mozilla :: #ionmonkey

7 Sep 2017
18:35jandemnbp: I don't see how the embedding story is jit related
18:36jandemi also think that's more of a policy decision
18:36jandemdo we allocate resources for API documentation, releases (including testing, RCs)?
18:37evilpiejandem: lol v8 reversal on for..in hasOwnProperty again: http://benediktmeurer.de/2017/09/07/restoring-for-in-peak-performance/
18:38jandemevilpie: ugh
18:39evilpieI didn't realize that they have this iterator just for receivers, that obviously makes it easier if you don't have to care about properties on the prototype at all
18:41jandemevilpie: oh right
18:57nbpjandem: embedding can be helpful for us, if embedders are capable or reporting bugs.
18:58nbpjandem: If you don't think this is useful for a JIT because IonMonkey is dedicated to SpiderMonkey, wait for wednesday ;)
18:59jandemwell just saying that embedding is more of a "SpiderMonkey" issue than a "JIT" issue :)
19:00* jandem is happy to wait though
19:02nbpjandem: well yes, and no.
19:03nbpjandem: I can understand how hard it could be for anybody to embed IonMonkey.
19:04nbpand we never thought of ever making it possible.
19:06nbpjandem: On the other hand, I do think there is "some" value to get if we were capable of providing a JIT which is embeddable.
19:06nbpjandem: Though, I do not think that is worth the effort in IonMonkey.
19:06jandemok but an embeddable jit is independent of spidernode et al
19:07nbpexactly.
19:10nbpjandem: that's part of the "Think Large!" comment ;)
19:11* nbp jandem Don't worry, this is a sub-part of a big plan, which seems hard to implement in IonMonkey, but fits perfectly in a long term vision which is combined with other ideas that I will present Wednesday ;)
19:12jandemalright :)
19:13nbpjandem: I just wanted to make sure everybody got as much opportunity as I do to present ideas, and not being the only one ;)
19:14jandemok :)
19:19evilpiejandem: in theory for Map/Set objects, which of the aliasing modes would be correct?
19:27nbpevilpie: which operation?
19:33nbpevilpie: sounds like for Map.prototype.set, we might want to have a new alias set named "classData"
19:33evilpienbp: so I am adding an operation for get
19:33evilpieprobably set etc
19:33evilpieok new class sounds good
19:36nbpevilpie: looking deeper in the impl, I guess privateData make sense.
19:37evilpieor maybe just add a new Type MapSetData ?
19:38nbpgetData -> getPrivate -> privateRef : http://searchfox.org/mozilla-central/source/js/src/vm/NativeObject.h#1295
19:39evilpieright
19:41* nbp off
8 Sep 2017
No messages
   
Last message: 13 days and 8 hours ago