Closed hashseed closed 7 years ago
Looks like the fifth option, March 30th, 6am CET, is the winner. The Hangouts link is here.
I'm looking forward to this!
Added to Foundation calendar, click here for this individual event.
Note that 6am CET is 9pm US Pacific the previous day (Wednesday).
Thanks for updating the calendar!
Wouldn't that be 10pm Pacific time?
@mscdex
Wouldn't that be 10pm Pacific time?
Till the last Sunday in March, yes. On March 26, summer time begins. :tada:
Some preliminary agenda items that we want to bring up.
If there are any other topics anyone is interested in, please add them.
Not to rehash this again and again, but the time is actually 6AM CEST and not CET, right? For the benefit of people using online convert-to-local-time utilities and whatnot...
Correct. CEST 6am.
Dumping meeting notes here. I may have confused who said what in some places. I was busy talking and typing.
CTC: @trevnorris, @jasnell, @cjihrig, @Trott V8: @bmeurer, @hashseed
The plan forward has been announced here as follow-up to previous meeting. V8 will introduce a GN wrapper to build the V8 build target. This can be used as alternative to the legacy GYP configs. The legacy GYP configs will then be removed from V8 upstream and need to be maintained by Node.js
@trevnorris: @jbergstroem, @indutny, @targos, and the build working group should be involved. @jasnell: What platforms does GN support? @hashseed: Ones supported by Chromium. In addition also PPC and MIPS. @jasnell: How hard is extending GN support? @hashseed: Have only gotten fuzzy answers from GN maintainers so far. May be fairly involved to set up GN configurations for platforms and toolchains. Devil is in the detail. @jasnell: Concern that niche platforms will require special maintenance soon. @jasnell: Any difference between binary produced by GN and GYP? @hashseed: There should be none. @hashseed: Maybe Node.js wants to get on track moving away from GYP altogether? @trevnorris: A lot of work would be involved here.
@hashseed: V8 has moved to a new compiling pipeline in 5.9. Some of the Javascript patterns in Node.js internal are tailored towards Crankshaft, and may regress with Turbofan. We want to address these issues by submitting PRs. This only applies to 5.9 and later. The window between Node.js officially updating to 5.9 and release is rather short, and we want to start working on this sooner. @jasnell: You could fork the Canary branch and hold off PRs to master until update to 5.9. @bmeurer shows example in the EventEmitter implementation. @bmeurer: We want to make sure Turbofan does not cause any regressions. Partly by replacing "Crankshaft-script" used in Node.js.
Some discussions on difficulties this brings to back merging fixes. Performance fixes due to new pipeline will need to be marked to not back merge.
@jasnell: Node 8 will stick with 5.7. Node 9 will get 5.9 or later. @hashseed: We could also simply accept regressions for now and fix them on master in the long run. @jasnell: Not preferable. Regressions should be fixed as early as possible. But waiting for 5.9 is what module implementers will need to do. @hashseed: V8 also maintains a vee-eight-lkgr branch. But unlike Canary it does not track Node master as closely. So it might make more sense to fork off Canary for these fixes. @jasnell: Make sure not to change the API signature during rewrite of internal JS. E.g. function length.
@hashseed: There is a huge list of micro-benchmarks that run very long, unsuitable for CI. Is there anything else? @bmeurer: We track AcmeAir and Typescript compiler. Is there anything else representative of real workloads and not micro-benchmarks? Webpack, UglifyJS? @jasnell: @mscdex, @mhdawson, and the benchmarking working group are the right contacts
A list of things to improve in Node.js and V8 wrt performance:
AFAICT the plan has two parts:
.gyp
files in sync with .gn
filesWho is responsible for (2)? /CC @targos
/CC @seishun
This should likely be opened on the tsc repo 😁
On Nov 3, 2017 2:42 PM, "Refael Ackermann" notifications@github.com wrote:
/CC @seishun https://github.com/seishun
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/nodejs/CTC/issues/76#issuecomment-341807047, or mute the thread https://github.com/notifications/unsubscribe-auth/AAecVywbiZ3JEEXsl_S93pcZnGPb-QdCks5sy2ywgaJpZM4MQFYj .
Agreed a new issue probably makes sense.
I've been told that I should open a new issue for this.