perf_prof_branch

[restive pilot]

@inland dew do you use the maxMem parameter?

[inland dew]

no

never used max mem or vram params

[heavy vortex]

out of interest, what are your page file settings?

[inland dew]

but I'm using mimalloc 2.12 with lock pages in Windows, for 2 MB chunks of memory, instead of 4 KB chunks of memory of the default Intel malloc

[inland dew]

auto
but it's not using it, because of 24 GB vRAM and 32 GB RAM
only minor stuff in the paging file, mainly from Windows

[heavy vortex]

Yeah it doesn't actually use it but it allocates it for texture backing, I think.

probably unrelated anyway. If that was an issue then it wouldn't just be Arma that had issues.

Usually Windows knows what it's doing :P

[inland dew]

HWiNFO says that it uses 0.6% of the paging file
but since it's auto, I don't know 0.6% of what 100% size it is

[heavy vortex]

I just noticed the 12000m infantry view distance on the right of the screenshot, so 16GB RAM usage makes a lot more sense now.

[inland dew]

even with like 4K view distance, RAM usage isn't lower

[inland dew]

using default Intel malloc with 4 KB memory chunks, doesn't change anything

[inland dew]

btw, this is a capture with patch 1.98.146597 on our PvE server with -limitFPS=1000 and 2 HCs
this way, server FPS went down like a meteorite within like few hours or less, without server FPS recovery, despite all the garbage, like corpses, wrecks and inventory stuff cleaned +- aggressively within reasonable intervals

with default 50 FPS and even without the 2 HCs, we can play +- ok for like 6 hrs full server, with like 200 AI, jets, helis, armor, arty etc

[halcyon iron]

but you're using a 60hz monitor though right

[vivid rune]

Don't play with our feelings. This was a little rollercoaster ride.🥺

[vivid rune]

The 50Hz are reffering to the server. Other games would call it "server tick rate" (e.g. Battlefield) and would describe more a "response rate" of the server regarding the data exchange with the connected clients . It has nothing to do with client monitor refresh rate.

[empty goblet]

32GB (or more?) RAM cases or same bug exists for VRAM too ?

[restive pilot]

it doesn't have anything to do with the amount of RAM (or VRAM)
In fact the amount is detected correctly but it's not "used" properly

[empty goblet]

ok when you roll ticket or report on it, toss me link on slack, shall be interesting read 🤣

[restive pilot]

Well it should already be hard to believe that a 10 yo issue was fixed in 10 min

[inland dew]

been flying very low on the server for few hours in a Cessna, with -maxMem=22528 -maxFileCacheSize=6144 and I'm already at ~20 GB

will continue to fly, since I haven't covered even half of Altis yet

flying over each village many times, almost touching the roofs, also over every field and forest

[restive pilot]

So you mean so far performance has not degraded yet?

[inland dew]

will tell you later
frametime is spiky, but not that much and since I'm changing directions and locations a lot, FPS also changes, so difficult to say for sure, for now

12k view distance extreme settings 1440p

vRAM 14 GB

[inland dew]

performance has degraded (see high and uneven frametime and low FPS)

but!

the server restared, with the saved state of the mission, so "everything" is like before restart and ta-da! - FPS is like new and frametime flat and low, with me still having 20 GB RAM and past 15 GB vRAM occupied, when I reconnected

so either the server or me needs restart, to restore client FPS, in both cases

when I restart and reconnect, with the server still running for hours - good client FPS

when the server is restarted and me still in the game and reconnecting to it - also good client FPS

server FPS before restart and after 6 hrs of online and me playing on it also close to 6 hrs, was between 37-46

[inland dew]

client FPS gets restored both ways

[analog acorn]

What happens if you just DC and reconnect without any restarts?

[inland dew]

crap FPS and frametime still

so I restartcor the server, to restore client FPS in both cases

if the server is still running and I only disconnect and reconnect, without restarting the game, it's like I haven't left the server - same crappy performance

[inland dew]

so, you restart Arma, to reset RAM and vRAM and reconnect as clean client to a not fresh at all server, that's been online for many many hours - good client FPS

with your client, with pretty filled RAM and vRAM, after hours of playing, without restarting Arma, you connect to a freshly restarted server - also good client FPS

go figure...

this is good FPS and flat and low frametime, thus buttery smooth gameplay

[inland dew]

FPS was as high as one can see here and nice frametime, when RAM was at 20 GB and vRAM at 15+ GB, just because of server restart, despite tired client

maybe it has something to do with simulation or synchronization of stuff...
I have no ideas anymore

[halcyon iron]

yes but his never ending quest for fps is strange when he is using a 60hz monitor

[fickle geyser]

it's not, he plays PvP, games can feel a lot more responsive at higher frame rates.

[analog acorn]

Performance issues can affect everyone. The person investigating here and now may only have a 60hz monitor, but that doesn't mean people with higher refresh rates aren't also impacted.
Frame rate also affects script and simulation performance, and generally leads to smoother gameplay when it's high.
Also, having a high base frame rate means you have more room to eat into when things start to crunch.

[inland dew]

don't feed a troll that has nothing to add and doesn't contribute in any way to a solution

lets focus on the problem instead and ignore him

[vivid rune]

Mh. Idk where your focus of searching is, so i would share my guesses after thinking about these regular spikes in frame times: What if the problem not allocation itself but somthing with a growing array or something similar? Maybe the program is has to allocate more and more space for a growing object and that slows the graphic renderer(e.g. a sorting algorithm takes more and more time on finding or providing infos). Just my two cents.

[halcyon iron]

That’s very disrespectful, I type 2 comments and you’ve determined that? I’m surprised dwarden even wastes his time responding to you. Good luck

[restive pilot]

I'm not "searching" for anything. I just try the repros we have for the issue, and we do have the tools to detect where it occurs
the problem is that most repros take hours, which we don't have time or resources for

I figured out a repro that takes less than a minute, but whether that's the same problem as the one that occurs after long gameplay is still too early to tell (once we hopefully post a fix, the community can test and tell us if it fixes the long gameplay issue too)

[inland dew]

will gladly test that
you limited -maxMem to something really small?

[restive pilot]

no. that's not a repro 😅 (well it is but that's a bit silly)
it has something to do with growing hashmaps

[inland dew]

ok

[restive pilot]

actually maybe limiting maxMem to a small number could give us a quicker repro for the long gameplay issue (hopefully less than an hour)

[inland dew]

don't think so
must be something else than memory

but lets check the fix(es) and see

there is something between server and client going on

[inland dew]

this problem was always there, even before hashmaps existed in A3

[restive pilot]

I know. I mean it gives us a quick repro

[inland dew]

ah, ok

the thing is that one can't repro this in SP and while trying to repro it in MP, stuff also evolves on the server at the same time...

it also happens faster for the client, if the client has higher FPS

same for the server - the higher the FPS, the faster it will start to happen, with no possible recovery, unless restarted, like the client also

[heavy vortex]

Happening faster with higher FPS sounds more like a leak.

Although you can get increasing memory fragmentation with alloc/free spam even if nothing is technically leaking.

[inland dew]

the higher the FPS, the sooner it will be very low )

easier to repro on a server, since no image to display, so very high FPS is easily obtained vs. client

[heavy vortex]

Depends if the server needs to do anything as well.

I guess I can test that easily enough...

[warm vault]

How do i talk to a dev

[inland dew]

1000 FPS for server, to make it easier/faster

[heavy vortex]

My server does not go that high :P

[vivid rune]

1000FPS means "unlimited" or "as fast as it can".

[inland dew]

it won't do 1000 FPS anyway
just a number as high as possible

[heavy vortex]

You get 300. I can run that for most of a day and see if it slides.

[inland dew]

depends on the mission, how well scripted and heavy it is and whether PvP only, or PvE and how many players, AI, vehicles and cleaning stuff on the map

[heavy vortex]

That's kinda the trouble. It probably depends on one specific thing out of that list.

Unless there are multiple leaks :P

[inland dew]

yes

in our mission, there are many many bots, they're flying many jets, helis, driving a lot of armor, reinforcements arrive to the villages constantly

each of dozens of players can also recruit several bots and buy them armor or air and send them to attack

[heavy vortex]

hmm. Surprised it even maintains much more than 100fps then.

[inland dew]

it doesn't

or you're talking about your server?

[tribal bay]

Ive got crashes with perf_branch client when using a certain mod (CBA and BWMOD). The games crashes when useing the turn out on mod vehicles of the BW Mod. In the stable aka normal branch there are no issues. Whom Im reporting this issue to? Mod devs or perf branch devs since the issue only exist with this arma build

[heavy vortex]

I would lean Arma. If something crashes with perf but not stable then that's surely worth checking.

[restive pilot]

do you happen to have any crash dumps?

[tribal bay]

sure i did some testing so have a conclusion, what do you need? .rtp, bidmp or mdmp

[restive pilot]

rpt and mdump. plz send via DM

[silk pewter]

like to any other person, open your voice and make coherent sounds 😄

[hushed seal]

https://i.imgur.com/4Lku5i4.jpg
anyone having this fun issue with players that are just ... empty?

[restive pilot]

is it new?

[hushed seal]

nah, been talking to dedmen for a while about it, not sure if perf branch or mod related or something else. Just wanted to hear if anyone else has seen this and people in here have seen the most whack stuff.

[heavy vortex]

Is that just on the client or the server as well?

[hushed seal]

on the server as well, had a debug.exe that was logging players without a group

[boreal wigeon]

Yeah had it happen some times. Also once when a player left, the unit was still there being a player, and couldn't delete it.

[heavy vortex]

Now that one I've seen.

[plain trout]

Yes, I had it too very very rarely but also no idea of a repo.

Actually you can spawn them in manually by importing an sqf mission file, exported by the editor with the debug console if that mission file has a unit set to player (not „playable“). But I’m not sure if their state compares to what happens during an actual game

[hushed seal]

Well we get it every OP if players JIP into a mission^^ Unfortunately can't repo it with just two people so like always you need 10-20 people to test.

[plain trout]

We have 30-40 people regularly and it doesn’t happen ever. It’s only when people disconnect that this issue shows up. Maybe locality of the initialization is broken and init.sqf or initserver.sqf have some special setup?

[hushed seal]

https://github.com/16AA-Milsim/MissionFramework/tree/master
Is all the scripting, for the most it's been the same for ages.

[empty goblet]

any chance for the 'fix' test to manifest this week into profiling ?

[restive pilot]

I don't know

[whole cloud]

Can you send your RPT from that run? superflush also prints new diagnostics

[inland dew]

will see if I have it

[inland dew]

this?

or you need the while RPT?

[whole cloud]

Ah profiling exe has more data about script memory, forgot thats prof only.
Well its atleast doing something

[rain moth]

in tank coop event, half the crews have 2 fps and contant object flashing

[heavy vortex]

Replicable? :P

[restive pilot]

for the time being tell them to use -maxMem=80% of their ram (e.g. -maxMem=24576 if they have 32GB RAM). it should "postpone" the object flickering issue

[rain moth]

testing now

[rain moth]

no more issues so far for me

[restive pilot]

you should be ok with up to 100% of your RAM too but I didn't want to risk it 😅
hopefully we can fix it properly soon ™️

[whole cloud]

Oki leopard found one issue definitely.
The game's default maxMem parameter works like this

physicalMemory = 1024MB;
virtualMemory = 1228MB; // This is minimum required to run Altis on 32bit

physicalMemory = SystemStats.TotalPhysicalMemory * 75%;
if (physicalMemory > SystemStats.CurrentlyFreePagefileMemory)
{
      Rpt("Warning: memory usage limited by a page file. Current limit %d MB, wanted %d MB.", SystemStats.CurrentlyFreePagefileMemory, physicalMemory);
      Rpt("Increasing your page file size might improve game performance.");
      physicalMemory = SystemStats.CurrentlyFreePagefileMemory;
}

if (running on a 64bit system)
{
      virtualMemory = 4GB;
}

if (maxMem parameter is set)
{
      physicalMemory = maxMem;
      virtualMemory = maxMem * 75%;
}

if (virtualMemory > 8GB)
{
      virtualMemory = 8GB;
}

So the game thinks, it can only allocate 4gb of memory. unless maxMem is set. That is just hardcoded there.
I looked up the reason for this. Back in 2016 "limit is purely artifical an could be freely increased over 10 gigs if necessary."
"64-bit is currently limited to 8 GiB which is a rather pessimistic choice. However, it can be increased in time after we test our x64 game properly."
Okey then, I think we'll just get rid of it.. I think we tested it properly enough over the last 7 years.

I looked where this virtualMemory is actually used..
Handful of usages.. one of them is unloading blocks of terrain objects when the free memory gets critical.... Flickering trees.

We definitely still have some memory bug. Because it gets worse and worse over time, so something is still wrong.
Fixing this will give us a couple more gigabytes of headroom before the critical limit is reached and you feel it. But we still need to find that leak.
From my testing, I couldn't find any leak.

[silk pewter]

dang!

[random isle]

i had this issues few nights ago on (i think previous pref), like 4-5 hour game and the flicker started and fps were in 20s, but every couple of minutes it looked like it cleared the ram fps where back to 60s but then again back to 20s after another 30 sec, so was going back and forth

anyway i can get a good log w/o crash ?

[whole cloud]

no

[random isle]

ok possible to force the crash 🙂

[whole cloud]

no, and would still be useless

[hushed seal]

omg, might that be the reason that my server was trying it's hardest to stay below 4GB reported usage despite having a quarter terabyte of RAM? All that weird behaviour went away when I added -maxMem=16000

[whole cloud]

yes thats it

[hushed seal]

Feels good knowing changing that actually did something 😄 And good work finding that!

[woven loom]

Crazy stuff! Curious why 75% is chosen as a multiplier for things.
Also wondering why virtualMemory = physicalMemory*0.75 wasn't just done always if it was safe to do when maxMem was specified (they wanted to test more?)
It's also strange because iirc virtual memory in windows = RAM+page file, so virtual memory pool should be more than physical memory, afaik, but not a dev so could be very wrong about this.

[whole cloud]

I am also wondering that.
code is now
virtualMemory = physicalMemory.
Because that works

x64 linux server is hardcoded to 4GB limit. Again, unless you manually specify -maxMem

It actually does know how much ram it actually has.
But there is a "Throttle" limit, that it "tries" to stay under. And thats 4GB.

[woven loom]

Amazing

What effect does this throttling have on the server? It's not doing rendering and stuff so what would it unload etc?

[heavy vortex]

In my experience of -maxMem=3GB, not a lot.

It just keeps trying and freeing barely nothing, hence <5fps.

[whole cloud]

In windows in the Rendering code, it unloads blocks of terrain chunks.
I would expect that on server that rendering code doesn't even run

[hushed seal]

FPS take a shit once it nears 4096 MB RAM usage

or rather decrease and become a whole lot more unstable, whole reason I've started using maxmem

[woven loom]

Thanks! Will pass this on. Exciting times 😁

[whole cloud]

Prof v11 shall be tomorrow morning

[random isle]

dedmen this game still alive because of you 🙂 thank you for your care

[whole cloud]

Also the "free memory" considers all of the game's allocated memory.
So a lots of data stored in script variables, can also cause the terrain to flicker and go into "emergency mode"

That is the repro Leopard found, which creates a huge hashmap and causes the flickering.
So chances are, even if I look more for a memory leak. I might not find it because it might simply be scripts storing lots of stuff.

Simply creating some couple million element long arrays in a while loop, should pretty quickly lead to chaos.
And there's not really anything we can fix about that

[heavy vortex]

Could kinda do with a script memory diagnostic command.

[whole cloud]

profiling build has it-ish.
On superflush it prints script memory usage.

[heavy vortex]

mmm, need to try that.

[whole cloud]

But that just shows you that there probably is a problem.
Not where it is

I guess I could build a memory profiler with ArmaScriptProfiler.. but that's so much work :U
And the profiler isn't that well suited for memory

[feral harness]

please use appropriate channel #dedmen_praise_channel

[fickle geyser]

We need #LiveOps_Praise, easy to forget about other people as Deddy has biggest online presence due to all the writeups he does (which is great), we luv u all

[restive pilot]

yes virtual memory is the "committed memory" in Windows, which is the sum of RAM and page file

[restive pilot]

Simply creating some couple million element long arrays in a while loop, should pretty quickly lead to chaos.
And there's not really anything we can fix about that
actually, I had an idea to increase the limit gradually once its violated several times after cleanup. worst case scenario is that we get a crash if the user doesn't have enough page file, but at least that's better than a low FPS torture 😛

but anyway yeah we should find that leak, which is what I meant by "fixing it properly"

imo we should add some additional profiling stuff to detect these leaks. it's very hard to find them in "synthetic" tests

most people encounter these issues after several hours of gameplay, which is not a feasible repro for us

[vivid rune]

Is there a (easy) function in C++ to measure the size of objects/structs? So maybe a frequently scan over all structs to identify the biggest objects/structs over time?

[restive pilot]

size of a struct is constant

[whole cloud]

Also should test what happens if we just disable page file. And disable all the checks in Arma that say "you must have page file" and see if that actually is a problem

[restive pilot]

but anyway afaik no. you should manually detect allocations and deallocations

[whole cloud]

Memory profiler can run in retail.
But then Arma quickly uses like 16gb of ram to run

[vivid rune]

is there someting equvalent like lists in C#?

[whole cloud]

C++ does have linked lists yeah

[vivid rune]

This 3FPS Topic is only a A3 thing, right? So what was functional introduced in A3? Maybe this can also identify some spots to investigate.

[restive pilot]

well for starters it doesn't have to be so detailed 😅
first we should find what the source of the leak is. Groove_C mentioned that even after disconnecting from a mission the FPS was still bad and RAM usage didn't go down. so that's a start. we can monitor all stuff that should be cleaned when a mission has ended but don't

[charred holly]

Could memory fragmentation be a problem? Or is that handled on the OS level?

[restive pilot]

it can be a problem but this sounds more like a leak

[empty goblet]

similar like with VRAM, one of those reasons was there is always something in memory and not reliable to assume it's all free

[vivid rune]

Mh. Interesting: I run A3 with MaxMem=256 (yes 256MB) but MSI Afterburner show a usage of 2500MB (allocated 17GB). It runs with 3-10 FPS with graphical glitches but it doesn't crash. (i also used hugepages)

(tested with YAAB)

[vivid rune]

I can remember there was some memory thing for terrain loading on the server a few weeks ago and a memory limit was set higher. So there are some hard coded "memory budgets" for different "systems" as i understand. Would they also dehardcoded and become more dynamic?

[inland dew]

and vRAM as well?

[inland dew]

yeah, had like 21 GB and 19 GB vRAM "used", the server (on Altis) restarted, so I found myself in Arma main menu, but RAM and vRAM values were still same
connected to another server, with another mission (on Malden), RAM and vRAM values continued to grow )))

no mods used

also today, as admin, on Malden, stayed for hours only at main island airfield, without going anywhere and RAM grew to like 20 GB and vRAM to like 14 GB, just staying AFK
so no flying or going anywhere on the island and FPS was very very bad
that was with -maxMem=23552

but tomorrow we'll also set -maxMem manually on the server (-maxMem not used on the server) + tomrrow's patch

[restive pilot]

you won't need -maxMem with tomorrow's patch anymore

[obsidian relic]

Did anyone had an issue where they don't see that player joined (server/squad) and you can't hear/see them on tfar, DUI, CTAB too? You can see model moving, get in vics, fight etc but player itself does not exist for you.
I got it twice in the row, while playing on the server, a lot of mods but it does not happened (yet) on stable, that's why I am asking here

[woven loom]

Random thought — is there any difference in how nvidia and AMD GPUs behave? Just wondering if the potential leak isn't due to an nvidia driver interaction or something

[heavy vortex]

Doubt it tbh. Seen people posting with similar issues with a wide range of GPUs.

Nvidia obviously more common but then so are the cards :P

[inland dew]

so it will still exist, but won't work anymore?

[heavy vortex]

I'd assume that it still exists as an override.

Server providers with their -maxMem=3000 are gonna freak out otherwise :P

[quaint flame]

I noticed that ram and vram usage stayed nearly the same when switching maps. For some reason I expected the mem usage to go down significantly if a new world is loaded and stuff from the old world might no longer be needed.

[empty goblet]

speaking of , do we have equivalent of superflush console / RPT output la #perf_prof_branch message for dedicated server ?

[whole cloud]

Even after maxMem parameter, it clamps to a minimum of 1024mb

[whole cloud]

vRAM would probably not flush. It won't unload textures unless it needs to. And not much harm in not cleaning up stuff at mission end.
But it will regulate itself if it needs more space for other stuff

[whole cloud]

Yeah RAM not sure what's going on there. Some stuff like Object configs will stay cached forever, once loaded once (Unless free memory is needed and it needs to get rid of it, which superflush would do)
But thats not alot of memory I'd think

[inland dew]

when I superflush, RAM values are back to pre-superflush values not even 10 mins later )

[empty goblet]

with the upcoming build the remaining limit for virtualMemory will be the maxmem * 75%?
ie. just the last 8GB (hardcoded ceiling) value cease to be or there are other changes ?

[inland dew]

understandable on mission end, since mission end, doesn't automatically mean server restart, but mostly server lobby

in Reforger, when you're back to game main menu, RAM and vRAM drop drastically, whereas in A3, everything stays cached in RAM and vRAM

not complaining, since why not use RAM and vRAM, if more than enough available, to keep as much stuff as possible cached, for as stable/high FPS and stable/low frametime as possible, since all the stuff is still/already cached

[whole cloud]

virtualMemory = physicalMemory
8GB limit increased to 16GB

2.14.151225 new PROFILING branch with PERFORMANCE binaries, v11, server and client, windows 32/64-bit, linux server 32/64-bit
- Tweaked: Player leaning sync in multiplayer (it can potentially also fix other animation desync; this is a server-side tweak) - https://feedback.bistudio.com/T177317
- Fixed: Vehicles' VisualSensor could see targets through buildings/objects - https://feedback.bistudio.com/T176526
- Fixed: Crash when deleting UI elements inside an OnDraw Event Handler - https://feedback.bistudio.com/T177796
- Fixed: EPE events were not returning Simple Objects - https://feedback.bistudio.com/T177763
- Fixed: Default system memory detection on 64bit systems was limited to 4GB (now it is 16GB)
- Fixed: Default system memory detection on 64bit Linux servers was limited to 4GB (now it is 8GB)

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

[quaint flame]

So the 16 gig is a Windows limit?

[empty goblet]

can't we get rid of the hard limit via another command-line option ?

[whole cloud]

We can remove it also without command-line option. But I don't know what purpose it serves so I'd rather leave it.
16GB gotta be enough for everyone

[rain moth]

640K ought to be enough for anybody

[empty goblet]

ye well, it would be of use if you have >32 (64/96/128) GB RAM ... 😁

[quaint flame]

We should just test what happens 😈

[rain moth]

there's always -maxMen if you're on a workstation rig with 1TB RAM

[calm mantle]

Is there anything on VRAM limits? I've seen object flickering without my VRAM being maxed.
Of course, it's possible the flickering isn't VRAM related. I'm just curious.

Nowadays I only see trees/buildings pop when flying. And I bet the RAM change may help that.

[patent dome]

#perf_prof_branch message

Or even more detailed: #perf_prof_branch message

[inland dew]

so I still have to use -maxMem, since RAM reaches 20-21 GB
because 16 GB default is a problem then

[inland dew]

if I disable paging file, would that be a problem for Arma?
does Arma really need it or it should be fine without it, as long as there is enough available RAM?

[empty goblet]

that will do nothing if the maximum virtualmemory cap is 16GB

[rain moth]

unless #perf_prof_branch message flow is changed it would override both physical and virtual?

[empty goblet]

{
      virtualMemory = 8GB;
}

if (maxMem parameter is set)
{
      physicalMemory = maxMem;
      virtualMemory = maxMem * 75%;
}

if (virtualMemory > 16GB)
{
      virtualMemory = 16GB;
}
``` this is the change how i understand it from the changelog

[whole cloud]

The previous 8gb, now 16gb limit is after the maxMem parameter

physicalMemory = 1024MB;
virtualMemory = 1228MB; // This is minimum required to run Altis on 32bit

physicalMemory = SystemStats.TotalPhysicalMemory * 75%;
if (physicalMemory > SystemStats.CurrentlyFreePagefileMemory)
{
      Rpt("Warning: memory usage limited by a page file. Current limit %d MB, wanted %d MB.", SystemStats.CurrentlyFreePagefileMemory, physicalMemory);
      Rpt("Increasing your page file size might improve game performance.");
      physicalMemory = SystemStats.CurrentlyFreePagefileMemory;
}

if (running on a 64bit system)
{
      virtualMemory = physicalMemory ;
}

if (maxMem parameter is set)
{
      physicalMemory = maxMem;
      virtualMemory = maxMem * 75%;
}

if (virtualMemory > 16GB)
{
      virtualMemory = 16GB;
}

Thats the new flow

[whole cloud]

Yes.
I have no idea if it needs it. But it will limit your memory usage all the way down. See the flow ^

[empty goblet]

thus doesn't matter if i set the value of maxmem to 64GB, it will still cap the virtualmemory on 16GB

[whole cloud]

It isn't. I found it and it was the virtualMemory cap

[empty goblet]

hence why i want to get rid of the cap or increase it significantly (96GB RAM, 256GB SWAP ... oh cmon it ain't that crazy, even i have it)

[rain moth]

yes please, all my RAM just sitting there, teasing me, while game is crying

[inland dew]

so one can't use maxMem to tell Arma there is more than 16 GB?

so it was better before then???
because one was able to use maxMem, to tell Arma, there is more than 16 GB RAM available

so it should now be worse?
because it always goes towards 20 GB RAM for me and now one can't even use maxMem anymore to override new default 16 GB...

I prefer to have default 8 GB RAM and being able to force it to be the value I set than just 16 GB and deal with that

unless I don't understand todays change correctly

[patent dome]

maxMem still has the same behaviour. Compare with this explanation: #perf_prof_branch message
Just the hardcoded values of 16GB and the x64 check were changed.

[inland dew]

Dedmen put a thumbs up under that...
so I'm lost now...
#perf_prof_branch message

[whole cloud]

It was 8gb before

[inland dew]

I know.
but I still can use maxMem to tell Arma there is like 22 or 24 GB, instead of new default 16?

[whole cloud]

You can use maxMem to set whatever, but it applies the 16GB limit AFTER maxMem is read

Refer to the pseudocode above

[vivid rune]

OK. Now we have possible 16GB for virtualMemory. I played the campaign a bit but it doesn't load more than about 5GB.
I guess the different systems (terrain loader, script thing, ai thing,...) have also there different caps, right?

[whole cloud]

Maybe. Or maybe it only needs 5

[calm mantle]

Before, no matter how high you set max mem. It would get pulled to 8. If you didn't use maxmem, then it defaulted to 4.

Now the values are essentially doubled. Still a big gain.

Did I get that right dedmen?

[whole cloud]

Previously default was 4.
And with maxMem capped to 8.

Now default is however much RAM you have. And with maxMem capped to 16.

[vivid rune]

It has some more, some a little bit less. I asume it deloads something.

[whole cloud]

Well if you play the campaign, then units spawn, units die, units get deleted..
So....

Ofc it doesn't keep deleted units in memory, deleted is gone.

[inland dew]

is it to prevent people from using maxMem command?

[patent dome]

Even without maxmem it is capped to 16, right?

[whole cloud]

Yes
I was thinking that when writing. Deleted a bit to rewrite and then decided I'm too lazy to specify that.
The code is above

[whole cloud]

I don't know.. maybe I guess.
It limits maxMem so.. I think yes?

[vivid rune]

Is there a reason to limit it to 16GB? Does Reforger has a memory limit, too?

[whole cloud]

Don't know

Reforger has a memory budget configured somewhere

[calm mantle]

Yeah I don't see the point in the hard limit, but I'm far from qualified enough to argue against it.

We still have double the ram we had before.

[whole cloud]

I don't see it either

But whoever implemented it a long time ago, surely had a reason for it

[inland dew]

I launched Arma now and connected to our server and instantly see almost 17 GB RAM
only Arma and Steam running (no maxMem used)

[whole cloud]

Well you also previously saw it use more than 8gb right?
The virtualMemory is limited.
I don't know why it can use more, but yeah it does. It did so previously and it still does so now.
Nothing was changed besides the limits.

[inland dew]

PC was freshly started, before launching Arma

[vivid rune]

On 32bit time it would make some sense. For testing the 64bit exe on the early days okay. But if the different systems are clamped to some values, it would not make so big difference.
I tested YAAB with vanilla (some around 4GB usage) and with heavy mods (around 6,5GB usage). So in heavy mod scenarios there is a some kind of advantage (the fps in v11 vs. v10 are the same - so no gain)
(I use MSI Afterburner for the memory values, it shows the allocation (around 17GB instantly and the "usage" (slowly gained over time running arma).

[inland dew]

sure, but I don't remember if before it was also 16-17 GB instantly (maybe?)
although I'm using mimalloc 2.12, with 2 MB memory chunks (locked in memory), instead of default old Intel malloc, with 4 KB memory chunks (not locked in memory)

[vivid rune]

I checked the different documentations of the maxMem parameter. Maybe it is by the next update of A3 the time for an clarification:

In the tooltip in the parameter folder of the arma launcher:
Something like: "sets the memory allocation in the range of 256-<all your RAM> - limited by your installed RAM"

In the wiki:
1024MB is a hard-coded minimum...

• 64-bit Windows + 64-bit game: (physical memory * 4) / 5
  -> this means if you have 64GB RAM, it uses 51,2GB

[calm mantle]

From what I understand, we are mostly talking virtual mem, which is only page file?
I guess it depends on how Arma chooses to use virtual and physical ram.

[vivid rune]

When i think about this two definitions. it means that arma can use all your RAM. What if this "RAM clamping" was only introduced to test the "unload mechanism"?

Would it not make sense to have the maxMem parameter for setting the virtual Memory value to every value till maxRam*0.75 and clamp the default to 16GB as a compromise?

[inland dew]

not done that?

[heavy vortex]

Isn't this line backwards?
virtualMemory = maxMem * 75%;

[vivid rune]

I hope it is the same as
virtualMemory = maxMem * 0,75;

[restive pilot]

No. If you mean the syntax it's pseudocode

[heavy vortex]

Is this incorrect then? I don't understand why you'd set committed memory smaller than physical memory.

yes virtual memory is the "committed memory" in Windows, which is the sum of RAM and page file

[calm mantle]

'''
physicalMemory = SystemStats.TotalPhysicalMemory * 75%;
if (physicalMemory > SystemStats.CurrentlyFreePagefileMemory)
{
Rpt("Warning: memory usage limited by a page file. Current limit %d MB, wanted %d MB.", SystemStats.CurrentlyFreePagefileMemory, physicalMemory);
Rpt("Increasing your page file size might improve game performance.");
physicalMemory = SystemStats.CurrentlyFreePagefileMemory;
}
'''
So if someone has 64GB physical RAM, and say 16GB page file size, then Arma could only use 16GB of physical RAM?

[vivid rune]

I have a question: The "virtualMemory" is not the only thing that Arma claimed from physicalMemory, right? FileCacheSize is also claiming RAM on top of this, correct?

[whole cloud]

You are asking if we tested that? no we didn't

[whole cloud]

no, thats what it says

[whole cloud]

I don't understand either

[whole cloud]

yes

[whole cloud]

I think so yeah

[vivid rune]

and maxFileCacheSize is clamped to 2GB max, right?

Then for correctness this line should be edited to (i know, the maxFileCacheSize can be seen in the 25%, but so it would also be possible to set maxFileCacheSize higher than 2GB safely):
physicalMemory = (SystemStats.TotalPhysicalMemory - maxFileCacheSize) * 75%;

[restive pilot]

No. The 75% part is still nonsense imo

E.g. If you have 64 GB RAM it makes no sense to "reserve" 16 GB for Windows and other stuff. 4 GB is more than enough imo

[heavy vortex]

I don't know. That's a pretty scary default.

[vivid rune]

I looked on the Steampage. Minimal Memory Requierement are 4GB RAM. So you should define for lower than 8GB a special case with 25% value for OS. Equal to and over that 8GB RAM it can be a hard-coded value of 4GB.

[restive pilot]

Well it doesn't take into account the page file at all

So it should be safe

[heavy vortex]

Discord, VSCode and Steam are all ~500MB each these days. Windows stuff is what, 2GB? And then web browsers are monstrous once you have a few tabs up.

[restive pilot]

This, and:
You shouldn't do other things when you game

[heavy vortex]

You need a web browser for Arma so you have something to do when you're ACE unconscious :P

[empty goblet]

looks at his 20GB browser ... 😁

so glad for 64-bit ... now i can game arma and browse w/o running out of memory before weekend 🤣 {reader warning, this applies only perfect scenario w/o memory leaks}

[heavy vortex]

If Arma's not clearing anything up until it hits the limit then 75% of physical memory is a bad default IMO. A flat 6GB would be preferable.

maybe 8GB at a push?

[empty goblet]

it's mainly ensuring Arma isn't using ALL the memory because like i mentioned before sometimes value returned from OS / drivers was unreliable (pure case example is VRAM)

so while you knew how much memory was there (after lot of fixes and tweaks), it was much harder to know how much is used vs free (available) and in modern era, there was always something eating VRAM

@whole cloud @restive pilot would be possible get RPT entry if the 8 GB (linux) or 16GB (windows) is reached (and flushing happens or similar), so we can actually tell how often that happens / when ?

[restive pilot]

When it happens you get object flickering

[empty goblet]

that isn't reliable log entry

[whole cloud]

Yes. but won't do

[vivid rune]

But when the flickering is the flushing, then the flushing not able to real "flush" ALL memory or the new data is so big that flushing is immediately necessary again.

[restive pilot]

I think instead of forcing that "16 GB" limit, we should do something like this. i.e. when the cleanups are useless several times in a row, we should gradually increase the limit up to max available RAM (well minus a few GBs)

[vivid rune]

That was also my idea to have the maxMemSize unlimeted for testing reasons.

(I don't see a performance gain by using more RAM so far)

[heavy vortex]

Yeah, something like this is a lot better. Start with a relatively low soft limit. Run cleanup when it hits. If the cleanup doesn't find more than ~500MB to get rid of, increase the soft limit.

[vivid rune]

Other point that could be investigated: CacheFileSize is also using RAM but is an "independed" system as I understand. Is there all right and no leak?

[whole cloud]

Flickering is not flushing

[ashen helm]

I'm on perf branch currently, been playing some missions set on Sumava map and I noticed that sometimes when zooming in (hold RMB) my GPU % usage goes up to 100% and while there is no visible lag or stuttering in game, it causes my OBS encoder to go down to 30ms per frame or even worse. I don't remember if it ever happened on stable but if it did, I would just see lower FPS in game when zooming in, not like here. Would limiting video memory through parameters help or should I seek a solution elsewhere?

[whole cloud]

Do you get higher fps when zoomed in? Maybe VSync or fps limit

[ashen helm]

in this specific case they didn't go higher but stayed on the same level as normal. I was playing with Vsync disabled, no frame cap but they were between 42-55

[plucky lichen]

the marker error popup is still there on v11, been there since v10
16:06:25 Warning Message: No entry 'bin\config.bin/CfgMarkerColors.'. 16:06:25 Warning Message: No entry '.color'. 16:06:25 Warning Message: Size: '/' not an array 16:06:25 Warning Message: Size: '/' not an array

[inland dew]

ok, could test current perf build on my end, on our server also running it, me and our server not using maxMem and having page file on auto in Windows - same/similar result in FPS degradation as it always has been

restarted the client and reconnected to the still running server and FPS went up by a lot and frametime also went down by a lot and was flat again

[restive pilot]

1. Are you using mods? If so try in vanilla first
2. Did you verify the integrity of the game files?

[plucky lichen]

No mods, just king of the hill and verify files every perf update

[whole cloud]

"just king of the hill"
KotH is lots of scripts. that could cause exactly that.
You could run with -debug parameter and it might tell you what script causes it

[ashen helm]

tried it again with vsync enabled, still in certain spots when zooming in the encoder starts to lag while the game itself is still smooth. Not sure if this is some sort of object issue that makes the latest nvidia drivers to choke or what

[inland dew]

@feral harness can you, maybe, have a look at this?

when setUnconscious true gets applied to AI while the healing animation is played (so healing animation gets interrupted), if you apply setUnconscious false, AI is stuck on his belly, with weapon behind his back, unable to stand up and do anything

only if you injure the AI and order it to heal itself (so finish or redo the healing animation, interrupted previously by setUnconscious true), only then is AI usable again
https://www.youtube.com/watch?v=_Gj9AT_pM8s

[analog acorn]

Possibly related: https://feedback.bistudio.com/T177333

[feral harness]

looks like it and @whole cloud has made the ticket

[inland dew]

this is what dedmen said back then, but it didn't help AI
#arma3_feedback_tracker message

[feral harness]

the ticket we have is open

[inland dew]

then, maybe, this is something you can check?

https://feedback.bistudio.com/T154433
Tank AI commander, when 0 commander turret machinegun ammo, makes player as gunner or AI gunner reload 7.62 mm coaxial machinegun after each shot (despite gunner coaxial machinegun having full belt of bullets), once/only if enemy AI gets spotted

concerns all types/models of armor with commander turret that has a machinegun

https://youtu.be/epiMGFElpuo?feature=shared

[whole cloud]

That is different.

The unconscious thing has been known for two years I think. And we know the solution. Just don't have the resources to do it

[vivid rune]

I've experience some familiar thing: If one player is reviving an other player and would be shot down in this animation, the reviving player cannot move but rotate and sometimes not revived. Same thing on reloading: it plays the animation to end and then character is laying on ground and cannot move, only rotate.

[brazen orchid]

Hello, the 3 FPS bug has disappeared after we have disabled BattleEye, confirmed by 2 players already @empty goblet @worldly badge @whole cloud

[worldly badge]

Why did I pinged

[plain trout]

We do get the 3fps bug in non BE environment :/

[ashen helm]

out of spite

[silk pewter]

@worldly badge yeah

[worldly badge]

I hate you

[quaint flame]

Has anyone else issue with Arma 3 being forced in the background if right clicking on an empty desktop? Is that a Windows issue or has this been introduced with the recent prof update. Never had that before.

Seems like a Windows issue 🤷‍♂️

[hot hollow]

I dont and havent used BE and I plus others have gotten it so i dont think that is it

[ashen helm]

is this a perf branch thing now? shows up when I give an EOD order to disarm a mine while he is in the Stealth combat behaviour. Haven't seen this before.

[charred holly]

I'm not certain about this, but I think rotor blades can now kill virtual curators.
It happened in two missions today and yesterday, but I'm not sure what version the players and the server are running.

[heavy vortex]

uh oh. What about other logic objects...

[feral harness]

looks pretty same to me

[feral harness]

it has to physically collide with an object to kill it and not just every object, it can be person parachute or rope

what is virtual curator?

the change was/will be removed from prof as there is better implementation now but do create a ticket so we can test against

[analog acorn]

VirtualCurator_F - playable logic object used for Zeus slots that don't have an associated man unit

[feral harness]

i'll check tomo if even possible to hit it with rotor

[charred holly]

What he said
The curator was killed when an occupant pinged him and the curator teleported to his position to find out what was going on...
I should have noted in the first post that we are playing modded, but we have never had this issue before

[charred holly]

Do you still want that ticket?
If yes, I'll need to test it in vanilla, to ensure it isn't a mod causing the issue

[feral harness]

yes please

[analog acorn]

If your server is running the Performance/Profiling branch of Arma 3, and this doesn't happen on the Stable branch, then this is probably the right place.
If you're not running Performance/Profiling, try #server_admins or making a ticket at https://feedback.bistudio.com/project/view/1/
e: this referred to a deleted message

[charred holly]

Server is running 2.14.151210 profiling branch.

I managed to repro in vanilla, latest profiling branch: https://feedback.bistudio.com/T177898

[feral harness]

player you set pos is a virtual curator?

[charred holly]

Yes

[feral harness]

Revision: 151232 Cannot say it was because of the rotor tweak could have contributed, but it is getting removed from profiling, instead the new patch will be added you would be able to test it on dev first probably, so next dev could you try to reproduce it please?

[patent dome]

2.14.151229 new PROFILING branch with PERFORMANCE binaries, v12, server and client, windows 32/64-bit, linux server 32/64-bit
- Reverted: Fix for player getting desynced after hitting a rotor - https://feedback.bistudio.com/T126837

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]

Oops missed a pin

[charred holly]

Was it removed in version 2.14.151229 of the profiling branch?

[feral harness]

2.14.151229 new PROFILING branch with PERFORMANCE binaries, v12, server and client, windows 32/64-bit, linux server 32/64-bit
- Reverted: Fix for player getting desynced after hitting a rotor - https://feedback.bistudio.com/T126837

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

[charred holly]

Virtual curators can still be killed in 2.14.151229, in case anyone was wondering

[feral harness]

yeah most likely was always possible, the new patch should take care of that

[tribal bay]

how do you open the minidump to try and figure out an issue ^^

[ashen helm]

You don't

[whole cloud]

You send it to me. In two weeks

I'm the opener

[silk pewter]

he can
he can opener

[stark falcon]

I think this is a profiling only thing, but when arma crashes I get a extra popup to create a "full crash dump"
That little window doesn't show mouse cursor for me (so it's hard to click No). Not sure if it's just me?

[whole cloud]

Arma might not release the cursor when it crashes yeah.
Not sure if I can figure out how to do that.

But the dialog has keyboard navigation.

I don't know if No is the default button though

Easy fix, just stop crashing

I need a way to release mouse lock for other things anyway. I never found the time, maybe now I will

[pallid island]

yeah, no is default

happens every time after a session of Arma and closing the game normally for me though

and not sure why, but Arma seems to delete rpt logs and mdmp files if the last one was close in time

[pallid island]

actually, this seems to stem from a mod 🤔

[inland dew]

play long enough to cry )))
https://youtu.be/rP_12Qx5nPI?feature=shared

[heavy vortex]

What was -maxmem for that one?

[inland dew]

no -maxmem

[heavy vortex]

I lost track of what prof is defaulting to now.

[analog acorn]

Going by the pinned notes, looks like 16GB default limit

[inland dew]

only if -maxMem used

just by the fact of using it

no -maxMem - no problem

[restive pilot]

it clamps to 16 GB max

the default is 75% of your RAM, or page file size if page file is smaller than 75% of RAM

[plain trout]

That video is 1 year old

[inland dew]

nothing has changed in that regard )))

[whole cloud]

Thank you for providing such valuable information all the time. We really appreciate you!

[empty goblet]

now all we need is ship @inland dew with his PC into @whole cloud laboratory at BI and fix it all 😁

[kindred radish]

How is the syntax for sendChatMessage now ?
https://community.bistudio.com/wiki/Arma_3:_Server_Side_Scripting
I get an error now when i try this

params ["_playerId","_WaitTime"];
private _AllowJoining = "DELAY";
if(serverTime < 120) then
{
    _AllowJoining = "DELAY";
    _playerID sendChatMessage "TEST";

} else
{
    _AllowJoining = "ALLOW";
};
_AllowJoining

19:47:18 Error in expression <en
{
_AllowJoining = "DELAY";
_playerId sendChatMessage 'You are Queued for Load>
19:47:18   Error position: <sendChatMessage 'You are Queued for Load>
19:47:18   Error Missing ;

[whole cloud]

missing semicolon error == that command doesn't exist

#dev_rc_branch message
it works like in my example back here

[kindred radish]

Ok, thanks. So it doesn't work like before then anymore i guess.

[whole cloud]

what before?
I think thats how it was since it was introduced

how did you format your script in server config?

[kindred radish]

onPlayerJoinAttempt                 = "call (missionNamespace getVariable ['CWZ_JoinQueueHandler', {'DELAY'}]) ";

params ["_playerId","_WaitTime"];
private _AllowJoining = "DELAY";
if(serverTime < 120) then
{
    _AllowJoining = "DELAY";
    _playerId sendChatMessage "You are Queued for Loading!";
} else
{
    _AllowJoining = "ALLOW";
    _playerId sendChatMessage "Queue finished, loading Server!";
};
_AllowJoining

Worked last time i tested without the sendChatMessage tho.

[whole cloud]

sendChatMessage is a server script command.
You are trying to use it in normal script, its not available there

[kindred radish]

Nvm then, sry 😄

[whole cloud]

The script must be compiled as server script.
If its from missionNamespace then it won't be

[restive pilot]

__IS_SERVER__ wen?! 😅

[vapid plank]

is there a "best guide" for learning how to use profiling branch to understand what is hurting performance?

[heavy vortex]

Nah, there's almost nothing written down. Some things are fairly obvious in the frame capture output. Some things aren't.

You can spot scheduled & unscheduled script costs. Individual vehicle simulation. Networking. Total renderer cost.

[vapid plank]

so arma runs poorly by itself, mods make it worse, and the tools to debug it are poorly documented? how does this game even perform as well as it does? must be a tremendous amount of dedication.

[ashen helm]

Performance/profiling branch is a new thing and was released to public only because Dedmen wants to flex around

[woven loom]

Hmm? Hasn't it been around long before dedmen joined BI? I remember dwarden posting perf/profiling stuff long before 🤔

[hushed seal]

Use the profiling.exe, take frame captures, look at frame captures.
Or use Arma Script Profiler to look at unscheduled mods (or even in engine if you enable the option)

[rain moth]

First public builds are from 2013

[woven loom]

Right. That's even earlier than i remembered, I was thinking 2015 😄

[spiral pond]

@vapid plank https://community.bistudio.com/wiki/Performance_Profiling

• https://forums.bohemia.net/forums/topic/211626-arma-script-profiler/

The biggest problem is the non scripting impact (simulation, rendering, AI, sound, etc) - very hard to impossible to judge.

Basically down to A-B testing, narrowing down, trial and error.

[vapid plank]

AI is a massive one

[plain trout]

Nobody should be complaining in here. The game is old and still being actively developed. While I‘m not a programmer and barely understand this, dedmen and the other that are still working on arma are digging into issues that sound like lazy programmers from the past. And it’s a lot old code that he hasn’t made. He’s actively engaging with the people here that take this seriously and implements things that are being suggested by the community. Show me any other game/company that hires such people after a game was released so long ago. It is what it is and we got to make out of it what we can and that doesn’t work if people complain.

Also I think we are still missing a reliable repo for the 3fps bug

[analog acorn]

....who was complaining?

[spiral pond]

A3 is still selling very well @plain trout

[plain trout]

I know. People are asking on Reddit and other platforms for advise or ask if and how it is worth getting the game and a scarring amount of people give bad advise how to optimize performance based on the old recommendations still on steam and the forums.

That complaint got removed. Someone complained about a supposedly bad documentation.

[analog acorn]

There is a lot of bad documentation (or bad absence of documentation). The wiki crew have done a wonderful job, but there is a lot of information that's not written anywhere, or it was written somewhere that's now lost, or was written down but has changed since, or no one even knows to write down. There's stuff even the devs don't know. Some parts of this game were developed at a time where internal documentation wasn't anywhere close to the priority it is now, let alone public documentation.

It doesn't mean the devs don't care, but we do have to deal with it.

[spiral pond]

Better performance will lead to happier players, and customers - thus more sales. As such it should and is a priority. Giving better tools and documentation to the community to either achieve better performance on their own, and help the remaining devs to isolate performance problems is very useful.

To give you a specific example - we are currently developing a tank mission for SPE. Its only 10-15 active tanks. Most of them even stationary. However their performance impact is HUGE - if you disableSim on them, you get 20-30 FPS more . (with still other vehicles types and a good amount of AI remaining simulated)

Something bugs out with either PX or other parts of tank simulation.

We profiled it down to individual tanks (at times also planes) taking up to 10x the usual calc cost - something is wrong for sure. (1 to 5 ms per tank/plane vs the usual 0.1 to 0.5 ms)

It could be some constant collision triggered (terrain or objects) by a tank.

One can see this happening at times if you use the EPE EHs

[quaint flame]

Make a suggestion in #community_wiki maybe we can add more documentartion

[heavy vortex]

Seems it was the same as the 3fps bug you get from setting maxmem too low. Just that it defaults to 4GB and that can be exceeded by the working set with high detail settings.

Whether there's also a memory leak contributing to that is unknown.

[inland dew]

whatever you set -maxMem to - it will be 16 GB in the end, on current perf
no -maxMem used = working correct

[heavy vortex]

...on the current perf branch.

On stable it's still going to be 4GB.

[inland dew]

forget about previous builds of perf, since it's past already

• we're in perf section
  or if you write something - specify that what you're writing, concerns stable branch only, if that's the case, to avoid any confusion, imho

[vapid plank]

My post never got removed. And the hilariously bad documentation is still there for all to see.

[vapid plank]

I think it’s accurate to say the devs don’t have it prioritized because they’re not being incentivized to deliver good documentation, they’re being incentivized to deliver an entirely new game for a deadline, that’s just how developers are managed. They have to meet the deadline to be rewarded.

[analog acorn]

I was talking about the historical record. BI has evolved and is following modern development practices; documentation is now being built as the game is built. There is a dedicated docmaster and things are documented as they're released. Documentation is now a priority - it's just that this doesn't automatically retroactively create documentation for things that were developed a long time ago, and it's difficult to do that because the information is scattered or outright unavailable.

[vapid plank]

100% agreed, this is exactly what would be done in any modern software engineering firm. Instead of fixing the old undocumented app that is very large and complex, build a new one from the ground up that you document as you build it.

@silk pewter I want to say thank you for all you do and also a thank you to all the devs you work with, for documenting Arma!

[rose moth]

that sound like lazy programmers from the past
Not denying the awesome work that Dedmen and others are currently still doing, but this is inaccurate to draw as conclusion. You can't compare the situation of prime time development and now 🙂

[plain trout]

Ok. Sorry

[silk summit]

Impressive to say the least... These devs are truly giving lots of love back to the community 👏

With semi-confirmed new CDLC in the works, current CDLC patches coming (as mentioned in the YEAR IN ARMA 3 2023 official article) and these devs backing the CDLC creators up with these much needed improvements 2024 surely looks bright, perhaps just like DayZ, A3 still has some records to break even after all these years 🥇

[quaint flame]

Don't get confused. I posted it here by accident. Post is now in dev rc channel.

[silk summit]

I wasn't talking about your post specifically, but this ---> https://arma3.com/news/year-in-arma-3-2023

Quoting the last paragraph of that official post:

"So, what is the plan as Arma 3 enters its 11th year? Even with Arma Reforger, reaching its 1.0 milestone this past November, and with development of Arma 4 underway, we still have some plans for Arma 3!

We will continue our maintenance and support of the game in a similar way as last year. We expect to publish more updates for existing third-party Creator DLC, and potentially some entirely new Creator DLC projects. This cannot be confirmed, however, until such time as these projects near their final milestones, so stay tuned for more information."

[plucky lichen]

Details on new perf build 151273?

[whole cloud]

No

Nothing to see here

[tribal pasture]

What was done with - Tweaked: Improved rotor collision handling exactly? Do they still kill? Does it go through HandleDamage now?

Or only the locality was fixed?

[whole cloud]

@feral harness ^

[feral harness]

fixed, should not kill as before

For purity of experiment it has to be tested with client/server on the same wavelength

Also, it takes 3 hits from 100% to kill with rotor

[tribal pasture]

Also remote projectiles sometimes remain in the world

[feral harness]

on v5 (2.14.151058)
That's from September, why?

[tribal pasture]

Yeah, we reverted there after rotor issues since it was randomly killing players

not sure why v5 exactly, but it is what it is

Investigating, gonna update to v15 now

[feral harness]

Damage related changes came later

[tribal pasture]

Yeah, thought we had v15 or something later, but turns out its v5, still very weird

Turns out client is on 2.14.151273

16:27:19 Received 203, expected bool
16:27:19 Unexpected message data from 1: message struct NMMA_247, item N/A
16:27:19 Before (0x00000034): 68 73 5f 77 65 61 70 5f 6d 34 61 31 5f 62 6c 6f 63 6b 49 49 5f 67 72 69 70 00 23 c4 40 3a 00 01
16:27:19 Current (0x00000054): cb
16:27:22 Received 203, expected bool
16:27:22 Unexpected message data from 1: message struct NMMA_247, item N/A
16:27:22 Before (0x0000002c): 72 61 63 65 72 5f 52 65 64 00 00 52 48 53 5f 77 65 61 70 5f 67 61 75 31 39 00 f0 09 23 3d 00 01
16:27:22 Current (0x0000004c): cb
16:27:22 Received 203, expected bool
16:27:22 Unexpected message data from 1: message struct NMMA_247, item N/A
16:27:22 Before (0x0000002c): 72 61 63 65 72 5f 52 65 64 00 00 52 48 53 5f 77 65 61 70 5f 67 61 75 31 39 00 0f a2 22 3d 00 01
16:27:22 Current (0x0000004c): cb
```Lots of such spam

Looks like non-PERFs can't play on latest PERF server

Same RPT spam, but server-side now:

2024/01/15, 15:42:49 Unexpected message data from 2017015023: message struct NMMA_247, item N/A
2024/01/15, 15:42:49 Before (0x00000011): 07 00 42 5f 37 36 32 78 35 34 5f 54 72 61 63 65 72 5f 47 72 65 65 6e 00 e9 05 3b 46 db 39 00 01
2024/01/15, 15:42:49 Current (0x00000031): 2e
2024/01/15, 15:42:50 Received 46, expected bool
2024/01/15, 15:42:50 Unexpected message data from 780409929: message struct NMMA_247, item N/A
2024/01/15, 15:42:50 Before (0x00000059): 00 00 00 00 00 00 00 49 ca 9f 3f 00 00 00 00 fe ef 01 3d 00 00 00 00 00 00 00 00 5f 08 28 39 01
2024/01/15, 15:42:50 Current (0x00000079): 2e
2024/01/15, 15:42:50 Received 46, expected bool
2024/01/15, 15:42:50 Unexpected message data from 2017015023: message struct NMMA_247, item N/A
2024/01/15, 15:42:50 Before (0x00000059): 1b be 3c 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 3e fa 3d 3c 23 5a 53 3f d1 00 5b 39 01
2024/01/15, 15:42:50 Current (0x00000079): 2e
2024/01/15, 15:42:50 Received 46, expected bool
2024/01/15, 15:42:50 Unexpected message data from 2017015023: message struct NMMA_247, item N/A
2024/01/15, 15:42:50 Before (0x00000059): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 92 bf 9e 3c 5e 13 bd 3f 5e 05 05 37 01

[whole cloud]

[tribal pasture]

Is this a bug or some engine change that makes these versions no longer compatible?

[whole cloud]

Don't know yet, we'll revert

[feral harness]

new message added so yeah

[tribal pasture]

forceHitPointsDamageSync I guess?

[whole cloud]

Out of 19 changes in 30 files, I missed one line of code

"Fixed: Repaired static buildings did not sync hit points for Join In Progress" that one is the baddy

[tribal pasture]

New message was added in the middle of the list instead of end of it? (not sure how this works exactly)

[whole cloud]

There is no new message

[kindred radish]

As we say in german "Wo gehobelt wird, da fallen Späne".
Rather some broken updates from time to time, instead of no updates/improvements at all.

[jaunty juniper]

German had a profiling branch back then?

[fast hornet]

i mean ... pushing a build that makes people invincible is a bit more than that 😒

[feral harness]

yeah

[whole cloud]

Changelog is loading, give it a half hour or so

[whole cloud]

Or a couple hours oops

2.14.151284 new PROFILING branch with PERFORMANCE binaries, v13, server and client, windows 32/64-bit, linux server 32/64-bit
- Tweaked: Improved rotor collision handling
- Tweaked: Config errors because of the wrong weapon name in addWeaponX/addBackpackCargoX/addMagazineX commands are now suppressed and RPT entries are added instead - https://feedback.bistudio.com/T177653
- Tweaked: Removed hit point update messages for entities without hit points
- Tweaked: Blood drop effect was improved
- Fixed: Repairing a building in Eden Editor was not working - https://feedback.bistudio.com/T133265
- Fixed: Zeus curators could be killed by rotors - https://feedback.bistudio.com/T177898
- Fixed: Group name was not shown in the Eden Editor entity list - https://feedback.bistudio.com/T177915
- Fixed: Repairing static buildings destroyed with effects was not always working correctly
- Fixed: Dedicated Servers with autoinit and persistent=0 were not created properly
- Fixed: AI characters might move away while getting healed (failing the heal) - https://feedback.bistudio.com/T158514
- Fixed: Explosion sound effects were not audible if the explosion occurred beyond the view distance - https://feedback.bistudio.com/T171589
- Fixed: Crash when attaching to sling load

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]

AI characters might move away while getting healed (failing the heal)
btw it should also improve player healing (increased max healing distance a tiny bit)

[tribal pasture]

Did something change regarding healing amount recently? We started setting weird setdamage logs numbers, previously all heals with FAKs did flat 0.250000 damage setting

[analog acorn]

https://feedback.bistudio.com/T172860?
* dev 11 Jan, perf idk

[tribal pasture]

Looks like this damage goes through BE's setdamage.txt filter, players using latest perf started getting triggering it.

Why its not possible to send this damage through normal entity damage means? Have ammo as #rotor or something?

This way it would also go through HandleDamage on reciever side

[weak panther]

Fixed: AI characters may move away while getting healed (failing the heal) - https://feedback.bistudio.com/T158514
Seems healing in AI team was improved, will be great to get some improvements and fixes for rearming in AI teams
There are some issues 'low ammo' and 'rearm' order with AI https://feedback.bistudio.com/T175631

[feral harness]

nothing to do with setdamage, the message is now sending proper value instead of rounded value

[feral harness]

what the log says?

how much damage

[tribal pasture]

#perf_prof_branch message

Looks like these are rotor damage logs, you can see guy piloting a helicopter and getting kicked because it

[feral harness]

possible, might be able to filter it

[tribal pasture]

Yeah, but damage is different every time, unlike say FAK which does flat 0.25

Can this rotor damage be implemented as usual damage that goes through HandleDamage?

[feral harness]

the helicopter is doing damage it is handled helicopter side

i am going to exclude engine damage from BE

should have been done long ago

rotor damage should not exceed 0.34 I think try that for now

[tribal pasture]

That filter checks against final damage number, not added one

so it can be anything

[tribal pasture]

Yeah, but can't the helicopter side pack it into a normal damage message similar to gun fire, crew damage, etc.?

[feral harness]

no

it is hybrid physx damage and is handled differently

this will remove engine healing from be too

these damages should not go through be but they do

[vagrant zodiac]

Might be completely unrelated to the rotor collision handling but...
Would it be possible to elaborate on exactly what's been done?

The reason I ask is I've made a micro UAV mod that's been around for nearly 3 years now. I've not done helicopter/drones before then or since so don't consider myself an expert and it's likely there's some stuff I'm not doing properly.
However, it's no longer possible to land this drone with the latest performance version - it explodes with the slightest contact with the ground.
Given the frequency of performance branch updates, I'm not sure if it's just been the latest version or whether this has been introduced a while back.
But stable branch still behaves as expected.

Like I say, it's likely that I'm not doing something properly but it would be handy to know what's changed so I can fix what I used to 'get away with'.

[vale shoal]

"- Fixed: Crash when attaching to sling load"

Does this need to be on all clients or only server-side?

[feral harness]

99.9% probability it is unrelated to rotor collision tweak, it has to be something else

[feral harness]

client

and server if server uses scripts. Don’t worry about this, it only fixes possible crash

[vale shoal]

Well I'm worrying about it because it happened again with latest perf.

The server runs the script in this case.

[feral harness]

does it use attachto? then update the server

[vale shoal]

Yes, and the server runs on v13 already.

[feral harness]

then it is unrelated and is something else

[vale shoal]

Already wrote Rabito as he made the ticket (we're the guys who found this issue)

[feral harness]

if you can repro this crash just ping me

[vale shoal]

Willdo.

[vale shoal]

Yep, it repro it. With latest perf running on client and server. When I do it as probably Rabito wrote it in the ticket, my game crashes with status application hang. And the server is in a frozen state.

[restive pilot]

Which mod?

Maybe we can test it and tell you what's wrong

[feral harness]

what version client and server you have will double check @vale shoal

[vale shoal]

Server 151284, Client 151284

attachTo runs server-side, the rope things are probably client only

to fix the issue for now I destroy all ropes before doing attachTo, Advanced Sling Loading I modfied that if helicopter is attached to something ropes wont extend

@feral harness

[vagrant zodiac]

SPS Black Hornet
Like I say, it’s very likely that it’s something that I’ve done incorrectly that’s just now being exposed with the improvements being made all the time in all the updates.
So any info as to what’s changed could be helpful to diagnose and fix at my end.
One thing that is dodgy is that I don’t have a hitpoints LOD but inherit hitpoints from the vanilla darter in config. I was expecting this would cause problems when I first made the mod but it didn’t at the time so I took the ‘if it’s not broken, don’t try to fix it’ option which in hindsight might be wrong!
Given the perf branch will at some point become the stable release, any help so I can fix at my end (assuming it is something I’m doing wrong) would be hugely appreciated

[feral harness]

Tried it again, cannot reproduce the crash, I will need detailed repro steps

possibly with video

[kindred radish]

Works fine in the last builds, no server crash with vanilla slingloading.
The problem is he uses Advanced slingloading, so he has to restrict that inside the scripts.

[vale shoal]

Still no. Rabito slingloads the vehicle before he attaches the helicopter onto a vehicle.
In my test I slingload the vehicle after attaching it.

[feral harness]

repro please otherwise this is a waste of time

[feral harness]

Reproed

[feral harness]

What is the advanced slingloading? I'm tempted to abort any slingloading operation on cargo with ropes attached to avoid millions of misuse cases

[fickle geyser]

https://steamcommunity.com/sharedfiles/filedetails/?id=615007497

one of most populars slingloading mods.

and quite old

[feral harness]

Revision: 151306

all illegal ropes will be destroyed

[plain trout]

launching the hornet also insta kills you on launching it too.

[feral harness]

always test in latest dev, there are fixes that cannot go to perf

[vagrant zodiac]

That should have been fixed in the update I sent yesterday?
It only happened if you had your weapon raised previously?

[vagrant zodiac]

Just checked Dev branch and same behaviour as latest profiling, sadly. (So likely I need to fix something but not sure what)

[spiral pond]

oldSize-1 != list.Size() 36 ,39
Offending nonprimary object ebe65240:SPE_GER_Soldier_XXX_Boots,ww2\spe_assets_m\characters\germans_m\ger_wehrmacht\spe_XXX.p3d
oldSize-1 != list.Size() 32 ,34
Offending nonprimary object c684c080:SPE_SdKfz250_1,<null>

is that a new logging? havent seen these before

Deinitialized shape [Class: "C_Soldier_VR_F"; Shape: "a3\characters_f_bootcamp\common\vr_soldier_f.p3d";]

also got this yesterday within a session - usually its only at the end of the rpt

but had also very severe fps problems with vegetation flickering at distance

down to <1 fps phases

seems something gone amiss with RAM usage (task manager showed still ~10% free tho)

-maxmem=2048

i tested with that first time. probably came from that. 2gb too low in general? even if, should the game not be able to manage it still?

[whole cloud]

Not new.
Dunno what thats about, but I know it doesn't matter.
You're describing the 3fps bug.
2gb ram only? on client thats too low

[spiral pond]

alright

might be good to adjust the available values in the A3 launcher then?

thats the default and you can go even lower than that

[whole cloud]

Probably, but touching launcher

[vagrant zodiac]

Has anything changed in how land contact is calculated with regards to Heli updates?
Looking more closely, the issue I get appears to occur slightly before touch down. (I have the points in the LandContact LOD slightly below the drone)
The nature of the crash seems to be that there's suddenly a lot of energy released - the vehicle cartwheels almost as if it's detecting contact on a rotating element e.g. the tail rotor (even though it doesn't appear to have touched the ground)
Adjusting the position of the LandContact LOD (e.g. offsetting it lower below the drone) just makes the drone explode at a higher altitude. I.e. it's suggesting something goes wrong the instant the points in the LandContact LOD touch the ground.

[feral harness]

could you go to the

utils 2

in debug console and paste type of the drone into the search field and then generate config with
onlyclasses - off
include inherited - on

then paste the result here

I have pretty good idea what happened

[vagrant zodiac]

Sorry - being a bit dumb here - utils 2?

Ah - might have found what you mean - standby

Hmmm. Hitting generate just seems to default back to Classes Only ticked and is super short
/*
INCLUDE INHERITED ENTRIES: false
SHOW CLASSES ONLY: true
UNLOCALIZED TEXT: false
CONFIG PATH: bin\config.bin/CfgVehicles/sps_black_hornet_01_F
SOURCE ADD-ON(S): sps_black_hornet
*/

class sps_black_hornet_01_F
{
class UserActions; (1)
};

If I don't hit generate and untick Classes Only and tick Include Inherited, it's quite a big list!

Want me to DM it to you as a file?

Or just look for a specific detail for you?

[feral harness]

yes, you can zip it and dm or post here since anyone can do it anyway it’s not a secret, it’s up to you

[vagrant zodiac]

DM-ed as I don't seem to be able to post it here

[whole cloud]

2.14.151313 new PROFILING branch with PERFORMANCE binaries, v14, server and client, windows 32/64-bit, linux server 32/64-bit
- Added: Additional context to the 'cycle as first waypoint' warning message - https://feedback.bistudio.com/T178338
- Added: Modded compositions now have a scopeCurator property that can be set to 0 to hide from Zeus - https://feedback.bistudio.com/T174947
- Fixed: "PlayerViewChanged" Event Handler was not firing for remote vehicles - https://feedback.bistudio.com/T127838
- Fixed: Another possible crash connected to sling loading
- Fixed: Bullets were not creating splashes when hitting water (broken since a recent revision), and fixed the water impact sound for bullets - https://feedback.bistudio.com/T176022

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

[twin orchid]

Does ArmaScriptProfiler with the -profilerEnableEngine flag still work?

Whenever I launch it with arma3profiling_x64 the game crashes

[whole cloud]

I heard of crashes. Already hotfixed one.
If it crashes then the answer is probably no

[shell harness]

im just trying to find a good server

[heavy vortex]

Then you are most definitely in the wrong place.

[shell harness]

dang

[heavy galleon]

Hey there, as far as capital letters in pbo name for linux - it works just fine with no issues, as long as the "addons" folder is all lowercase and not "Addons" as some mod creators like to name it... Would it be possible to have this fixed too?

[whole cloud]

Yeah possible to fix that but it would be so ugly.
Basically check both names 😢

[heavy galleon]

It would be really appreciated
I spent way too much time trying to find out why it didn't work

[hybrid cradle]

may be worth a pin in #server_linux regardless

[autumn timber]

I second that. It's sometimes annoying to have to contact other modders so that they fix&release a new version because you're the first person who wanted to run their mod on Linux 😅 (on a server, for example)

[heavy vortex]

However, if Arma doesn't force every reference to lowercase, you're now dependent on modders getting their casing correct.

While currently you just have to change all the names to lowercase with a one-line script.

[heavy galleon]

Perf build doesn't anymore as far as pbos go. The addons folder has to be lower.
WHY would you even do that... 3CB Factions has it like this and it's

[heavy vortex]

Wait, so lowercase forcing doesn't work anymore?

This is a really bad idea :P

Like do you really trust mods only tested on Windows to match their filepath cases with the actual filenames...

[rain moth]

Still works

[heavy galleon]

Perf has a workaround for linux so the case of the pbo name doesn't matter anymore for loading, but the game still sees and treats everything lower case.

[heavy vortex]

Ah, but it still cares about addons/Addons, right.

[loud condor]

I have experienced when using the latest performance build that players and ai become invulnerable, regardless of side relations. Although when a zeus spawns an ai it can damage them (or himself), but not anyone else. I tested this with mods and without (full vanilla server and new server profile). I used the manual .exe from the google drive, as the one from the opt-in beta build.

[heavy vortex]

Is this with everyone running perf or just some players?

[tribal pasture]

If its build 151273, then the only option would to be to forbid players from using it

Use productVersion select 3 to check build number

[whole cloud]

Check the server log for errors

[ashen helm]

is the pop-up about creating a full dump a prof branch feature? Because the full dumps can be quite big and their size often match the RAM used by the game during its generation, it hurts my puter. Had to kill process through taskmgr so not sure if it's even a truly full dump

[hybrid cradle]

are you on prof or main

its not even a feature to generate a full dump on main afaik and popup is prof only yes

dedmen will see at some point to probably correct me on what i got wrong

[whole cloud]

Yes.
If you don't want full dumps, then don't click yes in the popup

[ashen helm]

I don't mean I don't want them, I wonder if they are supposed to be of this size

I already thought normal mdmp contains ton of data, what's in the additional 9,4GBs?

[hybrid cradle]

everything

all of it

yummy data 😋

[ashen helm]

can I change its default creation location to a different partition? And does it have to take so much RAM it almost looks like it's a leak

[plain trout]

That’s a new perspective. Usually people complain about arma not being able to utilize enough memory.

[hushed seal]

I think that's the whole reason for the pop up, to click "no" if the size is an issue.

[ashen helm]

I am more concerned about ram it's taking with every second rather than how much space it takes, unless that <10GB is barely a part of it

If it's supposed to help debugging stuff then hell yeah I will click "yes", all I'm asking is if it's supposed to, well, be like that. 14GBs of RAM taken after less than one minute of work.

[hybrid cradle]

iirc not really needed aside from obscure crashes

[whole cloud]

normal mdmp's actually contain very little.
Full dumps are full, all of the game's memory/state

[whole cloud]

No. Arma uses what Arma needs

If I can see, from looking at the code at the location where it crashed. How the crash happened then a small dump does just that.
But if I need to see context, like which entity type did it, or what line of script ran or such. That info is usually not present in normal dumps. Its luck based

[restive pilot]

you can symlink the whole Arma 3 appdata folder

[loud condor]

Apologies for late reply, it was server side not player side loaded

[loud condor]

Sadly it's the server sided perf build that is the issue so as the administrator I won't use it atm

[loud condor]

I will see if I can get you some data soon

[tribal pasture]

Which build is it?

[subtle yoke]

Guys, Hello!
Seems to be data passed to extension limited to 8kb, is it bug? Wiki says there is no limited?
Biki says passed data to RvExtension unlimited, but practice says it is 8kb max

@whole cloud @feral harness Help!

[feral harness]

you must be doing something wrong

[autumn timber]

FYI, Biki says:

The limit may or may not change in the future and is currently 10240 bytes

Oh wait, sorry. I'm wrong. That's for receiving

[subtle yoke]

Someone told me that was "format" command limit which I've used before passing args to extension, now trying to do it without "format", 10 mins...

[autumn timber]

Looks like you're right about format:

The output of this command is limited to ~8Kb. joinString on the other hand doesn't have such limit and is a bit faster.

[subtle yoke]

Yep, that was format

[feral harness]

here you go

[whole cloud]

#rules number 2. Thank you

[flint ravine]

Is there a way to run the diag commands on Dedicated Servers and still getting the menu to show up?

[whole cloud]

No

You can run them on server and print it to a file then open menu on client and import file

[wraith jungle]

Hey team,
Just wanted to double check as it's been a while since I've looked at this and some of the processes seemed to have changed.
It seems like you dont need the beta code anymore to unlock the profiling branch. So that's all sorted.
And tracking that the mod doesnt have anything in it atm.
But do you still need to rename the launchers in your base game folder to switch over to profiling completely?

[heavy vortex]

To profiling? Yes. The default-named executables in the perf/profiling branch are performance not profiling.

[wraith jungle]

Yeah sweet as. Assumed as much. Just wanted to make sure. Thanks for that. 🙂

[spiral pond]

there was a fix for "edit terrain object" for MP state issues, wasnt there?

possible to apply same/similar kind of fix for "hide terrain object" module/hideObject?
aka the server hides objects globally, yet these dont get applied at times for clients as the [terrain] object has not yet been initialized on the client // might be same for mission objects

people says its even more of a prob with JIP. especially when you hide 5k+ objects

[quaint flame]

Can confirm. Even though hideObjectGlobal is used on the server. Players end up with different states.

The Objects are hidden early via mission parameters.

[weak panther]

I just started mission on dedicated server with 2 clients and module hide terrain object to hide 29646 objects in 1 km, the 2nd player was JIP
All objects was hidden in selected area, even if hide code was executed with 'local' checkbox

[weak panther]

note: current dev, not perf/prof

btw when module area were set to more than 2 km, after admin clicked the “next” button in briefing the mission did not start and froze player stuck on the 2D map, although the server console said that the game had started;

[feral harness]

so it is about map object streaming then?

[spiral pond]

@feral harness would this mean a difference if an object is located where the player unit is placed/location of the unit when JIP happens vs further away from that location/area? (aka outside view or rendering distance)

[feral harness]

don't know how objects are streamed and why sometimes they exist even if far from player view

[inland dew]

despite culling, when I join our server and respawn and turn around, frametime spikes a lot for many many seconds, so it loads a lot of objects, although I can't even see most of them, since where I spawn there is a building and walls

[spiral pond]

alright. we will try to organize a focus test. is there a way to verify those "desync" states via sqf? isObjectHidden may or may not i guess

[feral harness]

make sure you test for object existence first as it will return false on objnull

or true, don’t remember

[whole cloud]

2.14.151437 new PROFILING branch with PERFORMANCE binaries, v15, server and client, windows 32/64-bit, linux server 32/64-bit
- Added: ENDLOAD cheat code that forcefully ends a loading screen - https://feedback.bistudio.com/T177862
- Added: New Aspect ratio settings for Ultra-Wide monitors (Disables the triple-head detection)
- Added: "Restart required" dialog when changing aspect ratio's triple-head state
- Changed: Ultrawide is now the default for high aspect-ratio screens (instead of triple-head)
- Tweaked: Small rotor UAV handling
- Tweaked: Improved reliability of healing injured soldier in MP - https://feedback.bistudio.com/T158514
- Tweaked: Further refined and improved helicopter rotor collisions
- Tweaked: Additional inventory item type validation for setUnitLoadout from an array
- Tweaked: Improved allPlayers script command on the server - https://feedback.bistudio.com/T166966
- Tweaked: Performance optimization for script boolean values
- Fixed: Sling-loading hooks were duplicated on savegame load - https://feedback.bistudio.com/T177659
- Fixed: Structured Text image tags would lowercase image path (Caused issues with uiEx procedural textures)
- Fixed: Missing instigator in some of the shell damages - https://feedback.bistudio.com/T178697
- Fixed: onUserSelectedPlayer returning null player - https://feedback.bistudio.com/T179027

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 turtle]

@whole cloud question if im using profiling would that require everyone to be on profiling branch if i use it on dedi server?

[obsidian relic]

no

[whole cloud]

no, its the whole point of prof branch to be compatible. Unlike dev branch

[restive turtle]

gotcha thanks

[fast hornet]

the aspect ratio dropdown menu in 151437 has some missing string errors btw... otherwise it seems to be working fine.
See r9Ze3KV.png on imgur, not allowed to post screenshots or links in here.

[silk pewter]

you are allowed, it's [] links that are forbidden

https://i.imgur.com/r9Ze3KV.png

[fast hornet]

ah ok.. that's what i get for trying to beautify my links 😄

[silk pewter]

security :p someone could post https://google.com ](http://evilwebsite.com 🙂

[fast hornet]

i mean.. discord displays the real link when you hover over it.. but yeah, alright 😄

[whole cloud]

missing strings are expected because profiling branch has no data

[tough escarp]

you have to change the names of the arma 3 executeables to truly run pofiling/performance branch? I never had to do that and it says im using that branch. am i mistaken? I just changed to that branch in steam without any code or anything special

[heavy vortex]

If you don't change the name then you're using the performance exes.

If you want profiling then you need to change it.

[obsidian relic]

found a bug: you can't use turrets in drones, ie you can click it but camera doesn't move and you lose control over the (any)drone, it works fine on stable

how to replicate:
place any UAV operator and a drone, try to move camera after clicking "take uav turret control"

[ashen helm]

once in drone's turret you don't get access to the scroll menu anymore and can't disconnect from it

[random isle]

^ same for remote AI, our remote ai controls broke as well on latest pref

[hot hollow]

same for remote AI
Just normal controlling or do you use something fancy cause mine works fine

[random isle]

digging deeper not sure how this related to latest pref change log but crew command no longer returns anything. ```sqf
crew (vehicle player)

i get [] while in vehicle

[whole cloud]

Thanks! The crew thing was a good hint.
There was one change about Crew stuff, and seems that broke everything
maybe can even get fix out today

Found it.
It was a missing const in one line of code. Broke all crew checks

[silk pewter]

but crew is not constant, they can move in and out?

[woeful ridge]

1. Have the same problem, with AR-2, Stomper RCWS, and others.
   It displays the wrong turret view (gunner instead of driver, or driver instead of gunner), while the camera cannot move anymore, and the beloved mousewheel menu is gone.

On the topic of crew, this didnt work anymore:

if (player in (crew _this))

While this worked:

if (player in _this)

2. In multiplayer AR-2 Darter has problems going the set altitude, when not using workarounds and constantly checking. (Likely also in non profiling.)

3. When going forwards, then lowering weapon and trying to stop, the character continues to walk forward for multiple meters. 😅 (Likely also in non profiling.)

4. UAV's stay connected to dead player terminals, blocking access, basically making it a lost asset.

[whole cloud]

Mh still have a weird issue.
All the reported bugs fixed, its building and will be done in an hour or two.

Weird thing in Darter turret, the "release control" action I need to press it twice for it to work 🤔
That can't be my bug that broke everything. It works fine on second click. Weird

[restive turtle]

nope you clearly broke the entire game!

[silk summit]

Can't make an omelet without breaking a few eggs 😉

[restive turtle]

such a good sentence in regards to coding 🤣

[orchid mesa]

odd new bug with the latest profiling branch. friendly vehicles (including your own) don't show names of the people inside them. not sure if it's something SM needs to tweak because of a change, or an actual bug https://imgur.com/a/ZwKXLW8

[feral harness]

related to what dedmen fixed

[orchid mesa]

sweet

[whole cloud]

Eh ?
Seems better but not much 🤔 Would you count that as better?

[plain trout]

7% uplift in fps is what Intel usually grants their customers as “new generation with the best performance ever.”

Assuming it’s 44 to 47 average

Or as we say in German: Kleinvieh macht auch Mist

[feral harness]

works fine on internal, could be because of weird merge to prof or corrupt build

[whole cloud]

2.14.151459 new PROFILING branch with PERFORMANCE binaries, v16, server and client, windows 32/64-bit, linux server 32/64-bit
- Fixed: Crew checks did not work since v15 (broke UAV control, unit remote control, crew command, and more)

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

[obsidian relic]

@ashen helm could you check turret camera movement in drones? for me it's not smooth as before(almost like it has velocity now), it's quite visible when trying to swiftly move camera around ie left right

it's hard to record it but here in both cases I am doing circles with my mouse, starts with slow and small ones and switches to fast and large ones
https://streamable.com/6rxrm1

[woven loom]

Well look at old run 1 and 2...
YAAB does have a bit of variance

[inland dew]

margine of error

[hollow canyon]

Can someone on this branch who is familiar with the H-60 project, try and interact with the H-60 using only the required mods? (H-60 Pack, Hatchet Framework, Ace and CBA)

(Please reply to this message if you try it, A3 server notifications are off for me)

I've currently got 2 confirmed cases of the current profiling branch causing an error that bricks the hatchet mod. However, on Linux, Proton Experimental on Profiling, I am able to still interact with the bird, so I am at a loss for what is happening

[ashen helm]

does the stable camera snap to ground without Ctrl+T? Because prof one does

tbh I thought it's been always like that and only driver's 2nd camera view does not have that auto snap thing.

[obsidian relic]

no, only with ctrl t

[ashen helm]

so maybe that's a feature then

[obsidian relic]

nvm, it does

the thing on perf isn't like snapping

it seems like camera has velocity with some limiter

[ashen helm]

I would call it acceleration, like old games did it.

and some angle snap too perhaps. Had to check it's not my mouse software

[obsidian relic]

ye, it's better word for it

[vapid plank]

Nah

This was a trend when they released the same cpu 4 years in a row

10 to 12 was a solid uplift (we don’t talk about 11)

12 to 13 was even better

As you no longer had to disable e cores

[silk summit]

Lets be honest guys, for A3 there is nothing better than the 3D cache from the AMD CPUs

My 5800X3D gives me 130 FPS easily on most maps while my friend's 13600 can't go past 90

[heavy galleon]

New version of C++?

[vapid plank]

Ok bro

[restive turtle]

hes not wrong

3D Cache goes crazy

tbh

[orchid mesa]

intel is very dependent on also having fast, low latency ram

xmp cl30 6000 vs 7000 cl30 tuned is like 25-30fps difference

[patent dome]

This chat seems more relevant for #hardware_vs_arma

[whole cloud]

"an error" what error?

[hollow canyon]

They lose all interaction, the only error in one of their rpts relating to hatchet was this

12:15:13 Error in expression <
if (_canOpen || _vehicle ammoOnPylon 3 > 0) then {
[_vehicle, 3, 1 + vtx_uh60_c>
12:15:13   Error position: <> 0) then {
[_vehicle, 3, 1 + vtx_uh60_c>
12:15:13   Error >: Type Bool, expected Number,Not a Number
12:15:13 File z\vtx\addons\uh60_cas\functions\fnc_updateOverlayList.sqf..., line 18

[whole cloud]

_vehicle ammoOnPylon 3
returned a boolean 🤔

Well yep.. That function returns boolean if _vehicle is null object...
The script function that only returns Scalar.. Returns false.

I don't know why that vehicle would be null.

But we should probably someday fix that wrong return. I'd expect a zero instead

[patent dome]

The wiki actually says it returns false if it fails

[whole cloud]

[feral harness]

unbelievable, I thought I got them all

[feral harness]

ah that might be the reason

[ashen helm]

I assume this is related to that weird FPS bug (that caused vegetation to pop in and out when moving fast), but it's my first time seeing this "that" way. I was running a dozen seconds on 4x speed and noticed the character is gone, and now it shows up only under certain angles like on video. Tried doing the same numerous times on stable build and can't replicate, on perf this happened twice out of five tries

[whole cloud]

no thats not related to that

[restive pilot]

it looks like a fov issue to me

[ashen helm]

but I didn't alter fov, it's just ultrawide screen

it happened just like that after running for a good while on 4x speed while in 3rd person

[restive pilot]

could this be related?

[ashen helm]

I'm set to auto stretch 21:9

[plain trout]

Sometimes that own texture flickering is predecessor/predictor for tree flickering.

[heavy vortex]

Does it only affect the player model here?

I've seen something like that before in Arma but with corpses. Essentially it's like the renderer is culling the object as though it's in a different place.

[tribal pasture]

Pre-A3 corpses?

[heavy vortex]

Nah, ragdolled into the floor of a building, IIRC

[tribal pasture]

Oh true I remember seeing ragdolls disappear earlier than they should at the edge of the screen

[ashen helm]

Yes

No matter what I do then only player's unit shows and hides on camera movement, everything else in 3D is just fine

[solar root]

Anyone knows why the perf branch actually decreases my FPS? I remember back in 2017 the perf .exe increased performance quite a lot.

[ashen helm]

I know only a few assets that "disappear" in that way when close to the screen border and it's mostly gates from CUP and Livonia, but in case of my video it feels like that the relative position of my 3P camera to the character got "miscalculated" or something and game cannot figure it out.

[feral harness]

It is the same exe why would it drastically increase your FPS? It might have some improvements that are not yet available in stable but eventually stable catches up and the difference would be negligible

[empty goblet]

back your claim with some results ... decrease FPS where ? in static scene in editor ? in benchmark ? in multiplayer ?

by what amount 1% ? 10% ? 100% ?

[solar root]

I mainly play koth right now. Used to play on a3wasteland, too. After coming back to ArmA 3 around 6 months ago, I have tried the perf executables 3-4 times. With the same settings, the game would dip below 30 fps in the ao with the perf exe every single time, which never happened with the vanilla exe.

[whole cloud]

Arma 3*

[weary sun]

I wish I had more specific things to report, but the latest profiling branch build broke several mods. Hatchet Framework interactions no longer work and WebKnights Avionics hud pop ups no longer show. Switching to main branch fixed both of these issues.

[hot hollow]

Hatchet Framework interactions no longer work
Unless their was an update that wasn't posted it worked fine for my unit 24 hours ago

[feral harness]

what profiling version are you on?

[random isle]

not sure if this related to pref but ill just drop in here

Error in expression <
private ["_uid"];
_uid = _this param [0, "", [""]];
private "_unit";
File A3\functions_f_bootcamp\Misc\fn_getUnitByUid.sqf..., line 15

[whole cloud]

Error. But what error

If you're gonna drop an error please drop the full one

[random isle]

thats all she wrote

2855:20:44:51 Error in expression <
2856:
2857:
2858:
2859:
2860:
2861:
2862:private ["_uid"];
2863:_uid = _this param [0, "", [""]];
2864:
2865:
2866:private "_unit";
2867:>
2869:
2870:
2871:private "_unit";
2872:>
2873:20:44:51   Error Type Number, expected String
2874:20:44:51 File A3\functions_f_bootcamp\Misc\fn_getUnitByUid.sqf..., line 15

[worldly badge]

And what have you done

[random isle]

ill try to repo it tomorrow, was just looking at logs

[silk pewter]

seems like the method was called with a number, that's it

(uid is a string)

[whole cloud]

run game with -debug and itll tell you who did it

Probably just a bad mission script

[whole cloud]

https://devblogs.microsoft.com/cppblog/msvc-backend-updates-since-visual-studio-2022-version-17-3/
Looking at what kinda improvements microsoft did just inside VS 2022.. Makes me think about what Arma might gain upgrading from VS2013 to latest 2022

"Do a better job of devirtualizing calls to class functions." this alone could be huge for Arma. Depending on what they mean with "better job"

[quaint flame]

Only one way to find out

[silk pewter]

https://tenor.com/view/yes-freedom-jonah-hill-scream-shouts-gif-17286088

though a question remains - can it break anything? like, retrocompat with old gear (hmmkay), or old code's compatibility with new gear/OS?

[worldly badge]

I'm guessing Ded is certain of having no such concern, or at least the big ones

[whole cloud]

That's the question 😄

I hope noone plans to run prof branch on windows 7 next week

[feral harness]

Actually...

[whole cloud]

Oh do you have it? I need a tester 😄

[feral harness]

I have 2

[cloud nacelle]

Steam doesn't support win7 anymore, so would it be ok for A3 to not support it either?

[whole cloud]

Ehhhh.. I can't decide that and I'd rater not decide that. So we'll do what's needed to support it
I'd love to drop 32bit on prof though :harold: Maybe I get to do that someday

[cloud nacelle]

What's the requirement for 64bit, AVX2? Speaking of AVX2, is it possible that newer versions of Visual Studio assume AVX2, breaking old CPU support?

[whole cloud]

SSE2
No they wouldn't assume

[vapid plank]

In theory yes they could improve it like this.
In practice, Microsoft admits to creating 1000 new bugs in the process of patching 5000 old bugs. We don’t know how many of those bugs were critical to how Arma worked, and we don’t know how the new bugs will affect it.

[whole cloud]

in practice I've been running Arma with VS 2022 for over a year and didn't find any issues

[vapid plank]

The answer is always yes something can break even if we don’t change anything, but the chance of it breaking goes up by several magnitudes when we make big changes.
The real question is how severe is the break? How hard would it be to fix? Hire many ways does it break? How much time and effort can we apply to fixing the new breaks?

[vapid plank]

What kind of slideshow are 32bit players getting???

[restive pilot]

but if Steam doesn't support win7, and A3 requires Steam to verify its license or purchase or whatever, then it means A3 doesn't support win7 anymore either 😅

[vapid plank]

Stream will work on windows 7

But they died explicit support for it

They don’t care much if it actually breaks

[fast hornet]

The reason was that Valve uses chromium in steam and that doesn't support Win7 anymore. It might still run, it might not. But Win7 support from MS ended January 2020 as well.
Steam Hardware Survey from january 2024 has Win7 listed as 0.05% for 32bit and 0.59% for 64bit.
32 bit versions of Win8 and Win10 seem to be even below that 0.05% mark as they are not displayed at all.

So yeah.. probably really safe to say that noone needs Win7 support and/or 32bit anymore.

[cloud nacelle]

I believe that 32bit is mostly used by basic servers, e.g an arma 3 dedi running on an old system. Those are not represented in the Steam HW survey

[fast hornet]

well.. the Athlon 64 launched over 20 years ago. Guess it's safe to say that every system newer than 10-15 years does support 64bit

[cloud nacelle]

Only Bohemia knows if it'd even be worth it though (removing 32bit), I imagine it's mostly about making maintenance easier and not performance improvements.

[silk pewter]

it means that newer versions are not Win7 compatible
it doesn't mean that Steam uninstalls itself on Win7 🙂
do we want to alienate a small portion of customers on the basis they run an outdated OS, I think not

[hybrid cradle]

can you not just have a beta code for reverting to pre changes

just like the 2.12 legacy etc

[fickle geyser]

I think yes, the minority staying on outdated systems and hardware should not hold improvements to the engine back.

Sometimes you need to break things to move forward.

good example from Dota 2 where they've dropped support for outdated GPUs and CPUs:

As we mentioned when we launched Battle Pass: Part II, this update has removed support for OpenGL. For macOS customers, the minimum requirement is now macOS 10.13.6 with a Metal-capable GPU (AMD GCN-based GPU or newer; NVIDIA Kepler-based GPU or newer; or Intel HD 4000 or newer). For Linux customers, AMD GCN+, NVIDIA Kepler+, or Intel HD 5000-series or newer is required.

In a future update this year, we intend to upgrade our SIMD support from SSE2 to SSE4.1 to allow higher performance in physics, particles, and general gameplay code. Almost all processors released in the past 10 years support SSE4.1, so this should affect a very small portion of Dota players. If you have a computer that does not support SSE4.1, you will need to upgrade to a newer CPU once that update releases.

The difference being that A3 can also be played singleplayer warrants giving these people a way to play last compatible build so you can freeze last build that works on them on a branch.

[silk pewter]

I believe Windows 7 can run VS2022-compiled exe
let's just not "not care" if it's a matter of settings

also, let's check the % of peeps still having Win7, it's not exactly "super small"

[fickle geyser]

Less than 1% on Steam Survey

[silk pewter]

I always think of peeps with their no-Internet gaming laptop, even about Steam

Well, if < 1% is not 100000 players then… why not!

[fickle geyser]

I'm thinking about them too, and I believe they do not need the latest updates and can stay on older versions ;)

[silk pewter]

unjust yet fair enough :p

[ashen helm]

a tiny little bit of machiavellianism never hurt anybody, right?

[whole cloud]

I don't expect windows 7 to hold us back.
The problem is we use a newer SDK version. That adds API functions that were added in windows 8/10.
If something tries to use these, it won't start on win7.

I'll just have to fix and make it not use them and all will be good.

Supporting Windows 7 doesn't do us much harm. It's just that we would need to test and make sure we didn't break it there.
And we're currently not testing Windows 7.

The only harm is the effort to QA test it.

32bit on the other hand. Doubles our build times.
Due to that we can only push 2 prof per day, and only if the timing is right.

Without it we can do 4. So if someone notices an issue we have more margin to push a quick fix.
Maybe even more if new VS improves build times.

[fickle geyser]

https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam?platform=pc

looks like you could safely drop it, for profling I wouldn't even sweat about it.

I guess you should have more detailed data specificaly for A3

[thin wyvern]

VC++ 2022 version 14.32.31332.0 is the last version compatible with Windows Vista. Newer versions support win7 to 11.

Microsoft ended support for Windows XP on April 8, 2014. Current versions of the Visual C++ Redistributable for Visual Studio 2015-2022 only support Windows Vista, 7, 8.1, 10, and 11. The last version of the Visual C++ Redistributable that works on Windows XP shipped in Visual Studio 2019 version 16.7 (file versions starting with 14.27). The Redistributable is available in the my.visualstudio.com Downloads section as Visual C++ Redistributable for Visual Studio 2019 (version 16.7).

[silk pewter]

I am somewhat ok with dropping Vista 😄

[heavy vortex]

It'd be nice to switch to a version of VC++ that didn't have a installer that bugs out and breaks a lot of fresh installs.

but that's probably not gonna help if it's just profiling.

[empty goblet]

the vc++ 2015-2022 even latest version (14.38.33135.0) still supports W7, W8.1 and Server 2012 R2

[thin wyvern]

I'm using this AIO Repack https://github.com/abbodi1406/vcredist

[blazing wagon]

Does #captureSlowFrame 0.3 1 total have to be run on a dedicated server environment? Running in local multiplayer does nothing.

the function diag_captureSlowFrame appears to not exist

[blazing wagon]

Ah well it seems like the A3 launcher just doesn't like to launch the profiling exec, so that's probably my first problem

Memories flooding back to having the same issue last time I had to use profiling branch..

[woven loom]

@blazing wagon it's not a problem per se, it's just how it's intended. Rename arma3profiling.exe to arma3.exe and it should work. By default it uses the performance binary.

[blazing wagon]

Would locally running diag_captureSlowFrame on a server cause issues? I just ran it and it ended up kicking me off the server after the profiling gui popped up

[woven loom]

No idea, I've only run it locally

[heavy galleon]

It's next week

[blazing wagon]

woo profiling did a thing... showed me that the vanilla SAM systems were causing a silly amount of lag

lesson learned, don't spawn 10x SAM systems in Arma

[heavy vortex]

hmm. I remember something like that with the "ships".

[blazing wagon]

These were the naval SAM sites (ship-mounted turrets), so maybe related?

Server went from 10 FPS to 45 FPS immediately upon deleting them

Meanwhile we had hundreds of other AI on the server with no problem.

[fickle geyser]

It's the sensors/radars they can kill frames if they have big range

IMO they could use an optimization where they do whatever they're doing over X frames to not kill the frame rate

Right now it looks like the scanning is done periodically as this kind of issue usually happens in periodic stutters

[tribal pasture]

Having this controllable would be the best. If you care about FPS, set it to slower sensor simulation, if you care about precision, set it to faster sensor simulation

I recall something about shooting missiles at large distance and killing FPS right after Jets DLC, I think they were scanning for targets through the whole map through a huge cone each frame.

[plain trout]

The amraam missiles had the issue during when they released and occasionally froze servers until they hit a target and allows the server to „breath“ again

[blazing wagon]

so it looked like per SAM system, I was getting 100ms delay every few seconds

with 20 SAM's it was getting very delayed lol

Either way, glad profiling exists, and I won't be allowing the aircraft carrier in our mission file to have any air defense any longer 🤣

[woven loom]

This is good to know... i once made a mission with the aircraft carrier and AA turrets... on tanoa... performance was sketchy

[blazing wagon]

Wierdly it was the naval SAM's that were causing issues

the trailer SAM (can't remember what it's called) weren't popping up with any big frame delay

You wouild think they call the same functions but idk

[woven loom]

interesting 🤔 do the naval sams cause issues when they're on land too?

[blazing wagon]

I didn't go that far in testing

I solved my issue and then washed my hands

[woven loom]

fair 😄

But yeah might be worth devs checking that out, seems like pretty big latency spikes

[heavy vortex]

Hmm. When were the "ships" added?

The trailer SAM's from planes DLC, right?

[whole cloud]

I think theyre also especially bad if their scan range reaches outside of terrain bounds

[heavy vortex]

what, is it perlin-generating terrain out there to collide with?

shudders

[whole cloud]

I think so yeah

For vision checks. And there was also something about looking up units. because the quad-tree ends at terrain bounds

[fickle geyser]

not really that hard to achieve with 16km range

[woven loom]

that's sorta why i'm wondering if it's checking with terrain below sea level too

I don't know how it's being done, but maybe worth checking if the terrain LOS checks are happening in the end after a valid list of entities is filtered?

i,e, i hope it's not checking LOS first before checking distance

(because it sounds like it is)

[whole cloud]

2.14.151494 new PROFILING branch with PERFORMANCE binaries, v17, server and client, windows 32/64-bit, linux server 32/64-bit
- Changed: Arma 3 now requires the Microsoft Visual C++ Redistributable version 2022 (Steam should automatically install it)
- Tweaked: Performance improvement when using many R2T textures
- Tweaked: More helpful error messages when servers "keys" folder is missing or empty (thanks to @plucky field )
- Fixed: Desync in remote player gunner animations on static turrets - https://feedback.bistudio.com/T124694
- Fixed: deleteVehicle was not deleting dead crew
- Fixed: setUnitLoadout would not put UAV Terminal items into the GPS slot - https://feedback.bistudio.com/T179631

Note: CAUTION! Starting with Arma 3 version 2.16, we will not offer a 32-bit version of profiling branch anymore. The 32-bit binaries on profiling branch will be non-profiling/performance builds.

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
Note: There are separate Dll files that also need to be placed into Game folder.

[heavy galleon]

What about the AdDoNs/addons on linux?

[shell lion]

Hey,
I have a problem after the last update.

With the mod 3CB Faction the game doesn't start anymore.

For know I found out, only with this mod activatet arma doesn't start anymore.
Without mods or with other mods (ACE,RHS,CBA,TFAR,BW Mod, etc.) it works.

• There is no error message in RPT or elsewhere.
• The game loads an the screen is black with the Arma 3 Logo and then it cut off

I hope you know the issue.

[heavy galleon]

Loaded with 3CB Factions and required mods no problem.

But I do get imediate crashed with other addon presets

(Linux server loaded no problem with same preset)

[whole cloud]

We know of a crash with RHS. One of their helicopters has a configuration error.
I only tested mods without configuration errors, so didn't notice that we weren't handling that correctly.

Hotfix is going to come today

Looks to be 3CB Factions + RHS maybe

[heavy galleon]

I loaded game with 3CB and RHSes no problem

[whole cloud]

But it crashes while loading RHS_AH64D_noradar_AA
How can RHS be fine

[heavy galleon]

loaded normally, can't post screenshot's

https://ctrlv.cz/xfFP

But I am crashing with a larger preset, yet my linux server is up and running no problem. (Or at least didn't crash, didn't try joining yet)

[charred holly]

I also managed to do that before, however when I tried again it failed

[heavy galleon]

two times in row now no problem

[charred holly]

I loaded the mods manually, so something might have changed in the order in the launcher

[heavy galleon]

3*

[whole cloud]

The problem is when the Advanced helicopter sim configuration contains an error
Don't worry about reproing

[heavy galleon]

But it works on my machine

[restive turtle]

https://tenor.com/view/cat-cats-orange-cat-orange-cats-cat-weird-gif-17010424884748616697

[trim ether]

I'll investigate the issue from the 3CB side, tonight.

[whole cloud]

I'll send where the problem is once I got to it.
We also don't report config errors anywhere, its only in internal logs, so modders aren't even told about it. Will be fixed too

[whole cloud]

The error is in "rhsusf\\addons\\rhsusf_c_a2port_air\\AH64\\RTD_AH64.xml" / "rhsusf\\addons\\rhsusf_c_a2port_air\\UH60M\\RTD_UH60M.xml"
And I can repro it with just RHS. Don't think 3CB factions is involved 🤔
RHS has been informed ✅

[orchid mesa]

rip, rhs devs havent done in update in a year or so now

[whole cloud]

ETA is about 90 minutes from now

[fickle geyser]

Anyone did any benchmarking with new profiling?

Wondering if there's any perf gain from new compiler

[whole cloud]

Fix is up
I did some benchmark with YAAB and it wasn't much. Smaller scale tests (like just running SQF script) could have more effect than a test of the whole frame stuff

[fast hornet]

what's the RTDynamics_64.dll it now needs/loads about? 🙂

[fickle geyser]

Advanced flight model

not possible to compile with vs2022 so it was moved to external DLL

[fast hornet]

ah ok

[empty goblet]

if you updating server to profiling just remember to add the dll too 🤣 otherwise you will observe nice crash popup complaining about the lack of the dll

[fast hornet]

yeah, that's how i found out about it 😄
It's just the dev box though..

just copying it over from my client installation worked fine

[empty goblet]

• Fixed: deleteVehicle was not deleting dead crew
  that needs just clarify it solves problem of some invisible dead inside wrecks not being cleaned up (ie garbage collection), deleteVehicleCrew still needed in scripting !

[hybrid cradle]

going to try sqf benchmarking later

[tribal pasture]

Speaking of this, weren't dead bodies moved out before deletion before?

I remember it is buggy but its a thing, you can loot crew once vehicle is deleted

[spiral pond]

isnt vehicle clean-up before crew leading to floating bodies without any gear? (in DS env - cleanup by engine garbage collector. dunno if same with scripted cleanup)

[random isle]

i think that was only case where you trying to delete alive vehicle with alive crew inside

so like had to use

{
    if !(isNull objectParent _x) then { //--- In vehicle check
        deleteVehicleCrew _x; //--- Delete crew
        deleteVehicle (objectParent _x); //--- Delete vehicle
    };
    deleteVehicle _x;
} forEach units _x;

[empty goblet]

this is unrelated , hence that's why it needed the clarify, you still need deleteVehicleCrew

[heavy vortex]

Why though?

Like generally deleteVehicle appears to work.

[fast hornet]

spectating players with switchCamera is glitching out like crazy in v17. Only when they are on foot funnily enough, it still works fine when they are in vehicles.
It's caused by the client apparently, a stable client connected to a v17 server still works fine.

[heavy galleon]

Don't seem to be able to find it on Linux, but the server runs fine. Did Linux server not get this treatment?

[feral harness]

before? this hasnt changed

[feral harness]

remains collector did not delete invisible dead crew before, now it should (corpse collector that is)

[feral harness]

how does this manifest?

[fast hornet]

the camera position seems to jump around a lot.. like if some prediction went to far and then it jumps back to the real position when it gets the next update or something.. i'll pm you a video

[feral harness]

ok

crew dead vehicle shows [] but alldeadmen shows [someunit]. deleting vehicle makes unit visible on teh ground

this hasn't changed.

[fast hornet]

PM'ed you 🙂

[feral harness]

thanks, i will try to repro locally

[fast hornet]

strangely enough it's really bad for some players and not that bad for others.. maybe related to packet loss or other networking parameters? Their latency didn't seem to play big role. The guy in the video i sent you had a ping of about 30ms

[feral harness]

was it very recent? Like recent as world wide internet problems due to cut cables in red sea?

[fast hornet]

around an hour ago and ongoing. Also as i said, the stable client doesn't have the problem even when connected to a v17 server. The server is located in germany and the clients i tried are all land based.. germany, UK, russia, france.. all the same
That traffic won't take that route

[feral harness]

oh the very recent one? Hmm not seeing any change that could have directly affected this, anyway will try to repro first

[fast hornet]

v17 was the first client it noticed it with yeah.. pretty sure it wasn't there before with v16. But i can try if you want to.

[feral harness]

if you still have running v16 yes please this will help to narrow down this

[whole cloud]

sounds like issue with remote animations. third person probably shows it better

got report about remote plaxerr anims brokn in mnp

[fast hornet]

that could explain why it won't happen when the player is in a vehicle i guess

[whole cloud]

• Fixed: Desync in remote player gunner animations on static turrets
  There is that. should only be for inside turrets.. But sounds plausible that it somehow causes issues for units not in turrets
  Will test tomorrow, probably push v18 in afternoon

[fast hornet]

Just tried: can't reproduce it with v16 / 151459. Seems to really be only the v17 / 151512 client.

[whole cloud]

Is it

Spectator with prof client, sees other players jittering.
Or
Spectator, sees other players with prof client jittering.

?
Like prof needed on observer, or prof needed on observee?
Is it a sender or receiver issue of the animation/position data

[trim ether]

@whole cloud Is intercept working with the latest perf build? I've just pulled the latest build from the GIT and it is crashing on load ?

[fast hornet]

Spectator/Receiver needs v17 for it to happen.

don't know what my clients are using, most of them are probably on the stable 2.14

[whole cloud]

Likely to be broken.
Do you mean github releases? they are old.
Intercept Minimal Dev on my workshop has been updated yesterday, but not tested. Please throw me crashdumps (If youre using my workshop build)

[trim ether]

I'm using the one from the intercept GIT (with latest commit yesterday).

[whole cloud]

self-built?
Then send me crashdump and your intercept dll and pdb, I'll check tomorrow

[fast hornet]

steps to reproduce would be something like:

1. start 151512 client
2. connect to a server
3. run something like switchCamera cursorObject; on another player running around on foot in 3rd person
4. spazm-mania 😄

[feral harness]

It happened on internal as well, looks like some unintended change of code made it in.

[median cipher]

May be related, we're seeing some strange anim/movement interpolation issues on the latest client profiling build. Visual weirdness disappears swapping to stable branch. Server and HC are both on the latest profiling build: https://youtu.be/17saY4Se2QY

[stark falcon]

matches what I saw, it's like there's no transitions between animations (just my client on profiling, server on mainline): https://youtu.be/R0kxGaehUbs?t=29

some other possible oddities from rpt

Observer O OpFor AH1:2 (x) REMOTE (UNIT_OPF_AH1_CP) in cargo of O OpFor AH1:2 (x) REMOTE (UNIT_OPF_AH1_CP); message was repeated in last 60 sec: 466

Error: EntityAI SubSkeleton index was not initialized properly (repeated 47x in the last 60sec)

File /z/ace/addons/common/functions/fnc_changeProjectileDirection.sqf..., line 328
Error in expression <private _vlat = vectorNormalized (_vdir vectorCrossProduct [0,0,1]);

I'm not exactly sure what the problem in ace is but it's not something we've even seen before

[heavy vortex]

I think you cut off the Error Params: line there.

[patent dome]

Also @unborn urchin since you seemed to post about Prof branch things in #dev_rc_branch
I'm not able to reproduce any issues with the Dll's. They are there on both the Windows client and server and I can launch the 64-bit and 32-bit of both without an error. Could you provide more information on the issue?

[unborn urchin]

ah this was regarding what ben and dedmen were talking about

dedmen sent a modified exe

[patent dome]

As for running Linux server, it shouldn't need the library. It also doesn't have the PhysX ones and it runs, whereas the Windows one does have those.

[orchid mesa]

anyone else noticed very laggy looking player animations on this latest build?

[thin wyvern]

Read few posts up ☝️

[orchid mesa]

ah ok

[heavy galleon]

Alright, was wondering why I don't have rotolib in there, but the server ran just fine.
I doubt there's any server side physics checks being done, but is just weird that Windows requires it while Linux doesn't.

[whole cloud]

Linux never had rotorlib
In general servers never had it.

Ah I see why. Won't be fixed in v18 but in v19.

v18 with the animations fix is on its way

[whole cloud]

2.14.151524 new PROFILING branch with PERFORMANCE binaries, v18, server and client, windows 32/64-bit, linux server 32/64-bit
- Fixed: Player animations issue in multiplayer since v17
- Fixed: AnimDone event firing constantly after unit's departure from life - https://feedback.bistudio.com/T175658

Note: CAUTION! Starting with Arma 3 version 2.16, we will not offer a 32-bit version of profiling branch anymore. The 32-bit binaries on profiling branch will be non-profiling/performance builds.

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
Note: There are separate Dll files that also need to be placed into Game folder.

[tribal pasture]

• Fixed: Player animations issue in multiplayer since v17
  Was related to
• Fixed: Desync in remote player gunner animations on static turrets - https://feedback.bistudio.com/T124694
  ?

[whole cloud]

yes

[tribal pasture]

Hope the fix is still there, very gamebreaking in RHS in multiplayer

[whole cloud]

yes is

[feral harness]

yes and no, the actual fix didnt cause any problems but the way it was merged into project did (looks like, really weird issue)

[fast hornet]

can confirm the v18 client fixes the camera/animation glitching for me.

[river flame]

Did something break on profiling with RscGroupIndicator in relation to displays and procedural textures? For example using this little snippet in normal branch seems to be fine, but using it on profiling freezes up arma completely. It was fine 1-2 days ago.

test = createVehicle ["UserTexture10m_F", getPos player, [], 0, "CAN_COLLIDE"];
test setPosWorld (vehicle player modelToWorldWorld [0,10,5]);
test setDir getDir player;

teststr = "display" + (random 1e6 toFixed 0);

test setObjectTexture [0, format["#(rgb,1024,1024,1)ui('RscGroupIndicator','%1')", teststr]];

    
onEachFrame {
    private _new_display = findDisplay teststr;
    displayUpdate _new_display;
};```

[feral harness]

adds another glitch though

[green minnow]

Shooting on SOG CDLC with ACE and CBA got this error:

https://i.gyazo.com/9d9b4a3e59620ab44497578d46a698fc.png

Someone has the same error?

[feral harness]

_vdir has illegal float?

[heavy vortex]

private _vdir = vectorNormalized velocity _projectile;

Not sure how that could break. Would need to dismantle it.

vectorNormalized [0,0,0] just gives [0,0,0]

At least on stable :P

[stark falcon]

I got this as well, just in mp sesssion and only occasionally.
I think it starts at ace_scopes_fnc_firedEH but I'm not sure exactly where the bad data is coming from
If you have any way to reproduce it would be a big help (like a specific gun or some other condition)

[whole cloud]

And its definitely new with prof build?
I could make a custom build with a trap to dump memory when it happens, then I can see where it probably came from.
That could be related to our compiler upgrade

I made some changes to projectile velocity, but that was quite a while ago

[whole cloud]

Its fine on our latest internal build. So should be fine on prof too 🤔

[whole cloud]

2.14.151539 new PROFILING branch with PERFORMANCE binaries, v19, server and client, windows 32/64-bit, linux server 32/64-bit
- Fixed: Broken animations in vehicle FFV seats (since v17/v18)
- Fixed: Server binaries required RTD dll's (They are not actually using them)

Note: CAUTION! Starting with Arma 3 version 2.16, we will not offer a 32-bit version of profiling branch anymore. The 32-bit binaries on profiling branch will be non-profiling/performance builds.

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
Note: There are separate Dll files that also need to be placed into Game folder.

[calm mantle]

Has a ticket ever been made, or the issue brought up, that you cannot bind controllers (flight sticks etc.) to addon bindings?

I have now read cba documentation. I now understand how to do it

[restive pilot]

if you mean CBA, it's not a bug

[calm mantle]

Oh man. I've played with cba for so long I've forgotten what is or isn't vanilla. My apologies.

[whole cloud]

But base game now has a modded keybinding system, that supports controller bindings.
Just CBA is older than that.

[calm mantle]

I was just trying to avoid using joystick gremlin for Arma anymore. Looks like I have options. Thanks

[river flame]

hm...weird, i tested it with friends without any mods and they had he same issue (also on 2.14.151539). would it help if i provide .rpt and/or .mdmp files?

[whole cloud]

A freeze dump would help probably

[river flame]

i guess directly to you or per ticket?

[whole cloud]

DM is fine

[whole cloud]

That freeze, actually fits with #perf_prof_branch message
A number was negative indeterminate causing the freeze, that would also call that illegal float error

But just opening map screen using M doesn't freeze? 🤔

[river flame]

in the same way using RscMap doesn't freeze

[green minnow]

Any SOG weapon make me this error with profiling, ACE + CBA.

[whole cloud]

@river flame Its caused by the UI control.
Is that a internal thing that was never used in the retail game?

It is setting ptsPerSquareSea=0 config entry.
Which is very very very unexpectedly bad. I'm surprised it didn't freeze before.
It should've always been freezing like this

[river flame]

idk to be honest, it was just something that i experimented with to make a map interface inspired by squad aswell as do something like a motion sensor. the latter worked until a few days ago using RscGroupIndicator, not sure if retail uses it though. both are under rsctitles config.

[whole cloud]

I fixed that one.
The SOG weapon thing will be something similar. I don't know if there is enough time to finish a prof build today :/ If only we didn't need 32bit 😄

[whole cloud]

Cannot reproduce with XM177E1 weapon from SOG.
Are you sure its "Any" ? Which one did you try?

Also not with M16A1 with scope (If its related to ACE scopes)

[whole cloud]

Do you have repro?

I can see "/z/ace/addons/common/functions/fnc_changeProjectileDirection.sqf"
calling vectorCrossProduct, but its passing valid numbers
_left 0.0158068594, -0.999843419, -0.00795947760
_right 0.00000000, 0.00000000, 1.00000000

I guess I just don't have the right weapon to repro with, but which one is it?

[whole cloud]

Welp time's up, need to build now to get that other fix in 😢
SOGPF one probably not as bad as its just a script error? We'll fix that on monday then if I get a repro till then

[green minnow]

I tested with 5-6 weapons and I got the error, m60, m3...

[whole cloud]

maybe it requires some special ACE settings to trigger that code the right way?

[green minnow]

I just start the editor, put one unit and shoot, the settings are standard

[whole cloud]

I tried the same

[inland dew]

see if there are no ther mods enabled when testing that
who knows, might be due to another mod?

[green minnow]

Nope, no other mods.

[violet elbow]

Dump your ACE debug info to clipboard/RPT and check for any non-default settings, and give classname of your weapon+optic combo

[whole cloud]

@river flame fix is up

[river flame]

jup, seems to be working, thank you!

[river flame]

can also reproduce error, but no idea if my ace settings are default, need to compare new profile with no non default ace settings does not reproduce error, error only appears with advanced ballistics on

[whole cloud]

I didnt have AB on in testing

[stark falcon]

parseNumber "-nan(ind)" was 0 and now is -nan(ind) e.g. finite x went from true to false
(also effects parseSImpleArray)
imho it's not wrong, so it might just be something we have to fix in ace (we're going to redo extensions stuff anyway)
probably due to some change in atof between versions?

[whole cloud]

The other bug with the freeze was that previously x mod 0 was 0. Now its also -nan(ind) as it correctly should be
So not only string parsing

But its correct behavior indeed.
One bug fixed, the cause for the other gotta find.

https://learn.microsoft.com/en-us/cpp/porting/visual-cpp-change-history-2003-2015?view=msvc-170

Floating point conformance
And
Infinity and NaN Formatting
For the parseNumber part.

TLDR; Old Visual Studio was bugged and not behaving according to standard (Even going back as far as 1999). Now fixed.
Would've been a problem anyway because GCC for our linux builds probably doesn't have these same bugs.. And mac port probably.
Specifically for parseNumber that'll mean that linux and windows builds don't behave differently anymore. We'll surely find some more bugs in our code that are to be fixed.

[heavy galleon]

Freezing? Will that sort my occasional freezes I get when testing mods (hemtt launch with filepatching and debug). Also occasional crashes when quitting game? (Windowed mode, clicking X)

[whole cloud]

no. It was a specific freeze when you spawn a specific UI display