perf_prof_branch

[rain moth]

https://forums.bohemia.net/forums/topic/160288-arma-3-stable-server-208-performance-binary-feedback/?page=101

[full nova]

Performance_Profiling is the full name

[autumn timber]

They contain patches and changes that may be released in an upcoming Arma update, ahead of time

Since 2.08 has just been released, the current difference is minimal, but that will change over time

[full nova]

see you guys on 2.10 in summer (rough estimate)

[autumn timber]

Are you boycotting 2.08? 😁

[full nova]

[silk pewter]

wait, Steam says Cloud is up-to-date 👀
are save and settings synchronised now??

[restive pilot]

afaik they always were
it just shows it in the game page now

[rain moth]

This app does not have Steam Auto-Cloud configured, which likely means save files are in [Steam Install]/userdata/{Steam3AccountID}/107410/ folder.

[restive pilot]

[silk pewter]

at least on Profiling yeah!

[rain moth]

it's there on public as well

[silk pewter]

w00t w00t 🙂 @rose moth the move was made?! :3

(next up, userprofile & controls??)

[rain moth]

the appid folder does not exist for me in userdata

[whole cloud]

no

[silk pewter]

what's up with SteamCloud all of a sudden?

[restive pilot]

what is it storing then?

[whole cloud]

There was something because steam did some change for something something that broke something so we had to do that to something something fix something

[silk pewter]

…thanks, I will be documenting that feature right now!

[empty goblet]

let me fix that to something

[silk pewter]

it's fine, the page has been created somewhere already

[tawdry gazelle]

Hi I have encountered severe CPU usage after 2.08 update. 12% before on a 8/16 CPU, but now above 20%! Will give more tests tomorrow, gotta sleep now... Just asking...

another player has the same problem, he has a 6/6 CPU, 68% utilization.

[whole cloud]

if that corresponds with higher server fps, thats good 😄

assuming you mean server

[vivid rune]

I have a question: Is the 64bit linux version compiled with AVX flag?

[whole cloud]

no

[autumn timber]

#perf_prof_branch message
Called it! 😂

[rain moth]

😄

[warped rain]

:D

[whole cloud]

These changes are not on stable tho, not even on prof now

[vivid rune]

After this great new update I have some long standing bugs that annoys me and i would ask if there is channel here in discord where i can this post? Idk if there are redundant and well known. But maybe it is good to address it new?

[silk pewter]

#arma3_troubleshooting mayhaps

[vivid rune]

#arma3_troubleshooting sounds for me more like "quick help for user problems" but i try

[silk pewter]

or wait tomorrow for 🙂

[hexed gorge]

The clients performance improvement, when we don't use AI functions like knowsAbout, reveal, targets, etc, is in version 2.08?

[tawdry gazelle]

CPU usage increases both on server and client, however not only one players have reported stuttering with v2.08, with "abnormally" high CPU usage.

My PC is sooo high end that I could not produce the stuttering on my side...

They says, "I feel much more stuttering when shooting and killing AIs than before."

[eager vault]

Steam does this weird thing now (only just started happening) where it tries to download/update workshop items while the game is running, fails because the game is running, but then repeatedly reattempts. This causes CPU resource usage of the game to skyrocket. If I manually disable workshop updates it goes back to normal levels.

[uneven kernel]

I was experiencing the same thing, I read something in here saying it was because the game checks all your compositions to see if they have updates or something. I unsubbed from a bunch of compositions and it fixed it for me

[eager vault]

Ahh okay thank you

[rose moth]

No, this was forced to be enabled because of a Steam change, even though we do not support it properly. It was announced some time ago, but unfortunately the client is misleading and we cannot prevent it from showing 😦

[whole cloud]

Report them on feedback tracker, and generally feedback tracker stuff goes to #arma3_feedback_tracker

[whole cloud]

no

[hushed seal]

Am I doing something wrong, server is on perf branch, client is on perf branch, both are on Version: 2.08.148925-

https://imgur.com/a/ijdRYxr

[whole cloud]

no youre not

known issue. Already told probably a dozen people today to ignore it

[hushed seal]

oh okay 😄

thanks!

[rain moth]

https://twitter.com/Arma3official/status/1501231607298297859

Arma 3 (@Arma3official)

There may be a glitch in the in-game server browser that falsely marks 2.08 servers as incompatible. Try connecting any way, or use the Launcher server browser instead. We will be looking into this and anything else reported.

Twitter · Mar 8, 2022, 4:21 PM

[restive pilot]

this is what happens when you code like this: 😛
#perf_prof_branch message
https://cdn.discordapp.com/emojis/742599803606925332.gif?v=1

[hushed seal]

I naively assumed I did something wrong 😄

[rain moth]

avoid the 32bit server atm as well 😉

https://twitter.com/arma3official/status/1501531174937804804

Arma 3 (@Arma3official)

@21stRangerRGT We have identified an issue with the 32bit version of the game. A fix for the crashing will be deployed on profiling branch in a few hours, and we are updating our processes to prevent this in the future.

Twitter · Mar 9, 2022, 12:11 PM

[whole cloud]

2.08.148977 new PROFILING branch with PERFORMANCE binaries, v1, server and client, windows 32/64-bit, linux server 32/64-bit

• Added: Automatically add Creator DLC tags to compositions that contain CDLC objects
• Added: New -noPauseAudio start parameter to keep audio running while the game is not in focus - https://feedback.bistudio.com/T163147
• Tweaked: Limited length of script errors so that on-screen error messages are always shown completely
• Tweaked: Improved multithreaded code performance and fixed bottleneck on Dedicated Servers (Thanks Kohbrax)
• Tweaked: Very experimental multithreading improvement on Linux
• Tweaked: Minor performance improvement if there are many vehicles on the map
• Tweaked: Improved networking performance (update checking for players/objects is now spread over multiple frames)
• Tweaked: skipTime script command is now limited to max +/-1000 hours per skip - https://feedback.bistudio.com/T68952
• Fixed: File reading errors when there are multiple PBOs with the same prefix loaded
• Fixed: It was possible to see illegal game dates before 1970 - https://feedback.bistudio.com/T163269
• Fixed: More stray entities bugs in addItemCargo and setUnitLoadout - https://feedback.bistudio.com/T163205
• Fixed: Incoming missile warnings were played through the radio instead of the effects channel - https://feedback.bistudio.com/T163245
• Fixed: 32-bit executables crashing
• Fixed: Servers falsely shown as "incompatible version" in the game server browser (needs this update to be running server-side)
• Removed: Network reports (networkDiagInterval, NetStrStatsExport) (still available in Profiling build)

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[rain moth]

time to update all servers again then

nice with the -noPauseAudio!

[whole cloud]

There is also a issue with network compression. If you regularly send large variables, the compression takes time and might cause server fps issues

[kindred radish]

What means large in that context ? like variables with arrays and loads of information/data ?

[whole cloud]

like 10+KB size variables via pubVar or through remoteExec

[rain moth]

like that reported bug to ACE about 2GB strings? 😄

[empty goblet]

prepares ☕ and sips it silently while waiting for v2

[whole cloud]

v2 tomorrow, v3 monday :gottagofast:

[silk pewter]

any hotfix planned for the server-client mismatch? (& some stutter issues perhaps?)

[whole cloud]

version display thing was fixed in v1

second line from bottom

stutter issues will be next week. Stutter issues seems to only happ... like I wrote above.
so most normal gameplay not affected too much

[silk pewter]

I mean will it be ported to stable?

[whole cloud]

yes sure

if someone encounters issues on v1 please tell

[empty goblet]

if anyone has time, please switch to PROFILING branch https://community.bistudio.com/wiki/Arma_3:_Steam_Branches#Arma_3 then join KOTH stress server at 85.190.148.52:2102 and report problems in #perf_prof_branch (or DM me)

[whole cloud]

Dwarden be spamming again 🛃

[empty goblet]

desparate times demand actions, not passivity 😁

[whole cloud]

goes to bed

[kindred radish]

joined a bit on the Koth server and while watching the tower with a sniper it still looks the loads of desync (the towers were notorious for that tho)

[vivid rune]

For me on ground look at Jets they moved laggy/rubberband

[plain trout]

descyn is very bad

[gritty wasp]

is this 208 or only perf issue? Should I revert server to main branch?

[fickle geyser]

framerate also seems to be pretty bad during firefights, both stable and profling.

Seems weird that it got worse with 2.08 as all of the perf changes were supposed to be in 2.06 perf binary too, right?

[whole cloud]

both are bad

[whole cloud]

The compression was not

[fickle geyser]

Ah, so the network messages for bullets firing is the reason for frame drops?

[whole cloud]

I was running compression on my server for 3 months and didn't have any issues.
But now I have two people who had extreme issues,and they are mostly gone when switching to a non-compression build

[whole cloud]

no

Scripts and stuff are sending large stuff over network like remoteExec or pubVar.
And the compression of that uses too much cpu time

v1's

Removed: Network reports (networkDiagInterval, NetStrStatsExport) (still available in Profiling build)
Should help quite a bit tho

[fickle geyser]

I'm on profiling and still dropping to 30fps

where I usually run 60fps

and it feels like a stutter and framerate jumps a lot.

[whole cloud]

yush

[vivid rune]

Is compression an own thread?

[whole cloud]

yes/no

[fickle geyser]

I guess you will disable compression in hotfix?

[whole cloud]

yes

and in v3, tomorrow afternoon or friday

[rain moth]

🗜️

We'll be participating in large event on Saturday so will be interesting how that fares with 2.08

[vivid rune]

maybe a not so aggressive compression tech with less overhead is the solution?

[whole cloud]

Maybe.
For now just disable and maybe try again some day

[rain moth]

could it be opt in via server.cfg? 🤔

[whole cloud]

Doesn't help that Arma is stupid about it. Each message gets compressed atleast twice.
And if you send a copy of the same message to many players, its repacked/recompressed for each player, even though it stays the same message

[rain moth]

oops 😄

[whole cloud]

I may want to fix that first before reattempting compression

[vivid rune]

The requirement for the compression is not size. More important is stable and short compression time for all kinds of data that has to be compressed.

[whole cloud]

we only compress publicVariable, addCuratorAddons and remoteExec parameters

Stuff that is usually small and not sent often

but if you do a pubVar of 20+KB every second or two to 80 players..
That doesn't happen on my server, which is why it ran fine for 3 months in my testing

[empty goblet]

some scripts/scenarios do dozens of pubvar to all players nearly nonstop ...

[fickle geyser]

we're just running standard OP without any fancy scripting. But ACE might be adding some overhead.

[restive pilot]

they deserve the bad performance then 😛

[empty goblet]

just add publicvariable.txt and publicvariableval.txt BattlEye server-side filter with 1 "" and you will be like while watching the log

and if you want more madness add setVariable.txt and setVariableVal.txtand remoteExec.txt and mpeventhandler.txt with 1 "" to the server's \BattlEye\ folder where active beserver_x64.cfg is

after seing several PB of those logs in past 9 years ... my memory needs some very effective text compression {do not tell anyone i flushed it all}

[whole cloud]

2.08.149000 new PROFILING branch with PERFORMANCE binaries, v2, server and client, windows 32/64-bit, linux server 32/64-bit

• Tweaked; Several minor performance tweaks
• Fixed: Server did not fully disable VoN if disableVoN config parameter was set

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[thorny pumice]

Seems like this latest binary has fixed the 5fps issue 👍

[whole cloud]

v2 doesn't have the compression fix yet

[hexed gorge]

Who does that? 😮 🤔

[whole cloud]

[hexed gorge]

I'm a bad programer and never did that.

Who do this should be a bad+ programmer.

[hexed gorge]

There is a public link to watch the event?

[rain moth]

Good question, I think there will be some streams

It's an Arma Finland hosted event with a couple of different European Arma communities participating

[whole cloud]

2.08.149024 new PROFILING branch with PERFORMANCE binaries, v3, server and client, windows 32/64-bit, linux server 32/64-bit

• Tweaked: Submunition _triggerOnImpact now only triggers for actual hits, not for deflections
• Fixed: Counter Measures did not provide a projectile in Fired/FiredMan Event Handlers (always passed objNull)
• Fixed: Several game crashes
• Fixed: remoteExec failing when JIP parameter starts with numbers and a colon, while also not being a valid netId which influenced BIS_fnc_createTask (Thanks @shell lion)
• Removed: Network Compression (disabled for now)

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[rain moth]

update not out on steam cdn yet

[sharp bloom]

Doesn't look like it's in the Google Drive either

[rain moth]

I call fake news!

[whole cloud]

Ahh I thought I'd push changelog 10 minutes early. but it failed

[rain moth]

(╯°□°）╯︵ ┻━┻

[whole cloud]

┬─┬ ノ( ゜-゜ノ)

[vale shoal]

is it live now?

[rain moth]

no 😦

[patent dome]

Yes

[rain moth]

now, go go go!

but only for main arma 3 app, not the server app 😦

[patent dome]

Server should be up in couple more minutes

GDrive one minute after

Nevermind, server is already up

[rain moth]

now!

[plain trout]

do you need testing on that koth server?

[empty goblet]

getting reports of VON not working anymore on servers with 2.08.149029

[kindred radish]

this newest prof update broke a dll we use that has always worked before 😦

https://github.com/maca134/ExileLootDrop

[silk pewter]

and how do you know the dll did not break at the exact same time of the update, hmmm?

[kindred radish]

because it worked an hour before and i could replicate it on two different servers
looks like its not getting loaded but no error in the rpt connected to it

tested v1 & v2 prof and the dll still works

[silk pewter]

(I know, I joked about the dll "breaking" by itself right on time 😄)

[whole cloud]

what does broke mean

[kindred radish]

looks like it doesn't get loaded
because there is no loot spawning

and this message does not appear in the log

19:32:55 CallExtension loaded: ExileLootDrop (C:\Arma3Server@ExileLootDrop\ExileLootDrop_x64.dll) [1.0.0.0] [1.0.0.0]

[whole cloud]

Well then its not the extension being broken, but your script that loads it somewhere

[gritty wasp]

I think our replay dll not loaded too. ACE also not reported in RPT

[whole cloud]

🍿 what the what

[restive pilot]

battleEye issue?

[gritty wasp]

disabled

[fickle geyser]

WTH, Dev branch was there for quite a while and there seems to still be quite a bit of issues what slipped past

[whole cloud]

battleye doesn't block on server

We had one month of RC

The 32bit crashing was there from day one and noone found it

And issues just keep on comin

[restive pilot]

tbf almost noone uses 32bit these days

[gritty wasp]

I am stupid. ace dlls called only on client side

[whole cloud]

well is clientside broken too?

[gritty wasp]

No

22:11:42 CallExtension loaded: ace_advanced_ballistics (C:\Program Files (x86)\Steam\steamapps\common\Arma 3\@WOG_ACE\ace_advanced_ballistics_x64.dll) [3.12.0-8ddde18]```

[whole cloud]

great

not

If something was broken with callextension, or script execution that would be easy.
But apparently everything is working just not on server? 😄

[fickle geyser]

We played mission yesterday and OCAP saved the replay just fine.

[whole cloud]

v3 was released today

hopping from issue to issue, next one tomorrow

[gritty wasp]

Our OCAP implementation seems have limit on player count. I was alone on server that is why it was not loaded. So I think everything is OK here for us. Tomorrow will check at weekend games.

[whole cloud]

Well I can't wait till weekend. Can't just leave weekend broken 😄

[gritty wasp]

CallExtension loaded: task_force_radio_pipe (C:\Program Files (x86)\Steam\steamapps\common\Arma 3\@TFAR\task_force_radio_pipe_x64.dll)
TFAR also seems loaded.

[kindred radish]

just some dll's seem to be affected other dll's work fine (server).
for example we use Infistar, A3Log, exdtb3 and they work without a problem.

[whole cloud]

:u

Would be very hard for me to diagnose why some DLL's I don't have seem to have semi-randomly stopped working for you oof

[kindred radish]

i can imagine 😆

[gritty wasp]

What if call in console example from github readme to force callExtension?
'table' call ExileServer_system_lootManager_dropItem;

[kindred radish]

i will try, this is the rpt after the first loot spawn request with prof V2

19:32:55 CallExtension loaded: ExileLootDrop (C:\Arma3Server\@ExileLootDrop\ExileLootDrop_x64.dll) [1.0.0.0] [1.0.0.0]
19:33:01 Warning: 6790 ms spent in callExtension calling name: "ExileLootDrop", function: "Factories|1"

[whole cloud]

7 seconds

[violet elbow]

Dedmen's getting old

[whole cloud]

I am already old

[violet elbow]

[violet elbow]

Sure, gramps.

[kindred radish]

• 2.2k items in the loot table`👀
  loot spawn is still fast enough in-game and i think its only that slow once (cache)

_
With v3 and executing 'Factories' call ExileServer_system_lootManager_dropItem;

20:35:59 CallExtension loaded: ExileLootDrop (D:\AEOG\ArmaServer\ServerRoot\@ExileLootDrop\ExileLootDrop_x64.dll) [1.0.0.0] [1.0.0.0]
20:36:06 Warning: 6941 ms spent in callExtension calling name: "ExileLootDrop", function: "Factories|1"

No loot spawning tho with the same system that always has worked before (5-6 years) and no error's in the rpt.

I didn't change anything else. So still weird that it breaks with V3

[whole cloud]

but the extension isn't loaded normally, only when you do it manually?

[kindred radish]

looks like yea

[autumn timber]

@whole cloud am I missing something?

[whole cloud]

it should be right next to nopause

AH

its 2.10

launcher is not included in profiling

[autumn timber]

[whole cloud]

stop breaking me I'm already broke

[fickle geyser]

in case you're not aware you can use command line for now

[whole cloud]

Yes isn't that a nice feature?

Keep being in awe and don't think of all the stuff thats currently on fire ❤️

[fickle geyser]

Arma has been on fire so long it's burned to crisp anyway.

Tiny bit of fire won't change much.

[autumn timber]

I can lend you 10 bucks if that helps... 😬

[fickle geyser]

support him on patreon

[autumn timber]

...actually... you're actually right 😂

[whole cloud]

😠

nooo

[autumn timber]

You want to say broke? 😦

[autumn timber]

I wanted to say that this command works from the commandline and that it does wonders

[whole cloud]

oh no it makes you wonder? why? is it broken?

[autumn timber]

Makes me wonder why it wasn't implemented earlier

I can finally alt+tab and be aware of whether Titans or Voronas are flying at me, while we're having a chat right now 🙏

[green jackal]

broke: Voice Chat

[whole cloud]

[green jackal]

💀

Nah but like deadass

When is that shit gonna work

[whole cloud]

fix tomorrow

[cold vale]

Dedmen broke it so you have to install TFAR /s

[green jackal]

Also @whole cloud I just witnessed some guy instert a script or something so he left the zeus spot but had full zeus power and made everyone vote himself admin and crashed the server

Like 2 hours ago

Like everyone voted him admin when no one pressed anything

[whole cloud]

if you can get any more info that would be cool.
As of right now I don't really have much to go off... Basically nothing really

[violet elbow]

I hope BI pays for overtime.

[rain moth]

I’ve been using native extension with ”public” branch 2.08 Linux server, works fine

[thorny pumice]

must have been a combination of the binary and lowering the limitFPS param to 100

[autumn timber]

I must say that this is the best commandline flag so far

[glossy tangle]

cool. one big recommendation would be very cool to get like a month's notice that an update is coming ❤️

[upper sierra]

There was, in fairness. just a perf branch update that only happens server side.

[rain moth]

Almost exactly a month ago, https://twitter.com/Arma3official/status/1491067124047290373

Arma 3 (@Arma3official)

It's almost time for the first #Arma3 platform update of 2022. A Release Candidate build for 2.08 is now live for testing. Find support for #Steam Rich Presence, scripted lasers, and plenty of fixes.

💡Details: https://t.co/jKQDJP3aLP
📥Steam access code: Arma3Update208RC https://t.co/F6XhtKrAnB

Likes

206

Twitter · Feb 8, 2022, 3:11 PM

[whole cloud]

I only had to look at my VoN change for one second to see what I messed up...

[whole cloud]

You have that. RC branch releasing.

You can fix VoN by removing the disableVoN entry from server.cfg, instead of having disableVoN = 0

it only checks for the entry being there, instead of entry being there AND being set to 1

[rain moth]

does that include the v3 profiling?

nvm, part of v2 changelog 😛

[whole cloud]

oh it was broken since v2 and noone said ah

well v2 wasn't out that long :harold:

Is there someone else who has that exile loot not spawning issue that El'Rabito has who can help me with it?

[rain moth]

so VoN is still broken with v3 or...?

[whole cloud]

yes but ^

just do that

change disableVoN=0 to //disableVoN=0 and it'll work

[spiral pond]

any chance to get more detail on gsEva for diag_capture frame please - it seems to taken often a lot of processing time

[whole cloud]

MEventPrcs
Mission event processing.
EachFrame mission eventhandler

ArmaScriptProfiler will tell you more about them

[spiral pond]

ty

[rain moth]

just need my web admin config file generator to make an exception for now then 😅

[whole cloud]

Okey all issues that were reported for v3 are fixed, its building now, ETA ~3 hours

[silk pewter]

https://tenor.com/view/daaamn-daaamn-friday-daaamn-meme-gif-24222081

[whole cloud]

El'Rabito has been tortured but we figured it out

[kindred radish]

All good, feel free to contact me anytime for such debugging ^^

[silk pewter]

"torture me anytime" ❤️

[glossy tangle]

I think we can agree the "masses" are not on there... Social media and even better, the Launcher News tab would be a bit more effective 😉

[whole cloud]

We post on social media when RC gets released

[glossy tangle]

Yup, the day of... Talking in advance. Maybe a month would prepare us 😊

[whole cloud]

What?

it was posted ONE month ago

one month before. We try to have atleast 3 weeks of RC time

[glossy tangle]

Must of missed it. Got link for this old geaser? 🤣

[whole cloud]

Dahlgren already sent it to you ^

[glossy tangle]

❤️

[silk pewter]

if you miss even that, how do you expect to see the original post 😛

[glossy tangle]

https://tenor.com/view/ive-been-burned-so-many-times-im-equivalent-of-the-inside-of-a-roasted-marshmallow-dan-levy-david-david-rose-schitts-creek-gif-20773552

[fickle geyser]

What was broken? ;P

[silk pewter]

his ribs

[whole cloud]

I didn't know that netId's could have negative object IDs (They shouldn't, that should never happen, but it does)

Someone decided decades ago that the datatype for this ID that should never be negative shall be int :poutcat:

[fickle geyser]

well, at least you figured it out.

[restive pilot]

Now that it's uint you'll get 4 billion instead :P

[silk pewter]

50% perf boost confirmed!!1! 😄

[whole cloud]

It can be any number above 0, so that would be valid

[vivid rune]

Good, that you had this figured out. What me interested is what kind of code can produce negative IDs? What happened if this code now try to create negative IDs?

[whole cloud]

I have no idea

Nothing happens to it, it will keep doing that

[whole cloud]

2.08.149042 new PROFILING branch with PERFORMANCE binaries, v4, server and client, windows 32/64-bit, linux server 32/64-bit

• Fixed: triggerAmmo command not working on bullets, grenades, mines, flares, submunition - https://feedback.bistudio.com/T163630
• Fixed: Broken firemode selection on launchers with multiple flight profiles - https://feedback.bistudio.com/T163651
• Fixed: remoteExec not handling negative netIDs (bug from Profiling v3)
• Fixed: VoN being disabled if no disableVoN config entry was present, even if it was set to 0 (bug from Profiling v2)
• Fixed: Standing up after going prone with a launcher in top-down mode - https://feedback.bistudio.com/T128198

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[empty goblet]

anyone here runs dedicated server at 2.08.149042 with -noLogs and the servers randomly freezing ?

[amber glade]

I have a dedi server that is restarting all the time. on Malden

[limpid crane]

@whole cloud v4 client-side breaking our mission, no issue without it

[whole cloud]

┬─┬ ノ( ゜-゜ノ)

More details on "breaking" ?

[limpid crane]

all spawn points in the sea

[whole cloud]

And you are running profiling branch?

[limpid crane]

yes

[whole cloud]

Was v3 working fine?

[limpid crane]

yes, no one complained

just started to complain after v4

[whole cloud]

how do you set respawn points? some custom scripted thing, or editor placed markers?

No one complained, but do you actually know that someone has played with v3 and was fine?
Maybe just noone was on profiling at all and only now you have someone with it?

[limpid crane]

can't check it right now, I'm not the one who did it

[limpid crane]

the guy who complained had perf enabled and played yesterday without an issue judging by the logs.
will need to wait for the answer if you want to make sure.

[whole cloud]

Please send me your server info via DM, I would like to test myself

[vivid rune]

Maybe is not related to v4 but my server throw "Failed to receive from BE Master (11)
Disconnected from BE Master

[whole cloud]

not related to v4

[whole cloud]

Can repro
Will try fixing

[kindred radish]

Same for us, respawn is also broken (scripted)

Spawning near [0,0,0]

[whole cloud]

how do you define respawn points? :3

I hope editor placed markers or smth like that

[kindred radish]

with markers ye

[whole cloud]

in eden editor?

[kindred radish]

yes

[whole cloud]

Very good, then it can only be one change

https://tenor.com/view/blelele-cat-lick-gif-13137964

[empty goblet]

:ifixit: emotion not found :needthebuildfromfuture: emotion not found

anyone else observing issues with #shutdownaftermission being ignored ?

[autumn timber]

Profiling v<dayOfTheMonth>

[violet elbow]

I get my dopamine hits from reading changelogs. I'm loving this update.

[terse herald]

as an addition to the perf breaking spawn points:
I have no idea if or how this is related - but on our KP Liberation server basically all markers are missing too except a few (if that's any new information or helpful)

[woeful gorge]

Without profiling it does work, in the last version it worked to (played it multiple times).
For Altis Life Empty Markes do not work, same marker type I guess, getMarkerPos returns [0,0,0] but the marker attributes getting diag_log'ed right

[whole cloud]

Its the same thing. Respawn doesn't work because the markers are broken

[terse herald]

okay, great. wasn't sure 😉

[whole cloud]

Ok I confirmed my fix for the markers stuff.
Sadly our buildserver is being real big sleepy bum currently the fix will likely only go out tomorrow in ~10 hours

[pure berry]

What is the current recommendation for servers? server on profiling branch and headless clients on stable?

[whole cloud]

if headless clients don't need map markers they can also run on profiling

[pure berry]

What does map markers in this context mean?

[whole cloud]

Map markers that are placed inside Eden editor

[boreal wigeon]

I hate to be the bearer of bad news...
But this "Firing" bug is back: https://www.youtube.com/watch?v=8OeAn_4WihM
Warning lower/disable sound!
And it's worse!

When restarting mission impacts still occur...

Tested positive on:
Server (Linux) v3/v4 with client stable/v4

Stable version all fine!

[whole cloud]

Stable version all fine!
Well thats good that means hotfix will probably be good 😄

What is the repro? what is that selfhosted MP?
or just VR mission on dedicated?

[boreal wigeon]

Dedicated server

[whole cloud]

Can you repro with any gun?

[boreal wigeon]

let me try...

[whole cloud]

My unit just played 4 hour long mission with 20 people and noone had it

[boreal wigeon]

MX 3GL also has it.

It's like a plasma cutter...
https://www.youtube.com/watch?v=uyNZZH0TYC0

[whole cloud]

do you need to fire full auto, or single fire a couple times?

[boreal wigeon]

Last video (MX 3GL) I was in single, and only shot 3 times max

LMG was on full

[whole cloud]

please send me your servers basic.cfg

[boreal wigeon]

MinBandwidth          = 1000000000;
MaxMsgSend            = 640;
MaxSizeGuaranteed     = 512;        
MaxSizeNonguaranteed  = 256;    
MinErrorToSend        = 0.002;
MinErrorToSendNear    = 0.02;
MaxCustomFileSize     = 0;

[random isle]

Your custom script just cant find location and it defaults to [0,0,0] now, Pre 2.08 or probably after 2.04 it would spawn to position where you searching from but high in air so like [pos,pos,pos + 200]

[kindred radish]

It's not just a me issue and dedmen already fixed it i guess ^^

Still running V2 atm due to broken spawn and on a freshly wiped server its really laggy with only 40~ players.
Already got pending messages when not even close to the player amount / object count that we usually had without any issues.

[random isle]

i get pending msgs with 1 player so i dont think thats a good way of troubleshooting

[whole cloud]

but you have the fix :u

[kindred radish]

i meant player respawn sry

[whole cloud]

respawning is a clientside bug

you can still run the fixed server

[kindred radish]

true

[kindred radish]

its a good indicator tbh

if you can exclude that its only from one client ^^

[random isle]

saw it happen with just server and HC connected

[buoyant jungle]

I also had just one report of the "firing" bug previously mentioned. Server was running 2.08.149042, client that had the issue was on stable. Didn't get any repro info, but it at least seems rare. Only had it happen once during a warmup mission with smaller player count. Didn't happen at all during like 6 hours with 100+ people.

[peak pasture]

I'm still having issues with VoN not being disabled with v4 of the latest profiling branch build. Am I doing something wrong?

[rain moth]

@whole cloud v05 changelog? 🙂

[whole cloud]

nu

fix marker bug

[whole cloud]

"VoN not being disabled" ? explain what that means?

2.08.149044 new PROFILING branch with PERFORMANCE binaries, v5, server and client, windows 32/64-bit, linux server 32/64-bit
Fixed: Multiplayer Map Markers not working when client is using Profiling branch 😢

[warped rain]

Dedman has there been any reports of people not being able to join the server and just getting stuck on loading screen?

[rain moth]

I've had that issue recently with some servers. One of them had a corrupt keys folder, it was a dummy file instead of a folder. One had a rogue process which occupied the BattlEye port.

[warped rain]

Don't run battleeye but I'll look into the keys folder. Thanks

[peak pasture]

I can still speak using in game voice when using disableVoN = 1 in the server config on version 4 of the profiling branch

[kindred radish]

You gotta use disableVoN = 0;

0 = false
1 = true

[inland dew]

#perf_prof_branch message

[whole cloud]

Huh, but i didn't change that part. It was just not fully disabled but speaking was still not possible 🤔

[peak pasture]

I'm trying to disable VoN as we use tfar instead of ingame VoN. using disableVon = 0; would leave VoN enabled.

[inland dew]

once again
#perf_prof_branch message

[peak pasture]

setting disableVoN = 0; in the server config is still not disabling my ability to speak in game.

[inland dew]

reread Dedmen's message again

//

[empty goblet]

but you setting it to be enabled

=1; equals disabled

[inland dew]

not just disableVoN, but // in front

[peak pasture]

So I need to comment it out entirely to disable von?

[inland dew]

again, read it. it's all there, for the 4-5 time already
#perf_prof_branch message

[empty goblet]

you missed it was fixed already #perf_prof_branch message

[rain moth]

Set it to 1 to disable

[peak pasture]

I originally had it set to 1 to disable, I've set it to 0 to see what it would do and I've also commented it out entirely. All 3 of them do not work.

[inland dew]

oh...

[inland dew]

branch build version?

[peak pasture]

Profiling version 4. After it got fixed.

More specifically 208 149042 v4

[willow vault]

#arma3_feedback_tracker message
So, High Command disappears from 2.08perf v2 onwards up to v5 (still gone). v0 and v1 has it.

[hexed verge]

Just had my linux 32bit server binary use 700x CPU, all cpus and threads were at 100%
the players noticed nothing was happening anymore ingame, everyting froze, but no yellow or red chain.

the server rpt had tons of these, and then it just stops in the logs.

21:30:08 Message not sent - error 0, message ID = ffffffff, to 312407776 (Geordie)
21:30:08 NetServer: trying to send a too large non-guaranteed message (len=1348/1349) to 312407776
21:30:08 Message not sent - error 0, message ID = ffffffff, to 312407776 (Geordie)
21:30:10 NetServer: trying to send a too large non-guaranteed message (len=1348/1352) to 731190664
21:30:10 Message not sent - error 0, message ID = ffffffff, to 731190664 (leek)
21:30:10 NetServer: trying to send a too large non-guaranteed message (len=1348/1351) to 731190664
21:30:10 Message not sent - error 0, message ID = ffffffff, to 731190664 (leek)
21:30:12 NetServer: trying to send a too large non-guaranteed message (len=1348/1349) to 585742359
21:30:12 Message not sent - error 0, message ID = ffffffff, to 585742359 (headless)

[kindred radish]

V5 looking better
Metrics 1: https://gyazo.com/d29dc1f2922abb1d18c86f20210de186
Metrics 2: https://gyazo.com/1413dee3017c429d51bf06677608d0ac

[cold vale]

I've seen the same behavior on v5 (running windows) with projectiles spawned via script, switching to stable fixes the issue.

[eager vault]

Just making sure the config youre editing is the one specified in -config launch parameter

[peak pasture]

It is.

[uneven kernel]

Just had this for every single person who was running prof branch v5, server is on prof branch as well. Ppl not running prof didn't have an issue

[rain moth]

I think Bohemia managed to drive the adoption of 64 bit server quite a bit with the latest patch 😄

[kindred radish]

Good tbh, why still run 32 bit ?
It was the first thing i did when i started a server, updating Arma 3 & Exile to 64 bit.

[fickle geyser]

IMHO they could get rid of it. At least the server exe. And later on client one too.

[kindred radish]

V5 running for 24 hours now.
https://gyazo.com/cc84465e23cec469725886e6118ed521

CPU load is below 20 % again
Server FPS is stable
Players confirm less lag/desync but still not like before 2.08 and the prof builds.

[2022-03-13 14:13:43] [info] AEOG-ServerMetrics: Build:149044 | Runtime:133min | Players:50 | AI:16 | allMissionObjects:3545 | AllVehicles:303 | ObjSimulated:1755 | AllAliveEnt:587 | activeSQFScripts:11 | FPS:86 | Last10mAvgFPS: 87)
[2022-03-13 14:15:43] [info] AEOG-ServerMetrics: Build:149044 | Runtime:135min | Players:52 | AI:19 | allMissionObjects:3842 | AllVehicles:307 | ObjSimulated:2021 | AllAliveEnt:609 | activeSQFScripts:11 | FPS:87 | Last10mAvgFPS: 87)
[2022-03-13 14:17:43] [info] AEOG-ServerMetrics: Build:149044 | Runtime:137min | Players:52 | AI:19 | allMissionObjects:4033 | AllVehicles:305 | ObjSimulated:2235 | AllAliveEnt:626 | activeSQFScripts:11 | FPS:88 | Last10mAvgFPS: 87)
[2022-03-13 14:19:44] [info] AEOG-ServerMetrics: Build:149044 | Runtime:139min | Players:52 | AI:19 | allMissionObjects:3510 | AllVehicles:316 | ObjSimulated:1714 | AllAliveEnt:599 | activeSQFScripts:10 | FPS:88 | Last10mAvgFPS: 87.2)

14:13:45 Server load: FPS 91, memory used: 2881 MB, out: 3574 Kbps, in: 441 Kbps, NG:0, G:3038, BE-NG:0, BE-G:0, Players: 50 (L:0, R:0, B:0, G:50, D:0)
14:15:45 Server load: FPS 91, memory used: 2908 MB, out: 3527 Kbps, in: 465 Kbps, NG:0, G:6528, BE-NG:0, BE-G:0, Players: 52 (L:0, R:0, B:0, G:52, D:0)
14:17:45 Server load: FPS 91, memory used: 2923 MB, out: 3625 Kbps, in: 462 Kbps, NG:0, G:0, BE-NG:0, BE-G:0, Players: 53 (L:0, R:0, B:1, G:52, D:0)
14:19:46 Server load: FPS 91, memory used: 2905 MB, out: 6124 Kbps, in: 467 Kbps, NG:0, G:0, BE-NG:0, BE-G:0, Players: 53 (L:0, R:0, B:1, G:52, D:0)

[whole cloud]

32bit ? I missed that. Maybe the freezing issue is only on 32bit?

[hexed verge]

no idea

[whole cloud]

The stable 32bit crash thing can be either a crash or a freeze due to the same bug

[hexed verge]

today is another session, just hoping it does not hapen again

[whole cloud]

Why u running 32bit?

[hexed verge]

the box is too old

and no time to migrate

irl being a fuck atm

[slim steppe]

hey guys i have an big desync problem on my server. can anyone helps?

[whole cloud]

I told you to use #perf_prof_branch aka the branch, install it on your server, not to crosspost

[slim steppe]

ah

idk what this is

[slim steppe]

can you help me about this?

[whole cloud]

its a steam branch

[slim steppe]

ok wait

[whole cloud]

how do you install your server?

Can you just switch steam branch on it?

Alternatively you can just download the exe from google drive and use that

see pinned messages, you want the latest x64 performance server exe

[slim steppe]

no i have an server 91.210.224.218 20427

wait i stream

[whole cloud]

not sure what you mean, its hosted by some company?

[slim steppe]

Host Unlimited

[whole cloud]

can you change the server exe on there? if not then you can't fix it

[slim steppe]

?

im streaming for you say me what you mean im not good in english

[slim steppe]

yeah i can change ther server ext

[whole cloud]

https://drive.google.com/drive/folders/1PCjFnEFWa4Osz0Dftd6uZKcpQFMWC7e2
Get arma3server_x64_performance and replace your server exe with that new one

[bright flicker]

So…Irun a server and players weapons fire on their own, the server also freezes after a while, nothing in rpt got latest perf build v5

[gritty wasp]

So autofire issue is back?
We played weekend on v3 without issues reported. But fps was suspiciously high. 170+ as it was on broken 206 build with autofire issue.

[full nova]

wait so if server issues can be fixed by using Perf_branch why not putting it on stable?

[vivid rune]

NO! I running a Phenom II Server and 32bit is the only running server. 64bit doesn't start properly.

[rain moth]

is your OS 64 bit?

[rain moth]

it will trickle down to a patch for public branch eventually. perf and profiling is for testing fixes first.

[stray prawn]

since using profiling branch, my wasteland servers have issues with createTrigger, the trigger doesnt trigger sometimes after a restart. (havent had that issue with the public version)

[vivid rune]

Yes. It is Unraid OS and the minimum req is a 64-bit capable processor
This is what System Information reports: Kernel : Linux 5.10.28-Unraid x86_64

[vivid rune]

Other topic:
configuration: server: perf exe v5; clients vanilla 2.08
On Map there was many buildings destroyed. One Client left server and makes a JIP near these destroyed buildings. He said that he can observe that right after joining the destroyed buildings make the "destroying animation".
Maybe he can see this because of the 5-Frame-Update -Intervall(Maybe a little update delay)?

[whole cloud]

Who said thats not happening?

Maybe he can see this because of the 5-Frame-Update -Intervall(Maybe a little update delay)?
Shouldn't that should be handled by JIP, not by the later dynamic updating.
Every player that joins gets latest state via JIP

[empty goblet]

complex trigger or any/simple ?

[kindred radish]

👀

Almost 70 players and still ~90 server FPS (100 is our set limit)

[2022-03-13 19:55:10] [info] AEOG-ServerMetrics: Build:149044 | Runtime:111min | Players:67 | AI:26 | allMissionObjects:4097 | AllVehicles:343 | ObjSimulated:2216 | AllAliveEnt:703 | activeSQFScripts:11 | FPS:88)
[2022-03-13 19:57:10] [info] AEOG-ServerMetrics: Build:149044 | Runtime:113min | Players:67 | AI:26 | allMissionObjects:4300 | AllVehicles:339 | ObjSimulated:2431 | AllAliveEnt:713 | activeSQFScripts:9 | FPS:88)
[2022-03-13 19:59:10] [info] AEOG-ServerMetrics: Build:149044 | Runtime:115min | Players:68 | AI:22 | allMissionObjects:3935 | AllVehicles:339 | ObjSimulated:2097 | AllAliveEnt:691 | activeSQFScripts:18 | FPS:88)

Metrics: https://gyazo.com/c7b5f3fb8a4c92b334057777ab0d4a3e

[hushed seal]

running v5 now with ~70, lets see 😅

[empty goblet]

20:56:15 Server load: FPS 123, memory used: 4200 MB, out: 34020 Kbps, in: 1080 Kbps, NG:0, G:12444, BE-NG:0, BE-G:0, Players: 124 (L:4, R:1, B:8, G:110, D:1)

[strange badge]

memory used: 4200 MB
I've never seen our linux servers use anything other than 1.8GB, regardless of what's going on or how much memory is available on the server

[rain moth]

I know Linux server ran out of memory on 4GB VM when trying to start Mike Force on Cam Lao Nam

[kindred radish]

63 players (31 mins uptime)
Taskmanager says ~ 4.700 mb
https://gyazo.com/1eb22dbf96e7b6d27c6c5a3cb811d056

Contrary
21:29:54 Server load: FPS 91, memory used: 2753 MB, out: 6471 Kbps, in: 549 Kbps, NG:0, G:53, BE-NG:0, BE-G:0, Players: 63 (L:0, R:0, B:0, G:63, D:0)

So i think you cannot trust the memory stat above ^^

[vivid rune]

Yesterday Dynamic Bulwarks round for more than 7 hours: more than 8GB (thats what the docker reported) used RAM i think. (At the end the AI showed up one by one and not together. The time counter runs not in seconds anymore. It counts in 5 sec one sec. :D)

[strange badge]

We never tried mike force, but regardless of what we do it never goes over 1.8 in #monitor, on a VM with 16GB of ram

We run a server and one hc in a docker container, and the container's memory usage is reported as 3.7, so I have no reason to doubt the #monitor output

[gritty wasp]

Looks very good but suspicious. Last time high fps was associated with firing issue. After fixing it dropped

[hushed seal]

21:26:47 Server load: FPS 28, memory used: 4519 MB, out: 25608 Kbps, in: 680 Kbps, NG:0, G:8307, BE-NG:0, BE-G:0, Players: 70 (L:0, R:0, B:0, G:70, D:0)

task manager says 5837MB

[hexed verge]

I am unsure what to make out of this, my unit just played a mission with 40 players and all went well, then someone else tests their almost empty mission and this happens

23:49:10 Server: Update of object 2:510 arrived from nonowner
Setting invalid pitch 0.0000 for L Charlie 2-5:1 REMOTE
23:49:15 Object id 90037fa4 (1956) not found in slot 111,128
23:49:15 In last 3000 miliseconds was lost another 2583 these messages.
23:49:15 Link cannot be resolved
23:49:15 In last 3000 miliseconds was lost another 2584 these messages.
23:49:15 Server: Update of object 2:462 arrived from nonowner
23:49:15 Server: Update of object 2:510 arrived from nonowner
23:49:17 Server: Update of object 2:462 arrived from nonowner
23:49:17 Server: Update of object 2:510 arrived from nonowner
23:49:20 Server: Update of object 2:462 arrived from nonowner
23:49:20 Server: Update of object 2:510 arrived from nonowner

https://cdn.discordapp.com/attachments/258350836080574479/952699463598743612/2022-03-13_18-42-07.mp4

And this was with them being alone (with hc) on the server

[empty goblet]

64bit server, running for many hours, with 100+ players most of time ...

[stray prawn]

rather complex one, used in standard wasteland. the trigger is connected to a marker. when players walk into the zone of the marker, the trigger starts a script. most of the time it works, but sometimes it doesnt. havent had this issue with v2.06, some players told me that it showed up on v2.08 public also.

[cold vale]

I was able to repro similar behavior running both v5 on client/server/HC, I also got the firing bug on the client and it duplicated enough that my client is only displaying a new frame every 30 seconds. I have a vTune capture from the client and I also have some from the HC.

[cold vale]

This is weird, the firing bug also persists through a mission restart on the server.

[whole cloud]

23:49:15 Link cannot be resolved
That means it tried to get a specific object, which doesn't exist.
And apparently quite alot of times

[whole cloud]

with v2.06 you mean non-profiling 2.06?

[whole cloud]

I have a vTune capture from the client and I also have some from the HC.
Then send it to me 👀

[hexed verge]

the logs were spammed with this, 10+ each second

[whole cloud]

Can you tell me more?
It doesn't means players walk into the trigger and stay there and it doesn't fire even after X seconds? Does it fire delayed or just not at all?

Thats not present on 2.08, but on 2.08 profiling branch?
There was a change with triggers, but its also in 2.08 main build, and that one has been repeatedly fully tested in SP and MP by our QA since december 🤔

[whole cloud]

So seems like HC doesn't seem to trigger the spurious weapon firing. I retested all the old repros and also with a HC and just cannot get my weapon to randomly fire.

didn't test on linux yet tho

If anyone has more info about the freezes please send it.
A memory dump done via task manager might also help. I have no repro that works I have nothing to go on

[whole cloud]

So cannot repro with HC on windows. Any more info? is HC running on same machine? windows or linux?

[hexed verge]

Linux 32bit, hc running on same machine as dedicated server

can grab net config in a bit. currently sitting in a meeting

[whole cloud]

linux and 32bit

oki I'll try testing linux now

[hexed gorge]

😑

[autumn timber]

Didn't you pull out some statistics, a few months back when we were having a discussion about that, that said that there are more 32bit linux servers than 64bit (and the inverse held true for windows), or something like that?

[whole cloud]

Reproduced on x64 linux with HC 🥳

[whole cloud]

might be

Wow... killing the HC stops it wtf

ah no, only temporarily while the HC is loosing conneciton

Server is just spamming "You just fired your weapon" messages, why it do that.
Firing weapons or not isn't even part of the spread updates thing i did thats just positions and animations and such

[autumn timber]

Delegalization of 32bit servers and clients would make the life of some extension developers much easier.

You run 32bit arma? Bam! Go to jail! Do not pass Go and do not collect $200

[whole cloud]

Extension devs can already do that 😄
My newer stuff is all x64 only

[autumn timber]

True, but in the case where there are more 32bit servers, you're kinda going against the community, there 😔

We need an awareness campaign, first. "By running 32bit you're supporting Communism" or something like that 🤷‍♂️

[whole cloud]

But we played a 5 hour mission on that server on friday
wtf

[cold vale]

I'll have to grab the profile captures when I get home from work.

When I got it to happen I set ACE cookoff to a high probability coef (so it happens basically every time) and spammed AGS-30s in Zeus killing them over and over until it happened, I also used clumsy to simulate a bit of packet loss, etc. I think its far easier to repro when more clients are in the server.

[hexed verge]

yeah we had a 3 hour mission on the same server set up and everything, i was perplexed when i got the ping directly after the main mission ended and the mission maker trying some stuff. No restart in between. just like 20 minutes

[cold vale]

The fact that it persists through mission reloads is wild

[whole cloud]

packet loss makes sense..
But then again, during that mission on friday I was there for about one hour and my phone internet had extreme packet loss and dropouts so that really should've done it 😄

It seems like the server thinks the client hasn't acknowledged the message yet. FireWeapon is a guaranteed message that means the server re-sends it till the client acknowledges that it received it..
But my client clearly did receive it, hundreds of times.

And even when that happens, the client should compensate for when the acknowledgment to the server was lost, and ignore multiples of the same message, but thats also not happening

Its such a should be impossible situation

Well I guess just the same as negative objectID's should be impossible..

[stray prawn]

from 6 restarts a day, it shows up twice on each of my public servers. i cant reproduce it on my test server. it doesnt fire at all. the issue it shows up since using 2.08 public and profiling branch.

[cold vale]

I noticed my client RPT had a bunch of just "Overflow" message lines

[whole cloud]

Overflow is usually some object waaaay out of bounds outside the terrain

In the past it could be physx objects dropping through the floor, but that was fixed, they reset themselves to surface if they fall under

[cold vale]

Oh that makes sense, I blew up a bunch hundreds of TU-95 bears trying to repro and they flew fucking out into the middle of nowhere

[hexed gorge]

[∞,-∞,∞]

[whole cloud]

Mh.. I fixed the spurious weapon firing by turning some other feature off, that I did half a year ago.

[fickle geyser]

they reset themselves to surface if they fall under
you sure bout that?

I've seen it working when it was introduced (2.04?) but in recent months I've seen quite a bit of underground objects.

Happens with ground holders mostly.

[whole cloud]

wellllllllllll.....

The code that pops them back up is there, it also logs to RPT that it did it...

"ThingX %s was falling under terrain! Position corrected to ground [%.2f, %.2f, %.2f]. Disable simulation to keep object under terrain."

only runs if object is local, its simulation is enabled, and its vertical speed is less than -0.1m/s (so more that 0.1m/s downwards speed)

I guess yours was slower, because it fell into water...
But all objects fall into water below terrain

[full nova]

is VR even a terrain?

[whole cloud]

This code only works when object falls faster than 0.1m/s and is already 1m under ground.
So if ASL height of terrain surface is just very slightly above water, it may not exceed the 1m limit before it goes into water, which slows it down below the 0.1m/s

[fickle geyser]

Tried with a tank

in mountains. Falling pretty fast and water did not seem to slow it down.

Did not work.

[whole cloud]

uff

[fickle geyser]

yeeeeet

also this shows another bug.

Placement of objects in zeus is bad.

It does not care that object is upside down.

If you drag object that is not upright it still seems to use bottom (up at the time) land contact points.

Which makes it really easy to loose your stuff underground. In the above video I've not used ALT, just dragged the object.

[whole cloud]

I have ticketed it up for later consumption

[fickle geyser]

seemed to work for GroundHolders of dead units. Maybe it's limited to some simulation types?

[whole cloud]

its ThingX and ... should be everything that inherits from it like CarX/TankX

[fickle geyser]

welp,

[whole cloud]

Something for a different day, now I just gotta find out why my server does this

[fickle geyser]

not ackownledged firing message stuff bug?

[whole cloud]

I don't know how this causes it, but if I disable this half year old thing the bug is gone.
And if I log it and compare with how it was previously, it missmatches

[fickle geyser]

magic, gotta be something stupid

[whole cloud]

The old code certainly was stupid.
It wanted to know what kind of update a object got last time.
So every time it needed that info (every frame for every object for every player) it would iterate through all update types and check their last sent time to find the newest.

I changed it to just store which one came last when it comes in as it can simply look which was the last instead of checking all of them.
That change alone seems to be fine, the spread object updates is also fine on its own, but both together they go bonkers.
But both are kinda important performance fixes so I want to keep them

[hexed gorge]

May be it don't work on Editor, but work in game?

[fickle geyser]

Not really.

[vivid rune]

Is it possible that the server can advise the client to fire his weapon? If yes maybe there is some miss match between packets from client to server and server to client.
Something like:
Server: Client1! Fire your weapon!
Client: Server, I fired my weapon.
Server: Ah you fired weapon, i have to update the clients...Client1 fire your weapon!

[whole cloud]

Eh, no thats not whats happening

[stray prawn]

players walk into the triggers which are connected to markers on the map, usually they trigger and a counter begins. but after some restarts, those triggers do nothing. no counter, no msg, no log entry, like if they dont exist.

[fickle geyser]

Soft restarts or whole server process restarts?

In my unit we're always doing full process restarts as simple mission restarts were behaving weirdly sometimes.

[stray prawn]

whole server process restarts

[whole cloud]

┬─┬ ノ( ゜-゜ノ)

[whole cloud]

Well you'd think its probably a bit hard to find the "newest" message when literally every messages timestamp is 0 🙄
Arma!!!!!!!

[fickle geyser]

https://tenor.com/ZkLI.gif

[whole cloud]

Messages that the server sends to itself (so when server updates objects that are local to the server) always have a timestamp of zero
But that means even the old crappy code that was there before.. was broken too. For years, maybe decades.

[full nova]

basically Patch 2.08

[vivid rune]

No. basically all code around dedmen's work.

[whole cloud]

I found out why spurious weapon firing happens when you F'up your basic.cfg

The message is added from a queue, its flagged for removal by setting its error to zero, and if error is less than MinErrorToSend, its removed.

So if you F up your config and set MinErrorToSend to zero, it never removes it and re-sends it every frame.

I thought that was a great find as now I know whats going wrong but nope, linux server properly sends the message just once, and then removes it from the queue.
There is only one place that sends the message about weapon firing, and.. its not called while the client is receiving messages.
So client is receiving a message that can only be sent by one place, but that one place isn't sending anything. So where is it coming from if the only place that can send it, isn't whats sending it.

This is a real adventure dig

[rain moth]

support for basic.cfg removed in next patch? 😄

[vivid rune]

Sounds like a "Text problem" from a programmers learning book. The only thing that's missing is "Calculate the weeks, that's a normal programmer need to find the problem" at the end.

[gritty wasp]

Someone wants to be banned for being rude.

[silk pewter]

who that?

[cold vale]

Test server I replicated the bug on has MinErrorToSend set to 0.01 btw

[whole cloud]

I have now come full circle.

From that

its flagged for removal by setting its error to zero
to finding out that that code is not running so it cannot be the problem.
Over troubleshooting which other queues the message could be in, finding it in a queue, finding out what is inserting it.... landing back at
The message is added from a queue, its flagged for removal by setting its error to zero

The message's error is zero, its flagged for removal, but it has not been removed and apparently never is removed

But the code that does the actual removing later on after its flagged, has to check
if flagged then delete
else send out to players.

so it cannot happen that its flagged AND sent out.. but it does 😄

At this point i could start writing a drama story book about this bug

[cold vale]

Jesus, that's an absolute nightmare

The main question is, why the hell now is that breaking?

[whole cloud]

My gdb broke :omegalul:

[whole cloud]

Well its probably somehow caused by the new performance improvements.
But I really want to keep these as they are kinda really big

[cold vale]

Yeah, the last few rounds of performance improvements have been significant in testing.

[vivid rune]

At least one chapter in your memoirs.

[whole cloud]

And if I fix this, I can maybe turn that thing further up rather than just the 5 frames its at now.

So..
Only A can insert messages, B will send them after A inserted them and flag them for removal, and A will remove them next time it runs.

I now set breakpoints, A never runs. B runs everytime I get spurious weapon firing..
But only A can insert these messages, but A never runs, A never inserts any. What the frick

AHHAHAHAHAHHAHAAHAH

(╯°□°）╯︵ ┻━┻

I found it and i hate it

You know... When you have a comment in code that says
/// empty dummy result - returned when list not found
And you think "yeah sure, thats empty, it says there that its empty"

Is it empty? NO! Whats in it? 5 WeaponFired messages!

[edgy comet]

😮

[vivid rune]

WTF

[whole cloud]

So that thing i optimized, that checks every frame for all players for all objects whether they have updates to send?
Well if a player is not processed because it has no objects to update (because of my 5 frame spreading, there might be a frame with 0 objects for a specific player)
this dummy list is returned. And SOMEHOW this dummy list got the weapon fired messages inserted into it.

And it later tries to get the list of updates it should send, if there are no updates to be sent it just returns the EMPTY dummy list.
But ITS NOT EMTPY.

Thats also why @hexed verge's person got it on an empty server, its much more likely that a player has no objects to check for updates, if there are very few objects inside the mission.

This is technically not a new bug, this would always happen if there were 0 updates for a player.
But previously that extremely rarely happened, now it happens more often because it does less checks it overall sends less updates too.

I don't yet know whats inserting into that dummy list, nothing should be inserting into it.
Previously 0 updates was extremely rare, and you need to have 0 updates AND player firing a weapon together at the same time to catch this problem.

But something puts weapon fire messages into it.
So then every frame where a player has zero updates, it will fall back to the dummy list.
The fewer updates there are to send, the more often it will jump onto that dummylist which will replay weapon fire messages.
Thats why the firing is so sporadic, burst firing and then a bit of silence for a while

[vivid rune]

Maybe you can do a search for this dummylist and where the messages are added. Maybe the deleting saves some cycles. 😄

[whole cloud]

Found it.

over the decades there were 3 different systems on how to find out what update messages to create.
They were each time piled ontop of the old.

Instead of rewriting a thing, it was just patched to work with the new thing.

Old: Gets a empty list as input, processes all potential updates and adds them to list, grabs delayed messages and adds them too.
New: Gets a pre-filled list as input, skips potential updates because the new system already did them, grabs delayed messages and adds them to the list.

Now that list as input, is the list retrieved by the new and improved update calculator.. which may return the dummylist.
It will append to the dummy list.

That happens when you just ducttape new things over the old and miss one tiny thing

Now I just need to figure out the most efficient way to fix this.. Could clear the dummylist before its returned, but thats somewhat expensive for every player every frame
🤔

Uh interesting..
So that old code is still there, because this new WAY better method can be turned off dynamically in profiling builds with a # server command.
Well thats interesting atleast

[restive pilot]

And SOMEHOW this dummy list got the weapon fired messages inserted into it.
but why? maybe it's used when you're actually burst firing?

[whole cloud]

I wrote above why

😄

[hexed verge]

now that is an interesting error

[whole cloud]

New: Gets a pre-filled list as input, skips potential updates because the new system already did them, grabs delayed messages and adds them to the list.
It gets a list of updates, and to that it just appends these "delayed messages". Only weapon firing is a "delayed message" I don't understand why this delayed message system even exists, it seems like a extremely convoluted way to do.. nothing special.
I think its... no... No I don't know really.
It skips sending weapon fire message if a player is too far away, but you could just do that directly instead of going through this mess, maybe its a old leftover that was repurposed, its atleast 10 years old.

So it gets this list, and appends these messages.
But when that list is the dummy list (thats supposed to be empty) it inserts it into the dummy list thats supposed to be empty.
And that list is shared between all players, and its static its never cleared (because its always empty..) so after inserting it, every player that gets a "empty" list gets a list with random weapon firing messages

Now you know roughly what other programmers may do internally while publicly people say "we are working on it" for weeks with no visible progress
Luckily I'm a huge hater of not giving updates to people.

[kindred radish]

Great work, i knew some "hate" against your updates/fixes was a bit hasty.
So i say it again: THIS MAN NEEDS A RAISE !

[cold vale]

For the most part Dedmen's changes have been fantastic as they've usually only stayed segmented to the profiling branch reducing the chance of something huge breaking in stable

[vocal lichen]

Dedmens Dev Blog, I'm waiting

[whole cloud]

I have one on sqf.ovh 😄 Or I had before..

[whole cloud]

I introduced a extra system that lets me #ifdef parts of code to only be present on "experimental" builds like dev branch and profiling, so that we cannot possibly "forget" to remove a change before the next stable update (as has happened a couple times in the past before me)

[hexed gorge]

I'm exhausted just imagining the problem!

😓

[whole cloud]

Bam. Thank you, goodbye, don't come again slams door shut

[kindred radish]

More server fps porn then

Upped FPS limit to 150.
https://gyazo.com/c4fe69d0616639a56fe66029510584b0

[2022-03-14 22:21:42] [info] AEOG-ServerMetrics: Build:149044 | Runtime:81min | Players:57 | AI:34 | allMissionObjects:4351 | AllVehicles:351 | ObjSimulated:2353 | AllAliveEnt:702 | activeSQFScripts:9 | FPS:139)
[2022-03-14 22:23:42] [info] AEOG-ServerMetrics: Build:149044 | Runtime:83min | Players:58 | AI:34 | allMissionObjects:4550 | AllVehicles:351 | ObjSimulated:2546 | AllAliveEnt:714 | activeSQFScripts:9 | FPS:139)
[2022-03-14 22:25:42] [info] AEOG-ServerMetrics: Build:149044 | Runtime:85min | Players:57 | AI:34 | allMissionObjects:4237 | AllVehicles:357 | ObjSimulated:2249 | AllAliveEnt:701 | activeSQFScripts:11 | FPS:139)
[2022-03-14 22:27:42] [info] AEOG-ServerMetrics: Build:149044 | Runtime:87min | Players:55 | AI:34 | allMissionObjects:4444 | AllVehicles:348 | ObjSimulated:2432 | AllAliveEnt:703 | activeSQFScripts:8 | FPS:133)

[whole cloud]

This mysterious server freeze issue is still out there though, I still need more info about it. I have no clue what to do with it

The build ID for tomorrows profiling branch is 666 👀

[silk summit]

Just wanted to add that Dedmen is dealing with very old code which additionally was previously often updated/patched with "duck tape" to accommodate for newer requirements, frequently duplicating tasks and/or system resources usage or just being sub-optimal in general. Of course that improving these very old systems and code would understandably yield some temporary issues but it is important that we can provide constructive feedback and also understand that these situations are the price to be paid for a better overall arma experience.

I am incredibly impressed by just how much better Arma 3 feels compared to a few patches ago, never expected someone with the balls to do what was needed in spite of the dangers and I applaud Dedmen for that!

[cold vale]

Indeed, I have massive respect for anyone who has the mental fortitude to dive deep into the depths of legacy RV engine code

The fact that we're able to get > 500 server and HC fps with dips only as low as 100 is insane to me compared to the state things were in a few years ago.

[silk summit]

And I have reconnected with some single player missions and with the same hardware it feels just so much better! Old Man and Extraction in SP have suddenly become addictive after the performance updates. I know that most remaining issues are towards the MP department which is of course a lot more complex than SP but damn I love Arma's SP offerings now. (I used to only care about MP for the longest time but not anymore 🙂 )

[whole cloud]

frequently duplicating tasks
Hah! I was upgrading our CRC algorithm, and found that A3 had 3 implementations of the same thing, and all were used somewhere 😄

[hushed seal]

tell me your secrets!

[whole cloud]

Don't forget contact, there was a huge performance boost for attachTo I did recently that did like +10 fps or so on contact campaign. Which was notorious for being terrible performance before.

Actually I'm not sure if that change is on prof, but it should be

[silk summit]

Exactly! Things like that! And of course the engine would feel clunky with such things taking place all over the place but you are slowly getting rid of the clunkiness! I feel for the players that left arma 3 because the infantry felt like you were driving a tank because that clunkiness is no longer the case.

[whole cloud]

I actually did that by accident even.
Last year i spent a couple days trying to find improvements for contact campaign, I found one, it was tested on profiling, many complains, it was reverted again.
When I worked on making ViV cargo faster I just stumbled upon something that just solved that issue basically completely

[cold vale]

Current gen Intel i9 CPUs are nuts and years of tuning

[silk summit]

Thank you for remaining me of Contact! I need to try that again! Didn't have much performance issues when it was released anyway but my FPSs were on the 30's while they are on the 60s mostly. And we all know that playing Arma 3 at 60 Fps is orgasmic when compared to 30's (everything begins to work as intended 🤣 )

[warped rain]

Dedmen, love the updates, brings warm feelings to this cranky programmers

[whole cloud]

I shall now make some lunch oof

[silk summit]

I believe that since you are spending so much time so deep inside the RV engine's code that you will start to find LOTS of things that are duplicated or could be made in a much better way.

[whole cloud]

Already started doing that long ago

Now I'm almost excited to rewrite VoN 🤔

[cold vale]

Before you were even officially with BI even

I'm just glad VoN can officially be yeeted for those of us using acre/tfar

[silk summit]

Yes please! It is such a dead and unused feature (half baked) right now... and would be a much better user experience not to rely on so many mods to make Arma 3 feel complete and modern

[whole cloud]

I didn't mean that

[silk summit]

You meant optimize it but keep the same features then?

[whole cloud]

Move it from P2P to server

I guess you missed that whole thing about VoN on official servers

[silk summit]

I was dreaming of Tfar features in vanilla but moving anything from the old p2p system is a huge improvement. 👏

[rain moth]

I imagine the reason for the rewrite is user privacy 😉

[silk summit]

Yes, I wasn't following this channel so much because it is in great hands... no need to keep suggesting things when the management is on point 😉

Also, my poor english often got me into trouble. I don't know to express myself and seem polite in english yet 💩

[hushed seal]

sad i9-9900K noises

[whole cloud]

2.08.149057 new PROFILING branch with PERFORMANCE binaries, v6, server and client, windows 32/64-bit, linux server 32/64-bit

• Fixed: Spurious weapon firing in MP
• Fixed: VehicleTransport cargoBayDimensions/Dimensions/Exits positions were XZY format instead of XYZ
• Tweaked: Minor script performance improvements

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[restive pilot]

Minor script performance improvements
in where?

[whole cloud]

When a command returned a value, it would copy it into a temporary, then delete the original value, then copy the temporary onto the stack, and then delete the temporary.
Now its moved along no copying no deleting temporaries
So, everywhere.

I didn't bother testing how much difference it makes, probably not much

I needed it for the config stuff because there its quite a bit more significant

[hexed gorge]

"Fixed silly code"
"Hope this does not break other silly codes"
🤓

[silk pewter]

by the end of 2025 Dedmen will have rewritten 50% of the netcode by cascading effect

that's how we trick them

[mellow lark]

"We will now call it Arma 4 😎"

[rain moth]

... and that's how SKYNET was born? 🤔

[silk pewter]

that was when Microsoft bought Skype

[full nova]

Sky.Net

[silk summit]

Several minor script performance improvements on almost every patch end up being huge after a while 😉

[lime mauve]

Hello, how to fix red X in servers browser?

https://imgur.com/xOkJEGL

[rain moth]

update to hotfix RC or perf binary

[lime mauve]

Do you have the links? I'm really bad at programming and I don't understand much

[rain moth]

it's pinned to the channel 😛

hotfix RC info https://twitter.com/arma3official/status/1503721843114618881

Arma 3 (@Arma3official)

We're testing a hotfix for update 2.08. It addresses 32-bit crashing (please switch to 64-bit regardless ℹ️) and false version mismatches in the server browser. The hotfix should be multiplayer-compatible.

💡Details: https://t.co/sjgTO4hiMa
📥Steam access code: Arma3Hotfix208RC

Twitter · Mar 15, 2022, 1:16 PM

[lime mauve]

Thank you for your help

[vivid rune]

...and the executable would only have a size of 500 KB then.

[lime mauve]

Thanks but it didn't work, I applied the 208 and fixed the red x in the browser, but it was slow and the players couldn't connect, I went back to a previous version and it worked again but with this bug of the red x, thank you anyway for wanting to help.

[rain moth]

you should report that as feedback on the RC

perf binary contains additional fixes

[lime mauve]

my english is bad, sry

[kindred radish]

Loads of complaints about lag and kicks today with V6.
Same player amount, gonna switch back to V5.

Also lots of pending messages in the rpt (4000+ in 1.5 hours) while we had only 80 yesterday in 3 hours.

Edit: Might be related due to clients still using stable version ?!

[empty goblet]

share your basic server.cfg aka the one with networking settings ...

[kindred radish]

MaxCustomFileSize         = 0;                // Custom Face (Disabled)
MinBandwidth             = 500000000;          // 500 Mbit/s Server total guaranteed bandwidth in bits per second!
//MaxBandwidth             = 500000000;        // Slightly above total bandwidth in bits per second. (broken atm, dont use it !)
MaxMsgSend                 = 128;                 // Default: 128 Maximum number of messages that can be sent in one simulation cycle
MaxSizeGuaranteed         = 958;        
MaxSizeNonguaranteed     = 958;    
MinErrorToSend            = 0.003;
MinErrorToSendNear        = 0.01;

[eager vault]

Btw, what is the summarised rationale behind such a low MaxMsgSend default value?

[whole cloud]

Can you help us test the hotfix by putting that on your server?

[whole cloud]

on RC or on profiling branch?

I found the reason for the 100% cpu load server freezing.
I'mma try to be as fast as possibru

[silk pewter]

endless recursion or endless loop? ^^

[whole cloud]

endless loop.
Waiting for data that wasn't coming

[hushed seal]

ours is even lower at 48

rational being server seems to die less with high player numbers

[eager vault]

Hmm okay I've been going the wrong direction then

[edgy comet]

We use 320. 140 players.

[gritty wasp]

I think 640 and 180ppl
Maybe depends on CPU?

[hushed seal]

🤷 possibly, tested ours on 100 player liberation with 10900K and 9900K at 70 players

[whole cloud]

There are 7 worker threads.
First one is ID 1, second one ID 2, last one ID 7.
The code fills in work for 7 worker threads, the first one being 0, the second being 1, the last being 6.. oops.

When worker thread 7 starts first, it never grabs any work. But the flag is set saying that work is ready to be worked on.
So number 7 goes back to sleep, but in sleep it sees that the flag is set and work is ready, so it starts again, doesn't have any work, goes to sleep again.

There is a counter counting how many threads have started, to detect when all threads started up..
Threads 1-6 didn't even start yet. And thread 7 started 7 times and says "ok all threads started, all work is done, but oh also, I don't have any work to do so I'll just exit"

fun. I should've just used a bitflag instead of a counter, and I should fill in all threads..
So that means out of 7 worker threads, only 6 actually did any work

Meanwhile main thread is like "ok! I'm ready now, throw the data at me!" and waits for data that never comes
And that only happens when the last thread starts first, and the others are slow starting up

[silk pewter]

and I suppose the order was random because async?

[whole cloud]

ye

[silk pewter]

aaah, the best debug cases /ssssss

[whole cloud]

So thats solved now.. But next question, we have freeze detection, but freeze detection didn't detect the freeze 😠

[silk pewter]

maybe it froze - make a freeze detection freeze detection!

[hexed verge]

oh do tell, another dedmen blog post pls

and now i see that discord has not scrolled down

[whole cloud]

indeed

You could also say that you are faster at replying than at reading

[hexed verge]

I always am

shoot first, then ask questions

In the end i am glad this was caught and fixed.

[whole cloud]

more will be coming on the road towards better multithreaded ARma

https://cdn.discordapp.com/emojis/826517106308284438.gif?size=128&quality=lossless

of course now when I test freeze detection it works -.-

[hushed seal]

Schrödingers freeze detection

[autumn timber]

!namedebate

[frozen tundra]

The correct name is Arma 3. See: https://www.bohemia.net/blog/arma-2-the-name-tale

[whole cloud]

[silk pewter]

someone remove his image posting rights!!

[brave barn]

https://tenor.com/view/absolutely-not-nope-no-no-way-no-chance-gif-17243246

[silk pewter]

hue hue hue

[worldly badge]

!burgerpan Dedmen 10e34 Big frown face

[wispy shuttle]

Does the 2.08 hotfix includes the Linux multi-threading optimisations ?

[silk summit]

It seems that the recently released Direct Storage tech for Windows 10 and 11 greatly reduced CPU utilisation while enabled while also speeding the loading of textures and other assets. I wonder how hard/possible would it be for Arma 3 to exploit this new feature. DEMO VIDEO 👉 #videos_arma message

[vivid rune]

As i know there are only in the profiling/performance branch active for now.

[silk pewter]

5 minutes and a Dedmen on steroids

[silk summit]

😂 isn't he already on steroids like... all the time? 😆

[full nova]

if you give 1000L of steroids to him, he can make Arma 4 in about 2 weeks with RV4

[rain moth]

see dev channel yesterday I believe 😄

[worldly badge]

https://imgflip.com/i/68vfuo

[silk pewter]

I win

[hushed seal]

#ip_rights_violations!!

[silk summit]

https://devblogs.microsoft.com/directx/directstorage-api-available-on-pc/

And now there is also a GPU Direct Storage: https://www.youtube.com/watch?v=Aj1c4YaV-tQ

[silk pewter]

as long as it is DirectX11 compatible…

@silk summit ↑

[silk summit]

If it isn't Dedmen will just update A3 to DirectX 12 in like 4 minutes 😉 😜

[silk pewter]

prepares moar coffee/tea/drugs

#define DIRECTX_VERSION 11 // set to 12
```done!

[silk summit]

Ok so it seems that it does require DirectX 12 too... so bye bye Direct Storage for A3, so we should just focus on Dedmen Storage and DedmenX for the win! (Source: https://www.howtogeek.com/785339/what-is-directstorage-on-windows-11-and-xbox/)

[whole cloud]

no thats profiling branch exclusive

[whole cloud]

Arma 3 loading textures is already so fast that its basically irrelevant

[whole cloud]

DX12 doesn't run on windows 7. So thats already out

[silk summit]

Yeah, it was the CPU reducing part that was interesting... so that A3 gets more CPU and Direct Storage gives more "jobs" to the GPU... We all know how underused the GPU is for A3... 😆

[whole cloud]

Arma already does texture loading in extra thread, so that wouldn't change anything for Arma's main thread bottleneck

[silk summit]

75% of all steam users already run windows 10! Force the 10% (it is 4% actually) on windows 7 to upgrade 😉 (source: https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam)

I hope BI pcs aren't on windows 7 still 😅

[random isle]

they work on punch cards 😄

[silk summit]

And of all the windows 7 pcs, 90% of them are banned russians anyway, and the rest Chinese...

[whole cloud]

Tell that to paying customers who paid money for running it in windows 7

[silk pewter]

not the matter here
Arma 3 has been sold as Win7 supported, it stays so

[silk summit]

Ok found the detail... only 4.08 % of all steam users use windows 7

I would be too hard to implement anyway! Lets leave all that for Enfusion 😉

[random isle]

and there are 0.1% of steam users that play arma 3

[rain moth]

add new arma3_x64_dx12.exe binary 😎

[silk summit]

True but there are even bigger fishes to fry and Dedmen's plate anyway... and he shouldn't focus on competing with enfusion...

[rain moth]

I'm glad Arma 3 uses at least DX11 so that Auto HDR works 😄

[silk pewter]

what about Windows 98 and DX7 (ok, I am generous and aim for DX8)

[sick arrow]

Arent there few games on Win7 with DX12?

[whole cloud]

DX12 is windows 10+ exclusive

[silk summit]

Well that is also true... just because you make your game DirectX 12 compatible that doesn't mean it automatically isn't compatibly with older versions of the API.

[whole cloud]

I was very sad when I got the latest hitman, installed it and then learned it doesn't run on 8.1 :sad:

[rain moth]

what kind of weirdo still runs windows 8(.1)?

[sick arrow]

I know that feel when i had to upgrade pc from XP because Battlefield 3 in 2015

[silk summit]

what kind of weirdo still runs windows 7!?

[whole cloud]

Cheap nitro 😄

[rain moth]

lol 😛

[silk summit]

Windows 8.1 was't as bad as Vista. Base Windows 8 was terrible thou!

[rain moth]

vista was mostly crap drivers by vendors

[silk summit]

Windows 10 was supposed to be the last windows and that It would just get updates forever but then microsoft got greedy again. Thankfully no one is upgrading. The 15% of Windows 11s on Steam Charts is just new pcs and notebooks...

[silk pewter]

Windows Arma Edition when Dedmen?

[silk summit]

You meant Steam Deck Linux Arma Edition? Right? 😅

[silk pewter]

ARMacOS*

[hushed seal]

And people getting upgraded because they didn't read a prompt :D

[sick arrow]

Good chunk of people cant upgrade even if they want to. Hardware must be new post 2017.

[silk summit]

Thankfully I didn't met the requirements anyway...

[sick arrow]

Only 2 people i know can upgrade but they wont.

[silk summit]

And that end up being a blessing! 😅

[vale shoal]

@silk summit I'm one of these 15%, on purpose :D and really happy with it

[silk summit]

But you gained FPSs or you just like the new UI?

[vale shoal]

@silk summit I really didnt felt any difference between Win10 and Win11 - and the new UI is still buggy, but I installed it to be prepared for the future (I work as an IT Analyst so we might come across with it soon)

[celest sparrow]

I upgraded, mostly just so my win10 license turns into win11 license

indeed, UI is missing features is my biggest problem with it

I upgraded tablet, that doesn't have the swipe in from right options like lock rotation/etc

so that may need to get reverted

I am looking forward to the tabbed file explorer though

[silk summit]

Yeah, beta testing a rushed OS doesn't seem fun enough. I will do what I always do and wait for the performance to be finally better than W10 (Or Enfusion forces me to upgrade!) 😅

[rain moth]

Auto HDR was reason alone for me, then they backported it 🙃

[loud jungle]

👋 Moved to v6 today had really low server FPS but no crashes or pending messages (which we usually have with that)

Server was averaging at 10FPS but no reports of desync, things not working, or bad client fps

Snapshot of logs;

20:47:18 Server load: FPS 8, memory used: 4361 MB, out: 14466 Kbps, in: 573 Kbps, NG:0, G:25915, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:23 Server load: FPS 6, memory used: 4360 MB, out: 13907 Kbps, in: 558 Kbps, NG:0, G:26061, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:28 Server load: FPS 12, memory used: 4362 MB, out: 14939 Kbps, in: 563 Kbps, NG:0, G:31918, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:33 Server load: FPS 8, memory used: 4363 MB, out: 14197 Kbps, in: 546 Kbps, NG:0, G:11969, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:38 Server load: FPS 12, memory used: 4365 MB, out: 17442 Kbps, in: 591 Kbps, NG:0, G:33016, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:43 Server load: FPS 9, memory used: 4364 MB, out: 14473 Kbps, in: 577 Kbps, NG:0, G:40072, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:48 Server load: FPS 8, memory used: 4365 MB, out: 14942 Kbps, in: 585 Kbps, NG:0, G:12417, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)
20:47:53 Server load: FPS 12, memory used: 4367 MB, out: 15000 Kbps, in: 577 Kbps, NG:0, G:19856, BE-NG:0, BE-G:0, Players: 73 (L:0, R:0, B:0, G:73, D:0)```

[vivid rune]

How long was the Mission running?

[loud jungle]

At that point, 2 hours

1hr and 27mins with players on

[hushed seal]

How is that server not dying

[eager vault]

What's your basic.cfg?

[gritty wasp]

Can we expect v7 with 7 threads today?

[whole cloud]

7 threads?

no

v7 yes

[gritty wasp]

I mean fixed endless loop with lonely 7th thread that wait for its job

[whole cloud]

it may not be 7th thread depending on how many you have 😄
Yes of course, why would I not publish the fix for the critical bug I have been working on fixing?

[empty goblet]

on the 7th day the code was fixed and he could rest

[loud jungle]

MinBandwidth = 131072;        // Default
MaxMsgSend = 640;            // Official BI servers = 640
MaxSizeGuaranteed = 512;    // Default
MaxSizeNonguaranteed = 256;    // Default
MinErrorToSend = 0.001;        // Default
MinErrorToSendNear = 0.01;    // Default```

[loud jungle]

It was like this for 90 minutes and played as normal 🙃

[whole cloud]

"MinBandwidth = 131072; // Default"
Are you sure that your server only has 0.13mbit/s upload speed?!
Also thats is not the default

[loud jungle]

...we didn't remove the comments because we were lazy

We have spent months moving basic.cfg around and that's where we landed to reduce desync and network pending messages

We're on a 1Gbps line

[whole cloud]

If you are on 1gbps, why did you set it for 0.1mbit/s

[loud jungle]

When we were setting it higher we'd see a lot of pending messages which was causing desync/stutters

We've been down from 250mbit/s all the way to there

Bear in mind, probably not the best test case because of mods but that's something we're going through this month to remove unnecessary scripts

And have been doing for a while 😅

Dedman, what this? 21:27:27 Overflow

[whole cloud]

probably object way out of bounds, outside of map or way under map

[loud jungle]

Ah ok, first time I've seen that in our logs

[stray prawn]

neither the profiling v6 or the hotfix, fixes the trigger issue, i mentioned some days ago. the trigger issue also shows up, after about 2 hrs play time.

[whole cloud]

[empty goblet]

build eta?

[whole cloud]

time

[silk pewter]

hah! no

[empty goblet]

but he need to rest so he can resume next workweek 😁

[silk pewter]

hmmm… okaaay… but only because the Law says so!

[empty goblet]

so many more bugs to fix

server-side VON ... decoupled renderer and other wonders from code land

vanishes into shadows as puff of smoke ...

[silk pewter]

DirectX11v2, lightsource shadows, PBR, stable 144 FPS on GeForce2 MX… sorry, I was having a wet dream

[empty goblet]

you skipped to fairy code land

[silk pewter]

but honestly, "just" the MP performance improvements (AI local calculation etc), 3fps bugfix, crashfixes, it's already a big lot to many ❤️ so, yeah, ❤️

[empty goblet]

getting there ... sometimes i wish there was way where the engine would toss warrning into RPT like maxmsg too low / too high and same goes for some other basic cfg / network settings

[brave barn]

it would be amazing

[gritty wasp]

Or benchmarking tool like speedtest. Press X to win and make config

[empty goblet]

if it was that easy ... it would be done already 😁

[gritty wasp]

Sweet dreams

[whole cloud]

2.08.149072 new PROFILING branch with PERFORMANCE binaries, v7, server and client, windows 32/64-bit, linux server 32/64-bit

• Fixed: Crash when inter-creating ropes between transports - https://feedback.bistudio.com/T163714 -
• Fixed: Rare server freezing

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[whole cloud]

Comparison to the "before"

[silk pewter]

and without that code block? 😄```cs
Parallel.Run({ while (true) {} });

j/k, well done on MT!

[tawdry gazelle]

So what should we use as suggested value actually? According to https://community.bistudio.com/wiki/Arma_3:_Basic_Server_Config_File, MinBandwidth should be default to 131072 and the server runs well with it... Also the page is quite confusing... and we cant tell difference between different values at all...

[hexed gorge]

@tawdry gazelle this (131072) is also the default for some Game Server services i used, but i allways change it to something like 80% of the bandwidth.

[whole cloud]

The default is 8mbit/s for dedicated servers

[hexed gorge]

It's ok to update just the server to v7?
I don't want to force clients but if difference is big, i will force them 😂

requiredBuild = 149072;

[autumn timber]

I must have missed what is currently being tested here (I mean, besides the obvious core usage).
What is it that you managed to parallelize?

[silk pewter]

parking

[violet elbow]

funny

[wispy shuttle]

Hum, with arma3server_x64_performance_208_149072_v07_linux, I can't connect to the server (required version 2.06) (client has the latest stable 2.08 + hotfix)

[full nova]

how many Version until we get back to V1?

[tawdry gazelle]

That means I should place 8*1024*1024/8 = 1048576 as the default value? Ah that wiki page could be the most confusing one in the whole wiki...

May I ask what does official server runs, all of those basic.cfg parameters?

[whole cloud]

Profiling always has been compatible with stable, so yes ofc

[gritty wasp]

8mbit -> 8 * 1024 * 1024 = 8388608
or 8000000

[whole cloud]

That is the allMissionObjects script command

[whole cloud]

Required version 2.06? Well if your server is on legacy then you can't connect with current version

[whole cloud]

No... The default is what's set if you set nothing.
If you want the default you should comment out your whole line

Also the number is in bps not bytes

[tawdry gazelle]

Okay then... will use the default ones...

[vivid rune]

v07 Linux server 32 bit more than 3 hours same mission stable 45-50fps around 5-10 player

docker says 2.2GB, #monitor something around 1,5GB

(If you want stress your traffic, than use the minigun from niarms. Alone this weapon generate 1 mbit of traffic on firing) 😛

[hexed gorge]

Good day!

[whole cloud]

allMissionObjects
Singlethreaded: 23397 ms
Old Multithreaded: 3598 ms
New Multithreaded: 3202 ms

All of them do the same work and produce the same result.
But the best part is just looking at this image.

And the new stuff I can just reuse elsewhere, for example nearestTerrainObjects 👀

Trying to find all objects on whole terrain
without sorting
Old: 24.9512 ms
New: 202.6 ms

With sorting
Old: 47.7619 ms
New: 162.429 ms

:poutcat:

[restive pilot]

old was better?!

or is that typo?

[whole cloud]

ye, makes sense though.
This work is just too cheap.

[silk pewter]

more time but less freezing I guess

[whole cloud]

no more freezing

allMissionObjects spends most of their time filtering out, and it filters out about 99% of thingies (in an empty mission)
terrain objects though, filters out less than 1%, and filtering is extremely cheap.
The bottleneck is putting it into script values, and that can only be done by one thread.

[restive pilot]

btw isn't allMissionObjects too slow still?
I once tried it and it was like ~3 ms not 3000

or maybe my mission was too light...

[whole cloud]

my numbers are for 10k runs

[restive pilot]

oh then that's amazing

[whole cloud]

But I have a different idea for terrain objects, lets see what it does

Unsorted
Old: 74ms
New: 74 ms

Sorted
Old: 192.333 ms
New: 122.667 ms

Well

[silk summit]

Soon I will be able to buy that 140Hz monitor and make use of it in Arma 3!

[restive pilot]

as long as I can use my 60Hz monitor at full framerate all the time I'll be happy 😛

[hexed gorge]

Hope it's not a silly question. A server that is limited to 2 cores (Windows CPU affinity) will also benefit from that if i keep it limited to 2 cores?

[whole cloud]

if you set -cpuCount=2 yes

If you limit it via affinity, without setting cpuCount and Arma still thinks it has all cores, it will hurt performance alot

[iron saddle]

Noticed a regression for ropes: https://feedback.bistudio.com/T163780

[vivid rune]

Is copying/moving from one script value to another same expensive in time consume? If not, maybe one thread can prepare it in advance(at the end of calculations for frame 0 start it the preparation for frame 1 and if some change is happening for the terrain objects it discard the preparation and start again)?

[whole cloud]

no sense in wasting that much ram.
If you wanted to do that you might aswell predict the whole planet and write every possible state to a disk

[vivid rune]

If every second function needs the value, it makes sense. And something to do in advance if you have the resources is also not bad. I don't know how much RAM this consumes but for me it seems the only way to "accelerate" a single thread task. But i am not an expert in these things. Maybe there is an another solution.

[whole cloud]

If every second function needs the value
They don't.

[vivid rune]

OK. 🙂

[rain moth]

Will Linux check compute cores via nproc in the future?

[whole cloud]

I don't plan to switch that for now

[vagrant zodiac]

Anyone else reported or experiencing significantly worse performance since the midweek update as well as reduced stability?
I'm down over 20fps on last week's version and have had more crashes in 1 day (yesterday, the only day I've played since the hot fix) than I have all year.

Just doing some checking to make sure it's not something else other than the update but think that's the only thing that's changed.

[vivid rune]

Which executable has your server? Vanilla (with Hotfix) or Prof/Perf executable v07? And which OS do the server use? 32bit or 64bit?

[vagrant zodiac]

64bit and latest perf exe

[whole cloud]

If you have crashes then please report them

Can't fix what I don't know about

[vagrant zodiac]

Feedback tracker?
Just doing some more checks to make sure I've not done something silly first....

[whole cloud]

or just via DM to me the mdmp file

[feral harness]

next prof

[inland dew]

Same for me.
On current perf, 7 FPS min and avg lower than on 2.06.

On 2.06 5 FPS min and avg lower than on 2.04.

Each new patch FPS gets lower and lower

[gritty wasp]

Player fps or server?

Yesterday we had 246 avg fps at 132 players. It maybe was even higher but cap set at 255.

One month ago we had 64 avg at 141 players on same mission

[vagrant zodiac]

Interesting to know.
I’ve not properly benchmarked for some time but have used the same training map for a year or more so if I’m on that by myself, I have a rough guide to performance.
Thankfully, I’ve not been able to replicate the crashes although will do more work on Sunday. But hopefully the stability issues were something at my end.

[whole cloud]

Yesterday my server had issues too, but hard to guess what it might be

[whole cloud]

I’ve not been able to replicate the crashes
But you got crashdumps from the old ones right? so just send em

[vagrant zodiac]

Will do when I’m online properly. (Only on mobile at the mo.)

[inland dew]

Client FPS in singleplayer (YAAB)

[hexed gorge]

@gritty wasp even 64 avg one month ago is impresive for me. What processor your server have?

[gritty wasp]

Intel 8700 I think

[hexed gorge]

Hmmm. No AI or custom buildings?

[gritty wasp]

No ai. With 130-200 people you can play tvt without bots

[empty goblet]

i think the important detail is if you also tell us your actual fps, average fps and min fps ...

[inland dew]

~84 FPS min and ~113 FPS avg with 2.04 1080p standard video settings preset
~79 and ~107 FPS patch 2.06
~73 and ~ 100 FPS current 2.08 perf build

So at least 10 FPS min and avg lost from 2.04 stable to current 2.08 perf.

It's actually even slightly more than 10 FPS min and avg less

It's not Windows or drivers, because if I switch to legacy, I have same FPS as before

[inland dew]

and don't think that it's not much, because it's only 1300 m for objects and 1600 m for terrain in 1080p standard.
if you increase the view distance to something like higher than only 1300 m, FPS is much much lower and when FPS, because of the view distance is not high at all, 5-7-10 and even more FPS is not nothing

it's not like having 5-7-10 or slightly more FPS less, when you already have like 120 or even more

[vivid rune]

Can you run MSI Afterburner or comparable running for checking if CPU or graphics bottleneck? What are your graphics card?

BTW: I was playing a round on Antistasi and the Mission bottlenecked my potato CPU and not the graphics card (normaly in not so heavy scripted missions). In Windows Resource Monitor my CPU was really good utilised. (70-80%) and the graphics card around 70% (normally 100%)

[inland dew]

I was playing Warlords on current perf build for many hours, had good FPS, it was smooth and a lot of memory unused and Arma simply closed/crashed

nice...

all my money, all the defenses and jet and cover etc lost

no mdp, just rpt and that's the only stuff that's there, from 1-2 minutes before Arma simply closed/crashed and until it crashed/closed
nothing else
20:29:00 Client: Network message d916f (update info cc8ca9d8) is pending
20:29:05 Client: Network message d9311 (update info cc8ca9d8) is pending
20:29:10 Client: Network message d94bc (update info cc8ca9d8) is pending
20:29:15 Client: Network message d965a (update info cc8ca9d8) is pending
20:29:20 Client: Network message d97fc (update info cc8ca9d8) is pending
20:29:22 Setting invalid pitch -0.0000 for C Charlie 3-4:1 REMOTE
20:29:22 Client: Object 711:80 (type Type_124) not found.
20:29:22 Client: Object 711:80 (type Type_96) not found.
20:29:22 Client: Object 711:80 (type Type_461) not found.
20:29:25 Client: Network message d99a7 (update info cc8ca9d8) is pending
20:29:30 Client: Network message d9b7e (update info cc8ca9d8) is pending
20:29:35 Client: Network message d9d37 (update info cc8ca9d8) is pending
20:29:40 Client: Network message d9ee9 (update info cc8ca9d8) is pending
20:29:45 Client: Network message da0c3 (update info cc8ca9d8) is pending
20:29:46 Client: Object 728:13 (type Type_96) not found.
20:29:46 Client: Object 728:13 (type Type_124) not found.
20:29:46 Client: Object 728:13 (type Type_461) not found.
20:29:50 Client: Network message da259 (update info cc8ca9d8) is pending
20:29:55 Setting invalid pitch -7876488984526848.0000 for C Charlie 3-4:1 REMOTE
20:29:55 Client: Network message da3cb (update info cc8ca9d8) is pending
20:30:00 Client: Network message da52e (update info cc8ca9d8) is pending
20:30:05 Client: Network message da69a (update info cc8ca9d8) is pending

---

[spiral pond]

Warlords doesnt have persistence for a session?

[vivid rune]

6 hours bulwarks 5 Player #monitor: RAM 2090MB and dropped FPS to 12FPS at the end wit many placed Elements and AI (Linux 32bit server). At the End one Thread was 100% most of the time.
One Question: Can the linux 32bit address motre than something around 2GB of RAM or is that a limit?

[clear totem]

Hello
I have been using Custom Compositions in this branch
Sometimes the Comps I make are greyed out and unavailable in Pub Zeus matches
Shuffling around with other comps can lead to it either working or not

[rain moth]

2GB, 3GB with PAE. Use 64bit.

[vivid rune]

PAE Physical adresss Extension)? The server runns in docker container. Do the docker have to support this or the "real/host OS"? And would use 64bit but my Phenom II quits with error message the boot up. Do I have to activate PAE somewhere in the Arma parameters?
Edit: I found maxMem parameter. I will try.

[empty goblet]

it never had

[feral harness]

can have now with new namespace

[spiral pond]

uiNamespace on the server should have been good enough to handle people crashing/disconnecting, no?

[hushed seal]

does diag_captureFrame work with the performance binaries?

[silk pewter]

Only available in Diagnostic/Profiling branch(es).
https://community.bistudio.com/wiki/diag_captureFrame

[hushed seal]

I've read that and found it unclear, is that a yes or no?

[silk pewter]

I don't know 😄 I am unsure about which exe are available in perf branch
I'll make it clearer when I get confirmation (IDK between perf/prof)

[hushed seal]

Good to know that I'm not the only one :D

I'm guessing no but the wording can be interpreted both ways

[pallid island]

Only with profiling exe

Which is included with perf/prof branch

[inland dew]

"good" perf build

[inland dew]

after having played Warolrds on EU01 for not even 1 h

[fickle geyser]

Yeah, could've been done like always. New mission namespace just makes it simpler, was always possible.

[hushed seal]

thx 😦

[rain moth]

both, perf is default (replaces regular), profiling is separate binary which gets added to arma 3 folder

the branch is called profiling, not perf 🙂

[silk pewter]

okay, I'll add the exe distinction
thanks for the precision, I wasn't sure anymore!

is there any other way to have a diagnostic exe?

[whole cloud]

dev branch has diagnostic exe

But I don't know if that actually has profiling

[silk pewter]

rhaaa I am not thinking about it today, it's Sunday! 😄

[whole cloud]

thats a good thing to do

[silk pewter]

no
Elden Ring is

[inland dew]

and again, on current perf build, Arma simply clsoed/crashed with very good FPS and a lot of free memory while playing Warlords on EU01

can't play like this

checked the integrity and all files are fine

send already 2 times today rpts, bidmps and mdmps to Dedmen

won't play anymore until next perf build, because it's very frustrating and tiring to spend a lot of time to earn points, arrange defenses and gain superiority vs. enemy and being afraid every second of the gameplay for Arma to close/crash for no apparent reasons at any moment

and then restart the game, reconnect, regain all the points necessary and have the game crash again and like this many many times the same day

[warped rain]

It's a free choice?

[mellow lark]

Or play stable build?

[hexed gorge]

It's so sad on weekends, I come here and I don't find any bombastic news.

[restive pilot]

yeah

[kindred radish]

Today it was the first time i ever was able to air lift a vehicle with a helicopter without any desync for the ropes or the towed vehicle.
Ropes were perfectly attached to the vehicle and made no snapping attempts to desync (was always a nightmare).
Other players confirmed my experience from their view. I was amazed 👀

Weirdly enough for a AI heli the lifted crate does this -> https://gyazo.com/d7703a7dbba491711b2f67c5df01bdb6
It "desyncs" in front of the heli.

[whole cloud]

I added that when towing was added, everything you attach with ropes tries to change its locality to become local to you. Like when you get into a driver seat in a car.
Thus both the helicopter and the box are local to you and there is no desync

It "desyncs" in front of the heli.
Thats probably because the AI is remote and the crate is too, so you get what they're sending there

[kindred radish]

It was never that smooth before even when the vehicles was local to me. Even for other players watching it was super smooth while it was never like that in years (no matter how "heavy" the mission was).
I can live with the other "issue" 👍

Some more feedback to latest prof build.
It's running great for us, so far i didn't see/got reports about any major problems besides the more frequent kicks due to signature mismatch.
Most of the players report that it runs smoother for them and server FPS was never better.

https://gyazo.com/06940e3595ffe887104316afa56ea752

[hexed gorge]

@kindred radish what this server is running https://gyazo.com/06940e3595ffe887104316afa56ea752 ? Exile?

[kindred radish]

Yes

[bright latch]

Ran v7 on the weekend, no issues, all nice and smooth 🙂

[strange badge]

is profiling safe to run on linux again? haven't been keeping tabs on it for a few weeks

[vivid rune]

had no issues with v7

[strange badge]

thanks

[novel ravine]

@patent dome
Sorry for the ping, but I was talking to Dedmen earlier about server freezes related to 2.08 and he told me to get in touch with you but your DMs are blocked. He said you were experiencing similar problems but figured something out. We're on an OVH Canada machine and using the v7 Perf Branch (but had this problem on both v5 + v7 branches). Never had this issue until 2.08 and haven't uploaded anything or made any changes other than swapping out the Performance .exe's. If it's easier/better to sort out in private, feel free to DM me. 👍

[patent dome]

Ours aren't specifically related to any recent profiling branch update or even 2.08. We have been experiencing server crashes on one of our server boxes since 2.06v8 and we have been trying to figure out what the issue was without success so far. My main theory was related to SSE instructions that were introduced around that time increasing our server load, but even after our server host fixed a broken CPU fan we still experienced issues. At this point we are trying to figure out if it may be related to the OS being Windows Server 2016, while our other server runs 2019 and can run multiple profiling instances just fine.

We are currently trying to figure out some other issues before we can update the OS, but my best guess our issue comes down to corrupt drivers or something.

Not sure it is of much help.

[hexed gorge]

@novel ravine do you run Vanilla Arma 3?

[novel ravine]

Hmm, interesting. Not sure of how much help that is either. 😅 We are also on Windows Server 2016 though. The problem we experience is that Remote Desktop + IPMI both seem to suddenly die, but power is still clearly on and requires a hard reset. There's nothing in our RPTs that would suggest anything going haywire, it just stops writing because the machine locked up. If you've got any advice on diagnostics I should try running to get more info, I can do that. 🤷‍♂️

[novel ravine]

No, we're an Exile server with a few additional mods too.

[patent dome]

Exactly the same issue we experience.

We tried everything from network logging, process explorer logging, dbgview logging. All of them either suddenly stopped logging or the files got corrupted because it stopped in the middle of dumping to the logging file. Also nothing in the event viewer from the moment RDP stops working up to the point we perform the hard reset.

We have tried every form of diagnostics we could think off for a couple weeks, with nothing producing anything other than showing us it just suddenly stops

Good to know you also run 2016 though, at least something that could give a hint as to what causes it.

[novel ravine]

How often have you experienced this? We've only had this happen 4 times total so far. I'll post dates and rough times of when this happened in a bit to see if there's anything somehow related.

[patent dome]

For us it happened every op, and we run daily ops. There did seem to be some correlation between the amount of players and the time it took before it froze, but there were no signs of a memory leak and when our cooling fan got fixed there were no more overheating issues either.

[gritty wasp]

Memory corruption?

[patent dome]

Possible, but haven't figured out a way to check for it as we lose all access when the freeze happens.

[gritty wasp]

Maybe run Testmem5 with some presets. On GitHub you can find MemTestHelper article

[novel ravine]

Doubt this helps, but here's a rough estimate of when our freezes happened:

Freeze #1 = 03/14 around 1:20am EST (GMT-5)
Freeze #2 = 03/17 around 6:50pm EST (GMT-5)
Freeze #3 = 03/19 around 11:20am EST (GMT-5)

If there's anything you recommend that I get set up to start logging something let me know. I'm pretty clueless on this kind of stuff tbh 😅 I have not reached out to OVH yet though. I suppose I can ask them if they've seen anything weird around those times?

[whole cloud]

2.08.149116 new PROFILING branch with PERFORMANCE binaries, v8, server and client, windows 32/64-bit, linux server 32/64-bit

• Tweaked: Multithreading tweaks
• Tweaked: Slightly improved nearestTerrainObjects performance when sorting is enabled
• Fixed: Some modded AT launchers could not switch their fire mode
• Fixed: Cruise control would not disengage when the engine is turned off - https://feedback.bistudio.com/T163709

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[jagged minnow]

How much of a performance impact do the multithreading tweaks have in avg?

[restive pilot]

I guess these

there was another one... can't find it

[patent dome]

This one?

That is specific for the second point

[jagged minnow]

Oh kk 😂 I thought this was the possible fix for the CPU usage bottleneck of the engine. But I guess this would be too good to be true😩

Nevertheless good tweaking!

[hexed gorge]

Nice! allMissionObjects is just a example, the inprovement is in script in general?

[restive pilot]

¯_(ツ)_/¯

[mellow lark]

I believe it's script specific. Iirc you don't just flip one switch and make everything multithreaded lol

[hushed seal]

And IF that would ever happen I'm sure it would be more than a side note in a Changelog :D

[vivid rune]

"- Tweaked: Slightly improved nearestTerrainObjects performance when sorting is enabled"
37% faster. In Dedmens Eyes this is "Slightly"?

[iron saddle]

Not fixed in 2.08 v8

[hexed gorge]

Yes, for sure, but if all functions are already multithread, may be a fix or improvement can make all of then better.

[whole cloud]

(almost) nothing in SQF is multithreaded

[mellow lark]

Yes

[hexed gorge]

nearObjects, nearestObjects, nearEntities, entities worth a go to multithread?

To detect simple objects i use nearestObjects [_pos,[],_dist] what make it find all objects. Than i check the result for the ones i want.

🙏

[whole cloud]

basically only worth it for things that take over a millisecond to run

[hexed gorge]

Thanks for taking time to answer.

[vivid rune]

we played v8 on a linux 32bit potato server. It seems that we experienced sporadic stuttering weapon fire. clients are vanilla

(one time per player (2 players))

[gritty wasp]

Here we go again meme

[feral harness]

Trying to find out what happened

[whole cloud]

Dedmen forgot to merge it

[whole cloud]

puts on ignore list

[silk pewter]

https://tenor.com/view/the-northman-northman-reasonsimbroke-willem-dafoe-focus-features-gif-24200263

[full nova]

https://tenor.com/view/qa-cat-leak-flood-broken-pipe-gif-12195496

2.08 bugs be like ^

[whole cloud]

https://media.discordapp.net/attachments/896493185180782662/932097603586387988/Tumblr_l_46540248590577.gif

[vale shoal]

is anyone of you using Nextcloud on a machine where Arma 3 processes are running? I'll post it here because I use perf branch. We encountered a strange problem where the Arma 3 processes using 100% of the CPU when Nextcloud is performing syncs. After the sync, the cpu usage stays for some minutes and then it suddenly drops back to "normal" values.

[vale shoal]

Normal CPU usage while doing nothing:

https://abload.de/img/normal-cpu-usage32kpo.png

https://abload.de/img/normal-cpu-usage-monip8jh9.png

High CPU usage when Nextcloud is doing syncs:

https://abload.de/img/high-cpu-usage0rj1q.png

https://abload.de/img/high-cpu-usage-monitop2k8h.png

winver:

https://abload.de/img/winverwtjxx.png

(also occurs on Windows Server 2019)

[vivid rune]

what is nextcloud CPU usage occupying in this scenario?

[vale shoal]

Less then 5%

[gritty wasp]

Let me guess. Network bandwidth occupied by syncing. Arma lose packages and trying to fix desync generating new packages with extra cpu load

[tawdry gazelle]

Sounds convincing

[whole cloud]

Arma will try to resend old packets when not ack'ed
Sounds reasonable

[vale shoal]

Sounds legit. Is it fixable?

[silk pewter]

(I would guess Nextcloud-side perhaps?)

[hushed seal]

Limiting network usage by Nextcloud should fix that I guess, there's software for that on windows.

[whole cloud]

@novel ravine @patent dome
Maybe try setting -cpuCount= such that your server has two free cores?
That would help to tell if its related to the multithreading changes.

Other option would be that since recently game is flagged as a "Game" process, maybe that somehow iffed it up, but thats also on stable (I think?)

[patent dome]

Will try, also going to run Dev on the server to see if the issue persists on there. In case it is some Prof exclusive change.

[whole cloud]

2.08.149135 new PROFILING branch with PERFORMANCE binaries, v9, server and client, windows 32/64-bit, linux server 32/64-bit

• Tweaked: Disabled helicopters exploding when turned upside down
• Tweaked: Several adjustments for vehicle towing
• Fixed: Creating a rope from a vehicle and to the same vehicle is enabled again
• Fixed: Invalid "setting invalid pitch" log messages - https://feedback.bistudio.com/T163807
• Fixed: Cruise Control was not stopped when using handbrake - https://feedback.bistudio.com/T163813

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

[mellow lark]

Better would be to do set individual bandwidth limits for arma and nextsync, so that arma doesn't choke when it starts syncing

[whole cloud]

We are not talking about that

[mellow lark]

Oh yes sorry, wrong comment :/

[restive pilot]

• Fixed: Cruise Control was not stopped when using handbrake -
  what about the engine off bug?
  https://feedback.bistudio.com/T163709

[whole cloud]

😠