DDraceNetwork - developer

Kaffeine

[ddnet/ddnet] Pull request reopened: #3501 RFC: Refactor (from the upstream and own custom)

def-

[ddnet/ddnet] Pull request opened: #3503 Differentiate hook number by showing symbol

!screenshot-20210112@102255

Checklist

[x] Tested the change ingame
[x] 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 indexing
[ ] Changed no physics that affect existi...

complete reinstall couldnt solve the spike lags also. its sure now that the problem is in my pc

i found the source of my subtly stuttering image

10:53

i capped cl_refresh_rate to 300 (for my 144 hz monitor)

10:53

i now have it at 500 and my image is smooth

10:53

dont think that this is how it should be though

10:54

https://github.com/ddnet/ddnet/issues/1921 wow i even already had an issue for this were i said that i have it at 500, apparently set it to 300 at some point o.o

10:54

#1921

https://github.com/ddnet/ddnet/issues/1921

screen tearing even with vsync enabled · Issue #1921 · ddnet/ddnet

To have a smooth image I need to turn off vsync and up the cl_refresh_rate value. I have a 144hz monitor, and I keep the cl_refresh_rate at about 500 to fix this. Without any fps-limit I get at lea...

for smooth gameplay make it 0

10:56

better only cap gfx_refresh_rate

yeah, 0 is also the default because these slight stutterings are annoying and hard to track down

ah yeah i meant gfx_refresh_rate

gfx_asyncrenderold or smth like that on?

11:38

maybe ur GPU doesnt like swapping between throttling and fullspeed

gfx_asyncrenderold was on for me, though switching it off didnt seem to do much

that should be on

11:47

we only renamed it because the old gfx_asyncrender defaulted to 0, but setting it to 1 is better

yeah true

12:02

well if you want to find the issue, the best idea is probs to monitor clock speeds temperatures etc

Can we document all the different ways of throttling and it's effects?

13:12

People ask me about this stuff all the time and I keep telling them I have no idea

13:12

A framerate cap that's well above the monitor refresh rate should never ever cause stutters

13:13

I think we have like 4 different settings that essentially throttle the game in some way

I am interested in getting the game to run more smoothly on old underpowered hardware. Any specific settings that make a notable difference?

a better pc

What is the problem you see

13:20

@YupHio check your graphic settings

@Learath2 depends heavily on what u call stutter, but stutter can also be caused by hardware even if u only have 10 fps

13:22

@YupHio what hardware

13:22

if it supports GL 3.3 already u can enable that

13:22

turn on entities

def-

[ddnet/ddnet] Pull request opened: #3504 Display median time instead of average (fixes #3399)

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 indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--undef...

Learath2

A framerate cap that's well above the monitor refresh rate should never ever cause stutters

I disagree. If the internal framerate is not synced to the monitor framerate, then you will render some frames when they are quite old, others fresh, so it will stutter

You should at most get a doubled frame every once in a while, not large stutters tho, no?

deen

I disagree. If the internal framerate is not synced to the monitor framerate, then you will render some frames when they are quite old, others fresh, so it will stutter

So, vsync on then?

yeah, should be microstutters

13:33

Personally I've always played with vsync, yes. But some vsync implementations add lots of delay, which is even worse.

on windows

13:33

linux doesnt have vsync

well, I have a vsync setting on Linux and it does what it's supposed to

@YupHio show us your graphic settings

its still not the same

13:35

on linux its more a flush

13:35

on windows its actually waiting

As an example of polling frequency way over the hz of the monitor mattering: https://youtu.be/gOQNRvJbpmk?t=537

def-

[ddnet/ddnet] Issue closed: #3381 Memory leak in sqlite3

13:54

13:55

cheap 9 years old laptop with no graphics card, 6gb of slow old RAM, and a slow asf HDD

13:55

I get minor stuttering on feature-rich maps, like that novice sunny-side map with all the moving quads and music etc

@YupHio how many fps you get?

let me baseline that real quick

and try enabling OpenGL 3

you could try turning of texts on maps just amke sure u read everyhting beforehand incase it explains how to do a step, you can turn it off with PageUp buton

turn on entities turn off health and score hud

13:59

press "Page Up" on ur keyboard, or in ddnet settings make the overlay entities fully right or left

Sunny Side Up map actually was 150+ FPS so that's not a good baseline, but also the server was empty which may have an effect. I tried turning on vsync but immediately noticed a slight delay when hooking stuff.

14:00

Sounds good, I will try all that. Thanks guys

Don't use /showall

KaipTataPasake

you could try turning of texts on maps just amke sure u read everyhting beforehand incase it explains how to do a step, you can turn it off with PageUp buton

there is no text (except the chat) that is rendered as text(except gl 1.x)

i mean the backgroudn goes dark and every text (name of map on start, little words) alll of that disapeasrs

entities

14:03

u mean disabling the design or what

Ok I enabled OpenGL 3.3, unchecked "Use high DPI", turned off score and healthbar in the HUD, put overlay entities slider to off, and turned off text entities. Now I am getting 500-800 FPS depending on the map. Night and day difference

14:13

Also I turned off high detail

Btw the slight delay with vsync is usually so very slight that it isn’t much of a disadvantage, it’s something you can get used to within a day or two

well it makes the frames litteraly wait to e.g. 16ms per frame for 60hertz

14:15

+ all ur other input delays = rip

You would be surprised at how much delay your brain can just delete

delay is delay

14:16

doesnt matter if ur brain can recalculate it

So? If you can’t perceive it it’s a non issue

u cant predict the future

14:16

end of discussion

Lol, I love discussing things with you, it makes me want to take a swig of bleach

there is no discussion

14:17

its a fact

14:17

i dont discuss about facts

Did you know that it’s one of the greatest evolutionary advantages humans had over other competing species?

14:18

We are usually very good at predicting the future compared to fishes

i bet any wild animal would be better in reflex based games than us

I played quite fine at 90ms ping with vsync, it’s just a matter of getting used to. Even 200ms you can tolerate as long as your partner isn’t moving around in a stupid manner

but u dont want to be the best, thats the difference

Fine play with screen tearing, use a crt for all I care lol

at 1000fps and high refresh rate monitor u almost dont see it

95% of people are casuals, and as a member of the homo sapiens species they possess the ability to tolerate small delays in the order of miliseconds. Most people I know prefer playing without screen tearing and don’t own high refresh rate monitors

when somebody askes for best experience, i tell them how to get best experience

14:23

and not how to make ur laptop stay cool

Lol, ok. I’ll be triple checking your every contribution from now on for performance hogs

im fine with that

This game is supposed to run on toasters, we’ve already strayed so far from that and I have a feeling you’d be okay with going even further

i'd say the opposite

14:26

ddnet 10 runs like shit

@YupHio wait, what made the large difference? Was it OpenGL 3 or High DPI? or something else?

while 15 runs much better

Tell that to all the people who get much worse performance like @noby (or anyone else on macOS)

he gets better performance

actually its not "much" worse anymore, but its still worse

14:28

ddnet 10 is my fav

he just doesnt like his fans

14:28

so caps his fps

Oh sorry, I forgot we should be using 100% of the cpu and gpu for or AAA title, lol

its not my fault macos doesnt like anything

14:29

vulkan, opengl

i use the same fps cap of 600 on cl refresh/gfx refresh

14:29

on both versions

@noby i already told ya

the highdpi fix really did remove 90% of the problem

the calculation changed and its not the same

but i still feel like version 10 is smoother

14:29

im not talking about teh fps in the corner

14:29

i know thats changed

no

14:29

not only that

14:29

if u type cl_refresh_rate 480 on ddnet 10

14:29

its not <480 fps

@noby if you disable highdpi the curved buttons look like shit

bcs it was buggy

Ofc reducing the resolution improves performance

learath ik but my fps is noticeably higher and it feels better to play on and it makes the fans shutoff so id gladly take those 3 advantages over slightly worse looking buttons and text

14:31

plus it looks fine to me anyway, wouldnt describe it as looking like shit lol

How was it with ddnet10? Atleast on my macbook that doesn’t use 200% of my cpu and lets me render in full res

14:32

(though I do need to patch it a bit to get it to render at full res)

im using an external monitor rn which decreases fps but its only using 20%cpu or less and its good fps

For the record, I can run ddnet10 without throttling even. Without throttling (and out of the box on steam) the current patch makes my mac a toaster oven

14:34

But ofc, it’s never our fault. Every other game runs just fine, but it’s obviously not our fault

aslong it outputs more fps i dont see something bad

14:35

>= fps

14:35

rest is just config

I really love how console games handle this, there should always be a target fps

14:36

Just rendering a bajillion fps using the entire cpu and the gpu is just an insane proposition, doing it out of the box even more so

well depends

14:37

best out of box experience sounds good to me

14:37

rest is config

Anyway, as always, you know best. I don’t even know why I ever talk about any subject as none of you here respect my opinion as much as a doormats opinion on any subject except maybe heinrich

there is no perfect default

14:38

we talked like 2 times

14:38

not my fault that i just think in facts these 2 times

vsync off master race

i have high refresh rate monitor and i can set it to 60hz with vsync

Yeah you just present your opinion as fact and say facts need no discussion

so can i judge on both

Would definitely be best if we can get good performance without 100% CPU/GPU utilization by default

also i play this game much more

14:38

so i can also judge there more

it doesnt use 100% on mine

14:39

cuz 100% would mean using all cores i guess

it doesnt use 100% cpu, it uses 1 core fully

14:39

and the graphic thread as much it gets

thats how it hsould be

14:39

games are ur primary focus

14:39

when u play

my gpu doesnt even reach 100%

14:39

bcs my cpu is too slow

rip

The burden of proof is on the one with extraordinary claims, so you need to show how you start from your claim and get to a fundamental truth. You can’t just say it’s fact without any support for it

14:40

Immigration is bad. It’s fact

14:40

There let’s deport every immigrant, no discussion needed it’s just fact

well i just dont care, i played this game so much, i know what i like, and if somebody wants best experience, he wants atleast what i would say is best experience or better

its a fact ppl disable vsync for any competitive game

in csgo they also dont cap

14:42

bcs they want their 300fps

14:42

xD

i use cl_refresh_rate 600; gfx_refresh_rate 600; gfx_vsync 0

14:42

and it works well for me on both clients i guess

14:42

ddnet 10 gives slightly hgiher fps

i think the cap of 600 in ddnet10 is not the same in ddnet 11+

ddnet 11-13 was fcky for me

well u use mac

14:43

so its ur fault

Ryozuki

its a fact ppl disable vsync for any competitive game

I argued that it doesn’t matter. Argumentum ad populum is just weak

14:43

Yes many people do it

Ryozuki

so its ur fault

no u

14:43

its not my fault if old versions work better than new versions

14:43

u might be able to say its my fault if no versions work correctly

i dont think thats true

Does that make it correct? Many people smear yoghurt on themselves in turkey as a cure to sunburns. Are they correct?

the definition of vsync means delay

14:44

this is not a belief

vsync is bad

its math

feels shitty to play with

Delay is bad is a belief

?????

I think we can all agree that reducing delay is better

well before we destroy our teeworlds experience we could destroy bitcoin

14:45

thats more energy waste

bitcoin will grow cuz now the big holders went in

14:45

they expect it to reach 150k

1

It’s almost imperceivable. Do you notice the text input delay introduced by the X Server?

people notice it

u can certainly feel the delay introduced by vsync

and i do

u cant see it

specially on 3d games

but u can feel it when u play

on ddnet u notice the mouse delay

i know for a fact because it feels like i accidentally toggled antiping off when i try to play after forgetting to disable vsync

oof almost forgot to backup etc/hosts and my ssh key any other places i could search files before wiping my harddrive?

search .config/

14:47

and most importantly

14:47

ur vimrc

14:47

the sacred file

vimrc is on git :)

14:47

found 1 in .config/

make sure to backup any gpg key u have

Easy solution, back everything up, check what you needl ater

xd

14:48

true

hmm

it shouldnt take much space

smart

14:48

900gb of hentai

i would ignore stuff like node_modules

14:48

that shit is always big and so many files

ye

14:48

i think ill risk it to not backup everything cuz its mostly trash and idk where to put it

14:49

i should have ervything now yikes

wipe the hdd to clear all traces of chillerbot and other malicious software

In the endless pursuit of more and more fps we have lost our sight of the real goal. You want the screen to render a new frame every single refresh, nothing more

oof

btw i dont care about much more than 1000 fps, my gpu can do around 10k

u will definitly forget something

i also capped it

14:50

but the CPU cap is killer

also the fact that new copies of ddnet open like 3 random blank windows behind it

14:50

is very annoying

14:50

on a small screen

i even notice it when i set it to 1000 or more

@noby thats macos stuff

i dont care

on my linux its clean

noby

also the fact that new copies of ddnet open like 3 random blank windows behind it

It’s also an sdl bug iirc

its new ddnet stuff

14:50

old ddnet never does that

Oh, I guess. But it’s definitely not our fault

14:51

You know how I know that? Because that would imply there is a mistake in something Jupstar wrote

i could do the bisect thing if yall want

14:51

me to narrow down the commit that cuases the black windows to appear

That’s a fundamental contradiction

edg-l

[ddnet/ddnet] Pull request closed: #3472 Add discord rich presence

You can try, I couldn’t reproduce it reliably, so I didn’t bother bisecting

it happens everytime for me on new ddnet x_x

when did i wrote sdl code? except ones

the reason i havent bisected is because certain versions seem to not compile

and that isnt released yet

Then please do bisect for it

will yall help me compile the ones that dont work

also u cant really trigger me. I do bugs, thats normal, i just try to fix them

im lazy to debug each error in each versin

Oh yeah, you didn’t touch SDL, I guess if you do find a bug we don’t risk the universe collapsing into a blackhole

i mean our sdl code

14:53

xD

do u mind reminding me how to bisect pepeH

Well first find a version it doesn’t happen in

seems to be fine in 12.7.2

14:54

ill try 13x

14:54

13.2 fine too

14:55

14.0 fin

14:55

e

14:55

15.1.3 not fine

I guess that’s close enough. Bisect works in logn steps

14:55

Do git bisect good 14.0

14:55

git bisect bad master

okok lemme pull

Oh you might need to first do git bisect start

ok i did these cmds now what

14:57

git bisect bad master Bisecting: 882 revisions left to test after this (roughly 10 steps) [2a3b254b4793930a0d19efaf8494a521d6668fca] Merge #2890

https://github.com/ddnet/ddnet/pull/2890

Finish /showothers 2 implementation on server by def- · Pull Reques...

Ok now test this revision

oh it already updated ok

If it’s bad you do git bisect bad

14:58

If it’s good you do git bisect good

ok its compiling

14:58

prepare for errors

Fingers very crossed ;P

make sure to do a clean compilation

fatal error: too many errors emitted, stopping now [-ferror-limit=]

14:58

lmao

remove the build folder

there was no build folder

14:59

i made it

14:59

[ 45%] Building CXX object CMakeFiles/DDNet.dir/src/engine/client/backend_sdl.cpp.o In file included from /Users/nobody1/ddnet/src/engine/client/backend_sdl.cpp:15: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../include/c++/v1/cmath:317:9: error: no member named 'signbit' in the global namespace using ::signbit; ~~^

14:59

should i call this bad or good

Oh this garbage

14:59

Just clean it up do the cmake and wait

idid

I’ll send you the “fix” I found for it

ok

macos

https://discord.com/channels/252358080522747904/293493549758939136/788467380682489886

u

15:00

wheres the dlib

15:00

dylib

I really would love to properly get this fixed but idk who to even annoy about it

15:00

You just need the sed command

ahh

15:01

In file included from /Users/nobody1/ddnet/src/base/system.h:18: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../include/c++/v1/stdlib.h:93:15: fatal error: 'stdlib.h' file not found #include_next <stdlib.h> ^~~~~~ uh? xd

If that doesn’t work you need to grep for sdk in cmakecache, it might be some other sdk overlap not the exact one I have

id rather just download precompiled 14.5 or smth to narrow it down more broadly

15:03

ok 14.5.1 works fine

15:03

how do i quit the bisect and start over from that point

Yeah this is what happens whenever I want someone to debug anything on macOS... I seem to be the only person patient enough to debug xcode internals

lmao

git bisect reset

ok nowwhat

git bisect start && git bisect good 14.5 && git bisect bad master

oh i just found

15:05

15.0.5 from site doesnt work

I honestly am not sure how well you will be able to debug like this

15:05

The version from the site links to one sdl2, you link to another sdl2

,_,

15:06

easier to just use an old version right

For you yes, but it’s sad that we never get to find these bugs

i narrowed down the problem is after 14.5.1 and before 15.0.5

Even I with my almost infinite patience couldn’t fix our cmakefile

oh between

15:08

14.5.1 and 14.7.1

15:08

actually

15:08

since .7.1 makes balck windows

On macOS I can’t get it to link to libraries I want. It always wants to link to the bundled ones

14.6.2 makes the error too

15:09

14.5.1 doesnt

15:09

i cant narrow it down anymore from precompiled

On macOS two different versions of the sdks includedirs can end up in the compilation line. I even figured out why it happens. I just don’t know how to fix it. I’ve decided cmake is just very meh on macOS

bring back bam

noby

i cant narrow it down anymore from precompiled

Are these both precompiled?

yes i went ddnet.tw/downloads

15:11

14.5.1 no black wndows

Also what are your gfx settings?

deen update the build stuff and libs

15:11

@noby https://ddnet.tw/downloads/DDNet-14.7-osx.dmg

15:12

that is without the patch i think

noby

bring back bam

Bam is even more confused at this btw, it had 3 versions of the sdk in my linker line

rip lol

15:13

@Deleted User bug

have u tried without ffmpeg

15:14

that was probs the biggest change in 14.7

how

Compile with the videorecorder option off

cmake .. -DVIDEORECORDER=OFF

compiling doent work

too bad

Can you paste your cmakecache somewhere?

15:16

I’m sure I can fix it for you

have u tried gfx_opengl_major 1 in newest client @noby ?

15:17

maybe its the gl driver

15:17

or sdl trying to get gl 3.x context

15:17

and just openes alot of windows xD

wpah wtf

15:18

it worked thx

nice and sad :p

no more black windows on any version

Sadly you don’t get the 3.3 benefits anymore :/

i wonder if its SDL or the video driver

15:19

bcs SDL could be fixed

15:20

@noby what does GL 3.3 in version 14 or smth do?

I could debug it if I could reliably reproduce it

or better 12

do in what sense

does it also open black screens?

gfx_opengl_major 1 no

no

15:21

in ddnet 12 enable gl 3.3 in graphic settings tab

no i have cl_svae_settings 0

and see if u get these weird black screens

how to do it in cfg

wait gonna see

15:22

gfx_opengl3 1

ok

15:23

no windows in 12

alright

or in 15

not even in 15 or what?

15:23

i hope its in 15

15:23

else its no SDL bug xd

gfx_opengl_major 1 fixed it for all versions

15:25

opengl3 made no difference

15:25

cus it was already fixed

in 15.x u need to use gfx_opengl_major 3; gfx_opengl_minor 3

why

bcs we allow different gl versions now

i mean what is the diference ingame

i mean for the windows

15:26

ingame it gives more fps if ur driver isnt bad

15:26

it lowers CPU time basically

15:26

it wont give more fps if ur GPU is the limiting factor

those settings on ddnet 15 actually make it better

15:27

nice

@Deleted User would it be possible to only start rendering like 3 frames before the screen is ready to accept one?

accept what?

a new frame, idk if you can get this info somehow

ah

15:28

without vsync it will probs replace half of the buffer of ur monitor

15:28

(or any devisor)

15:28

it also sends chunks or smth and if u get new input it just sends that

15:29

offscreen it can theoretically render anything

15:29

double buffering is doing that basically, and the driver decides if it uses the content or not

do we do double buffering?

yeah

15:30

but probs the driver even does tripple buffering

15:30

so it can swap two offscreen buffers

15:30

and has one "ready" buffer

can we somehow know when the driver uses the new buffer?

not in release mode atleast xD

15:30

maybe there are extension for amd and nvidia dunno

ah, too bad

u can disable double buffering

15:31

and u can also always render on the main buffer directly, gut there is no garantuee the driver actually does it

15:31

or just lets u think that u do it

this was just an idea to reduce the unlimited frame rate we have to something sane

15:32

without increasing input lag

gpu isnt the problem

but if we can't know when the driver will swap the buffers, that doesn't sound possible

its the mouse input mostly

15:33

the main thread is too busy and has to create all rendercalls

15:33

there is no visualisation / simulation difference

15:33

the main thread has to create everything

15:33

it would be possible by rewritting teeworlds

I don't follow. the problem is that we get a 16ms old mouse input with something like vsync because we render the frame 16ms early, right?

example: cpu renders no frame goes into sleep but 1 nanosecond later it could have rendered the frame so it wakes up some point has todo the simulation has to prepare rendercalls => delay

15:36

you might be able to use the FPS (if its stable enough) to predict when to wake up

15:36

but its probs never 100% reliable

yes, not 100% reliable

(oh and maybe to get it right, it sleeps long)

15:37

bcs it had less todo in that frame without a rendercall

but if you wake up 3 frames early, it should be quite reliable, no?

15:37

3 frame times early

well depends on what ur target fps is probably

hm?

15:37

say I have a monitor with 60 or 144fps

but there is more to it anyway, it also fetches the mouse relative state from SDL it doesnt use global mouse state

then teeworlds would render 180fps or ~450fps

i understand what u want

but not 1800 or so, wasting energy, battery, etc.

15:38

kk

15:38

do you mean we could wake up more often to fetch mouse input but not render?

it already does it if u dont limit gfx

15:39

or do u mean without any simulation?

15:39

just the mouse

ye, just the mouse

could work, but i dunno if it works

15:39

in the end its more about semi state

15:40

wait

15:40

i mean like u are lucky to either hit the exact moment u want to draw a frame or not

15:40

if u wake up at some point u still have todo the simulation

yes, but that's no different with unlimited fps

15:41

we just need to make sure we're continuously rendering frames around the point where it's important

yes and no

15:41

180 can still be 1000/180 in worst case

15:42

+ 20ms of ur moniitor + 20ms of ur mouse or whatever

I don't understand these numbers

15:42

180, 1000/180?

u could go sleep right before it would allow a new frame

15:42

and then u probs sleep for 1000/180 ms

you mean if we don't know when the frame gets drawn?

yes we dont know that

ah yes, that's unlucky

15:43

then my method won't work

as said, u can predict it, it might improve it over what we have now, but i dont think without a clean rewrite it would get much better

how can you predict it?

15:44

if we have no insight whatsoever into when the frame gets put onto the screen, we have no data to predict it from, no?

well we have the FPS

15:44

and know the last time we triggered a rendercall (edited)

that only tells us the frequency with which the monitor accepts new frames(?)

15:45

not the exact moments

well we know the exact moment we prepared the graphic buffer

15:45

when we kick it in

15:45

and from then it should (if the FPS are stable) 1000/fps ms to the next frame xD

15:45

thats what i mean with prediction

okay, let me rephrase what I want to do, it seems to me that you're not quite saying the same thing as I do

15:47

the way "unlimited fps" helps reduce input lag is that it continuously renders new frames, so it also renders a new frame just before the graphics driver swaps the output buffer

15:47

we're wasting a lot of renders that we basically know won't get displayed anyway, namely right after that swap, we know that there won't be any more swaps for 16ms

15:48

so if we knew the moment of the swap, we could pause our rendering until just a couple of frame times before the next swap, rendering a couple of times until the swap comes and then pausing again

thats not true tho

15:48

there are more swaps

15:48

else tearing wouldnt occour

this would, in my understanding, lead to the exact same input lag as "unlimited fps"

15:49

tearing occurs when we render to a buffer that is swapped while we're rendering to it

yes

I'm not talking about tearing or solutions to tearing

15:50

I'm talking about the same setup we have for "unlimited fps" but reducing the amount of renders we do

but for a non constant fps i already said the same as u

Deleted User

thats what i mean with prediction

here

but that's wrong, no?

15:50

just because we gave the driver a new frame doesn't mean it will put it on the screen immediately

well u could substrace the time it requires for the simulation part

15:51

a fps on simulation part basically

there might still be time to render another frame

it should without vsync

15:51

it just updates the current stream to monitor

15:51

ur hdmi cable doesnt have instant bandwidth xd

yes, I understand that

15:52

maybe I need to draw pictures, I don't think we're on the same page

u want to swait for the swap and then start the new frame?

no

but using 180fps on 60hz doesnt fix issues

that's vsync, in my understanding

bcs in worst case u have 1000/180 delay

I think I really have not made myself clear enough

15:53

1000/180s delay would not happen in my understanding in my model

15:53

I'm not just doing 180fps

well i can only tell ya u render the full rendercalls offscreen, and as soon as u swap the buffer it is probably updating internally to the monitor stream

15:54

so its never wasted

say we have a 60hz monitor

15:55

then the monitor can only display 60 frames per second

yes

15:55

but that doesnt mean u dont want half images too

15:55

that are "more" updated

so every frame we render above that is wasted. we might not be able to eliminate that waste because we're not able to predict the future, but it's still wasted

no

if you're going to tell me something about input delay

u are litteraly describing vsync xD

no

15:56

I think we're just not talking about the same thing

15:56

you think about it in practical terms, with a specific algorithm in mind

15:56

namely vsync

its not wasted, ever

and then you say the extra frames aren't wasted because they help reduce input latency

fokkonaut

[ddnet/ddnet] Pull request opened: #3505 Fix dummy not sending DDNet version

Checklist

[ ] 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 indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssanitizer--u...

they are also displayed on screen

are they?

just "only" partially

15:57

yes, thats tearing

a 60fps monitor can display more than 60fps?

ofc there is no garantuee that absoluty none is wasted

I thought tearing happens at a different level

it can render half of the old frame and half of the new

15:59

in theory u could also say, ur eyes follow from top left to bottomr ight and it only updates a small potion and then u would have somthing that is more than 60hz on 60hz monitor xd

bors[bot]

[ddnet:master] 2 new commits

7e169d9 Fix dummy not sending DDNet version - fokkonaut bf17072 Merge #3505 - bors[bot]

16:23

bors[bot]

[ddnet/ddnet] Pull request closed: #3505 Fix dummy not sending DDNet version

yikes so if my os crashes after fresh install its probably hardware right?

you could try running memtest

yeah maybe its the memory

17:19

get a ECC card monkalaugh

@heinrich5991 have u seen https://github.com/launchbadge/sqlx ?

launchbadge/sqlx

The Rust SQL Toolkit. An async, pure Rust SQL crate featuring compile-time checked queries without a DSL. Supports PostgreSQL, MySQL, SQLite, and MSSQL. - launchbadge/sqlx

17:42

features compile time checked sql instructions (if they are not dynamic)

anyone up 4 an ATH 0.7 rewrite?

xush'

3

1

2

17:58

no api features for accessing player input this time around >.>

@GutZuFuss you can start by helping to update ddnet client to 0.7 :)

k lets go

19:07

where are the whitespaces i should delete

haha

19:07

the skin system is probably a nasty one

Would someone mind to approve my refactoring backport from 0.7? https://github.com/ddnet/ddnet/pull/3502

@Kaffeine we don't always review same-day, you'll need some patience

20:01

but we still appreciate cleanups and other contributions

deen

but we still appreciate cleanups and other contributions

I don't force anyone

The ask for "0.7 refactoring" CR supposed to be in the context of "anyone up 4 an ATH 0.7 rewrite?" and "you can start by helping to update ddnet client to 0.7

" I consider "helping to update ddnet client to 0.7" to include "review 0.7 backports".