v1.0.0 - #92
v1.0.0#92weenachuangkud wants to merge 73 commits into
Conversation
|
Proper review needed |
|
@Naymmmm Are you busy? |
|
@Per2iako Bro |
|
|
Yoooo it more readable LGTM |
Per2iako
left a comment
Per2iako
left a comment
There was a problem hiding this comment.
This can be maintain easier good
|
I just need to test stuffs |
I busy rn i'll report test when am home |
|
But I don't have a pc to use roblox studio, so I really need some help and I can't find anyone that actually understand how to do manual testing FastCast2 from .md instructions |
|
Okay ty, I'll working on SerialSimulation |
your welcome i still use this module now a day |
|
There must be a better way to implement SerialSimulation, because I don't like having per instance connection |
weenachuangkud
added
enhancement
- Add src/Signal.luau lightweight synchronous event dispatcher - Expose Caster events as multi-listener Signals with Connect/Once/Wait/Disconnect - Keep function assignment as backwards-compatible shorthand for :Connect - Batch parallel worker events into a single Output message per frame - Add tests for multi-listener and Disconnect behavior
Resolve ParallelSimulation conflict by keeping upstream's packed-buffer event queue and layering the per-frame event batch on top. Adapt the hotpaths harness to pass Signals to the simulation, and exclude its Lune runner from selene.
…ignals-replacing-bindableevents feat(events): replace BindableEvent caster events with Signal
Dispatcher:Allocate sent 'Init' exactly once, so a client ClientVM that bound its message handlers a frame late never initialized. Every later 'Raycast' message was then dropped and the caster produced no casts. BaseCastParallel.Init now sets a FastCast2Ready attribute and the dispatcher retries 'Init' until every actor reports ready.
Parallel casters sent one Actor message per shot, so a high fire rate spent most of the main thread serializing and crossing the Actor boundary. Shots are now queued and flushed once per frame, distributed round-robin across workers. The behavior is snapshotted when queued so per-shot mutations such as the raycast filter are preserved.
…hed-dispatch perf: batch parallel fire requests into one message per worker per frame
…l-init fix: initialize parallel actors reliably and document the client container
|
Dang, when @Naymmmm come in to assist with this PR everything got a lot better |
|
I've checked TODO.md, this looks ready to merge |
This workflow is triggered manually to run tests using npm after installing Rokit and necessary tools.
3 participants
Added
Fixes
Changes
Cancelled
(Because you can simply edit the FastCast2 code if you want to change the specific RunService event.)
(Might add this back)