Sender | Message | Time |
---|---|---|
30 Jan 2023 | ||
In reply to @tengkuizdihar:matrix.orgOn ground or in air? | 04:36:21 | |
In reply to @an_origamian:matrix.orgyeah, probably need to use a custom "deterministic" physics engine like Rapier | 04:36:24 | |
In reply to @an_origamian:matrix.orgon the ground | 04:36:30 | |
What's special about deterministic physics? | 04:36:43 | |
In reply to @an_origamian:matrix.orgI think Godot internally already use that. Rust should be used to get all the data need to be serialised and deserializing data. | 04:37:18 | |
Perfect. | 04:37:41 | |
In reply to @an_origamian:matrix.orgFair fair lol | 04:38:33 | |
Note: I recommend thinking about how you will delag everything before you write the code. What do you want delagged? What don't you want delagged? | 04:38:34 | |
In reply to @an_origamian:matrix.org two different machines can simulate the same exact physics behaviour given the same input now, most physics engine aren't deterministic because there will always be a slight floating point differences between machines. | 04:38:38 | |
I'm pretty sure there shouldn't be any floating point differences at all since most machines conform to the IEEE-754 floating point standard. | 04:39:25 | |
I think Q3 fits that description, at least for the most part. | 04:39:47 | |
In reply to @an_origamian:matrix.orgwell my main concern right now is reconciliation and prediction. seems like knowing a physics engine deterministic or not is the first step to do it in a server authoritative way | 04:39:52 | |
I've never messed with physics engines, so I wouldn't know. | 04:40:14 | |
In reply to @an_origamian:matrix.orgyes, that's also what I thought, but then someone made a demo to showcase that indeed it happens | 04:40:26 | |
:( | 04:40:35 | |
I'm on phone right now, maybe I'll show it later | 04:40:41 | |
04:58:02 | ||
an_origamian: not on the phone anymore, here's the example of how non-deterministic "regular" physics engine is https://github.com/feetuh/netdot-boxes It looks deterministic in the second gif not because the simulation was made deterministic, but because everything is now synced by properties lmao | 12:37:11 | |
Redacted or Malformed Event | 20:59:12 | |
> I think the server's view would just be delayed, i.e. you only start running on the server at frame 5. Don't quote me on it though, it's been a while since I've looked at that. It for sure never rolls back your position (that only happens for weapon delag). The server doesn't do interpolation IIRC, but it does extrapolate a bit if it doesn't get user commands in time (although not for long). | 21:00:02 | |
^^^ That is from the more knowledgeable source. | 21:01:12 | |
Izdihar: You should read Glenn Fiedler's articles about game physics and fixed time steps. He is specifically focused on game physics over a network and he talks about how developers can implement deterministic fixed physics steps into their games. His Fix Your Timestep! article goes back to 2004. | 21:43:48 | |
In reply to @an_origamian:matrix.orgthank you, that's a nice insight | 23:35:40 | |
In reply to @gh0std0g3:matrix.orgwill do | 23:36:08 | |
31 Jan 2023 | ||
Owo | 02:26:39 | |
15:43:12 | ||
Izdihar: This looks useful: https://fabiensanglard.net/quakeSource/quakeSourcePrediction.php | 17:48:10 | |
1 Feb 2023 | ||
15:19:03 | ||
15:36:36 | ||
2 Feb 2023 | ||
10:36:21 |