DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2021-02-15 00:00:00Z and 2021-02-16 00:00:00Z

Hey

00:14

Can someone help me changing small things in the teeworlds source? I already compiled it I just need some help please

hello also from over here ;) ask your question and you might get an answer within a day or so ;)

00:23

good night

Okayy so in which cpp file can I do unlimited ammo?

i found it out myself

02:14

but how can i make the gun autoshoot?

Heheee i got ittttt

Oh mann Wie kann ich die shotgun mehr spread machen?

i need the shotgun to spread more but it is acting weird i dont get it :/

i guess you didn't add new directions for your bullets (edited)

[ddnet/ddnet-data-svg] Pull request opened: #14 Add nanami & mermydon redrawn by mind

04:50

[ddnet/ddnet-data-svg] Pull request closed: #14 Add nanami & mermydon redrawn by mind

04:50

[ddnet-data-svg:main] 2 new commits

f4d7654 Add nanami & mermydon redrawn by mind - Jupeyy c07485b Merge pull request #14 from Jupeyy/pr_naname_and_mermydon_by_mind - Jupeyy

[ddnet/ddnet-data-svg] Pull request opened: #15 Fix paths

04:59

[ddnet/ddnet-data-svg] Pull request closed: #15 Fix paths

04:59

[ddnet-data-svg:main] 2 new commits

714881d Fix paths - Jupeyy c95da31 Merge pull request #15 from Jupeyy/pr_fix_paths2 - Jupeyy

ChillerDragon: can we also get a script to remove paths automatically xd

bruh

Image attachment

intel driver

using nvidia tho lol

05:42

it is probably the site that is broken

man these botters that ask for source code help are increasing exponentially lately

Hey fstd

hi fokkonaut

heinrich told me you have a 512 player server? Could you show me how you did the collision between tees? My server sometimes need up to 100% of the CPU, I did some local tests and think it is related to the collision code, e.g. CCharacterCore::Move()

oh well he's probably referring to http://penenen.de/twparticle.png

Embedded image

09:20

in no way shape or form i run such a server, that screenshot is at least 6-8 years old

oh true, i remember this

09:20

ohhh

and i didn't do anything special collision wise, it performed horribly at 512

09:20

in fact the lesson learned was

exactly

many tees needs ac ollision overhaul

09:21

quadtree maybe

well, i probably dont know how :D

the general idea is to group tees together that are physically close

09:23

that way you only have to check collisions against your group rather than everything

09:23

of course maintaining the groups has some overhead on its own

uhm..

09:24

why is it even that expensive?

what collisions?

yea

well because the vanilla approach checks every tee vs every other tee

09:27

so N tees -> N^2 checks

09:27

that gets costly as N grows

hm

(technically it only takes half the amount of checks as after checking X vs Y you don't have to check Y vs X

09:30

so there's a /2 in there somewhere

09:30

but that becomes insignificant as N grows as well

but for 128 players it should be possible to improve it, right?

IIRC 128 was right at the edge of becoming noticably nasty but still kinda playable (i only tested with dummies though, not with 128 real clients)

09:32

but yes for 128 players the naive approach really gets to its limits

well for my server i can say it usually works, i even had 128 clients connected 2-3 times, but in the past few days im facing problems more and more, where CPU usage goes up to 100%

09:33

i dont really know why it suddenly started to do that so often

the easiest you could do without changing source too much is replacing distance checks with squared distance

isnt distance already checked by squared distance?

oh in 0.7? i dunno ^^

let me check it

fokkonaut: it's a little involved but a profiler would tell you what part is eating the CPU

@Deleted User sqrtf(a.xa.x + a.ya.y)

thats not squared

@Deleted User sqrtf(a.x*a.x + a.y*a.y)

its rooted

09:35

root function is expensive

ah lol

see this is the kind of thing you'd probably want to profile before changing

[quakenet] fstd

fokkonaut: it's a little involved but a profiler would tell you what part is eating the CPU

ChillerDragon tried to do that, but couldnt get it to work

[quakenet] fstd

see this is the kind of thing you'd probably want to profile before changing

i did some local tests with visual studio, and it looked like the distance function in CCharacterCore::Move (inside of the player <-> player loop) is very expensive

you could also compile with Ofast, then it optimizes floating calculations

okay

fstd: Jupstar: do you think you can help?

sqrt(9) == 3 becomes 9 = 3*3 so basically remove sqrt and instead increase the checks by ^ 2

09:40

but this is really only about optmizing the math a bit, fstd's approach of grouped physics will still be faster

09:44

last time i checked with Ofast vs O2 was about 10% performance increase, tho i didnt dig much deeper into what exactly is better probs the static sized loops like MAX_CLIENTS and maybe the floating point arithmetics

:D

09:44

creating a mod gets more complicated every time

I would really appreciate any help

10:20

with this specific lag problem

10:22

@Deleted User i will try the squared way<

[ddnet-web:master] 1 new commit

2a30324 Add donation by Mekrioz.' - def-

@Deleted User sqrtf(a.x*a.x + a.y*a.y)

10:33

What do I have to replace this with? (edited)

We can all suggest optimizations but doubt anyone will know what to do without a profile

Locally with some server dummies I can see that its the distance function

What jupstar meant was to compare distances without sqrt which would work because sqrt is monotone increasing

I understood what he means, but not how to do it

(As long as we don't have negative coordinates, that will crash and burn)

oh wait, thats very important

10:59

I want to try it anyways

I don't think it would help significantly, check perf see how much of the time is spent on the sqrt

11:02

fstd was considering a quadtree which would help with much larger N but it's not trivial to implement.

11:05

I considered a precalculated BSP or spatial hashing (chunking) both are much easier to implement

11:08

I guess most engines now use a Bounding Volume Hierarchy though so that's also an option you can look into

Learath2

(As long as we don't have negative coordinates, that will crash and burn)

Actually this will work just fine, sorry, just woke up

:D

11:11

Could you help me implement any of those ideas

If I had time to code anything I'd finish my own PRs

@fokkonaut You can try the trivial optimization of checking squared distance but I have a feeling gcc/clang already figures this out on it's own

hmm

I doubt that will do much though. I'd go straight to spatial hashing. You basically split the map into chunks and a hash function tells you which chunk you are in. When checking for collision you only check against other tees in your chunk

makes sense, but i doubt i can implement such thing

fokkonaut

@Deleted User sqrtf(a.x*a.x + a.y*a.y)

make a 2nd function called distance_squared

11:29

and change all affected calculations to the squared versions

i still dont know what i should calculate in there

e.g.

float D = distance(Pos, pCharCore->m_Pos);
                    if(D < 28.0f && D >= 0.0f)
                    {
                        if(a > 0.0f)
                            m_Pos = LastPos;
                        else if(distance(NewPos, pCharCore->m_Pos) > D)
                            m_Pos = NewPos;
                        return;
                    }

gets

float D = distance_squared(Pos, pCharCore->m_Pos);
                    if(D < 28.0f  * 28.0f && D >= 0.0f)
                    {
                        if(a > 0.0f)
                            m_Pos = LastPos;
                        else if(distance_squared(NewPos, pCharCore->m_Pos) > D)
                            m_Pos = NewPos;
                        return;
                    }

i mean inside of that function...

remove sqrt

Well doesn't the name tell you what it is?

but better create a 2nd function

Definitely create a 2nd function, changing it everywhere at once won't be fun

just retuirn a.x*a.x + a.y*a.y

lol

11:32

and that works without problems?

up to a certain point

fokkonaut

makes sense, but i doubt i can implement such thing

Shouldn't be too hard btw, in CCharacterCore::Move instead of going thru all the tees, you only go thru tees in your and neighbouring chunks

how would i define where a chunk is and ends?

Deleted User

up to a certain point

what are the boundaries?

fokkonaut

what are the boundaries?

When the squares get way too large

11:37

At some point floats will become too inaccurate

u shouldnt care too much i think, in theory u could also use doubles, bcs x87 uses 80bit precision internally

Also I think there is a conversion to an integer there so floor(D) + 1 can't be larger than INT_MAX

11:38

So that's another limit

11:38

Ooooh now that I think about it you shouldn't replace that first distance with a distance squared e.g.

11:39

It will square the amount of iterations you do along the line

yea

yeah only inside the MAX_CLIENTS loop

I think you can skip thru tees aswell

11:40

didn't know this

Deleted User

just retuirn a.x*a.x + a.y*a.y

it performs better i think, but not yet perfect of course

is ur server on linux?

11:46

for the release build u could deffs try out Ofast and lto the good thing here is, if they dont work as expected, you can just remove them again without changing any code

ChillerDragon:

11:52

you here for this? :D

11:52

@Deleted User https://github.com/fokkonaut/F-DDrace/commit/c2017bcd2e4016c4171bbd2deb1a2016d9753b6e

i think the reference looks wrong

why?

&v => v

11:54

bcs u take a reference of a tmp object

just as the length function would do

no, it accepts a const reference

i never understoof this xD

just remove the reference

11:55

the compiler is smart enough to not produce slower code

i dont even know what that is

its like a pointer just that it cant be null xd

what is it used for

11:55

like

11:55

idk

e.g. if you have an object of class CPlayer and u take it by reference, you modify the object itself

11:56

not a copy of it

but isnt that the same with a pointer?

11:57

https://github.com/fokkonaut/F-DDrace/commit/1ac880a085ecd7376d3570fd60da52c1444ed72f

yes

11:57

references are basically pointers

11:58

its not directly true, but they are implemented like that

11:58

the rest looks fine to me

okay, thanks

if its broken you'd notice anyway, bcs the collision wont work

it all works fine, as it seems

11:59

I wonder if my mod has more of these "bugs"

12:00

performance issues

12:00

I think I should replace the distance by distance_squared in gameworld too, right?

12:00

all these FindEntities etc

12:00

ClosestCharacter

[ddnet/ddnet-data-svg] Pull request opened: #16 Add mermydon-coala redrawn by mind, and update some others

u can replace it basically everywhere, where the distance is not used directly, but only compared against others

12:04

but it wont improve it everywhere as much

makes sense

12:05

why?

12:05

i guess a single call is not that bad? :d

12:05

:D*

[ddnet/ddnet-data-svg] Pull request closed: #16 Add mermydon-coala redrawn by mind, and update some others

12:05

[ddnet-data-svg:main] 2 new commits

8a74afd Add mermydon-coala redrawn by mind, and update some others - Jupeyy e8e313c Merge pull request #16 from Jupeyy/pr_add_mermydon-coala_by_mind - Jupeyy

well the FPU might not even profit from a signle call, if it waits on another call anyway (edited)

12:06

its really hard to predict what happens inside the CPU

12:06

ask an intel or amd dev xd

xd

12:06

so basically you lol

12:06

what you've done to ddnet is immaculate xD

but i havent done anything FPU related xD

12:07

i also dont know how GPU drivers are implemented in detail, its more guessing

also dont know what that is tbh

floating point unit

Deleted User

i also dont know how GPU drivers are implemented in detail, its more guessing

xD thats why its experimental?

its like a CPU just for floating points

lol

ah no, opengl is well defined

12:08

its experimental, bcs we had quite a few bugs on Mac and older hardware on windows

12:08

we even have an unknown bug in 6+7th gen intel

12:08

tho it also affects older GL versions

12:09

but i'd say GL 3.3 is stable, i didnt really encounter an issue lately

12:13

anyway fokko, if u have the motivation to change every distance check do it. it wont hurt^^

Deleted User

anyway fokko, if u have the motivation to change every distance check do it. it wont hurt^^

:D

something something premature optimization something something

12:27

Profile your code before making small insignificant optimizations a modern CPU has so many tricks up it's sleeve, small changes usually mean nothing

i ran my fng server with 256 slots, 255 dbg dummies which randomly moved and jumped etc. and it worked fine without code modications, but maybe i just used a stronger CPU

I had trouble with cpu usage so I'd guess that's what helps

12:33

Maybe just throw money at the problem @fokkonaut

ChillerDragon said that we have the best CPU already

12:33

i wondered too

12:33

he said anything better costs > 100€

12:34

p/m

heinrich5991

idk why it only happens in your repo. you can do something like this for -Wstringop-truncation

https://github.com/fokkonaut/F-DDrace/runs/1903237972

12:41

doesnt work

12:43

ah

12:43

i think i know how this works

12:43

yea i fucked it up lol

12:48

still doesnt work

[ddnet-web:master] 1 new commit

c514561 GER3 -> GER1 - def-

Image attachment

cant u use smth like push "-Wno-..."

13:03

ah nvm that looks right indeed

13:03

maybe outdated compiler?

not sure, its github actions test

13:04

the first error just occured out of nowhere

13:05

and the fix doesnt work

its also missing a pop isnt it?

13:06

but probs not causing this issue

13:07

but ubuntu latest doest fail?

13:07

then its probs the compiler version, i'd say

13:07

gcc 5.5 is pretty old xd

added the pop later on

13:08

how do i update it?

i think you can just check the gcc version somehow

i am not familiar with all of this xD

13:09

at all

#if defined(__GNUC__) && .... (edited)

13:10

/* Test for GCC > 3.2.0 */
#if __GNUC__ > 3 || \
    (__GNUC__ == 3 && (__GNUC_MINOR__ > 2 || \
                       (__GNUC_MINOR__ == 2 && \
                        __GNUC_PATCHLEVEL__ > 0))

13:10

https://gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html

13:10

so a && __GNUC__ > 7 might do the job

13:10

i dunno when they added that pragma support

well, i do have a fix already in f-ddrace

13:11

for other false positives

maybe it just fails bcs the actual warning is unsupported

no idea, but how can it suddenly pop up?

13:11

like the whole warning why i want to add this ignore thing

yes, but like this: warning was added in gcc 8 warning causes werror in gcc 8 so u only want to remove it for gcc 8

i mean

13:13

Image attachment

13:13

from one day to another

13:13

this error appeared

13:13

Image attachment

yes, maybe ubuntu latest was updated

oh

13:13

true

that warning is considered not very useful anyway, i already read it in the gcc mailing lists, and they were devided, if its useful or not, in the end, not warning might cause more issues

@fokkonaut haha we do not have the best single core cpu on the market but i think the best we can afford at the current hoster

meh

anything i missed i did not read all?

yes

13:15

Can you add a script to remove the paths for svgs? XD

sure

nice thanks

make an issue please

13:16

wait i can make one :D

ok

oh no then i do not get a mail

13:16

can you do it ? :D

ok xD

nice thanks

[ddnet/ddnet-data-svg] Issue opened: #17 Meta/Extension data removal

Add scripts to automatically remove meta data we don't want to have in the repo @ChillerDragon

ofast and lto is just to boost performance right? not to solve the actual issue but more compating the sympthoms?

yes, but it heavily relies on defined behavior

13:22

so better write good code with it xD, there are edge cases you never heard of, before you googled them xD

i wrote "compating" sometimes i wonder how people understand me

i just ignore words i dont understand xd

https://9to5linux.com/linux-kernel-5-11-officially-released-this-is-whats-new

Linux Kernel 5.11 Officially Released, This Is What's New - 9to5Linux

The Linux 5.11 kernel series is now officially available with improved hardware support, new features, and numerous updated and new drivers.

13:49

poggers

[ddnet/ddnet-data-svg] Pull request opened: #18 Add whis redrawn by mind

13:53

[ddnet/ddnet-data-svg] Pull request closed: #18 Add whis redrawn by mind

13:53

[ddnet-data-svg:main] 2 new commits

562c36d Add whis redrawn by mind - Jupeyy 8ce32b6 Merge pull request #18 from Jupeyy/pr_add_whis_by_mind - Jupeyy

[ddnet/ddnet] Pull request opened: #3621 Fix chat refind skins

It should always search all lines(reprod. by changing windows size, which causes text containers to reset, which then causes skins to be not refound, when skins actually changed)

Checklist

[x] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test if it works standalone, system.c especially
[ ] Considered possible null pointers and out of bounds array indexi...

[ddnet:master] 2 new commits

00e22d8 Fix chat refind skins - Jupeyy dafb120 Merge #3621 - bors[bot]

14:31

[ddnet/ddnet] Pull request closed: #3621 Fix chat refind skins

@Learath2 splitting the map into different chunks doesnt make mush sense if all players are in the same chunk, or in chunks close to it

You make the chunks rather small

14:40

There is a version of this scheme that makes the density of the grid variable at a given point (quadtrees), quite a bit harder to implement though

https://www.drdobbs.com/jvm/an-algorithm-for-compressing-space-and-t/184406478

An Algorithm for Compressing Space and Time

Making a slow program fast can lead to both joy and frustration. But sometimes a new approach yields amazing improvements.

15:29

15:34

Image attachment

15:34

u have to scroll all the way down

15:34

to reject all their shit

15:34

but allow all is up there

Image attachment

15:35

i wonder if u can law suit them

15:35

doesnt it have to be clear?

15:36

gdrp should make this a opt in feature, where it is disabled by default, and the user has to explicitly go and activate it by his own will

@Deleted User @fstd: it probably is not even related to that part of the code, i did some tests with real clients, and it seems the DoSnapshot function is just taking so much

17:35

and all the entities in general

17:36

i did some intense debugging with chillerdragon but idk, i cant really find anything, but i also never had anything like this before, sucks really hard

17:38

I dont know why it ran so good for months now, I dont want to remove 128p support again

Can i for example thread some specific stuff?

17:52

for example the snapshotting which will go up to 60% iirc, or some other stuff?

Eh the gun is broken xD

18:05

You cant see if you shoot with grenade or smth (edited)

@Trafalgar Law you made a bad update. I think the version you got is already removed

18:07

Check out if you can download the latest version

So its not the problem from the srv?

18:10

Ah I see

@fokkonaut would it kill you to learn how to profile the code? Speculating about where the issue might be and randomly moving things to threads won’t help

I profiled it

18:11

DoSnapshot is using 50-60% of the CPU on my local server

18:12

CGameWorld::Tick and Snap takes a lot too

18:12

iterating over all entities

18:12

CSnapshotBuilder::Finish is also very expensive due to the tl_swap

18:12

and all the loops in CCharacterCore::Move and ::Tick

Snap on what?

CDragger takes 10%, character just about 2-5

18:14

but creating the snapshots takes a lot

There is a tl_swap in ::Finish?

3

18:16

well not in finish i think

18:16

let me see

18:17

yes

18:17

3 times per index

18:17

in the loop

Also is this profile from a live server? People moving around and chatting might change the profile drastically due to the way collision is handled

i tested it with clients locally because chillerdragon said its not possible to get a profiler on the live server

There is no loop in CSnapshotBuilder::Finish?

i am on 0.7

18:18

my mod

ChillerDragon learn to use perf nobo

his argument was that the one he used didnt give any information

Oh 0.7 now sorts the snapshots?

and valgrind or some others are way to expensive

Learath2

Oh 0.7 now sorts the snapshots?

i have no idea

18:19

maybe? it seems like

18:19

why tho?

Well bubble sort has awful performance characteristics as n grows. Maybe try replacing that with std::sort?

I see, I should have based on 0.6 xD

Learath2

Well bubble sort has awful performance characteristics as n grows. Maybe try replacing that with std::sort?

how would that work?

18:21

std::sort(pSnap->SortedKeys()[0], pSnap->SortedKeys()[NumItems]);?

18:22

ah no

You need a custom swap function to swap the offsets too

why is that even needed

Because of the way split 1 item into 3, (data, size, offset) all go into separate arrays

18:25

So when sorting the data by key, you also need to swap around the size and the offset so they stay correct

Oh, yea of course! i mean why does 0.7 sort them?

Oh, so they can do binary search on the client

18:26

Avoids the costly linear search 0.6 does

ohh

But I don’t like the idea honestly. The client sorting it would distribute the cost. The server pays a huge cost per client all on it’s own like this

And why dont you need to do that for 0.6?

18:28

i mean for the bridge servers

I think I did, didn’t I?

not sure

Hm I actually don’t remember

you sounded surprised

I did but then I do remember discussing this before

18:29

Well let me check

i dont think you sort them, tbh

Hm, okay than idk what the bubble sort on the server is for. Maybe to reduce the cost for the client?

fokkonaut: if nothing about your build changed then i guess something changed on the host (duh)

Well I push changes every day almost

this would be a good time to have been running munin for a few months :)

and Mr. Anderson says nothing has changed on his side

so you can revert to a commit that has okay performance?

18:31

in that case use git bisect

https://github.com/teeworlds/teeworlds/commit/0c4b3ab62bb3ef798c77f8d591f921bffe6b30b0

Sort snapshot items, binary search · teeworlds/teeworlds@0c4b3ab

[quakenet] fstd

so you can revert to a commit that has okay performance?

thats the point, i dont know when it started

This is indeed why they bubble sort, I wonder how our bridge even works at all now

finding out at what point it started is the whole point of using git bisect

Learath2

This is indeed why they bubble sort, I wonder how our bridge even works at all now

x D

?

[quakenet] fstd

finding out at what point it started is the whole point of using git bisect

how does it work?

18:32

i reacted to learaths message fstd

you tell it a good and a bad commit

18:32

and it does a binary search

no idea what any of those means

binary search is rly simple

18:33

just look it up

it makes find teh bug fast

18:33

is what it means

But I would like to understand xD

or let's say the commit that introduced the bug

binary search discards half

well

every iteration

18:34

there's this magical network of information

18:34

i'd explain it to you but only if you genuinely can't figure out out by yourself using google or wikipedia etc

yeah

18:35

literally wikipedia can tell u all this

18:35

u just a lazy bones

so how do i use git bisect

google it

18:36

or run "git bisect"

18:37

git bisect --help

wtf

18:37

it runs git-bisect.hmtl (edited)

on linux it opens a man page

Ah the client when unpacking the delta also uses the snapshotbuilder so it sorts too

even the .html file will do. i think you can download .html viewers somewhere

okay, but git bisect wont help me much i think. since i need about 60-80 players for it to trigger

@fokkonaut you can probably abuse that to distribute the cost of sorting. Just send unsorted snaps

you obviously need some way to tell whether a given version is good or bad

[quakenet] fstd

you obviously need some way to tell whether a given version is good or bad

i dont really have that

18:40

because it just starts to lag very hard, there is no value that pops over :D

Learath2

@fokkonaut you can probably abuse that to distribute the cost of sorting. Just send unsorted snaps

so removing the bubble sort should work?

well but you notice whether it lags very hard or not...?

18:40

so you can tell when it's bad

yea, true

now with the same test load if it doesn't lag very hard, it's good, no?

but its not always happening

okay that sucks

its not just, oh okay 70 players i will start to lag now

18:41

yea

18:41

it sucks so bad

well bring 70 players then every time?

its at 72 right now and doesnt lag

sigh

There is actually a way to handle this. perf is perfectly capable of saving to a ring buffer and you can send a SIGUSR2 to it to get it to dump the buffer to a file

its pure randomness so the only thing i can do is improving everything

So to profile transient issues like this keep perf running and send it a SIGUSR2 when it gets laggy

Learath2: nice idea

ChillerDragon

18:43

u there :/

Learath2: what's perf tho? not gprof?

fokkonaut

so removing the bubble sort should work?

@Learath2 ?

@fokkonaut I think so, try it it should break very obviously if my hypothesis is wrong

okay, thanks

fstd: perf is part of the kernel

18:45

it can do the kind of sampling profiling gprof does

works @Learath2

Again not sure how much it'll help but it's one less bubble sort per player per snap less

18:51

O(nk^2) where k is the number of snapitems which I guess is fair to assume is of the same order as n so an O(n^3) process gone

18:52

(atleast I think, I'm usually not very good with the theory part of CS)

well, just locally already it is a huge improvement

fstd: perf works without instrumenting the code so a little less information but it's still very useful imho, you should add it to your toolkit for profiling stuff

ulan ziyaaa

rider is better than vs but running trying to run 32 bits on it sucks

19:25

you gotta download some of shit (edited)

19:25

sad

@Learath2 why is the CDragger::Snap so expensive?

19:50

because it moves all the time?

Because of their very complex behaviour with respect to teams and solo

19:56

The way we fake entities being separate in different teams is actually quite costly. That’s what I’d suspect to be the culprit

for (int i = 0; i < MAX_CLIENTS; i++)
    {
        if (m_SoloIDs[i] == -1)
            break;

        Server()->SnapFreeID(m_SoloIDs[i]);
        m_SoloIDs[i] = -1;
    }

stuff like this belongs to the Tick function, no?

19:56

that would save us quite a bit already

19:56

on fuller servers

I honestly don’t remember the logic there at all, it’s been that way for a long time

19:57

All I know is that snap function is called once per player and it has a large costly loop that is run once per player. That’s O(n^2) and O(n^2) bad

yea

19:58

snaps every 2nd tick for all active players

I have no idea why for snapping for a single player it needs to go thru alll players. You’ll have to read the code to figure it out

i believe for spectators

19:59

idk

19:59

nah probbaly not xD

that looks like it checks if people are in solo

yeah thats just a small snippet

20:00

well, we have a SnappingClient already, I also dotn understand why we would need a loop inside of the Snap function

20:01

the Snap function basically IS A loop

oh, yeah, that's really strange

But it really is expensive

Well maybe check the commit messages? At the end of the day to optimize something you need to understand it first

the handling of m_SoloIDs is also very, very strange (edited)

My theory is that it’s snapping multiple times because it can be attached to multiple people in different teams

thats what i thought too

So a way to optimize could be to keep track if connections instead of going thru every tee to see if we are connected

20:05

But again, I’m only speculating

fact tho: i need to change it

20:05

the more players are online on my server the more it eats

20:05

locally, it went up to 17%

20:06

just the dragger snap

Sounds like a great place to optimize

i already improved another thing to not loop over all players, works fine

20:06

also the snap sorting was quite expensive

20:07

if i understood the profiler correctly

I did some tests now

20:24

1. You will only see draggers from your team

20:24

2. using /showothers you can see the draggers of another person from the same team who is in solo

20:26

i dont really think its needed to see other draggers than the one that can affect you

20:27

thats what the loop is for, as you suspected, showing multiple draggers

Are there some resources on the Maths behind Tw physics, best i could find is some old Forum post https://www.teeworlds.com/forum/viewtopic.php?id=10603

More tuningdetails ? (Page 1) — Support — Teeworlds Forum

More tuningdetails ? (Page 1) — Support — Teeworlds Forum — Everything Teeworlds!

wow it even creates 1 CDragger for every team

Before I embarrass myself on the rust books issue tracker: This section from the book explaining lifetimes doesn't make sense, right?

``` fn longest<'a>(x: &'a str, y: &'a str) -> &'a str { if x.len() > y.len() { x } else { y } } ``` The function signature now tells Rust that for some lifetime 'a, the function takes two parameters, both of which are string slices that live at least as long as lifetime 'a. The function signature also tells Rust that the string slice returned from the function will live at least as long as lifetime 'a.

Rather it should be:

The function signature now tells Rust that for some lifetime 'a, the function takes two parameters, both of which are string slices that live at least as long as lifetime 'a. The function signature also tells Rust that lifetime 'a will live at least as long as the string slice returned from the function.

(edited)

20:48

Where are you rust crew? :D

Isn’t that an iff?

20:50

The two sound equivalent

no!?

(Disclaimer I’m not much of a rust person, just thinking)

As an argument that it's not equivalent (ik it's super shallow but idk how to talk about lifetimes idk if I understand them good enough): The sentence is "x lives at least as long as y", so x >= y. You can't change x and y here.

20:57

No?

Ah I see, hm

I guess you could say that the lifetime of 'a should really be the same as the lifetime of the string slice returned from the function

20:59

That would make sense in my understanding of lifetimes

So your argument is that it's the lifetime 'a that is constrained by the lifetime of the string slice?

Exactly that

but from what I understand (even more shallow than your knowledge) lifetimes exist separate from the objects that they constrain

@Learath2 omg

This is important context from later in the paragraph:

[...] only that some scope can be substituted for 'a that will satisfy this signature.

21:01

So it will compile only if there is any scope 'a that works

@Learath2 if i dont create the draggers, it doesnt lag that heavily, because one dragger creates MAX_CLIENTS draggers, 1 for each team, so in my case 128 draggers which alle have to be snapped

21:02

it only snaps draggers to the same team tho

Yeah the proper solution is to get rid of this silly idea of entities having multiple personality disorder

but just alone the calls to CDragger::Snap, 128 calls to 128 draggers, so 128*128 takes already 5% alone

GameWorld per team is the future

thats so insane

21:03

sadly, i dont know what a good fix for this would be.

@timakro doesn't that signature basically say that the lifetime of the returned string slice is the same as the lifetime of the inputs?

i think i will rework the dragger for my purposes xD

Yeah you could say it like that and I guess it would be unambiguous what you mean.

21:05

But that wouldn't be precisely right because the lifetime of the inputs can be longer than that of the output, right?

I see your confusion though. I think the original sounds more correct to my understanding of rust lifetimes but the only one that understands the concept well enough to comment on a technical matter like this is probably @heinrich5991

Heck, I don't even know if the lifetime includes when the variable comes into scope or only means the time when it goes out of scope

21:06

From the ascii art stuff they draw in their code it seems like they mean both, where lifetime A is "longer" than lifetime B if A comes into scope before B and goes out of scope after B

21:09

For a moment think of a' as the lifetime which starts right before the function is called and ends right after the function returns. That would work with the original sentence, right?

[ddnet/ddnet] Issue opened: #3622 Rewrite CDragger

I think it would be a great idea to rewrite CDragger class. On my server I have noticed insane laggs, which are partly caused by draggers, thats because placing one single dragger tile from editor, will create MAX_CLIENTS new draggers, so in DDNet 64, for each tee one, and in my mod 128. All of these draggers need to get snapped, which is not good at all, it creates many snap items and requires many calls to CDragger::Snap which will eat some ressources. Also, I think there is still a me...

timakro

Before I embarrass myself on the rust books issue tracker: This section from the book explaining lifetimes doesn't make sense, right?

``` fn longest<'a>(x: &'a str, y: &'a str) -> &'a str { if x.len() > y.len() { x } else { y } } ``` The function signature now tells Rust that for some lifetime 'a, the function takes two parameters, both of which are string slices that live at least as long as lifetime 'a. The function signature also tells Rust that the string slice returned from the function will live at least as long as lifetime 'a.

Rather it should be:

The function signature now tells Rust that for some lifetime 'a, the function takes two parameters, both of which are string slices that live at least as long as lifetime 'a. The function signature also tells Rust that lifetime 'a will live at least as long as the string slice returned from the function.

(edited)

The function signature also tells Rust that the string slice returned from the function will live at least as long as lifetime 'a.

21:54

both have to live as long as 'a

21:54

note that rust only knows stuff at compile time

21:54

there is no runtime

21:54

so it needs this

21:55

also len() on a str is the length in bytes

21:55

use .chars().count() if u want the count in chars

Ryozuki

so it needs this

u dont know if it will return x or y at runtime

21:57

ur best bet is to return a owned str

21:57

aka a String

21:58

altho i find it hard to find this function practical

21:58

or just do it this way and deal with the lifetimes

Ryozuki

both have to live as long as 'a

Both parameters? Yes, both parameters need to live at least as long as 'a. But the returned value? I'm arguing 'a has to live at least as long as the returned value. Not the other way around.

the returned value has to live as long as 'a too

This example is not my code, it's from the book

cuz its a reference still

Why would you want it to live longer than anything? It could be dropped immediately after the function returns, no? (edited)

it has to live as long as the parameter

22:01

cuz what u return is the parameter

22:02

they are references to the same thing

No, the returned reference can live at most as long as the thing it's referencing, not any longer

yea

22:04

it has to live as long as 'a

Or shorter

no, it has to be 'a

22:07

Image attachment

22:07

Image attachment

22:08

the lifetime of both parameters has to be equal or longer than 'a

Right

monkalaugh

22:08

lifetimes reminds me of learning pointers

22:08

in c++

22:08

all confusion

And to be more specific, the lifetime of the things the parameters are pointing to has to be equal or longer than 'a (edited)

but u once u get it its fine i guess

22:09

ye

I get it but the explanation in the book is still wrong xD

22:09

I think

22:10

And the lifetime of the reference that is returned has to be equal or shorter than 'a

The function signature now tells Rust that for some lifetime 'a, the function takes two parameters, both of which are string slices that live at least as long as lifetime 'a. The function signature also tells Rust that the string slice returned from the function will live at least as long as lifetime 'a.

22:10

its not wrong

22:10

check my image again

22:10

u can see the value returned will live as long as lifetime 'a

22:10

check how i marked 'a

22:10

and where it gets dropped

22:13

btw, i think i get what u think, if a was a reference to b in that context, it will only get dropped once all its lifetimes drop

22:13

thats how i would say it

22:14

i dont think the compiler drops x directly as u would think, it drops x in that scope and b in that scope

In the image you sent 'b would be the argument and 'a would be the string slice that's returned from the function. And now I can't go on because we have two different 'as ^^

??

22:15

no

22:15

'a is the argument

22:15

and the returned

Yeah ok whatever, this is just hypothetical hypothetical ...

let b = "a";
{
  let x = "c";
  let a = longest(b, x);
  drop(x) // a is implicitly dropped by this
}
drop(b)

22:17

u could say a is just a alias

22:17

idk

22:17

im a bad teacher

22:17

i guess i couldnt explain it to u

22:18

pepeH

a is dropped after the 4th line because it's never used xD (edited)

no

22:19

a is not dropped directly

22:19

cuz its just another references

22:19

i imagine it as the compiler only dropping the original references

22:19

idk

22:19

im rly bad explaining it

stuff is always dropped implicitly after it's last use

yeah thats not what im trying to say

if reference or not

a in this case can be x or b

22:20

a isnt dropped as is

22:20

x or b will be dropped

22:20

when their lifetime ends

ah ok yeah what i just said was wrong, just noticed they all have the same type

22:22

the brackets are just confusing, they don't change anything, right?

they define a scope

but in this case they don't change anything, right?

no

22:23

xd

22:24

Image attachment

22:24

Image attachment

I think what makes the example dumb is that x and y have static lifetime

yea

22:25

Image attachment

22:25

his works

22:25

cuz it has static lifetime

22:25

but

22:25

Image attachment

22:25

Image attachment

22:26

if u put the print inside it works

22:27

the compiler knows this thanks to lifetimes

22:27

xd

btw this let b = "a"; does something similar to my brain as

Image attachment

yeah sry

22:27

my examples xddd

what r u tryna do btw?

I was just reading the book and thinking this sentence is wrong xD

this book is actually reviewed

22:29

there even is a beta version of the book iirc

22:30

the book is released for stable beta and nightly

22:30

i dont think they would miss something that important

22:30

specially in the lifetimes section

All your examples have only static lifetimes, they can't possibly get to the core of my problem with the second sentence (edited)

my last

22:34

doesnt have

ah sry

22:35

Okay so my problem with that second sentence will also be a problem with only one argument, so I'll make a function with only one argument

btw with 1 argument u can omit the lifetime

22:35

but its there

Yep, I will keep it for now just so we have a name for it

@timakro btw i think 2021 will be the year a new edition of rust is released

What's new?

https://github.com/rust-lang/rfcs/pull/2966

Edition 2021 and beyond by nikomatsakis · Pull Request #2966 · rust...

This RFC proposes revised rules and plans for editions: Announce plans for a Rust 2021 Edition, and for a regular cadence of editions every 3 years thereafter. We will roll out an edition regardl...

22:39

this is the rfc

22:39

well wait

22:40

https://doc.rust-lang.org/nightly/edition-guide/rust-next/index.html

The Next Edition - The Edition Guide

22:40

this section here keeps track of changes

22:41

its there in nightly already

22:41

xd

22:42

https://github.com/rust-lang/rust/pull/79576

Exported 630 message(s)