Guild icon
Teeworlds
IRC / bridge
One-way IRC channel bridge. If you want to be able to send messages to IRC, contact @Dune or @heinrich5991. https://www.teeworlds.com/?page=docs&wiki=rules/irc_rules
Between 2020-06-09 00:00:00Z and 2020-06-10 00:00:00Z
Avatar
[quakenet] chococoala BOT 2020-06-09 00:50:28Z
.
00:50
you could always mute channels you don't want to see
00:52
i also wouldn't say discord has a "monopoly" as it's free and there are other options if you choose. there's not too much barrier to entry for online chat programs
Avatar
Just because discord doesn't technically doesn't have a monopoly doesn't mean they don't have enough users to practically be one
Avatar
Still the other tees stutter a bit when they are moving
10:24
What about some machine learning to predict the input of the other players? :D
Avatar
[quakenet] ZillyHuhn BOT 2020-06-09 10:33:24Z
lol
10:33
we should use blockchain
Avatar
10 years ago it'd be big data
10:38
ZillyHuhn: I don't see any difference. Isn't it supposed to be unnoticeable on a LAN server?
Avatar
You can especially see it when he's hooking his dummy up. Without the prediction there is a noticable delay.
Avatar
ah ,right
Avatar
[quakenet] ZillyHuhn BOT 2020-06-09 10:44:06Z
its not a lan server this was recorded on a 200 ping server
Avatar
Gj @Oblikumquat
Avatar
It is sometimes really annoying that the own tee is not in sync with projectiles and other tees. When the server has high bandwidth I sometimes even play without prediction at all :O makes it a bit easier to avoid nades
Avatar
ctrl shift d can help since you shoot from the non-predicted tee
Avatar
While it's not working perfectly, it should be considered to merge this... but disabled by default and no advanced stuff like rocket jump prediction. That would just require too much shared code and cause problems with modded weapons
Avatar
should comment that on the PR
Avatar
I commented it already, an option to predict specific weapons only. That would be good if you want to play on modded servers, then special weapons are not predicted.
Avatar
blergh dissect everything in a dozen of options
Avatar
The code needed to predict weapons/entities is just bloat that does not belong into the official game. The stuff included in this PR would be okay but anything beyond that should go into custom clients
12:00
Instead of using client side options for this, the server should tell the client to activate/deactivate weapon prediction. The server knows best if it's physics are compatible with the prediction.
Avatar
@redix why do you draw the line at weapons/entities?
12:10
if one found a good way to share the code between client and server, I wouldn't really see issues with it
Avatar
Well I haven't seen a good implementation for that yet. A problem is for example merging the entities created locally with the new ones from the snapshot. The ddnet implementation compared the properties of both. This kinda works but isn't really a nice solution for the official game
12:14
A solution would be a predictable id allocation for entities. But that would require a new major version I guess
Avatar
ddnet implementation is really the best I've seen yet but it's still way too much code duplication imho
Avatar
Yes absolutely
Avatar
Unreal engine deals with this issue by having a "core" for each entity that is the only thing that's shared
12:17
but they have the advantage of having the server and client be the exact same binary to begin with 😛
Avatar
Let's implement the core in lua or wasm and send it to the client :D
Avatar
you could compile the C++ code we have to wasm 😉
Avatar
matricks worked on something similar iirc, I think he went a little overboard trying to write an entire VM and come up with a bytecode though
12:34
The core being sent over the network will never happen in vanilla anyway, might aswell try to look for solutions that don't involve that
Avatar
why not vanilla? 🙂
Avatar
Can you imagine oy letting us link with a wasm vm? Any solution would probably need to be custom made which is a big no no when you are letting remote code execute on your own machine
Avatar
well, the vm would definitely need to be a well-maintained library, yes
Avatar
how big are wasm vms?
Avatar
depends on what you need... interpreter, ahead of time compilation, JIT
13:52
there are ones that are based on clang... probably faster but way bigger
13:53
this one is probably a good compromise: https://github.com/bytecodealliance/wasmtime 8 to 14 MB depending on the platform
13:53
but still big compared to tw itself
Avatar
I did start integrating WAMR into teeworlds but the gamecontroller kept getting in the way so I started cleaning that up instead, then I got bored of it 😛
13:57
Well got bored because of scope creep tbh
13:57
I wanted to ship something actually usable for modders if I cleaned up the gamecontroller
Avatar
data exchange between the host and the VM is still hard =\ once the interface types thing is ready, it will hopefully be easier to design usable APIs
Avatar
it's inconvenient but at least it's a clean interface
Avatar
Does the gamecontext actually need to know whether this is a team game?
Avatar
@redix how did you get the sizes?
Avatar
downloaded the latest release from here: https://github.com/bytecodealliance/wasmtime/releases an embeddable version might be smaller though
Avatar
I get 14MB, 5MB stripped, 2MB stripped+gzipped for a simple wasmtime example application
Avatar
Okay 5MB stripped sounds better. Still bigger than tw itself
Avatar
tw is 10MB afaik
Avatar
the binary i mean
Avatar
I can put it into a dynamically linked library if you want to have the executable size down 😛
Avatar
1.5 - 2 mb depending on platform and compiler
Avatar
the interesting thing is the total size of tw, no?
Avatar
well I would not mind a few mb. but oy might 😄
15:03
but i guess he won't like the idea of shipping a vm in general
Exported 62 message(s)