DDraceNetwork - developer

I am trying to run DDNet client as fast as KoGAI can handle (x10, x100). I am facing problems with it though. How often does game/client/prediction/entitites/character.cpp run?

01:42

I speed up the server and the client here return std::chrono::nanoseconds(1s).count() / 10; under int64_t time_freq() in base/system.cpp (edited)

01:43

Is this correct or am I missing something?

Mr.Gh0s7

Is this correct or am I missing something?

TICK_SPEED

ddnet-maps

[ddnet-maps:master] 1 new commit

bc36a09 M Barren, M Behemoth, M Cerulean, M Exhale, M Flux 2, M Flux, M Lotus, M Polarity, M Apex, M Momentum, M Smash, M Stellar, M Binary, M Grim Reaper, M Crimson Woods, M Slippy Slide Up, M Spoon, M oco, M Orion, M Sketch, M Sunny Side Up, M Disposition, M Dodge, M Rapid, M Zap - ddnet-maps

Mr.Gh0s7

Is this correct or am I missing something?

Have u tried multiplying the time

Matodor

TICK_SPEED

xD doesn't work like that

Jupstar ✪

Have u tried multiplying the time

nah this wasn't the problem. It was using std:chrono::otherstuff to toher places too so I had to change it there accordinly also.

Mr.Gh0s7

nah this wasn't the problem. It was using std:chrono::otherstuff to toher places too so I had to change it there accordinly also.

But dividing sounds wrong tho xd

09:53

1 second will take 10 times longer

Mr.Gh0s7

nah this wasn't the problem. It was using std:chrono::otherstuff to toher places too so I had to change it there accordinly also.

Ah u doing it in time freq

09:55

Soz

ye xD

09:56

in other places was *10

Better change it the other tho

1

09:56

Time freq is not used everywhere

09:56

But besides that . Multiplying it at that one position should be enough actually

Hmm didn't seem to be though idk xD

Mh. I'm pretty sure it is

09:58

Can u show Ur diff

https://pastebin.com/kVa2GvyD

Diff of AI_SPEED_UP - Pastebin.com

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

Mr.Gh0s7

https://pastebin.com/kVa2GvyD

Don't change it in other places. Simply use cl refresh 0 Have u compiled the server with that change too?

10:04

And yeah remove it from freq call.

Yeah I have. Ok will try it out ;)

Dividing can be scary (10*10)/10/10 is 1

10:06

It can quickly happen that u divide your result instead of the freq valley

10:06

Values

Jupstar ✪

Don't change it in other places. Simply use cl refresh 0 Have u compiled the server with that change too?

That seemed to do the trick ty. Idk why the other way also "worked". My metric for working or not is it/s (iterations/second). This time it's 480-490. Without speedup is 60-70 it/s.

10:14

I would expect it to scale lineary though (should be 600-700 it/s) with that line of thought

Sounds a lot. It can probably not keep up

10:16

A better design would be to not rely on a server anyway

10:17

And just run as much ticks as possible

ah lemme check something. I (zogtib) had put a limit to only run once per tick so it's closer now to 500 (as it should be). I wanna see what changes if I remove this limit

10:21

Yeah it runs with 500 now too. I'll put it back up.

Jupstar ✪

A better design would be to not rely on a server anyway

How would that work though?

Make server part of the client source. Increment the time every time u finished the current loop cycle

10:24

Or sync the server time with a network packet

10:24

But don't use system time

10:24

That sounds very unreliable

Jupstar ✪

Make server part of the client source. Increment the time every time u finished the current loop cycle

Hmm sounds a lot of work and it works well rn. If I face problems I'll do so ;).

10:25

Yeah it's unreliable but reliable enough ig. For speed of x100 or smt bigger it's starting to rise as a problem

Ok. Good that we have 1 trillion fps xd

1

But my cpu is already at it's max with x10 and 20 envs

@Mr.Gh0s7 here is the thing. it catches up if the server somehow takes too long and slows it down if it goes too fast. https://github.com/ddnet/ddnet/blob/master/src/engine/server/server.cpp#L2741 while(t > TickStartTime(m_CurrentGameTick + 1)) (edited)

ddnet/server.cpp at master · ddnet/ddnet

DDraceNetwork, a free cooperative platformer game. Contribute to ddnet/ddnet development by creating an account on GitHub.

13:08

if you were to replace it just with while(true) it would not check for map changes or anything else and just go as fast as it could. ofc you couldnt join the server anymore if you were to do that but im not sure. (edited)

He has to keep server and client in sync tho. So the time workaround is legit xd

just remake ddnet physics and maploading in the language you are dealing with xd (edited)

Teero

just remake ddnet physics and maploading in the language you are dealing with xd (edited)

im pretty sure it would be faster and easier since it would just be translating. (edited)

13:16

exept with python....

https://anthonyhobday.com/sideprojects/saferules/

https://www.worldofbs.com/minimize-state/

All Programming Philosophies Are About State | World of BS

15:56

I recently realized that all the various programming philosophies are concerned with state, and can be boiled down into a simple statement about how to work with state.

>

Object-Oriented - Modifying a lot of state at once is hard to get correct; encapsulate subsets of state into separate objects and allow limited manipulation of the encapsulated sub-state via methods.

>

Functional - Modifying state is hard to get correct; keep it at the boundaries and keep logic pure so that it is easier to verify the logic is correct.

>

Declarative - Modifying state is hard to get correct; use a language to describe how state should be manipulated and allow the machine to determine the steps required.

>

Imperative - Modifying state is hard to get correct; minimize the number of conceptual layers between how the computer operates and how the code is written.

>

Monolith - Modifying state distributed among many services is hard to get correct; keep it centralized.

>

Service-Oriented-Architecture - Modifying all of the state in one service is hard to get correct; distribute it among multiple services.

>

MicroServices - Modifying any state in a service is hard to get correct; have many services that are primarily stateless.

15:56

makes sense tbh

15:56

xd

data oriented?

16:23

seems like it mixes different levels

Well a CPU is at it’s core a state machine. Instructions take us from one state to another. At the end of the day programming is technically navigating inside the state space of the cpu, in that sense technically all programming is “concerned with state”

I don’t think mixing different levels is an issue when you think about it like that. At every level/layer of abstraction we constrain the transitions of state allowed, which gives rise to smaller and smaller subsets of the state space of the entire cpu.

Hello, what i'm doing wrong?

char *CAccount::GenerateID()
{
    char aBuf[8];
    FormatUuid(RandomUuid(), aBuf, sizeof(aBuf));
    return aBuf;
}

?

17:46

dbg_msg("GenerateID()", pSelf->m_pAccount->GenerateID()); print nothing

did u verify that the code path is reachable

17:48

also a char is not directly a string

17:48

what u want is to print it as hex values probably

Jupstar ✪

did u verify that the code path is reachable

i think yes

Jupstar ✪

also a char is not directly a string

if i do same thing but inside GenerateID it print right

newlesstee

Hello, what i'm doing wrong?

char *CAccount::GenerateID()
{
    char aBuf[8];
    FormatUuid(RandomUuid(), aBuf, sizeof(aBuf));
    return aBuf;
}

?

u also return a stack variable

17:50

this is wrong, how does your compiler not throw a warning?

Jupstar ✪

u also return a stack variable

what?

char aBuf[8] is allocated on the stack

17:50

as soon as you call return its dead

How do I make it correct?

pass aBuf as parameter

17:51

GenerateID(char(&aBuf)[8])

17:54

note u can still return aBuf then

17:54

its just important that u increase the lifetime of the array

@Jupstar ✪ oh thanks, thats worked

it will probs not work for all cases tho

17:56

ok nvm formatuuid is probs what u want anyway lol

newlesstee

dbg_msg("GenerateID()", pSelf->m_pAccount->GenerateID()); print nothing

account = for ddnet?^^

Jupstar ✪

account = for ddnet?^^

my server

any cool mod?

i'm trying to code an city

ok xd

what do you think?

i hoped it was fng gores or ddnet xd

newlesstee

Hello, what i'm doing wrong?

char *CAccount::GenerateID()
{
    char aBuf[8];
    FormatUuid(RandomUuid(), aBuf, sizeof(aBuf));
    return aBuf;
}

?

ur returning a stack allocated array

Jupstar ✪

i hoped it was fng gores or ddnet xd

xD

Ryozuki

ur returning a stack allocated array

yes

Ryozuki

ur returning a stack allocated array

3 2 1

xd

17:59

in rust u would get a error

xd

Ryozuki

in rust u would get a error

mania of others languages

Jupstar ✪

xd

Hi! I texted u in dm. I need a little help.

i hate dev related dms

18:01

its better to ask in public

18:01

everyone helps

18:01

u dont depend on 1 person

oh okay im sorry

others see the answer

Raks

Hi! I texted u in dm. I need a little help.

i am also not very good with ddrace code (edited)

How can I create projectiles that don't move?

Jupstar ✪

i am also not very good with ddrace code (edited)

xd

but yeah just ask here

I can ask it over here?

newlesstee

How can I create projectiles that don't move?

tune their speed i guess

set the speed to 0 or smth

tuning params

Raks

I can ask it over here?

u can ask anything here

Raks

I can ask it over here?

if its not about bot clients or coding bots

its about voting system

i'm trying with new CProjectile but dont work like expected

Jupstar ✪

its about voting system

It's about ye..

best is u tag heinrich or learath or robyte

18:02

they probs know about this kind of stuff

I try to make almost the same voting system wich one is the ddddnet using.

18:03

Like if i choose Novice vote

18:03

that will only show up my novice maps

18:03

I tryed like this:

check the ddnet repo