#perf_prof_branch

1 messages · Page 2 of 1

autumn timber
#

The only "reasonable" way that I see to do that, that would actually work, would be to install a proxy allocator that checks if we're calling free on memory that has been allocated using the old allocator or the new one.
This is all academic discussion, though, because such proxy would probably completely defat its own purpose by adding a time overhead for each free() call

#

But technically that should be able to work

gritty wasp
#

@tawdry gazelle what happened to thread on forum?

whole cloud
#

2.10.150123 new PROFILING branch with PERFORMANCE binaries, v9, server and client, windows 32/64-bit, linux server 32/64-bit
 - Fixed: Possible multiplayer desync by deleting a remote player's uniform/vest container, which would then be deleted locally but not remotely - https://feedback.bistudio.com/T167483
 - Fixed: Visual glitches with wrong fogDecay values - https://feedback.bistudio.com/T123262
 - Fixed: doStop in an init script could break AI behavior in multiplayer - https://feedback.bistudio.com/T167206
 - Fixed: It was possible to move objects below the terrain by using Alt+LMB in Eden Editor

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
analog acorn
#

objects below terrain in Editor
Wait, this wasn't supposed to be possible??? ASL mode should let you do whatever you want for altitude, shouldn't it?

fickle geyser
#
  • Fixed: doStop in an init script could break AI behavior in multiplayer - https://feedback.bistudio.com/T167206
    I wonder if that also fixes the issue where player is not initialized (stays as a seagull) on dedicated.
whole cloud
whole cloud
fickle geyser
#

I will test. Basically doStop this when done on dedicated and on player caused him to stay as a seagull.

whole cloud
#

Well that does indeed sound exactly like this 🤔

empty goblet
#

that's very old issue, great if it was resolved 👍

inland dew
empty goblet
#

maybe the 5 "seagul" for createvehicle.txt could become finally extinct 🤣 {workaround how fix stuck player with the issue}

empty goblet
whole cloud
#

2.10.150146 new PROFILING branch with PERFORMANCE binaries, v10, server and client, windows 32/64-bit, linux server 32/64-bit
 - Fixed: Eden Editor Translation Widget sideways movement could move objects below the terrain, even if surface snapping was active (it now follows the terrain)
 - Fixed: HashMap did not handle nil keys correctly (it considered two nil's to be different keys) 
 - Fixed: In modded keybindings, key combinations would block the main key - https://feedback.bistudio.com/T164496

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

Have a nice Christmas 🎄

whole cloud
vale shoal
#

is -bandwidthAlg2 still experimental and should it be used on live servers?

#

also we're running on v10 v9 and have some massive BattlEye related problems

spiral pond
#

17:22:16 Client: Object 2:262 (type Type_117) not found.
17:22:16 Client: Object 2:262 (type Type_118) not found.
17:22:16 ["BIS_Escape_fnc_OnPlayerRespawn",[2031068: seagull.p3d,<NULL-object>,1,30]]

#

could this be related to the recent fix? JIP client is unable to get into game - stuck in the loadingscreen

#

Type: Public
Build: Stable
Version: 2.10.150146

#

referring to this

#

(Escape doesnt use doStop - only CP)

whole cloud
#

More likely

Fixed: Possible multiplayer desync by deleting a remote player's uniform/vest container, which would then be deleted locally but not remotely

empty goblet
vale shoal
#

btw - RPT is printing this for ages now:

d:\bis\source\stable\futura\lib\network\networkserver.cpp ClearNetServer:NOT IMPLEMENTED - briefing!

you guys should do the briefing 😄 😄

#

while investigating the BE issue I found this error in several client logs:

14:27:52 (7335 [update._supplyEnt]): Network sync error - NMTReplaceContainer 14:27:52 (7336 [update._fromSupply]): Network sync error - NMTReplaceContainer

what does that mean? is this related to any perf update?
in all cases the error happens when a player is connecting

whole cloud
#

yes could be same as I posted above, that was related to containers

#

This is JIP into a existing unit? aka player taking over a AI unit?

vale shoal
#

yes

whole cloud
#

Something I haven't tested 😄
Good to know. Does it actually break anything tho?

vale shoal
#

well, it crashes the player multiple times and kicks him out via BattlEye Query Timeout

#

at least these are the symptoms

whole cloud
#

crashes? then send crash reports

#

better make FT ticket for them tho

vale shoal
#

well, I don't think that these crashes are related, can only tell that players told me

whole cloud
#

Report all crashes pleae

vale shoal
#

it seems no dump files were created, sadly

empty goblet
#

still FT ticket to make, attach RPTs from client and server with the errors

vale shoal
#

do I need to open a ticket for every user?

whole cloud
#

RPT's are useless for crashes, only need mdmp's

#

if you don't have any then I don't need a ticket

#

And no, I don't need ticket spam

weak panther
#

Seems class Land_Rail_Track_TurnOut and class Land_Track_01_turnout railway segments (terrain objects) draw a blue line from the object to [0,0] grids on the map after last DEV and PROF update
https://ibb.co/5Tnt9hF
https://ibb.co/ZWGsQxJ

  1. Open Tanoa
  2. move Eden editor camera to [10421.6,4976.46,39.1409]
  3. take a look at the map on the camera position
night mulch
#

New easter egg

whole cloud
#

Now the big question, how could I possibly have messed that up... There was dabbling in there for the new river rendering

spiral pond
#

Fixed: Possible multiplayer desync by deleting a remote player's uniform/vest container, which would then be deleted locally but not remotely
yep the clothing/gear gets removed (player setUnitLoadout (configFile >> "EmptyLoadout");) and replaced (by the server)

#

the clients had 2.10.150146
but arma3server_x64.exe is 2.10.149954 - will fix and try again

empty goblet
#

PERFORMANCE and PROFILING branch discusssion goes here

burnt tartan
weak panther
#

There no updates yet, mb fix already in internal prof branchaviator

night mulch
#

Rivers

heavy vortex
#

Bug in 2.10.150146, not sure how new: Have actions defined in CfgUserActions and mapped to Y and leftshift-Y. When pressing leftshift-Y, only the Y action fires. Works fine in 2.10.149954.

#

Same problem if binding is removed and re-added in the UI.

burnt tartan
full nova
#

oh yeah it also happens on ground vehicles, some ai or crews will desync outside standing like that when going at full speed

ashen helm
#

isn't that also an issue in the base version? Seen this in few Workshop scenarios already

burnt tartan
feral harness
heavy vortex
#

149954 is stable.

#

So no. Perf only.

whole cloud
whole cloud
heavy vortex
whole cloud
#

I'll check that tomorrow.
Planning next prof for wednesday

heavy vortex
#

Maybe it's different when only one of the actions is from CfgUserActions.

#

In this case it's both.

burnt tartan
burnt tartan
burnt tartan
pallid island
#

I had a similar instance happen on a modded vehicle yesterday, but wasn't running perf/prof

whole cloud
#

And just retested on profiling 2.10.150146, same thing. Both my actions eventhandlers fire when pressing shift+Y

heavy vortex
#

Oh I see, this is intentional...

#

Isn't that going to break a lot of other stuff as well? Like you don't always want the key action firing when you hit shift+key.

#

The order doesn't help either. In our case we have different menus on Y and shift+Y, so when you press shift+Y the Y action fires last and replaces the shift+Y menu.

analog acorn
#

Is this specific to the shift modifier? Because, for example, in the default controls I believe you have G for grenade and Ctrl+G for switch grenades, obviously if they both fire that's going to be a bad time

heavy vortex
#

It seems like any other bound combo with G doesn't throw the grenade, so apparently that has different rules.

#

I put shift-G on GPS and it only opens the GPS.

#

But otherwise shift-G throws the grenade.

#

This seems entirely reasonable to me and I'd prefer to maintain that behaviour for user actions unless there's a very good reason :P

whole cloud
#

all modifiers because it would block vanilla keys

#

guess I'll make it a config entry on the action whether its modifyer blocking or not?

restive pilot
#

I thought this feature already existed? think_turtle

whole cloud
restive pilot
#

isn't that what it's for?

whole cloud
#

ah yes-ish

#

Ah no

#

its only for UI display "hey this key is bound twice"

#

not for keys blocking eachother

#

this system is mess, need to find way to make everyone happy harold

spice dirge
#

inb4 Keyboard Input FSM

whole cloud
#

well prof tomorrow is gonna have the completely non-blocking thing.
I'll think about that more tomorrow to make a flag and choose some default

whole cloud
#

2.10.150195 new PROFILING branch with PERFORMANCE binaries, v11, server and client, windows 32/64-bit, linux server 32/64-bit
 - Fixed: PhysX vehicle suspension reaction on setDamage - https://feedback.bistudio.com/T167797
 - Fixed: Copy-pasted units in Zeus did not have their origin waypoints set - https://feedback.bistudio.com/T164049
 - Fixed: Linux server was not correctly handling large numbers in config files
 - Fixed: Random numbers on Linux servers were not always random - https://feedback.bistudio.com/T169575
 - Fixed: Railways on the 2D map would render blue river lines
 - Fixed: Crash when a HashMap was in a variable in a scheduled script while saving the game

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

xmasblob

whole cloud
light cargo
#

what are the changes compared to stable? what purpose/usecase do the profiling branch have over the regular version of the game?

light cargo
#

im having a issue where it would get stuck in the loading screen when joining a started multiplayer server turns out it was a mods incompatibility weirdness

rose moth
# light cargo what are the changes compared to stable? what purpose/usecase do the profiling b...

https://forums.bohemia.net/forums/topic/160288-arma-3-stable-server-210-performance-binary-feedback/
It started purely as special branch with a focus on profiling (multiplayer) performance and trying experimental optimizations or changes. Recently it has been used a bit more broadly to try certain engine fixes / optimizations / additions before hitting other branches. Many of these changes get to the next platform update, but some will never graduate. One rule is fixed: it must be multiplayer compatible with main / default branch.

Bohemia Interactive Forums
Arma 3 STABLE server 2.10 "performance binary" feedback

NOTE: 2.10. PROF/PERF up! you dont need a new client on the PERF/PROF server! there is a branch on steam you may use, but beware that sometimes data on dropbox are newer due to manual build e.g. on weekend: special profiling Steam branch for Arma 3 following this wiki page : https://community.bis...

kindred radish
#

Fixed: PhysX vehicle suspension reaction on setDamage

I wonder if that is one of the reasons for exploding vehicles in persistent game modes like Exile and co.
Vehicle gets setdamage from the server on creation, but since its not simulated that suspension reaction might be in the queue till a players gets nearby and it simulates the suspension then and boom (due clipping into ground/elements) ?
I saw this suspension reaction often as player even under other circumstances like using Enhanced Movement to climb on a vehicle.

feral harness
#

who knows the engine is mega complicated, you just have to try and see

whole cloud
light cargo
ashen helm
#

both are used in the same way, prof branch has to be mp compatibile and adds stuff that don't break mp, dev adds everything else

light cargo
#

got it thank you

#

for a regular user, is perf branch worth using?

ashen helm
#

it might run smoother than the base branch in some cases, it's compatibile with stable branch servers so you could use it, but it's frequently updated and can have bugs

whole cloud
#

it gets bug fixes and performance improvements first. Which would make it worth using.
But it also sometimes gets bugs first blobdoggoshruggoogly

light cargo
#

hm, how can we spot bugs or regressions?

whole cloud
#

when you notice something broken in game 😄

whole cloud
light cargo
#

that's not easy for someone getting started with arma 3

whole cloud
#

yup

light cargo
#

i experienced a issue where launching from the launcher to join a server keeping the mod selection wouldn't work properly

#

mods not being there properly

#

and i had the issue where the watch wouldn't stick when pressed

#

um what are the known issues so far?

#

oh good question, how do you switch to profiling exe if needed with the default launcher? which launcher would yall recommend?

hushed seal
light cargo
#

yeah but im asking about switching to the profiling exe after switching to the branch on steam with the standard launcher, it is not possible?

ashen helm
#

if you chose profiling branch in steam in the beta tab, launcher automatically detects new exe and starts it instead of the stable one

light cargo
#

performance branch replaces arma3_x64.exe.....

#

im asking if its possible to switch to profiling exe with the launcher

#

do i need to disable battleye to use the profiling exe?

patent dome
#

You shouldn't need/want to use the profiling exe, unless you want to measure the performance of the game and need additional debugging.

light cargo
#

yeah, im asking in the event a user ever needs the profiler/debugging, how would one do it from the commodity of the included launcher?

fickle geyser
#

You dont.

#

You can rename files.

analog acorn
#

Branch switching is not controlled by the launcher.

patent dome
#

He's not talking about switching branches, he wants to use the profiling instead of the performance exe when on the Perf/Prof branch. And as Veteran said, you'll have to rename the profiling exe

light cargo
#

thanks for the responses!

restive pilot
light cargo
#

it's 250mb for some reason

#

what else is there to transfer other than the executables?

#

is the profiling branch just including the profiling depot on top of arma 3?

#

oh i see, the combined size of all the profiling branch executables totals 266mb

restive pilot
#

Last I tried it only downloaded the exes. Dunno what has changed. Dedmen knows

light cargo
#

yeah i realized that

#

im seeing through steamdb that there's a separate depot for profiling exes

#

so i assume the branch includes the profiling depot on top of everything else

restive pilot
light cargo
#

oh really there's no asset testing on those branches?

restive pilot
#

There is. I meant when executables alone download

analog acorn
#

There's not much asset testing on perf/prof because changes to assets aren't very MP-compatible. There's not much asset testing on dev branch either really; it is possible but changes to assets are pretty rare at this stage of development (and new assets are even more rare)

empty goblet
broken scroll
#

awesome :)

vale shoal
#

When actually occured the error with NMTReplaceContainer?

#

I'm down to perf v6 and I still see it on many client logs

whole cloud
#

What?!
I thought v9, the uniform naked units thing

#

V6 on both server and client?

#

If server is older perf, it might still happen if a client has the new one?

#

If it's not that uniform naked fix, i need to know as that's quite a big Bugfix that i took back out to be safe in case it is causing that

#

I tried to repro that replaceContainer thing but couldn't. And also kju's report of JIP being stuck and also couldn't repro

vale shoal
#

the server is on 149973, the client who had this error is on 149954

vale shoal
#

Hope this error report gives you no migraine

vale shoal
#

Any new information about that topic? notlikemeow

whole cloud
#

next prof I'll revert that one naked units thing, and we'll see if that makes issue go away

kindred radish
#

If iam not mistaken the NMTReplaceContainer rpt error was there before.

whole cloud
#

First post about it was august 2016 🤣
But why did it appear so much recently

restive turtle
#

every arma bug ever.

weak panther
#

Oh no notlikemeow
ctrl + X don't work for Zeus
Vehicles and Props a little jumps and with AI nothing happen

weak panther
willow vault
#

not tested, but check your keybinds.

weak panther
willow vault
#

zeus are different I think.

weak panther
#

mb, but when I try to cut the vehicles jumps a little

willow vault
#

yup, doesn't work 😛

#

you can copy, delete and paste 😛

#

then ctrl+v. it has the same effect

#

i'm on stable btw. it just appears there's no ctrl+x

night mulch
#

Did it work before ?

weak panther
#

I'm not sure blobdoggoshruggoogly

whole cloud
#

2.10.150241 new PROFILING branch with PERFORMANCE binaries, v12, server and client, windows 32/64-bit, linux server 32/64-bit

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
#

Zeus also has no undo/redo. So I'm pretty sure ctrl+x is not a thing (Eden uses undo/redo functionality to implement that)

weak panther
spiral pond
#

the touch of god? (the damage to cause a PX bounce maybe?)

#

Reverted: Possible multiplayer desync by deleting a remote player's uniform/vest container, which would then be deleted locally but not remotely

Is this/could this be also connected to BI garbage collector? Like we are seeing floating and naked vehicle crew in MP/DS env - as we dont do vehicle cleanup ourselves, it must/should be the BI garbage collector we are using.

whole cloud
#

If that bug is now gone, I guess it could. Please tell me if it was

spiral pond
#

clients used a mix of stable and perf. server is on stable - next play-test is on Sunday

pale mica
#

Speaking of mix stable/perf. If the server is on stable, is it a good idea to force stable clients?

whole cloud
#

No

vale shoal
#

As it seems the error is still persistent (stable client, perf server)

whole cloud
cold vale
#

Hmm, discovered an interesting crash in 2.10.150241. Still working out the details on why (disabling code until it stops crashing)

Both lines cause immediate crashing

addUserActionEventHandler ["moveUp", "Activate", {}];
addUserActionEventHandler ["moveDown", "Activate", {}];
whole cloud
#

I got a report too, looking into it rn

median cipher
# 
addUserActionEventHandler ["moveUp", "Activate", {}];

is the repro, on vanilla

#

causes immediate CTD

whole cloud
#

I fixed that yesterday, after prof

#

:sad:

#

yep the report i got is that too, sad

#

new prof is on ze wai

whole cloud
#

2.10.150255 new PROFILING branch with PERFORMANCE binaries, v13, server and client, windows 32/64-bit, linux server 32/64-bit

  • Tweaked: Improved performance of nular commands and variables
  • Fixed: Crash when adding UserAction Event Handlers for non-modded keys - https://feedback.bistudio.com/T169879
  • Fixed: RscPicture texture paths were always forced lowercase

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

idap

whole cloud
whole cloud
#

I'd like to know if this breaks ArmaDebugEngine or ArmaSciptProfiler, I don't have time to test. But I assume it should be fine

restive pilot
whole cloud
#

that should be fine

#

I think base intercept doesn't have that stuff.

gleaming vortex
rain moth
#

wiki page there contains more info as well

restive pilot
# gleaming vortex is there someore in indept page or something on what intercept is / does/ how it...

simply put it allows you to make extensions and also call sqf commands inside your extension too (which is not normally possible with ordinary extensions)
with extensions you can perform operations that the game can't and then transfer the results to the game. e.g. you can do file operations, use SQL database, interact with another program (e.g. TeamSpeak), etc.
with intercept you don't have to "transfer" those results anymore. you can do it directly in the extension

gleaming vortex
#

or just function / command calling

restive pilot
#

wdym by layout?

gleaming vortex
#

so for ex use css so you can make a nice rectangle with edged corners

#

stuff like that

restive pilot
#

no. you can make overlays tho, but they're not "part of" the game

gleaming vortex
#

ok, ill read trough the github

vivid rune
# whole cloud <:meowsweats:707626030613135390>

The good thing is, that this fix:
Fixed: Possible multiplayer desync by deleting a remote player's uniform/vest container, which would then be deleted locally but not remotely - FT-T167483

seems to work. I can remember that i played some loot based round in Arma with >=V9 and we saw no things like you cannot take things from killed AI.

vale shoal
# vale shoal As it seems the error is still persistent (stable client, perf server)

I've tested it again now with perf client/perf server

got this warning vehicle connecting:

Warning Message: Can not create source for procedural texture - probably badly formed texture name: "#(argb,8,8,3)color(0,0,0,1)"#(argb,8,8,3)color(0,0,0,1)"
(this isn't present in stable client)

but also sadly still "Network sync error - NMTReplaceContainer"

whole cloud
vale shoal
#

As it seems like, but why is it showing up in perf and not stable? Which change made it looking this way? Or is it an error?

whole cloud
#

I don't know, I need to know which prof version it first appeared in

#

when you first posted about it, that was just after the naked units fix thing, which could cause exactly such issues. Now its removed again and it still happens I have no idea why

vale shoal
empty goblet
#

note, all the large official Arma 3 servers (KOTH and Warlords) run recent profiling , so if anyone observe oddities please ping me and dedmen at once, thanks

whole cloud
whole cloud
nocturne obsidian
empty goblet
#

no, the fix is addresing loading of htm,html files from inside .pbo, wasn't working for w/e bug reason

nocturne obsidian
#

oh, that is said. is it wanted that websites can't be loaded or is that something that will be patched as well?

empty goblet
tall iron
#

our server provider have the option for running Performance-branch or Profiling-branch. I am using the Profiling-branch myself client side after it being recommended but what is the difference compared to Performance-branch? I though they were the same thing bundled into Profiling

heavy vortex
#

They're the same branch. The default exes are performance, the ones called profiling are profiling.

whole cloud
#

You shouldn't use profiling if you don't need profiling

tall iron
#

ok, so what they call "performance-branch" would be the one to use then in our case, if any. is there any recommendation in general for running the server (+hc) on performance vs main? or do we just have to try and see if there is a difference in our case?

rain moth
#

you get early access to fixes and regressions Troll

autumn timber
#

And sometimes bugs 🙃

hushed seal
whole cloud
#

its still good

#

but less good than performance, because all the profiling stuff

hushed seal
#

Ah okay, well I like to do frame captures every now and then so probably won't switch to perf.

vale shoal
#

Actually I don't see any big difference on client side with performance (tested it on KOTH some days with each more then 6 hours playtime). But on server side there is a huge impact on CPU usage which is actually very positive.

dry zephyr
#

It's the difference between the server grinding to a halt with high player counts and AI on stable but on perf it doesn't care and can be pushed wayyyy further for me.

hushed seal
silk summit
dry zephyr
light cargo
#

how do you "unlock" limitFPS?

light cargo
heavy vortex
#

Command line, -limitFPS=300 or whatever

#

It's mostly a waste of electricity though. Only makes sense for benchmarking.

silk summit
heavy vortex
#

On Antistasi Community servers? Maybe 30 :P

silk summit
heavy vortex
#

Average, .1% lows would be garbage due to spawn hitching.

silk summit
hushed seal
dry zephyr
#

Having the FPS unlocked gives more headroom for the server to recover in a low FPS event, so instead of it dipping from 50 to 5-10 it dips from 700 to 100, and once the server starts to choke it's difficult to get it to recover (without cleaning up units, objects, scripts, etc) so its best to stay away from dipping too low. At least that's my experience.

analog acorn
#

Just for my clarity, is perf branch inherently always faster than stable because of how it is, or will that reported 70fps under load come to stable as well in a future update?

whole cloud
#

Can never guarantee anything

#

There are some performance changes that are currently exclusive to Prof and are not enabled on stable, because it's not clear if they are safe

#

Like the recent thing that made all scripts faster

#

This thing preventing AI freeze too i think

analog acorn
#

Right, but it's changes that could theoretically become stable depending on whether they work, not something about how the perf exe works that makes it Just Better™?

full nova
#

after testing, the weird looking road (looks like a static tv screen) also happens on stable with thermal

whole cloud
#

Yes. Everything on perf is just flick of a switch away from working in stable

light cargo
#

are any of the perf improvements validated as safe for stable?

hushed seal
light cargo
#

ah fair

novel ravine
# whole cloud There are some performance changes that are currently exclusive to Prof and are ...

Just curious...when there are crash fixes on perf build, are those fixes related to crashes caused by older perf builds or can they be related to the stable build as well?

For example:
Perf v8 mentions - Fixed: Several game crashes (helicopter waypoint, Unicode script error, UIOnTexture, dynamic simulation) - are those related game crashes still possible within the current stable build or only if you are using perf v1-7?

whole cloud
#

both

#

usually from stable

novel ravine
#

Thanks. 🙂 I ask because we've been suddenly experiencing very weird mass lag spikes which seem to lead to Arma 3 server not responding crashes over the past month or so every now and then. Lots of "network message is pending" and "network sync error - NMTReplaceContainer" spam, but they feel impossible to even determine the source of what is causing them. However, I do know that we recently made dynamic simulation way more prevalent so I'm starting to wonder if it's related to that. We reverted to the stable build to see if it would stop and everything was fine up until about 30min ago when I tried turning on our second headless client and it almost instantly timed out the server.

vale shoal
novel ravine
vale shoal
novel ravine
#

Yeah, we have similar issues with BattlEye Query Timeout and BattlEye Client Not Responding, but I agree they are likely not related to performance builds so this probably isn't the best place to discuss it further. Do you mind if I DM you about it? Just curious about your experiences compared to ours.

vale shoal
#

I meant that I don't think the crashed aren't related to perf.

The BattlEye Query Timeouts might related to perf.

#

Sure, just DM me.

whole cloud
#

If you could try old perf builds and see if that makes the issue go away that'd be great. Knowing which build introduced it would narrow it down a lot.

#

But i fear it's also happening in the oldest you try

novel ravine
# whole cloud But i fear it's also happening in the oldest you try

Correct. The crashes we have been experiencing also happen on Stable build too, which is why I originally asked. I intend to update to anything beyond Perf v8 soon since that addresses crashes related to Dynamic Simulation and we made some changes a while back that swapped a lot of global simulation for dynamic simulation and these crashes began occuring a little while after that. I'll follow up and let you know how that goes.

whole cloud
#

if you have crashes you could also report them

novel ravine
#

How? Just post RPT here?

whole cloud
#

feedback tracker ticket, upload crash reports

#

full crash report with mdmp, not just rpt, rpt is useless

whole cloud
whole cloud
silk summit
#

"- Fixed: Unable to pick up items from a dead character's inventory in some cases"

#

Some might think this is a small thing but for many it was a huge deal! I have had many new players not returning to my A3 sessions due to this bug because it could break the mission flow, specially in some instances where the mission requires scavenging due to limited resources. It can also happen for items on containers or items/weapons just on the floor so the issue is not exclusive to dead bodies as shown in the video: https://youtu.be/a7FLyLSqhN8?t=99

YouTube
DarkLiberatorZone
The Uncle Games | ARMA 3

TCs plays in the Uncle Games against each other. All hail Uncle. Enjoy this video of this event made by Zapper and Stormhammer.

The Cooler Server is an Arma 3 group. Apply here:
https://www.thecoolerserver.com/

TWITTER: https://twitter.com/darkliberator
FACEBOOK: https://www.facebook.com/DarkLiberatorZone
TWITCH LIVESTREAM: http://www.twitch...

▶ Play video
#

I hope you guys are able to hunt down all those instances where it becomes impossible to pick up items in MP because it can become really frustrating for players. Even worst is the fact that once it happens once (on one dead body for example) it was very likely to happen again for the same or other players on the mission, usually requiring a full restart to mitigate the issue

vivid rune
#

6h Dynamic Bulwarks (same session). No issues, 32bit Linux Server V15 on a potato. 👍
Only thing to be full happy is to get rid of the second start of the server every day because on first start per day the server crashes.

hushed seal
#

any chance of this 19:18:58 Tried to AccessTargetList for non-local AI group while non-local targeting was disabled! Engine will now enable non-local targeting which will hurt clientside performance also logging which script/mod/whatever is causing this?

whole cloud
#

no

vale shoal
#

@hushed seal disableRemoteSensors check that and you might find it

hushed seal
novel ravine
novel ravine
#

😐 Of course...30min after posting that, it shits on itself. I think this problem is related to one of our headless clients though since it was a severe lag spike that did not stop until I ended the task of the headless client. It never actually crashed.

I'll move any troubleshooting/updates elsewhere though since I I do not believe it related to Perf Branch. Sorry for the false alarm.

whole cloud
#

@empty goblet The latest perf build just saved our mission. Everyones arma crashed on mission start so i thought maybe perf build's crashfixes will help. I distributed latest perf to everyone and now everything is running fine

hushed seal
empty goblet
#

thanks for telling us @whole cloud yet it seems the new perf breaks someone else mod so expect new one soon with more fixes ;)

whole cloud
#

yeah now ive got one guy for whom it doesnt work but still about 20 people that can play with it

shell lion
#

Hi, i have a question about this command:

diag_captureSlowFrame ['total',0.3];

I use the A3 profile branch and start my mission in the editor as an mp game. After a certain time I want to run this command via the console. But this error comes up:

21:47:52 Error in expression <diag_captureSlowFrame ['total',0.3];>
21:47:52 Error position: <['total',0.3];>
21:47:52 Error Fehlendes ;

What am I doing wrong? According to the wiki it should actually work like this (https://community.bistudio.com/wiki/Performance_Profiling).

heavy vortex
#

Profile branch has two sets of exes. The default ones are performance branch and have no profiling support. You have to use the ones with profiling in the name.

shell lion
hushed seal
shell lion
#

Hi, can someone tell me what this rpt log message mean (Profile branch):

19:24:17 TimeLimiter, Category AICreateTargetList exceeded time limit, runtime 20036, limit 20000, denied requests 0
19:24:17 Ent 2189883f640, dbg , time 20036

spiral pond
#

only Dedmen/KK can/could 🙏

empty goblet
#

i too but i will need to remember what they told me 😁

rain moth
#

I was gonna say "when profiling update" but I guess I don't have to now 😦

#

40s ago 🏎️

full nova
#

wait there is already perf_branch for 2.12?

patent dome
#

2.12.150330 new PROFILING branch with PERFORMANCE binaries, v0, server and client, windows 32/64-bit, linux server 32/64-bit

  • Sync with 2.12 main branch update
  • Previous Profiling branch exclusive improvements are still present

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
whole cloud
ripe anchor
#

Could we get the V2 dx11 renderer as a gift for the 10th anniversary :)

spiral pond
whole cloud
#

repro would be good, but quite hard as usually its not even a reliable repro

#

I was able to get slow AI thingies, but never reliable enough to analyze it

vagrant zodiac
#

Would I be correct in thinking the "dotColor[] = {1,0,0};" functionality was only applied to non-IR lasers in the end?
I'd hoped it would affect IR lights too as a way of balancing dot brightness (not size)
But a quick series of tests now suggest no effect. (It's been a long day so I might have f-ed up somewhere - hence wanting to check...)

whole cloud
#

yes

vagrant zodiac
#

Thanks. Although it’s not the answer I’d hoped for, the confirmation is most appreciated as it stops me spending an hour or so fighting with configs this evening before coming to the same conclusion!

gaunt chasm
#

Hey Im trying to Follow a tutorial (https://www.youtube.com/watch?v=86d3VCpWPIE Link) however, at the 13 second mark, he tells me to go to the website to copy a line of code. I go to the website and theres no line of code for me to copy. (https://community.bistudio.com/wiki/Arma_3:_Steam_Branches) I've looked everywhere online for this code and I can't seem to find it. Can some one please send it to me, I just want to make custom Compositions

YouTube
Tiber Brutus
Arma 3 Adding Custom Compositions to Public Zeus

Link number 1: https://community.bistudio.com/wiki/Arma_3:_Steam_Branches

Link number 2:
https://steamcommunity.com/sharedfiles/filedetails/?id=2419451036

Quick details:
(1) A new review is on the way.
(2) This is the start of a series of short tutorials on A...

▶ Play video
Arma 3: Steam Branches
granite dock
gaunt chasm
#

oH

empty goblet
#

q, anyone has crashes from 1.52.133024 (or 1.52.132868) ? if yes please pm me url where i can download the dumps

whole cloud
#

perf linux 1.52.133.024 cant Load my ALiVE insurgency after i click okey in the lobby i go to loading screen and after that im back in mission selection. Nothing unusual in RPT. And with perf v2 (1.52.132.709) with the same startup parameters and mission it loads and runs just fine (Although perf v2 on linux server crashed twice for me in the last 3 days after about 4 hours of play)

golden crest
#

@whole cloud any RPT for ALiVE team would be useful

#

along with alive plugin log

whole cloud
#

via PM to you? @golden crest

golden crest
#

Thanks, looking at the logs now

golden crest
#

Hmmm weird looks like v5 breaks ALiVE badly

#

will need to test

#

ALiVE doesn't even init

#

any module init changes in v5?

hushed seal
#

I've been looking at the output of #monitords 5 for a bit and it seems odd to me how rare it is that Nonguaranteed and Guaranteed messages get reported on there.

Surely if there's fighting going on there's no way that there is zero G or NG traffic happening for 10 minutes?

rough pebble
#

Since traffic should get sent asap, that would make sense it’s always 0

#

that’s speculation on my part though

hushed seal
#

that would explain why the only time you do see values in there they're 5 digit numbers and up

#

the documentation on it is quite sparse

boreal wigeon
#

I see similar things, it's either 0 or a really big number.

hushed seal
#

whats your usual intervall?

boreal wigeon
#

30 sec I believe

#

I mainly use it for the ram and fps though

hushed seal
#

Same, was just looking at the logs and was wondering, idk why I never thought about it before

boreal wigeon
hushed seal
#

ahahahaha, I was wondering if #monitor accepts numbers smaller then 1 as argument

#

0.1 works

#

0.01 also works

#

lol

boreal wigeon
#

Do you want that though... The amount of writes...

hushed seal
#

100 kB for about 1.5 minutes of logging

# 
16:10:11 Server load: FPS 280, memory used: 2256 MB, out: 86 Kbps, in: 0 Kbps, NG:0, G:0, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 240, memory used: 2256 MB, out: 59 Kbps, in: 0 Kbps, NG:0, G:0, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 270, memory used: 2256 MB, out: 64 Kbps, in: 0 Kbps, NG:0, G:0, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 220, memory used: 2256 MB, out: 49 Kbps, in: 0 Kbps, NG:0, G:60, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 228, memory used: 2256 MB, out: 57 Kbps, in: 0 Kbps, NG:0, G:60, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 238, memory used: 2257 MB, out: 27 Kbps, in: 0 Kbps, NG:0, G:60, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 272, memory used: 2257 MB, out: 49 Kbps, in: 0 Kbps, NG:0, G:60, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 171, memory used: 2256 MB, out: 36 Kbps, in: 0 Kbps, NG:0, G:60, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)
16:10:11 Server load: FPS 142, memory used: 2256 MB, out: 81 Kbps, in: 3 Kbps, NG:0, G:60, BE-NG:0, BE-G:0, Players: 1 (L:0, R:0, B:0, G:1, D:0)

buut there's now small values in the guaranteed messages column

gleaming vortex
#

or where you found them

hushed seal
#

logFile argument in server.cfg, running monitords dumps into the previously defined .log file

gleaming vortex
#

thank you

empty goblet
#

otherwise i would need to get simple repro and new A3FT ticket with it

random isle
#

Any idea what this is dlcUpd; 1.46536; 21.28701; this is from diag_captureSlowFrame

#

hmm unchecked require DLC in editor and its gone

hushed seal
#

It seems like #monitords only works reliably on perf branch with the profiling.exe. Is that intended behaviour?

boreal wigeon
#

I have it working fine with the performance.exe? What issues are you experiencing?

hushed seal
#

with the arma3server_x64 on perf branch? It just ignores it, same modset and everything, all I'm doing is switching branches and trying the different .exes

#

lauch server, login as admin, select mission, run monitords 5 and observe

#

only prints to console window with profiling.exe for me

boreal wigeon
#

I'm on Linux though 🤔 might be an difference then.

hushed seal
#

something that works on linux but not on windows? what is this?? proper servers?

maiden wasp
#

Linux is the way arma was designed to be played /s

#

But tested it myself and appears to be fine but unknown if I'm on the latest, I'll have to check that later if you haven't figured it out yourself

hushed seal
#

Someone else reported it working for them. Mine is still borked. I'm just gonna sit it out.

empty goblet
whole cloud
#

ALiVE problem still persisting

hushed seal
#

captured on the server

whole cloud
#

I guess so

#

This works with script command hook, so it will only detect variables that are sent from the client you are running on

hushed seal
#

ah so if a client is receiving? a big public variable that wouldn't show up

heavy vortex
#

What tool is this?

hushed seal
whole cloud
#

its only intercepting the script commands

golden crest
#

Dedmen, have you tested a stock ALiVE mission with just CBA, ALiVE?

#

Just would be good to rule anything else out, weird ALiVE is not initing - its not the Params issue

whole cloud
#

ill try

warm sable
#

Hello Dedmen. I am the creator of that mission, if you send me what you have ill take a look at it in the next few days. -Dev Hazey

whole cloud
#

What exactly do you need? rpts and missionfile?

warm sable
#

Both would be fantastic.

#

However the RPT is fine now to start with if you have it.

#

I can tell you I have not visited the base mission for about 6 months due to taking a temp step back. So the base mission is broken as well.

whole cloud
#

well the mission works fine on perf v2 or lower

warm sable
#

Hmm. Grab me a V2 rpt and a V5.

whole cloud
#

see private messages

warm sable
#

Thanks

golden crest
#

@warm sable - might not be just mission issue with v5. Looked like ALiVE didn't init

warm sable
#

@golden crest @whole cloud, Yeah I don't see Alive even being stroked in V5

#

@whole cloud Have you tested just Alive with V5?

whole cloud
#

no v6 also not working same symptoms

#

Ill now test v6 just with cba and alive and with the mission provided in alive download

warm sable
#

Yes please, and thank you!

#

Let us know, ill hang out for awhile.

golden crest
#

I will just continue drinking

whole cloud
#

over 150sec loadTime but works... weird

warm sable
#

Check your PM in a second.

whole cloud
#

~~Okey was not ALiVE sorry for that :X. Now @empty goblet in perf v5 and v6 filePatching seems to have a problem
server has allowedFilePatching = 2; in server.cfg
but

1:24:26 Warning Message: Script \userconfig\asr_ai3\asr_ai3_settings.sqf not found
1:24:26 Warning Message: Script \userconfig\asr_ai3\asr_ai3_settings_ss.sqf not found

exactly same settings on perf v2 v3 and v4 work fine.
ASR_AI is using

ASR_AI3_SETTINGS = compile preprocessFileLineNumbers "\userconfig\asr_ai3\asr_ai3_settings.sqf";
ASR_AI3_SETTINGS_SS = compile preprocessFileLineNumbers "\userconfig\asr_ai3\asr_ai3_settings_ss.sqf";

to Load its configs~~

Wow perf v5 added the need for -filePatching startparameter on LinuxServer without real notice in the Changelogs. So after several hours hunting the error and half an hour searching the source of my servercontrol software i got my server back up and running...

whole cloud
#

2.12.150383 new PROFILING branch with PERFORMANCE binaries, v1, server and client, windows 32/64-bit, linux server 32/64-bit

  • Added: maxFileCacheSize command line parameter
  • Tweaked: Eden Editor OnHistoryChange event is now called after the editor is cleared (because history is cleared too)
  • Fixed: Unable to pick up items from a dead character's inventory - https://feedback.bistudio.com/T126030
  • Fixed: Crash if a default user name could not be determined (now it displays a critical error message with a workaround) - https://feedback.bistudio.com/T170710
  • Fixed: wheel_* animation sources would not animate for remote vehicles - https://feedback.bistudio.com/T150837
  • Fixed: Eden Editor/Zeus placing units into vehicles ignored locked turrets if they were also the primary gunner/commander turret
  • Fixed: On Linux servers error messages did not print a timestamp to the log (thanks @trim anchor) - https://feedback.bistudio.com/T171037
  • Fixed: Linux servers were limited to 2GB of RAM usage (thanks @rough pebble) - https://feedback.bistudio.com/T80680

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
rough pebble
plucky lichen
#

@whole cloudwhats the usage on the command line param maxFileCacheSize -maxFileCacheSize=1024?

restive pilot
vivid rune
#
  • Fixed: Linux servers were limited to 2GB of RAM usage (thanks @rough pebble) - https://feedback.bistudio.com/T80680
    I use maxMem=4095. Docker is showing a use of 3.648GiB, #monitor shows 2648 MB. Is there a explanation for the difference?
full nova
#

praise dedmen for new profiling

hushed seal
#

always been there afaik

rough pebble
vivid rune
#

Would it make sense to raise the maxMem Number higher than 4095MB on a 64bit Linux running the 32bit Arma executable in docker? I would say it would make sense if the file caching is a separate process, right?

empty goblet
#

@whole cloud you got some unpacked data ?

rough pebble
plain trout
#

Any chance to fix mixed tracer magazines so that:

multiple CfgAmmos per CfgMagazine are possible or find a way to show tracers on other clients and apply air friction.

#

I’m happy to post it on the official tracker if there is any chance to realize this

empty goblet
#

just repeat what i posted in the #community_wiki channel

for sake of curious experimenting if one has tens or more GBs RAM, i would love to see maxFileCacheSize option to support >2GB at 64bit OS under 64bit binary

inland dew
fickle geyser
inland dew
fickle geyser
#

I would expect as much as you want/os memory size?

inland dew
#

so it's like RAM disk, kinda

empty goblet
#

it's in the message, limit is 2GB atm

fickle geyser
#

?

#

Default size was either maxMemory - 448MB or if you had more than 2GB of memory available to Arma, it would be limited to max 2048MB.

hushed seal
#

Just had the same issue with the chongo map.

cold vale
#

I've had a couple CTDs seemingly due to running out of VRAM (on my 3080 Ti?) while running a large modset but I've only started to get these issues in the last 3 or so months.
I have crash logs while running 2.12.150383 if that would be of any use.

inland dew
cold vale
#

Yep, I get that it'll use as much as you can give it but I have to change my texture quality settings to clear it out and bring it back down to ~4-6GB after playing for several hours just to ensure I don't crash while operating a critical asset. Tried running with the custom mimalloc-v209-lock-pages to maybe see if that could help but I'd assume that's not related to VRAM management?

whole cloud
#

no i dont. But ASR_AI is using an sqf from \userconfig\asr_ai3\ as a config and server just throws back to mission selection if it cant load that... although i wonder why it doesnt just fail the script but fails to start the mission

woven loom
#

With GM assets i simply can't run very high or ulta textures because it overflows my 1660 Ti's 6GB VRAM at 1080p

#

Made a number of feedback tickets about it too

#

With vanilla arma i don't have an issue running ultra high

#

It's a shame because vanilla textures don't look great with high settings so i keep having to switch back and forth

gritty wasp
plain trout
#

I tried 20GB cache size and maxmem set to 25GB with a total of 64gb of ram and it would let me load on a server. Just complained about running out of memory. That’s with mimalloc v209. After changing to system allocator and no ram lines it would later use 19.3 GB of VRAM on 4K, all ultra 150% scaling.

I haven’t seen any performance changes in modded koth with 16GB of cache, mimalloc v209

inland dew
plain trout
#

Isn’t the -maxvram parameter responsible for that?

inland dew
whole cloud
whole cloud
hushed seal
whole cloud
#

In this example there was

hushed seal
#

Starting arma now to take another look at the framecapture 😅

whole cloud
#

looks to me like texture is LZ compressed.
Game loads from disk into temp buffer, then decompresses on CPU before sending to GPU. But why would decompression be so extremely heavy 🤔

hushed seal
#

here are my three framecaptures if you want to check them out yourself.
Two of them in the low FPS area, third someplace else.

restive pilot
whole cloud
#

I can speed up that pacLL, but that's only 200us, thats the actual disk load
2.1ms to decompress that file.. well how large is that texture file? :u

#

nvm can't speed it up, its already doing that

hushed seal
whole cloud
#

specifically only the lca layer masks.
That's the only thing that uses that code path 🤔
Maybe I can move them to thread

hushed seal
whole cloud
#

do the lags stay though? it should just do that load once and then it stays in cache

hushed seal
#

oh they stay

whole cloud
hushed seal
#

sure thing, let me just check if it occurs in the editor...

#

Alright, can't reproduce it in the editor 🙃 But I do have footage of it sooo

#

Location where I look down into the ground and get the bad FPS

#

X: 8090.88
Y: 8486.31
Z: 115.555

#

Chongo map from the Angola Maps mod

#

Would you like it as a ticket?

woven loom
#

There is a spot on Livonia too, one particular angle while in the radar station compound in the south

#

Absolutely tanks FPS with high view distance while looking at a wall

whole cloud
#

oooh

#

It has hardcoded limit to only cache 10 layer mask textures

#

if youre view distance is high enough that you need more.
It will re-load them every frame

#

heh yes. todo...

devenv_2023-03-27_12-17-20.png
silk pewter
#

"your mission, if you accept it: Ctrl+F TODO"

whole cloud
#

Also its supposed to asynchronously preload that texture. So then when it actually needs to load it would have it ready.
That was never implemented, so it does the full load then and there causing you the fps drop.
I'll increase the cache size that should help a bit

#

next prof will have it

hushed seal
#

I don't quite understand why it isn't showing in the editor tho, in MP it was really noticable

whole cloud
#

I also couldn't repro just now

hushed seal
#

I'm gonna give it a try on the dedicated server.

#

nope, standing in the exact same spot, same mission server and everything and nothing.

#

wPrep now doesn't even mention texture loading now

spiral pond
#

how exactly do you capture @hushed seal ?

hushed seal
#

profiling.exe and diag_captureFrame 1; in debug console locally executed.

fickle geyser
#

@shell wyvern should remember more about that issue.

rain moth
spiral pond
hushed seal
#

true, although during the OP itself wPrep was quite present in every framecapture.
Could be tho that it still happens with only one player, just less frequent so I'll try that command.

spiral pond
#

i would assume you need a combination of factors to trigger it - like you may need a higher base level of CPU/GPU/RAM/VRAM load to start off

#

so in your OP you had other conditions to start with

#

still an issue worth investigating, but may need some more "pushing" to replicate it via editor/lower scale setup

#

like a similar thing we experienced with Enoch objects and vegetation - in MP it could grind any system to <10 FPS even for powerful systems and rather low AI counts/else by just looking into specific directions (but not towns or other obvious trigger points)

cold vale
empty goblet
#

ofc i expect most of advantages to be for clients with excessive amount of ram ... (more reasons for those 48GB DDR5 sticks ... )

inland dew
#

buying 64 GB...

#

will also benefit since it's also dual rank

#

increased bandwidth also helps a lot in loading textures much faster from RAM

empty goblet
# inland dew buying 64 GB...

it's quite problem to find any good >32GB (DDR5-6400/6800/7200 (32/34/36CL) modules those days so i will be lucky to find 24/48GB ones

#

unless i can buy it locally in EU w/o headache of import tax and vat declaration

whole cloud
whole cloud
empty goblet
fickle geyser
#

idk, it was over 2 years ago at this point.

hushed seal
fickle geyser
#

and we were in a crunch and did not had time to do any detective work, we just fixed our pipeline to not compress these and carried on.

cold vale
#

It did cause an out of mem error when I first started having the issue, although the most recent crash was an access violation IIRC

whole cloud
#

ticket then

#

with crash report

rain moth
#

I wonder if it's at all related to the fps bug on weferlingen 🤔

spiral pond
#

RAM/VRAM running out like with the "3 fps" bug?

hushed seal
empty goblet
#

@whole cloud also use v6 not v5 (already outdated)

#

yes if you have some unpacked data, then you need use -filepatching, that's since 1.52

whole cloud
#

i know that it should be since 1.52 but it worked without -filePatching till v5

woven loom
hushed seal
vivid rune
#

I remember, that one of the RHS vietnam maps has bad fps if you looked in a special direction.

restive pilot
#

I even get that on Livonia blobdoggoshruggoogly

whole cloud
#

just because this may happen when you look in a specific direction, doesn't mean every specific direction issue is the same

rain moth
vivid rune
burnt tartan
#

So a map that isn't designed for long view distances?

hushed seal
wide field
#

Hmm

#

So is that engine bug, engine limit, I have to do something to fix that?

#

We did 100+ player ops on that map and didnt experience this... at all?

#

If someone tells me what to do, I'll fix it.

vivid rune
burnt tartan
vivid rune
#

I am not here to blame maps for it performance. What i understand is that we search for reproducible frame drops on a map to help dedmen to optimise the A3 engine. As i understand the point are mask layers. Maybe it can be reproduced by using the rhs map, maybe not (it is only too much objects).

plain trout
#

if you turn south on Angola something completely breaks the rendering. It is a different issue to the high density problems on some maps that appear on almost all machines. I saw arma taking up all 20GB of my vram easily and once arma has to load/unload vram, stuttering begins. But again on Angola, there is something different going on because it basically only happens when you turn south and no matter the viewdistance

hushed seal
cold vale
empty goblet
#

always exact repro is best, this is like the bug in Arma 2 i found on Chernarus with reeds in the south islands ...

#

each time i flew with airplanes over them, the FPS went from 100 to single digits

#

turned out , the models of reed failed to LOD and used the highest poly count only

#

so if you have situation where FPS goes to hell , make mission with repro, with exact fov / angle / placement / movement etc.

woven loom
#

I guess splendid camera helps with that

hushed seal
empty goblet
silk summit
#

I remember Leopard even posting videos about it... when driving or shooting near those big rocks caused FPS drops like crazy, but I can't find those videos or the ticket now thou

#

I north west malden and several workshop maps that use those rocks extensively all show FPS degradation when shooting/driving near them, including the many "cave" composition on the workshop that also use those assets so not only it is possible to replicate the issue 100% of the time but it is also very prevalent on many scenarios (AKA it is a common and recurrent source of FPS degradation and stutter)

plain trout
#

That happens with all rock formations. Issue get a lot less if you disable shadows

full nova
#

Wait disabling shadow? I thought it only lowers to Low

silk summit
# plain trout That happens with all rock formations. Issue get a lot less if you disable shado...

I remember someone suggesting that the issue was related to how collision was calculated and that those rocks had a "too complex" geometry thus causing those issues. I don't know why arma has to calculate collision/physics on rocks the bullets are not even impacting on when players driver near them and not over them but that was the explanation I was told before. Never knew shadows could also exacerbate the issue

#

All I know is that it can be fixed because the devs of SOG Prairie Fire also encountered this issue and they simply redid the big stones in a much more optimised way. I don't know why BI doesn't just buy those assets from SOG devs and just replaces those unoptimised rocks (because they clearly lack internal developers to do it)

whole cloud
silk summit
whole cloud
hushed seal
cold vale
whole cloud
#

Out of memory messages I don't care about, their dumps won't be useful.
But actual crashes without messages would be neat.. I mean not neat but neat neat

plain trout
#

modded or unmodded?

whole cloud
#

Arma is a modding focused game, so mods shouldn't crash it. So I don't care

plain trout
#

ErrorMessage: Fatal error: Out of memory in the file operation

Exception code: 0000DEAD at 6FE1CD29?

whole cloud
#

thats what I mean without messages.
0xDEAD, don't care. It told you what's wrong and I can't tell you much more anyway

plain trout
#

ok

silk summit
# hushed seal A mod could do that, right?

I don't envision a mod fixing it when even the official devs can't / are not willing. (And it is not about just replacing the rocks via modding with the SOG ones, because that fix would only work for players with SOG enabled, the SOG mod installed and/or only on maps that the mod supports and not all maps... so it would be a half baked fix for a limited number of players on a limited amount of maps)

#

But... getting the devs to acknowledge an issue, even if they don't fix it in the end, is already a win in my book 😎

#

If you ever get to that stage as a person who reports issues, that is as far a your responsibilities go. After that the ball is not in your court anymore

silk pewter
#

it's fair, but do also prioritise reporting issues that you know are most likely to be fixed at this stage of the game life
otherwise if done knowing it won't be fixed, it's more cluttering than anything

whole cloud
#

2.12.150430 new PROFILING branch with PERFORMANCE binaries, v2, server and client, windows 32/64-bit, linux server 32/64-bit

  • Fixed: selectionVectorDirAndUp would print error when querying proxy positions even if the proxy was found - https://feedback.bistudio.com/T167265
  • Fixed: Creating a new user profile now trim's whitespace from the profile name
  • Fixed: Terrain performance issue related to layermask
  • Fixed: RPT Error spam when loading mission that uses waypoints - https://feedback.bistudio.com/T171125
  • Fixed: Amphibious vehicles ignored obstacles on terrain above water during pathfinding - https://feedback.bistudio.com/T167742
  • Fixed: Map icons (drawIcon) were always drawn as 4:3 aspect ratio irrespective of real screen aspect ratio - https://feedback.bistudio.com/T170754
  • Fixed: Engine/Scripting caused changes of slider control positions did not trigger SliderPosChanged eventhandler
  • Fixed: UIToTexture displays that contain another UI Texture would not load the sub-UI when they are first rendered - https://feedback.bistudio.com/T170766

Note: Shotguns that use Submunition should not have changed their spread, if they did then please inform us.

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
#

Terrain performance issue related to layermask
is that the issue people reported here?

restive pilot
hushed seal
#

I guess/hope the cache size got increased.

#

I doubt it only took three days to make the whole thing async 😅

whole cloud
#

i doubled the cache

inland dew
whole cloud
#

no

full nova
languid heath
oak minnow
#

That was quick 😄

languid heath
#

Right

spiral pond
#

if you are on a different branch, but overwrite the exe with the perf/prof, will updates on other branches overwrite the exe again or does a branch not necessarily have also the exes in them? or is every branch always standalone (contains all pbos and exe)?

rain moth
#

You can apply the win/linux profiling depot on top of your other branch afterwards with some scripting

spiral pond
#

Thanks. Isnt it easier then to move prof/perf exe+dta into a custom modfolder instead?

rain moth
#

You can fetch the profiling binary depot to a separate folder if you prefer that

whole cloud
#

Forgot to mention, scaling of map icons was fixed.
That will be noticable on the ACE map tools, they'll stretch horizontally because they try to workaround the bug that now doesn't exist anymore

spiral pond
#

This was mainly noticeable with larger icons?

whole cloud
#

ye

sage zenith
#

@whole cloud i have since a long time a issue with my arma server, i know thats a lot of text but i have to explain all of the last ~9 month
**The entire dedicated box get a blue screen if more than are ~15 players playing I have this massive problems since August ~September of the last year. **

I'm hosting currently (since september) 4 Arma3 Exile server at one dedicated box (provider= hetzner.de)
It was now since ~9 month not possible to get more than 10-15 stable for one server period (4h), if there are to many players, the root machine just crashs with Kernel-Power ID 41 Reason
"The system restarted without first shutting down gracefully. This error can occur when the system has stopped responding or has crashed, or the power has been unexpectedly lost."

I tried in that last many month so much stuff to figure out whats wrong there. At fist i thought that the root machine is just broken and switched... At least i tried 3 diffrent AMD server (Ryzen 9 5950x) and between this time i also tried to remove AI script, use the latest available AI scripts, removed old ones, removed parts of my own overwrite mod and so one... then i thought maybe arma has a compatibily problem with AMD hardware ...okay then i tried intel... Now im on a i9 9900k and still the same problem dedicated box freeze complely on more then 10-15 players. The crahes happens all time if at one server more than 10 players are playing, so if on each server just ~6-7 players all its fine, it also happend while i just run 1 server and more than 10 players was online at the same time🙄
~1 Year ago i had up to ~50 players on the entire root machine with no crash problems, all runs fine on that AMD R9 5950x, sadly i dont made a 100% backup of this time where it was working perfectly

#

Then i asked @kindred radish after help what i can try, so he told me I should try the performence branches because since a arma patches there was problems with random crashs like mine but they should already fixed, according to his statement..
So then i tried all dev branches, still the same. I reduced AI count at all server on the same time to... still the same problem
The Rpt logs shows no critical error to me what can cause a blue screen of the entire dedicated root box... I have no crash dump files from arma because arma dont crashs directly it crahs the whole machine. Sadly Windows dont create any bluescreen crash dump reports on a Kernel power- applikation hang crash. I have just nothing, just my half rpt.log until the crash happens
I already spent, because of the often dedicated box changes, ~400 € without any improvements or changings of that problem, i just dont know whats wrong with that server, on weekend now i will try to just redownload everything again (but i already tried this some month ago) just to be sure thats not just a dump mistake what i did, maybe its a dump mistake by my own side, maybe not i just dont know.
So thats the current situation since 9 months and ~400€ later, I hosting arma exile server since 7,5 years now, so im not completly new to that.

Serverstats of my 3 AMD Servers (from past)

Ryzen9 5950x
128 GB DDR4 RAM
4 TB M2.NVme drive

Now im on an

Intel Core i9 9900k
64gb DDR4 RAM
1TB NVME

I use Windows Server 2022 Standard Version: 21H2
Mods what I use: @Exile;@CBA_A3;@PS46;@CUP_Units;@CUP_Vehicles;@CUP_Weapons;@NIArmsallinone;@BWMod;@Enhanced_Movement;
So no really weird or unormal mod stuff, many exile server use that too

So maybe you or someone other here have idea what i can try

whole cloud
#

friday late evening is bad timing for long text

sage zenith
#

no problem its okay if you can take a look on monday

#

i have this problems since 9 month so i guess i can wait some days longer

#

if i miss something just ask, i can also send rpt. log filles of times where the servers had crashed (last time yesterday while 10 players was at server 1 and 6 more on the other 3 server)

spiral pond
#

@sage zenith can you drop the rpt. curious to see how it looks
did you try to monitor cpu and ram usage?

sage zenith
#

the RAM usage of the servers themself are between 3,3GB (Namalsk) and up to 7gb for my main altis server (using arma + exile 64bit version)

restive pilot
#

well in any case drop one of the rpts here

sage zenith
#

fun fact in january i tried hosting over my 2nd home PC (I7 6900k @4,4GHz + 32gb DDR4 ram) all 4 servers runs without any crashs too
but bvecause of the prroblem that locally the ip gets changed every time i just got 1 day with ~9 player at one server and in sum ~13 players at the same time so i cant be sure if its also would crash on my home pc with mor players or not

#

okay

#

sorry

#

i want to upload that from crash yesterday, too but its to large for pastebin.com... only can 512bk??
Im wrong or was the limit in past higher, had in past never problems to post the entire rpt log... and with 6k lines its not so ultra large

#

here that one from yesterday, the last dedicated root crash happens 22.35.52sec (windows syshtem event display)
i removed some typical arma stuff like loading pbos, and error bone messages to can it post i made . inside this lines where i removed stuff so dont wonder
wait was the wrong one

spiral pond
#

pack as 7z

#

so after ~25-30min the server went down altogether?

sage zenith
#

sometimes he reach 2,5 or 3h

#

its random depends on player counts

#

here i have some historic log filles there happens just 3 blue screen crash within less than 30 minutes
i saved all .rpt from all servers

#

i pacekd them into .rar because pastbin for 12 logs need to much time^^

spiral pond
#

intercept_x64.dll

#

i think Dedmen just fixed some crashes related to it

sage zenith
spiral pond
#

may or may not help

sage zenith
#

(by uninstalling the entire custom loot spawner)

#

but has no effect still crash by ~10 player

spiral pond
#

hard for me to say if/how much Arma related

sage zenith
#

when i removed that and its still the same problem its not related for me

#

with my mind logic

#

but if someone or dedmen menas that could be the reason and maybe something other too, i will just uninstall/delete this... I will try everything what maybe could help, wasted 9 month in my free time and i dont got any good results

spiral pond
#

things i would try but likely unrelated:

  1. drop and let auto detection sort it out

-limitFPS=100 -malloc=tbb4malloc_bi -enableHT

  1. fix

18:09:08 Error in expression <"SQL",_parameters]joinString":";_result=parseSimpleArray("extDB3"callExtension _>
18:09:08 Error position: <parseSimpleArray("extDB3"callExtension _>
18:09:08 Error parseSimpleArray format error
18:09:08 File mpmissions__cur_mp.Altis\ГБВЩР\  \ЪЪАШБЯ\ \ЙПТ\ \  \  ..., line 1

  1. fix

18:09:29 Error in expression <}) then
{

_obj setPosATL _objPos;
_obj setVectorDirAndUp (_x select 3);
}
else

18:09:29 Error position: <setVectorDirAndUp (_x select 3);
}
else

18:09:29 Error 0 elements provided, 2 expected
18:09:29 File x\addons\dms\scripts\fn_ImportFromM3E.sqf..., line 63

  1. fix

18:15:25 Cannot create non-ai vehicle Land_Cargo_Tower_V4_Fv,
18:15:25 Error in expression <bleDynamicSimulation true;
_character setPosATL (_obj buildingPos (round(rando>
18:15:25 Error position: <setPosATL (_obj buildingPos (round(rando>
18:15:25 Error 0 elements provided, 3 expected
18:15:25 File q\addons\custom_server\Compiles\Units\GMS_fnc_placeCharacterInBuilding.sqf..., line 13

  1. [processReporter] Active SQF seems to have some "issues" with logging of scripts with non ASCI characters
sage zenith
#

to 1
i alread use that

to 2
mhh its a funny error just get display if i obusqf my mission file😅 but i will take a look
and not on all servers if i remember correct

to3
Is form a AI mission where some objects/ buildings could not be loaded can fix that but im sure the server get no blues screen because 3 buildings couldnd be loaded😅
but i will take a look there

to 4
should be the same basic issue like the 3. but with the diffrence that the server blue sceens happens too altrogh the entire GMS mission systhem is removed from the server

to5
could you give me more details like what log file you found that and on which time/line pls

spiral pond
#

ie

18:32:11 "V71 (23-01-2019) [processReporter] Active SQF: [""DyCE_fnc_convoyController"",""mpmissions__cur_mp.Altis\БЙИР\  \  "",true,1] @ 241.114"

sage zenith
#

because of this?

#

mp.AltisБЙИР  

#

its due to privat obfuscating the mission.pbo file

#

but that log line is from occupation extension (roaming ai script for exile)
i also tried completly whithout that, still crash with 10 players

plain trout
#

alright, I have a possible repo for you. Once the VRAM hits the max, the tree textures/Lods are flickering like crazy

whole cloud
#

I've heard of many possible repro's but never got any that actually worked for me

plain trout
#

if you fly around on livonia you get the vram filled up at some point, and if you then keep flying around it start becoming unstable

#

until arma dies

woven loom
#

It has worked 100% of the time, and I've tried on multiple occasions over the years

woven loom
#

Because what I've seen matches what KajSchak describes

plain trout
#

I have an 7900XT and once the 20GB are filled up the flickering starts kicking in

woven loom
#

Oh wow. My GPU has only 6GB VRAM so that's full almost immediately with GM assets + very high / ultra textures

#

So I assumed with more VRAM the situation would be much rarer / harder to trigger

plain trout
#

Depends on the map. This was on Gulf Coast. In the past I managed to get the issues on Livonia with 8GB of vram after 1-2 minutes. Just fly around enough and fill up the vram. System memory allocator handles this better then any other, although technically that shouldn’t affect the vram?

woven loom
#

Weferlingen will trigger it sooner too, since its assets also use 4K textures afaik

#

I also had this with Gabreta while in the editor

#

Probably setting viewdistance to full will accelerate the process too if you're just flying around

hushed seal
#

any chance of getting diag_captureSlowFrame a server version so it outputs to file?

weak panther
random isle
#

wtf you use editor for ?

calm mantle
#

I can't post a screenshot, but my vram was only around 3GB. during this. Testing in SP, I can get it up to 10 before my frame rate dies. Even then I do not see flickering.

obsidian relic
whole cloud
hushed seal
whole cloud
#

normal captureFrame and slow frame both fall into the same thing, and logs to RPT Capture Frame saved to file

fickle geyser
#

I've been trying it around 2y ago and slow frame variant never worked for me on dedicated server.

spiral pond
#

works only with prof exe, or not even with that?

hushed seal
#

I tried it with the profiling.exe, no bueno

spiral pond
#

darn. so we need to pray to Ded and KK 🙏

dire lotus
#

^^ same

#

-filepatching fixed it for me

#

didnt need it in v3

hushed seal
random isle
silk summit
#

Steam is officially dropping support for Windows 7 and Windows 8 soon, mostly due to security concerns but also due to those two windows being used by less than 2% of all users and dropping fast. Perhaps it is a good time for Arma 3 to also breaking free from those older systems. I remember some newer features not being implemented due to "having to support" windows 7 and since A3 requires Steam and Steam won't run in windows 7 anymore, it could be a good time to look into that 😉

whole cloud
#

I can't think of any feature besides thread naming which isn't public anyway

silk summit
#

Well, its been a long time since someone mentioned "legacy support" as a reason not to implement "x optimisation" so perhaps since then those previous restrictions have already been revised and if that is the case please disregard my suggestion ws

whole cloud
#

And even if. Steam dropping win 7 doesn't mean we can too

silk pewter
#

scarce resources that could be assigned elsewhere 👀

silk summit
whole cloud
#

Arma server doesnt need steam

silk summit
empty goblet
whole cloud
#

Steam dropping support also doesn't necessarily mean that you cannot run steam games anymore.
Maybe people will be allowed to keep running old steam versions or whatever.
Now is too soon to tell what steam does in a year

silk summit
empty goblet
#

most likely Valve may offer way to run the W7/W8 users in offline mode only (conserved state of client)

silk summit
silk summit
woven loom
#

Arma 3 DX12 update when? 😛

hexed verge
#

still prefer the dayz renderer upgrade first before dx12

full nova
#

what stopping us from getting it on arma 3?

worldly badge
#

Probably simply too much work for three-men team

full nova
#

but.. but... 60fps arma 3..... SaiaNLT

quaint flame
#

Buy a decent cpu and you'll have that today 😉

autumn timber
#

Sadly, he's right, you know? 🥲

worldly badge
#

Buy a decent game and you'll have that today 🙃

quaint flame
#

Minesweeper with 60fps

autumn timber
#

A 13600k gives you 2k@60fps on Ultra on Tanoa

#

Okay. 57fps 😂

rain moth
#

but but 4k??

quaint flame
#

I can confirm. I recently upgrade from a 3570k to a 13600 and it's insane

autumn timber
quaint flame
#

Sounds legit

autumn timber
#

Besides it's not like you need that 4k monitor to hunt pixels hiding in bushes, in Arma, right?
...right, guys? 🙄

rain moth
#

1080p on 48" monitor is not ideal 😛

autumn timber
worldly badge
#

Joris

autumn timber
#

Oh, didn't know that. I thought that he was a team leader, more of a managing person

worldly badge
#

I mean, true, but I count him as the member

rose moth
#

The running joke in the team is that I'm a part-time Junior Designer 😅

worldly badge
#

Mmmmaybe we can count reyhard as well, but anyways

analog acorn
#

I think NeilZar does some stuff for A3 too, but like as a side thing

rose moth
#

I'll also just shout out to @patent dome as our Deployment Engineer and making all pipeline magic happen, and Wulf our QA Tester. And indeed Reyhard as the main person remaining to do most data fixes and tweaks. As well as several other folks doing ad-hoc dev support.

autumn timber
#

There turns out to be quite a lot of people working on A3 still!

...DayZ renderer in Arma wen?

patent dome
#

I do dabble in A3 engine programming so now and then, if something catches my eye and I have the time for it.

silk pewter
vivid rune
light cargo
# whole cloud Steam dropping support also doesn't necessarily mean that you cannot run steam g...

valve will keep a version for the os for a while then it'll be dropped and no longer offered. there's no such thing as offering a way to run offline mode unless you got a client backup.
when xp support was dropped they froze the steam client build. they dont offer steam for xp anymore but i think the updater offers the xp build if asked from said OS (not sure of that). but anyways there's a bit of a special case there because some underlying changes to how the steam implementation works which are tokens requirement and manifest request codes enforcement. that happened after xp client stopped shipping making it non functional.
there are third party software that can emulate steam services that enable a client to run but its all work in progress of someone's hobby/free time @empty goblet

floral socket
#

not sure if this is the right channel, but what should i set arma3 exe affinity to in process lasso, a guy said take off cpu 0 , another said take off non physical cores , so im not sure

patent dome
#

If it even matters, I don't think there will be a definitive answer that works for everybody. Every CPU has different cores that perform best or worst, so which cores you need to set is different from what someone might be telling you. In the end, I don't see why you would do it, unless you need to lock it to specific cores for some reason.

heavy vortex
#

all non-physical is probably sensible but I'd be surprised if it made much difference.

calm mantle
# calm mantle I can't post a screenshot, but my vram was only around 3GB. during this. Testing...

In reference to this terrain object flickering.

The most recent event of this had my vram around 11GB out of 12. I've also noticed higher frame drops/ lockups with explosions while the objects are flickering. Almost like the high VRAM is making the standard explosion lag much more severe.

Also, it's worth noting that playing with low object textures does not delay the issue.

This is while running GM WS and SOG.

pale mica
patent dome
#

That is the scenario I was trying to give as "lock it to specific cores for some reason", but couldn't find the correct words to explain it. Although it seemed like he is trying to just get more FPS on the client.

woven loom
heavy vortex
#

It's usually not a good idea for an application to do anything other than decide how many threads to run.

#

Because otherwise you're fighting with the OS.

pale mica
woven loom
#

I don't know enough to comment on that really

heavy vortex
#

Headless clients are just hacks to make up for the server lacking sufficient threading.

#

I think server AI threading improved recently too, at least on perf?

pale mica
#

please let us know if we can ditch the headless clients. Scripting AI with HC is always a pain.

heavy vortex
#

you gotta try it and see, really.

woven loom
whole cloud
whole cloud
pale mica
#

Maybe on Arma 4 then

inland dew
#

@whole cloud do you know why when you're trying to change overall, objects or pip distance in video settings, by starting to enter a number with the keyboard, no matter what number you enter, it always writes 5 first, then you have to delete this number 5 and only then you can write your number?

wasn't always like this.
previously you could experience this only when trying to change pip distance by starting to enter it with the keyboard.
now it's also happens with overall and objects

really annoying

whole cloud
whole cloud
inland dew
#

before latest perf/prof, it was only the case with pip slider.
now it's also for overall and objects

#

don't know about stable, since never using it

silk pewter
#

View distance had this issue where typing 1 for e.g 1500 would automatically fill 500

inland dew
#

I see that with current perf you also can't change anymore separately from overall, the distance for objects and/or PiP, like at all.
you have to use overall slider to adjust objects and pip.
and pip and objects are now always the same value, namely ~50% from overall, for both...

#

that sucks big time

random isle
#

Lol fun things happen when you get into static weapon and then attach it to your self and play a move 😂

#

Don’t ask why lmao

fickle geyser
#

So I've tried maxFileCacheSize and not sure if it's working.

#

-setThreadCharacteristics -maxFileCacheSize=4096
I've set it to 4GB and my whole arma was using 1.6-1.7gb during 3h co10 Escape session.

image.png
tribal pasture
random isle
#

you become basically an air plane you fly around map and can actually turn left right up and down

whole cloud
#

Though i found it's rather.. eager in purging stuff from the cache sadly. Everytime you tab out, it does a flush and throws everything out that isn't currently in use

plain trout
#

where to dump the no message arma crash memory dumps/rpt?

#

finally created one

whole cloud
#

somewhere where I can get it

plain trout
#

I started moving the camera in eden editor when the mission file was still loading

whole cloud
#

preprocessLineNumbers scrpt ran. Loaded a.. apparently very big file? 🤔

#

inside a UI eventhandler 👀

#

Who is processing files inside a eventhandler 😠

silk pewter
#

:mild-panic:

plain trout
#

you got a mod name? it's a pretty big modpack. Could it be the USAF AC130 which we haven't updated yet and is throwing a lot of error messages

whole cloud
# 
  •    scriptpath    "z\\ace\\addons\\medical_gui\\functions\\fnc_modifyActionTriageLevel.sqf"

MouseMoving eventhandler

#

Seems its on the Eden display

plain trout
whole cloud
#

Yes its MouseMoving eventhandler, on Eden MainDisplay
That in line 33 of unknown file, does preprocessFile on that ACE function

#

I cannot find code that does this inside ACE (and tbh didn't expect to)

plain trout
whole cloud
#

no MouseMoving in there

# 
    if (!(missionNamespace getVariable ["cba_xeh_isProcessed", false])) then {
        diag_log text format ['[%1] (%2) %3: %4', toUpper 'cba', 'xeh', 'INFO', format["missionNamespace processed [%1]",  (missionNamespace getVariable ["cba_xeh_isProcessed", false])]];
        [] call CBA_fnc_preInit;
    };

This is the content of the MouseMoving eventhandler 🤔

#

Maybe its just preInit?

#

So all the mods are just initializing (at a unexpected weird time)
That will take a long time if your hard drive or PC is slow. But the game should just freeze for a bit (drop the freeze dumps) and then recover from it

plain trout
#

it's plenty fast, it does take time though. But on this load the game just vanished, on the second time the loading worked fine, just slow

whole cloud
#

Freeze dumps don't kill the game (Except on diag binary)
And there is no actual crash dump in the files, only that freeze
so blobdoggoshruggoogly

plain trout
#

oh well, I'll keep looking out

#

windows event logging was empty too

#

thank you for what you are doing

fickle geyser
#

Just felt bit weird that it used so little on bigger mod pack while it takes 4gb when I open mission in Eden on almost vanilla 😄

silk pewter
fickle geyser
#

Not sure if it affects me tho, I always run borderless and noPause thomp

hushed seal
silk pewter
#

we actually mine bitcoins since 2001

#

we're in your walls PCs!

hushed seal
#

Mining on all the unused threads would certainly make it look like it's multithreading 😄

whole cloud
#

2.12.150487 new PROFILING branch with PERFORMANCE binaries, v3, server and client, windows 32/64-bit, linux server 32/64-bit

  • Tweaked: Terrain Quality video options limits (higher view distance limits maximum available quality less)
  • Tweaked: UI to Texture now supports mipmaps - https://feedback.bistudio.com/T170799
  • Tweaked: Generic performance improvements
  • Tweaked: Added recursion protection to LBSelChanged eventhandler
  • Tweaked: Performance optimization for configClasses command
  • Tweaked: SQF Performance improvements (type check, array returns)
  • Fixed: Eden Editor Init script boxes did not correctly display compilation errors
  • Reverted: Fixed: Changes of slider control positions by engine/scripting did not trigger the "SliderPosChanged" Event Handler

Note: Please pay attention to unexpected/newly appearing script type errors like "Type ... Expected ...". If you think its wrong, please notify us.

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
#

@inland dew that should fix all your issues

spiral pond
#

awesome once again 👏

#

Tweaked: Terrain Quality video options limits (higher view distance limits maximum available quality less)
anywhere to learn more about this please?

restive pilot
#

Also what does that sentence even mean? 😅

whole cloud
#

The sentence means, the maximum view distance to use the high terrain quality settings, is now greater
The engine uses the view distance as input, to calculate approximate memory usage to store all these terrain squares.
And if it was too much for your VRAM, it will lock you out of the option, so if you increase the view distance, the higher options start to disappear

That calculation was waaaaay off though.

#

Like previously the algorithm calculated that for 16km distance on max quality you'd need ~125gb of video memory.
While.. If I bypass that check and run 16km distance on max quality on my 2gb vram gpu.. it.. works.

We wanted to allow for ultra quality which, with that old calculation was basically only possible if you reduce view distance to 2km or below
Wouldn't make sense to add "extreme" options, if noone could use it

hushed seal
#

How is that going btw? Currently looking at my notes regarding that 😄

whole cloud
#

Sounds like it's time to scramble them again meowheart

hushed seal
#

Noo 😦 It's so useful knowing what's wrecking server performance

#

It's still that public variable... but now with Arma Script Profiler working I'll find the mod and kill it.

whole cloud
#

"working" xD

hushed seal
#

Well... if you're not using this branch 😄

inland dew
#

now for overall and pip, when you start to type the distance value with the keyboard, it directly shows the first number you enter with the keyboard.

but when you start to type distance value for objects with the keyboard, it still always shows as first value "5", regardless of the first value you enter with the keyboard... (((

whole cloud
#

But is that new or old

#

Must be old

inland dew
#

or it was for both, pip and objects

#

then there was the previous perf, which made it also do for overall

#

now overall and pip good, but objects always 5, regradless of what you start typing first

#

so you have to first start typing, then it results in first value 5, then you have to delete it and only then you can type the objects distance you wanted...

restive pilot
weak panther
whole cloud
#

Ah dangit

random isle
whole cloud
#

Didn't do nuthin

random isle
#

saw 250 when zoomed all the way in lol (thats with draw system in place)

#

on game load would go to 1000 never seen that before

whole cloud
#

that configClasses perf improvement is a 2x speedup on "true" configClasses _cfgPatches btw :harold:

quaint flame
#

only cfgPatches ? notlikemeow

restive pilot
#

no wuv for configProperties? 😢

whole cloud
#

Thats just the example I was testing on. Others probably too much faster.

#

config properties only has that problem in the non-"optimized" case where it runs your generic script
I shall fix it, but when I put short circuiting into that it will fix it anyway

spiral pond
#

16:43:09 Unsupported language english in stringtable
16:43:09 ➥ Context:WW2\LIB_Missions_p\Arsenal_p\s
16:43:09 Unsupported language english in stringtable
16:43:09 ➥ Context:WW2\LIB_Missions_p\Campaigns_p\XXX\missions\XXX.YYY\s

#

getting some strange cut off for some logging

#

➥ Context:#(ai,64,64,1)f

#

havent seen this before

empty goblet
wintry basin
#

ya? coooooooooooool

dire lotus
#

installing now, got some big coop sessions on mon/tues shall let you knmow if we have any crashes

whole cloud
#

2.12.150512 new PROFILING branch with PERFORMANCE binaries, v4, server and client, windows 32/64-bit, linux server 32/64-bit

  • Added: Experimental script performance optimization "short-circuiting" (See discord #dev_rc_branch channel for details)
  • Tweaked: Improved performance of select/apply loops
  • Tweaked: Improved performance of nular commands and variables
  • Fixed: Eden attribute slider textboxes stayed empty (since v3)
  • Fixed: enableWeaponDisassembly effects were not correctly synced in multiplayer (Requires profiling branch running on server and client owning the vehicle) - https://feedback.bistudio.com/T171651

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
restive pilot
#

Speaking of these optimizations, they're unscheduled right?

whole cloud
#

short-circuiting may also run in scheduled (small array with simple checks, arbitrary unspecified limit, don't depend on it), but mostly unscheduled only.
the other stuff applies no matter how its ran

vivid rune
#

Are working all performance improvements only with 64bit or is "short circuiting" the only 64bit specific?

whole cloud
#

short circuiting is 64bit.
The others are just small-ish tweaks

tepid river
#

Hi all, I'm not sure if I'm doing this right but my server still gives out Object not found errors despite adding this to my server_config.cfg?

# 

LogObjectNotFound = false; 
ignoreMissionLoadErrors = true;

};```
#

We're using the performance branch and its worked wonders for performance,

whole cloud
#

It turns off server obj not found messages, but not client ones

vale shoal
whole cloud
#

Can you post example?

vale shoal
#

13:35:14 Error: Object(6 : 548) not found 13:35:14 Error: Object(6 : 549) not found 13:35:14 Error: Object(6 : 550) not found 13:35:14 Error: Object(6 : 551) not found 13:35:14 Error: Object(6 : 552) not found 13:35:14 Error: Object(6 : 553) not found 13:35:15 Error: Object(8 : 662) not found 13:35:15 Error: Object(8 : 663) not found 13:35:15 Error: Object(8 : 664) not found 13:35:15 Error: Object(8 : 665) not found 13:35:15 Error: Object(8 : 666) not found 13:35:15 Error: Object(8 : 667) not found 13:35:15 Error: Object(20 : 109) not found 13:35:15 Error: Object(20 : 110) not found 13:35:15 Error: Object(20 : 111) not found 13:35:15 Error: Object(20 : 112) not found 13:35:15 Error: Object(20 : 113) not found 13:35:15 Error: Object(20 : 114) not found 13:35:15 Error: Object(4 : 690) not found 13:35:15 Error: Object(4 : 691) not found 13:35:15 Error: Object(4 : 692) not found 13:35:15 Error: Object(4 : 693) not found 13:35:15 Error: Object(4 : 694) not found 13:35:15 Error: Object(4 : 695) not found

whole cloud
#

Neither server nor client, interesting

#

Ah. Ok. Yes.

devenv_47Zs6ZRbZb.png
vale shoal
#

I mean, it's not a big of a deal as the RPT is getting spammed with all kinds of error logs anyway. At least "object ... not found" is not that annoying as "Can't change owner from 0 to 2", as this doesn't stop when it occurs once

whole cloud
#

Actually the one you posted, only happens on ownership change
And that code has a todo from 2014 on it to stop doing that 😄

I'll work near that anyway, I'll look at it when I do

vale shoal
#

2014? Ah well, it's aging like good wine I would say 😄

#

If it helps you: some say forceAddUniform is the main factor of that spam

whole cloud
#

I heard about a bug with forceAddUniform that would trigger that spam, but didn't get a ticket with repro 😢

hot hollow
#

Hmm i can maybe see if i can get one as my unit utalises that

vale shoal
#

Maybe there are some repros in the bug tracker already. El Rabito also found something related to SupplyXX which triggers that, he wrote that here #arma3_feedback_tracker message (it's a direct link to the message)

whole cloud
#

I specifically want the forceUniform thing.
I know there are many other object not found things

vale shoal
#

I try to get a repro

silk pewter
kindred radish
#

But i guess run forceAddUniform on the server for a client once with local check and once without
The one without should cause the object not found errors.

kindred radish
#

Also found this a while ago, not sure if someone tried a repro

When closing client inventory the server sometimes logs this below.
Just repeatedly open and close your inventory and this error should appear in the server .rpt.

18:21:14 Error: Object(6 : 6) not found
18:21:15 Error: Object(6 : 7) not found
18:21:16 Error: Object(6 : 8) not found
18:21:18 Error: Object(6 : 9) not found
18:21:19 Error: Object(6 : 10) not found
18:21:20 Error: Object(6 : 11) not found
18:21:23 Error: Object(6 : 13) not found
18:21:24 Error: Object(6 : 14) not found
18:21:25 Error: Object(6 : 15) not found
18:21:27 Error: Object(6 : 16) not found
vivid rune
# vale shoal I try to get a repro

I tried it quick and dirty:

  • login in a server as admin
  • put this in the dev console: player forceAddUniform "U_B_CombatUniform_mcam";
  • execute it local -> error message on server
vale shoal
#

Ah, yes. I also can confirm this.

vivid rune
#

BTW: Can someone confirm this:
I tried to run on a vanilla server "Escape 10 Altis" but if i load into mission (after map screen) i have a SegFault.

#

I run ded linux 32bit server and latest perf/prof version (v4); client win10 64bit prof v4

#

on prof v3 server mission starts; client was win10 64bit prof v4

noble badger
#

i downloaded the performance binary from the google drive, but i also tested switching to the profiling branch on steam beforehand... now, will switching branches give me the prof or the perf version of the binaries?

whole cloud
#

steam branch is default perf

quaint flame
#

Have you looked into marker creation at some point? I noticed that createMarker becomes very quickly very slow once there is a decent amount of markers ~5k

#

I am wondering what it's doing.

restive pilot
#

5k?! meowsweats

whole cloud
#

it was touched in that marker draw priority thing

#

if thats a new problem

#

if a new marker is added at not end (usually always is at end, except with the new priority thing) it needs to rebuild the hashmap of all markers

blazing wagon
#

When I run diag_captureSlowFrame ["total", 0.3];, it gives me an error "Error missing ;"

#

What am I doing wrong?

#

Okay well I'm sure if I had actually selected the right executable I'd be better off...

quaint flame
#

HOLY COW

#

With priority. It takes like 50 seconds

#

without barely 2

#

Let me get some concrete numbers

blazing wagon
#

Still not workin 😦

#

do I have to run profiling on my client as well, or just the server side?

quaint flame
#

Does the debug console does some caching to executed scripts?

whole cloud
#

if you run script on client oyu need to run on client

quaint flame
#

I think I broke it somehow

whole cloud
#

it remembers the last executed if you mean that

quaint flame
#

I put > _start = diag_tickTime; into it

and it throws

for "_x" from _minX to _maxX step _resX >
21:46:25   Error position: <for "_x" from _minX to _maxX step _resX >
21:46:25   Error Missing ;`

Which is from a script I executed earlier 😄

blazing wagon
quaint flame
#

Yeah. Restarting Eden Editor fixed it

#

I thought I was going nuts

blazing wagon
#

Is there any easy to server exec the captureSlowFrame from debug console and have the window open back up on my client?

silk summit
# restive pilot 5k?! <:meowsweats:707626030613135390> 
VAL_fn_tileFog = {
    _tileSize = 100;
    private _tiles = worldSize / _tileSize;
    _smallmks = [];
    
    for "_x" from 1 to _tiles do {
        for "_y" from 1 to _tiles do {
            _pos = [_x*_tileSize - _tileSize/2, _y*_tileSize - _tileSize/2, 0];
            _pos = [(((floor(((_pos select 0)-0) / 100))*100)+50+0), (((floor(((_pos select 1)-0) / 100))*100)+50+0)];
            _mk = createMarker ["small_" + str _pos, _pos];
            _smallmks pushBack _mk;
            _mk setMarkerShape "RECTANGLE";
            _mk setMarkerColor "ColorBlack";
            _mk setMarkerBrush "SolidFull";
            _mk setMarkerSize [_tileSize / 2, _tileSize / 2];
        };
    };
    V_UPDATETILE = {
        {
            _mkpos = getMarkerPos _x;
            _near = allPlayers inAreaArray [_mkpos, 250, 250];
            if (_near isEqualTo []) then { continue };
            _near = _near inAreaArray [_mkpos, 150, 150];
            if (_near isEqualTo []) then { _x setMarkerAlpha .5 } else { deleteMarker _x };
        } forEach _smallmks;
    };    

    while { true } do {
        sleep 1;
        if (diag_fps > 15) then {
            call V_UPDATETILE;
        };
    };
};``` ----> this can create thousands of markers depending on the map size 😅
blazing wagon
#

Still not working 🤔 I'm running profiling Steam branch on client side, and profiling branch on server side

#

Still errors out with "Error missing ;" when I run diag_captureSlowFrame ["total", 0.3];

quaint flame
# 

Create markers without drawPriority
21:51:26 0.0839844
21:51:52 "Marker Count: 17856"

Create markers with drawPriority
21:52:39 1.54785
21:52:39 "Marker Count: 17856"

// Deleting 17856 markers
21:54:13 23.9067
21:54:13 "Marker Count: 0"

Okay, so the difference is not caused by drawPriority, but by actually deleting all these markers

#

So creating markers is that much faster than deleting them? That's weird

whole cloud
#

server will just dump into a file

whole cloud
blazing wagon
#

👍

#

I think it's still not running profiling branch

#

maybe cause of A3 launcher

#

says the .exe is still arma3_x64.exe even though I'm on profiling

whole cloud
#

if you delete from start of array, all other elements need to be moved up and the hashmap recreated.
if you delete from end, it should be fast
allmarkers thing returns in order

quaint flame
#

ah yes.

#

Might be worth a note on the wiki

restive pilot
#

I still don't understand what priority has to do with rebuilding a hashmap think_turtle

quaint flame
#

What hashmap?

restive pilot
silk summit
whole cloud
#

hashmap is [name,index]
to quickly find the index in the marker array

quaint flame
#

I wasn't able to find that without you

whole cloud
#

if all indices change, need to rebuild

restive pilot
#

wouldn't reassigning be faster than rebuilding?

#

the keys don't change after all blobdoggoshruggoogly

blazing wagon
#

My launcher shows the profiling banner across the top left, but when my game starts, I can see in task manager that it's still using arma3_x64.exe. Shouldn't it be launching the profiling_x64.exe?

restive pilot
#

no. you should launch the profiling exe manually. the launcher just means you're using the profiling branch, not the exe

whole cloud
quaint flame
#

22:08:07 25.4492
still slow

#

17k markers with forEachReversed

blazing wagon
whole cloud
#

if its important i can look at that marker next week

quaint flame
#

It's an extreme case I guess, so not really important.

blazing wagon
#

Anyone know an easy way to copy the A3 launcher startup modlist to command line?

#

Doesn't show in my RPT because it's too long and copying these manually is painful

quaint flame
#

-par parameter

#

sorry, misread

blazing wagon
#

all good 🙂

lethal trench
#

Task Manager? 🤞 nvm somehow I thought it had the params, it doesn't

restive pilot
#

it gives you the exact command lines and you can just copy paste them into a bat

blazing wagon
#

appreciate, will try now

quaint flame
#

😮

hushed seal
marsh gyro
#

Can the server run performance branch or profiling, and the clients run regular? I've enabled perf on my server and I can't find it any longer.

boreal wigeon
marsh gyro
#

When I launch the server with profiling, and I run stable, I can't connect to it.

full nova
#

how so?
most of official servers (like warlords) are using the perf_branch

nocturne obsidian
#

the client is compatible with any 1.52 stable server? in the forum i found the statement "(compatible with STABLE clients)" - but how vice versa?

boreal wigeon
empty goblet
#

both way

nocturne obsidian
#

thanks

#

played koth on some french server. 120ppl, w/o perf binary i had 25-30 fps, with the perf binary i got 30-35. besides the server is overloaded, thats a lot

#

they advice ppl to use the perf binary with BE messages. would the server load sink, if more players would use the perf binary?

spiral pond
#

can someone confirm the A3 server on profiling branch doesnt have the cdlc folders?

patent dome
#

Yea, Profiling and CDLC are different branches for the server app

spiral pond
#

i see. you can use the workshop mod versions of the cdlcs instead, i guess?

patent dome
#

If I'm not mistaken, they don't have the map data. So, yes to an extent. You could also get the Profiling binaries from the google drive and use those when the CDLC branch is selected.

fickle geyser
#

they do have map data

spiral pond
#

thanks 👍

fickle geyser
#

compat data must be equal to cdlc data for multiplayer to work properly.

rain moth
#

either apply CDLC depots on top of profiling or even easier, apply profiling depot on top of creatordlc branch

sinful smelt
#

I know this is probably just something I"ll have to pick through the "v03 crash fix" patche(s) and try to understand what's changed w/r/t type management of operators, but just in case someone could give me a point in the right direction --

Ever since the 2.12 v03 patches landed that changed the typing code for operators, custom-registered SQF functions from intercept plugins seem to return fully-functional game_values, except for somehow the type metadata. (i.e. causes script errors when passed to functions that accept that custom type, and typeName returns a concatenated list of all registered types). After a cursory look, I'm assuming there's something that ole' trusty register_sqf_function needs to handle differently now? or am I barking up the wrong tree?

Thanks in advance for the pointers... could save me loads of guesswork.

whole cloud
#

I thought i fixed it all 🤔
But i only tested that commands work and show correct type info in debug console. Which I'd expect to be enough

sinful smelt
#

@whole cloud - I ran a quick test, and it seems that the "dubious" return values with incorrect typeName returns and other bad behavior (despite definitely being the right value -- str _value and friends return as expected) all relate to functions that return custom types registered with register_sqf_type. Looking through the code, it seems like some extra data space was added to either sqf_script_type itself, or at the very least interleaved with each sqf_script_type in an operator definition.

Don't take this as gospel, as I haven't spent hours digging yet, but even a point in the right direction as to what the original intent w/ this was (assuming it relates to type-checking improvements for short-circuit), is a help.

As an aside, a comment showing the intended unary_operator/binary_operator layout differences in addition to the runtime "fix" would make that kind of thing easier to understand (even if just mentioned somewhere in the PR on GitHub or in the git log). I was originally only looking at the unary one, so until I saw the *2 in the binary_operator class, I didn't grok that we were offsetting once per instance of sqf_script_type, so I was quite confused 🤣

If I can do anything to help you help me, let me know. The other guys who don't use the GDrive all reverted to stable branch temporarily, and they're not happy w/ me about it 🤣

#

If the size difference is actually in sqf_script_type itself (which seems unlikely given that it's stored statically in many locations which did not have padding added in the patch), then that might explain it, but based on the patch it didn't seem like that was the intent, so I assumed not.

whole cloud
#

Ah its only custom types. yeah that may be broken

sinful smelt
#

I think so

#

don't take that as gospel quite yet though my test was quite cursory

whole cloud
#

it is in the type, type is now larger
Command registration accounts for it, type doesn't

sinful smelt
#

ahhhh... ding ding

whole cloud
#

Actually I wrote that in my intercept channel on my discord

Discord_ZV3DapTpVY.png
sinful smelt
#

😦 whelp, sorry I guess I didn't read back far enough

whole cloud
#

seems like register_sqf_type just needs to allocate more space, and zero it out
and... somehow return a type struct that's 8 byte larger, but only when running the new build

But only if you actually use that type struct, chances are you don't
I think none of my projects actually need that, so you may get away without it
And I think compound type is fine

Most likely just allocating more space and zeroing it is fine

sinful smelt
#

Honestly runtime binary compatibility isn't really an issue for me. I'm perfectly happy having that interoperability gated behind compile-time flags to avoid the runtime hack

whole cloud
#

Well you are, but I'll need to fix it properly for intercept itself. But if you could test if making a larger allocation is enough, that makes my fix easier

sinful smelt
#

Yea, ofc kekw - going through it right now... just gotta revert the offset hackery as well

#

Thanks for the help. Unfortunately ran out of time before work, so I have to go for a few, but I'll get back to ya tonight (NA) with findings

whole cloud
#

Actually very smart to solve the type passing by enum/index instead of passing the actual type structs. So changes like this can be done in backend only.
Good job me! meowheart
Bad job me for passing the type struct anyway. Seems noone not even me is using that anyway :u

empty goblet
#

clientside they get better FPS if they use PERF binary ... it will have neglible effect on server w/e client they use, ofc server uses the perf server too

thorny frost
#

just to let me (trying) to understand: if it works so well, why you don`t replace the normal files? like putting perf into testing, make it stable then?

empty goblet
#

it already works that way, just replace the arma3.exe or arma3server.exe files ... and it will become part of stable in future ofc ...

sinful smelt
#

Awesome, thanks. I'll check it out now. Just getting back to this now, as things were a bit on 🔥 yesterday so I didn't get to it last night.

👍 to moving to pointers for sqf_script_type, and having the host extension allocate it since it should know the correct underlying sizes, even if the client is less up-to-date.

On that same note of ABI compatibility, though, is there a reason we pass the client_functions struct by value to the assign_functions function for intercept plugins? This makes it so that if you add a new function to that struct @ the end, it becomes ABI-incompatible, whereas a pass-by-reference/pointer then copy in the plugin would support such additions upstream w/o recompilation of the plugins.

sinful smelt
#

@whole cloud - I'm still experiencing issues, but maybe because of the linux differences w/ script_type_info. Based off of the changes in intercept, it looks like that type changed size in addition to sqf_script_type? Is that the case? If so, would you mind quickly checking if the linux definition of script_type_info is still accurate w/r/t what's in upstream (performance v04) (i.e. that it's still missing description, _category, and _typeName)?

Right now, simply calling typeName on a game_value returned from a custom function results in "generic error in expression"

#

If that is accurate, then maybe the 2x space allocated is not enough to hold the new information on linux (because the base size is less)?

thorny frost
#

very cool

spiral pond
#

is it just me or something broke in perf/prof (the cut off)

Unsupported language English in stringtable
➥ Context:WW2\Assets_c\LIB_Characters\Core_c\s

Error in expression <getPos>
Error position: <>
Error Invalid number in expression
[] L140 (//temp/bin/A3/Ui_f/scripts/GUI/RscDebu

➥ Context: [] L5 (/x/cba/addons/common/XEH_postInit.sqf)
[] L21 ()
[] L17 ()
[] L19 ()
[] L77 (WW2\Core_f\LIB_System_TargetRadar_F\functions\radar\fn_animate.sqf)
[] L10 (WW2\Core_f\LIB_System_TargetRadar_F\functions\radar\fn_anim_flat.sqf)
[] L58 (WW2\Core_f\LIB_System_TargetRadar_F\functions\radar\fn_anim_flat.sqf)
[] L63 (WW2\Core_f\LIB_System_TargetRadar_F\functions\radar\fn_a

#

Original output filename: Arma3Retail_DX11_x64
Exe timestamp: 2023/04/18 13:19:44

Type: Public
Build: Stable
Version: 2.12.150512

whole cloud
whole cloud
spiral pond
#

have this in all for a while, yet only for these two types (as far as i can tell)

#

is there any other with of the context type?

silk pewter
#

anyone else on profiling has this?

image.png
#

using this opportunity to say I could enjoy a 0.1 or 0.01 precision here, as fog decay is set to 0.5% by default but just clicking on the number makes it 1.0%

plain trout
silk pewter
#

adding to the mix that typing negative values in Base does take the abs value

thanks for confirmation!

quaint flame
silk pewter
worldly badge
#

LouMontana in the sourcecode?!?!?

silk pewter
#

maybe in the credits, who knows

desert olive
#

is there anything to learn from this message ? the server straight up segfaulted

Apr 30 17:05:29 novak kernel: [30491973.501370] Code: 75 e7 c3 66 0f 1f 44 00 00 4c 8b 05 a9 cc 2a 01 31 c0 eb 0d 0f 1f 44 00 00 48 83 c0 01 84 d2 74 20 0f b6 14 07 41 0f b6 0c 10 <0f> b6 14 06 41 0f be 14 10 38 ca 74 e2 0f be c1 29 d0 c3 0f 1f 40```
hushed seal
#

afaik the dump files are what's actually helpful

whole cloud
#

but thats linux and we don't have good dump files on linux 😢

whole cloud
whole cloud
#

Inventory code in A3 is: (╯°□°)╯︵ ┻━┻

#

Networking code in A3 is: (╯°□°)╯︵ ┻━┻

#

Combination of networking and inventory:

#

The server is triggering error because its leaking a entity and not deleting it when it should.
And then it continues spamming errors because it deleted a entity it shouldn't have deleted.

whole cloud
#

And if you want some bonus.
The addGoggles command, which, while it only was added with Arma 3, is (according to our code) a Operation Flashpoint/Arma 1 script command, has a funky feature.
If you call addGoggles, with a uniform classname. It will create the uniform item, create the invisible uniform inventory container entity, then (because uniform doesn't go into goggles slot) fail adding the item, discard the item, forget about the inventory container entity that's now lying around somewhere outside of world bounds and will from here on till forever be network synced regularly.
Also addGoggles will delete your previous goggles, even though it cannot add new ones

It will not spam "object not found" errors on server and other clients though, because it actually is network synced properly and every client will have this bogus unused entity

You can actually find them using count (-1 allObjects 4)

full nova
#

hopefully will be addressed by Arma 3 Remaster in 2035

rose moth
whole cloud
#

One interesting trick for those that want to fix this inventory container leak stuff via script.

You can find it via allObjects by checking for "Out vehicles"
and if it doesn't have a parent, its a leaked uniform/vest inventory container

(A uniform that was dropped on the ground, will have its weapon holder as parent)
weapon holders won't be in that list, because they are not "out" vehicles.

so
(-1 allObjects 4) select {isNull objectParent _x}
should be pretty reliable I think

Arma3NoOptInt_DX11_x64_WjegBuSQok.png
hushed seal
#

Is there any way of figuring out what script is causing 22:00:01 Can't change owner from 0 to 1136400883?

whole cloud
#

You mean after Server: Object info... not found ?