:49, on Zulip):

so, defining would this interface looks like would be another action item for the MVP

mw (Mar 05 2019 at 12:50, on Zulip):

and then implementing everything :)

mw (Mar 05 2019 at 12:50, on Zulip):

I'd be up for implementing the string table and some of the profiling stuff in the compiler

Wesley Wiser (Mar 05 2019 at 12:51, on Zulip):

I'm happy to work on whatever needs to be done :)

mw (Mar 05 2019 at 12:51, on Zulip):

cool

mw (Mar 05 2019 at 12:51, on Zulip):

I think the concrete items will materialize once we write the MVP tracking issue

mw (Mar 05 2019 at 12:52, on Zulip):

I thought I'd create a draft of that some time this week

mw (Mar 05 2019 at 12:52, on Zulip):

@Wesley Wiser are you looking for work? or are you blocked somewhere?

Wesley Wiser (Mar 05 2019 at 12:53, on Zulip):

Not blocked currently. I have those in progress changes for the LLVM codegen timeline I'm working on. But I can also put that on hold and work on something else

mw (Mar 05 2019 at 12:53, on Zulip):

no that's fine then, that's important stuff

Wesley Wiser (Mar 05 2019 at 12:53, on Zulip):

Are you worried about removing that before we have a way of regenerating that data?

mw (Mar 05 2019 at 12:53, on Zulip):

I just don't want you to be blocked until I get to write the tracking issue

mw (Mar 05 2019 at 12:54, on Zulip):

no, that can be removed any time

Wesley Wiser (Mar 05 2019 at 12:54, on Zulip):

Ok

mw (Mar 05 2019 at 12:54, on Zulip):

cool

Wesley Wiser (Mar 05 2019 at 12:54, on Zulip):

I've also got some changes to that code I wrote at All Hands for the perf.html I'm working on in the background

mw (Mar 05 2019 at 12:54, on Zulip):

ok, great

mw (Mar 05 2019 at 12:55, on Zulip):

alright, I think that's all for this meeting

Wesley Wiser (Mar 05 2019 at 12:55, on Zulip):

Should we call the end of the meeting?

mw (Mar 05 2019 at 12:55, on Zulip):

I think we made some good progress

Wesley Wiser (Mar 05 2019 at 12:55, on Zulip):

(I don't have anything left to discuss)

mw (Mar 05 2019 at 12:55, on Zulip):

I'll write a summary some time later

Wesley Wiser (Mar 05 2019 at 12:56, on Zulip):

If you get busy with other stuff, let me know and I can do that :)

mw (Mar 05 2019 at 12:56, on Zulip):

you can certainly do that if you feel like it :)

Wesley Wiser (Mar 05 2019 at 12:57, on Zulip):

Sure! I don't mind

mw (Mar 05 2019 at 12:57, on Zulip):

excellent, just ping me on the PR to the compiler-team repo

Wesley Wiser (Mar 05 2019 at 12:57, on Zulip):

That will give you more time to write the MVP tracking issue :P

Wesley Wiser (Mar 05 2019 at 12:57, on Zulip):

Will do

mw (Mar 05 2019 at 12:57, on Zulip):

yes

mw (Mar 05 2019 at 12:57, on Zulip):

alright, thanks you two!

Wesley Wiser (Mar 05 2019 at 12:58, on Zulip):

Have a great day!

mw (Mar 05 2019 at 12:58, on Zulip):

you too

simulacrum (Mar 05 2019 at 12:58, on Zulip):

:wave: thanks!

Wesley Wiser (Mar 05 2019 at 14:02, on Zulip):

For future readers, notes from this meeting are available https://github.com/rust-lang/compiler-team/blob/master/working-groups/self-profile/NOTES.md#2019-03-05-meeting

Last update: Nov 15 2019 at 20:05UTC