4 ms·

V8 JavaScript Engine Release v6.5

50 points by petercooper 9y ago

hajile 9y ago
Since the explicit tail call proposal is no longer moving forward, how long until v8 allows proper tail calls as the spec requires?
- jkrems 9y ago
  Last I've heard no JS engine ships or wants to ship PTC. So I'd be surprised if V8 breaks away from the pack there, given that nothing changed.
  - ash_gti 9y ago
    JavaScriptCore from WebKit does PTC, see https://webkit.org/blog/6240/ecmascript-6-proper-tail-calls-in-webkit https://webkit.org/blog/6240/ecmascript-6-proper-tail-calls-... for more info
  - hajile 9y ago
    v8 did and then pulled support. JSC (javascript core) added support in June/July 2016 (a year and a half ago). Their team makes some great points about this. https://github.com/tc39/ecma262/issues/535 https://github.com/tc39/ecma262/issues/535 https://github.com/rwaldron/tc39-notes/blob/master/es7/2016-05/may-24.md#syntactic-tail-calls-bt https://github.com/rwaldron/tc39-notes/blob/master/es7/2016-... For all of Google's talk about their syntactic tail call proposal, they haven't cared enough to implement it -- only to remove their PTC code. FF and MS are in the same position. At this point, I don't really care one way or the other, they need to quit the excuses, quit the petty power struggles, and do something one way or the other. Every year they wait is another year before we can use it. EDIT: I'd also add that the performance impact in the real world seems to be close to zero. https://github.com/tc39/proposal-ptc-syntax/issues/7 https://github.com/tc39/proposal-ptc-syntax/issues/7
thepumpkin1979 9y ago
Off-topic: this page is really hard to read in a mobile browser, it freezes and when you try to zoom out it redirects to another post, I can't figure out what's wrong with it. Edit: Seems to work ok with content blockers disabled.
- aetherspawn 9y ago
  Works fine here, iPhone 6 w/ Safari.
  - thepumpkin1979 9y ago
    Safari iPhone 7 here, it seems to be caused by Firefox Focus content blockers. I disabled it and it works fine now.
- MBCook 9y ago
  I’m having he same problem when I zoom in. This seems like another page that tries to interpret swipes near the edge of the screen as ‘goto next page/last page’. iPhone 8, latest release of iOS.
Klathmon 9y ago
I'm glad V8 (and Chrome) are getting streaming compilation for WASM. That was an annoying pain point I saw when playing with it (fat binaries meant a long download and compile time, almost penalizing you 2 times for every byte). However I noticed that Firefox's recent update with their streaming and tiered WASM compiler is still almost a magnitude faster (using the benchmark at [0] it takes 717ms in Firefox 58, and 4283ms in Chrome 66 on my machine with a good gigabit connection and an i7 4930k) I tend to think of my PC as fairly "high end" (yeah, it's a few years old, but still is a powerhouse in my opinion!) but I'm just not seeing the speeds that they are talking about in the blog post. Is the difference really just in Firefox's 2-tiered compiler!? I noticed that the benchmarks are in seconds (!) of compile time, does this streaming compiler only really shine with much larger WASM modules? Or is that benchmark not a good representation of a general workload (it was made by a Mozilla employee, for what it's worth). [0] https://lukewagner.github.io/test-tanks-compile-time/ https://lukewagner.github.io/test-tanks-compile-time/
- kevingadd 9y ago
  Tanks is a unity game, so it's representative of the (probably) single most popular use case for WebAssembly right now.
danbolt 9y ago
I’m happy to see inlining on the common Array.prototype functions.

50 points by petercooper 9y ago

Since the explicit tail call proposal is no longer moving forward, how long until v8 allows proper tail calls as the spec requires?

jkrems 9y ago
Last I've heard no JS engine ships or wants to ship PTC. So I'd be surprised if V8 breaks away from the pack there, given that nothing changed.
- ash_gti 9y ago
  JavaScriptCore from WebKit does PTC, see https://webkit.org/blog/6240/ecmascript-6-proper-tail-calls-in-webkit https://webkit.org/blog/6240/ecmascript-6-proper-tail-calls-... for more info
- hajile 9y ago
  v8 did and then pulled support. JSC (javascript core) added support in June/July 2016 (a year and a half ago). Their team makes some great points about this. https://github.com/tc39/ecma262/issues/535 https://github.com/tc39/ecma262/issues/535 https://github.com/rwaldron/tc39-notes/blob/master/es7/2016-05/may-24.md#syntactic-tail-calls-bt https://github.com/rwaldron/tc39-notes/blob/master/es7/2016-... For all of Google's talk about their syntactic tail call proposal, they haven't cared enough to implement it -- only to remove their PTC code. FF and MS are in the same position. At this point, I don't really care one way or the other, they need to quit the excuses, quit the petty power struggles, and do something one way or the other. Every year they wait is another year before we can use it. EDIT: I'd also add that the performance impact in the real world seems to be close to zero. https://github.com/tc39/proposal-ptc-syntax/issues/7 https://github.com/tc39/proposal-ptc-syntax/issues/7

thepumpkin1979 9y ago

Off-topic: this page is really hard to read in a mobile browser, it freezes and when you try to zoom out it redirects to another post, I can't figure out what's wrong with it. Edit: Seems to work ok with content blockers disabled.

aetherspawn 9y ago
Works fine here, iPhone 6 w/ Safari.
- thepumpkin1979 9y ago
  Safari iPhone 7 here, it seems to be caused by Firefox Focus content blockers. I disabled it and it works fine now.
MBCook 9y ago
I’m having he same problem when I zoom in. This seems like another page that tries to interpret swipes near the edge of the screen as ‘goto next page/last page’. iPhone 8, latest release of iOS.

Klathmon 9y ago

I'm glad V8 (and Chrome) are getting streaming compilation for WASM. That was an annoying pain point I saw when playing with it (fat binaries meant a long download and compile time, almost penalizing you 2 times for every byte). However I noticed that Firefox's recent update with their streaming and tiered WASM compiler is still almost a magnitude faster (using the benchmark at [0] it takes 717ms in Firefox 58, and 4283ms in Chrome 66 on my machine with a good gigabit connection and an i7 4930k) I tend to think of my PC as fairly "high end" (yeah, it's a few years old, but still is a powerhouse in my opinion!) but I'm just not seeing the speeds that they are talking about in the blog post. Is the difference really just in Firefox's 2-tiered compiler!? I noticed that the benchmarks are in seconds (!) of compile time, does this streaming compiler only really shine with much larger WASM modules? Or is that benchmark not a good representation of a general workload (it was made by a Mozilla employee, for what it's worth). [0] https://lukewagner.github.io/test-tanks-compile-time/ https://lukewagner.github.io/test-tanks-compile-time/

kevingadd 9y ago
Tanks is a unity game, so it's representative of the (probably) single most popular use case for WebAssembly right now.

danbolt 9y ago

I’m happy to see inlining on the common Array.prototype functions.