Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes.
Set by lizmat on 24 May 2021.
Nicholas good *able6, #moarvm 06:05
timo goo goo 07:42
Geth MoarVM/fallback_resolver: 5d0f92c65b | (Stefan Seifert)++ | 2 files
Add missing :deoptallpoint flags to invokish ops

Since invoked code may do anything, including creating mixins, we have to mark all non-spesh ops that may invoke as deoptallpoints.		 07:48
timo vrurg_: seems to be about orchestration of gc runs 08:43
thread 5 is just coming out of the sleep and is waiting on cond_blocked_can_continue in orchestrate.c:331, thread 4 entered the gc from allocating in the thread pool scheduler somewhere and is waiting on cond_gc_start 08:47
nine And the tight loop is missing a GC_SYNC_POINT? 09:17
timo Thread 5 run 2 : GC thread elected coordinator: starting gc seq 2 Thread 5 run 2 : A blocked thread 6 spotted; work stolen Thread 5 run 2 : Signalled thread 4 to interrupt Thread 5 run 2 : A blocked thread 3 spotted; work stolen Thread 5 run 2 : A blocked thread 1 spotted; work stolen 09:23
without jit it completes
so yeah, that's quite possibly the reason, nine 09:24
jnthn There's already an issue about that, I believe; think it's that the expr jit doesn't emit gc sync points 09:31
github.com/MoarVM/MoarVM/issues/1230 09:32
timo well, that's not good :) :) :) 09:34
nine jnthn: do those :deoptallpoints look correct now? 09:58
timo jnthn: is there something that can be done with the commits from more-pea? last one is already three years old, wow... or perhaps relevant commits have been cherry-picked or something? 12:09
timo maybe we can merge partial-PEA :D 12:24
jnthn timo: My thinking was to cherry-pick parts of it focused on particular improvements I was exploring 12:26
And polish up each of them in turn.
Also a lot can be simplified thanks to the end of attrinnited 12:27
timo ah, yes
jnthn Which was quite an analysis complication
The p6o-setup stuff is probably the logical next step, since that really gets rid of the whole attr auto-viv stuff
That blocked on finding a solution to attrinitted, which has been done already. 12:28
After that I'd probably go for the transitive EA bits
timo was that already on your agenda for some point in the near future? 12:30
Geth MoarVM: Kaiepi++ created pull request #1714:
[WIP] Make some enhancements to the Semaphore REPR		 14:30
vrurg Anybody would mind adding optional $obj parameter to metamodel archetypes methods? First, it would allow $obj.^archetypes in Raku. Second, I would be able to complete nominalizable transparency because for a generic T:D archetypes do not report its genericness. 19:33
The problem with definites is that single metamodel instance serves them all via parameterization. 19:34

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!