#perf_prof_branch

1 messages Β· Page 37 of 1

rain moth
#

profiling includes todays hotfix changes?

#

or rather, can it still be used or do we need to wait for an update?

#

ah, profiling was updated 3 minutes ago πŸ˜„

whole cloud
#

You need data update, but can use old profiling binary

#

but yeah seems that was just pushed

rain moth
#

are those part of the update just pushed out?

#

πŸ‘

sterile jewel
#

any easy way to get the performance branch using steamcmd?

rain moth
#

yes, define the branch profiling

#
SET ARMA3_SERVER="C:\arma3"
SET STEAM_USERNAME=???
SET STEAM_BRANCH=profiling
SET STEAM_BRANCH_PASSWORD=CautionSpecialProfilingAndTestingBranchArma3
%STEAMCMD% +login %STEAM_USERNAME% +force_install_dir %ARMA3_SERVER% +app_update 233780 -beta %STEAM_BRANCH% -betapassword %STEAM_BRANCH_PASSWORD% validate +quit```
#

something like that

sterile jewel
#

Thank you!!

scenic canyon
#

What is the best server/client configuration for getting the most out of the performance/profiling branch? Like server on performance binary and client on performance binary? Or now with the hotfix should the client just go back to the stable branch and get the same effect?

tranquil basalt
#

Anyone else having less server FPS on the latest performance binary using the same server files and mission? Could be just me.

#

Also, tyvm to the team for the hotfix and getting the prof branch pushed out so quickly ❀️

whole cloud
#

binary is exact same code as last pre-hotfix perf/prof

tranquil basalt
#

@whole cloud Apologies, I think it was a simple mistake on my end after investigating.

whole cloud
#

Waiting for next prof to finish building be like

rain moth
#

its out πŸ˜›

#

what do I get to enjoy @whole cloud ?

whole cloud
#

1.98.146381 new PROFILING branch with PERFORMANCE binaries, v05, server and client, windows 32/64-bit, included linux 32-bit server

Sync with 1.98 Hotfix
Fixed: Memory Allocator issue causing "Too many virtual memory blocks requested" crashes - https://feedback.bistudio.com/T150427

Not on dropbox/gdrive yet krtecek

#

Only the high prio stuff. The juicy stuff comes next week

brave granite
#

uh juicy, can't wait

rain moth
#

πŸŽ‰

scenic canyon
#

My box could really use some new juicy binaries inside it. πŸ˜‰

brave granite
#

is it as juicy that I am forced to hurry with my ansible playbook?

whole cloud
rain moth
#

its just a different branch than public one, supereasy to switch between πŸ˜›

brave granite
#

sure, the Windows machine is already running on that...but that is not being configured by ansible

rain moth
#

why not? πŸ˜‰

brave granite
#

I use Powershell DSC for that

empty goblet
#

1.98.146381 available as usual on my dropbox/google drive, for manual download

sly mica
#

What was the difference between Perf and Prof again? As ofc the variant provided through steam is Prof.

whole cloud
#

prof has profiling

#

As ofc the variant provided through steam is Prof.
its not, its both

sly mica
#

is it both rolled together?

whole cloud
#

both binaries are included in the download yes

sly mica
#

So does the launcher utilize perf? As there is no exe named performance so I assume the default named is perfomance?

whole cloud
#

perf is default yes

brittle karma
#

Hi guys, new to the perf prof branch. Up to this point, I've managed my servers with TADST and FAST(ER)(2). Any chance I can continue using those mods for server startup? Or do I have to do [most/everything] by my own? Want to get started with the upcoming patch this week.

whole cloud
#

just continue using tadst/faster?

brittle karma
#

Was unsure about doing so, but thanks for the reassurance.
In FASTER there is a dev branch entry. Is this reffering to the Dev RC Branch? Or to the Perf Prof Branch?

whole cloud
#

dev branch I assume

#

aka not RC, not perf/prof.
But dev branch, the actual dev branch

marble mason
#

I already asked about the Prof branch over on their discord and it is a planned feature prob in the next release

#

(faster)

nocturne obsidian
#

After some bad performance last night I managed to track down the offending mods and scripts using the diag_captureFrame (62 fps to 107fps for 4 changes was not bad). However what I am now finding is ACE and other mods are utilising the CBA per frame stuff more and while I could see the advanced ballistics separately taking 2.5ms a frame and I can't see what is taking up the remainder using this method.

I have tried the last 2 versions of dedmen's ArmaScriptProfiler but the prior version crashes arma before it gets going and the new one fails complaining intercept is missing (bug raised).

Any way I can get in and see what CBA is actually running and the time they are taking on the per frame/clock stuff? Looking to at least see where that last 1.5ms is going and see if there is anything I can turn off and live without for a bit more FPS.

whole cloud
#

when did you check the advanced ballistics?
in the first 10-20 seconds of a mission it initializes the terrain.
"complaining intercept is missing" well is intercept missing?

brittle karma
#

Can you give me an invite to the FASTER Discord @marble mason? The one on the BI Forums is invalid

nocturne obsidian
#

I am certainly giving it longer than 10-20 seconds, its still there several minutes later. I'll give it a bit more and see if it stops.

Yes intercept and the PBO associated with it is missing, the release package is broken its missing bits.

brittle karma
#

Thanks alot @marble mason πŸ™‚

whole cloud
#

the release package is broken its missing bits
No its not.

#

Or do you call ACE release broken, just because it doesn't contain a copy of CBA in the download?

nocturne obsidian
#

Maybe update the docs then since its not in the wiki or in the package or on the readme in the github. I am sure you must have documented that dependency somewhere but I can't find it in the source repository. Want to help or keep jabbing at me?

whole cloud
#

Well it tells you intercept is missing when you are launching the game

#

that atleast tells you that there is a dependency missing

nocturne obsidian
#

Alright I am done. Have a lovely day.

whole cloud
#

its a dependency on steam and lists it on the BIF thread

#

Sorry but you are the first one complaining about that, everyone else that talked to me about ASP so far didn't make that mistake.

wide sparrow
#

@brittle karma or just get the binaries from the drive or dropbox and replace the normal binaries (rename the perf.exe after doing backup of the basic one)

fickle spade
#

Is there a sense to use a performance binary as a client? Can't find a command line parameters to edit and to add a -limitFPS flag

whole cloud
#

limitFPS is server side only, and why would you want to limit client fps :u

kindred radish
#

While looking for performance hogs i found this.(Exile server with 48 players online).
I guess the higher the number at the end means worse performance ?

diag_captureFrameToFile 1; (Executed on the server)

* Thread SendMessagesThread
nwSerSMT; 62.89392; 32.41064;""
netSSM; 62.89436; 32.36350;""
evGet; 62.89441; 0.00004;""
crUOC; 62.89856; 29.22024;""
mtPmj; 62.95535; 29.15987;"Lock wait: 0.000 ms"
memAl; 62.96492; 0.00043;""
mtWai; 85.75771; 6.35249;""```

Could this mean that our network config is not optimal ? 

Full log -> https://pastebin.com/yU7w4fxE
fickle spade
#

limitFPS is server side only, and why would you want to limit client fps :u
@whole cloud is it not limited by default?

whole cloud
#

no

#

that'd be pretty stoop

#

Could this mean that our network config is not optimal ?
Yeah could. What is your network config

kindred radish
#
//MaxBandwidth = 629145600; //DISABLED
MaxMsgSend = 256; 
MaxSizeGuaranteed = 512;        
MaxSizeNonguaranteed = 256;
MinErrorToSend = 0.004;   
MinErrorToSendNear = 0.04;        
MaxCustomFileSize = 0;```
#

1 GBit connection

whole cloud
#

minbandwidth is kind-off the most important setting you can set

#

and you leave it default?

#

default is like 130kbit/s

kindred radish
#

ive read somewhere that the server calculates it if you dont set it ^^

whole cloud
#

yes it tries to figure out bandwidth within a certain margin

#

by default that margin is still on 2008 kinda internet speed level

kindred radish
#

so guess i try to enable it for testing purpose ^^

whole cloud
#

also you could potentially increase MaxMsgSend, but first only set MinBandwidth correctly. correctly aka 900mbit-1gbit
its the bandwidth the server is guaranteed to have

kindred radish
#

allright, thanks for the answers πŸ™

brittle karma
#

Is the setting set as (k)bit or (k)byte?

whole cloud
#

bit, no k

brittle karma
#

So for 900 MB/s it would be 7,549,747,200 bit/s, got it πŸ™‚

whole cloud
#

huh?

feral harness
#

without , probably

whole cloud
#

no

#

900mbit/s is 9e+8
What you posted is 7550mbit/s

#

thats about 8gbit/s

#

or are you really on a 10gbit?

brittle karma
#

900 * 1024 * 1024 * 8
MB -> KB -> Byte -> Bit

#

No, on a 1Gb/s connection

rain moth
#

then you want 900 Mbit/s

#

not 900 MByte/s

brittle karma
#

Yeah, I've seen my mistake. I am already on Bit, not Byte

rain moth
#

πŸ‘

brittle karma
#

So it would be 943,718,400 Bit/s

whole cloud
#

That would be 943,7mbit/s

#

mega is the SI unit for million.
one megabit -> one million bits -> 1,000,000
900 megabit -> 900 million bits -> 900,000,000

#

or you can just type into google 900 mbit/s in bps
Though google will sadly shorten that to 9e8

#

though, I think config parser might even be able to handle that

brittle karma
#

Thanks alot dedmen. Been a long time since I've calculated those numbers. Need to fresh up my calculating skills ^^;

#

So I've set min to 900,000,000 and max to 1,000,000,000. Or should that value be higher?

#

Never seen the connection reaching anything higher than ~ 920 MBit on a stable though

whole cloud
#

also 1MB to 1 KB are also not 1024.
1024 is MiB -> KiB

#

no looks good

brittle karma
#

Thanks again πŸ™‚

whole cloud
#

max is above your max, I usually say 120% the max
But being honest, your Arma server will never ever reach that high anyway

brittle karma
#

Probably not, yes

#

20 - 25 concurrent players at the moment. Will take some time until we reach 50+ if we ever would become that big anyways

#

Do I understand this correctly that bandwidth setting can reduce some desync/lag? Or what exactly does it do to the player?

whole cloud
#

if the desync is caused by too many messages accumulating and the server not getting them out fast enough, then yes.
But it also depends on MaxMsgSend (max messages per frame) and accordingly the server framerate.
low fps -> fewer messages

brittle karma
#

Min/MaxBandwidth is working like a charm. Thanks again @whole cloud and all the others helping out with that. Next thing would be MaxMsgSend. IIRC I've read to not set it too high. Somewhere else I've read that it is bound to the bandwidth, what is only logical on my mind.
Server Bandwidth is on 900 MBit/s. We are currently on 48 fps/s without the perf branch with 25 Players. I would like to calculate it with 40 players. Packet size is on 1400 default. Which would bring me to 900,000,000 / 1400 / 40 = 16.071 ... but that sounds a "good bit" too excessive ...
Any best practice on that account?

whole cloud
#

Somewhere else I've read that it is bound to the bandwidth, what is only logical on my mind.
Somewhat. They are the max messages per frame.
it of course cannot send a GB of messages in a single frame if your bandwidth is just 100mbit, so its also limited by bandwidth.

#

set it too high, potential fps drop

brittle karma
#

Any recommendation for 900 Mbit/s and standards server files (switching to Prof after the update)

#

Or should I start with 256 and see if it impacts anything and then go up from there?

brittle karma
#

Set the value to 256 and let the player give me feedback. Also put the ErrorToSend Value each to +1. Looking forward about the feedback :)
Other thing: Any ETA for the perf/prof branch updates? Do you release stuff on the weekends as well? Don't want to brag about it. Just being super hyped for the update πŸ™‚

whole cloud
#

Also put the ErrorToSend Value each to +1
I hope you don't mean 1.0 ..
Any ETA for the perf/prof branch updates?
I wanted it out yesterday, but things don't always go as I want to. There's never a safe ETA.
I can only say what I intend, don't know if it'll work out

brittle karma
#

@whole cloud

I hope you don't mean 1.0 ..
Jesus Christ Dedmen ... No! Of course 0.00(0)2! kruemel

I wanted it out yesterday, but things don't always go as I want to. There's never a safe ETA.
Absolutely not! But thanks for your effort and great job! Gladly waiting some more time to get a (more or less) stable update!

I can only say what I intend, don't know if it'll work out
No hard feelings! Take your time, let us know how it is going.

jolly prairie
fickle geyser
#

Yes, but you must run profiling exe

jolly prairie
#

thank you @fickle geyser

rapid cradle
#

@whole cloud in the basic.cfg the "class socket" has the values "initBandwidth", "MinBandwidth", "MaxBandwidth" someone sayed there are in bytes/s is this right? because the server MinBandwidth is in bps or im to confused?

whole cloud
#

ye

#

yes

rapid cradle
#

what?!?!? ^^ ... "class socket" is in bytes/s and server MinBandwidth is bit/s?

#

@whole cloud also in the socket class are the values upload or download of the client?

whole cloud
#

yes

#

yes

nocturne obsidian
#

Are we still going to see server transmit data limit fixed by next perf update?

whole cloud
#

Why do you think that we won't?

nocturne obsidian
#

its not that i think you wont 🐺, is there any tracking to see planned releases/features outside of this channel?

whole cloud
#

no

whole cloud
#

1.98.146445 new PROFILING branch with PERFORMANCE binaries, v06, server and client, windows 32/64-bit, linux server 32/64-bit
Added: Linux server 64-bit
Added: Config setting to configure the Steam Query limit (limitation of the mod list in Launcher) - https://feedback.bistudio.com/T123655
Fixed: Inability to store Code type variables in profileNamespace ("Global namespace not passed during" error in RPT)
Fixed: Freeze related to UI Control creation - https://feedback.bistudio.com/T136596
Fixed: Inability to spawn P3Ds inside a scenario file via createSimpleObject - https://feedback.bistudio.com/T147191
Fixed: Server crash when trying to load a scenario with a dot in its filename - https://feedback.bistudio.com/T150767
Fixed: Squad/Arma 3 Units logo not showing up on Linux servers - https://feedback.bistudio.com/T80864, https://feedback.bistudio.com/T136477
Tweaked: Security improvements concerning Headless Clients
Tweaked: Forced whitelisting of 127.0.0.1 in headlessClients[]
Tweaked: Stability improvements
Tweaked: Performance improvements on network processing in MP versus AI

Notes:
Linux x64 Server is still experimental, please report any issues you encounter.
Linux x64 Server requires a recent version of libc, meaning it might not run on old Linux distributions.
Big thanks to our community testers for catching some errors early!
See https://community.bistudio.com/wiki/server.cfg for the new "steamProtocolMaxDataSize" setting.

empty goblet
#

files available on the usual dropbox/google drive location

boreal wigeon
#

steamProtocolMaxDataSize working like an charm! Good job!
oh and also good job on the linux x64 launch works splendid as well! πŸ˜›

nocturne obsidian
#

Praise to dedmen

fickle spade
#

Dedmen, nearRoads?

whole cloud
#

fixied

#

of course πŸ˜‰

fickle spade
#

Great! Thank you. You are the best!

whole cloud
#

That was a fun one to figure out

fickle spade
#

πŸ˜„

whole cloud
#

but in the end it was the same bug as all the other main x64 ones, and I didn't yet search if there are more left. So more bugfixing over the next weeks even if you didn't find the bugs yet πŸ˜„

fickle spade
#

This is a hard developer's destiny πŸ˜‰

spiral pond
#

Tweaked: Performance improvements on network processing in MP versus AI

#

some more info available on that one?

silk summit
#

πŸ‘† wanna know that too!

#

but in the end it was the same bug as all the other main x64 ones, and I didn't yet search if there are more left. So more bugfixing over the next weeks even if you didn't find the bugs yet πŸ˜„
@whole cloud From content creator to bugHunter... both equally important it just just that content can be more exiting πŸ˜‰ (Unless the bug you fix is a big one and/or it is a gamechanger!)

whole cloud
#

Adding features to allow content creators to even do the stuff they wanna do is also exciting

#

some more info available on that one?
Network messages checked AI's for validity, without any need

silk summit
#

Cool! And yeah, your bug fixes certainly inspire us to do more as we feel our needs are being attended πŸ˜‰

whole cloud
#

its not limitFPS's fault

#

probably just degredation the more frames you ran through

#

and due to the higher limit you run through more of them

#

ala diag_frameNo

#

If what i guess is a thing then i guess so

#

don't think so

#

servers rarely run for 6+ hours

#

do a diag_captureFrame on profiling on server to see why its so slow

#

maybe I can find the cause with a couple captures

#

yeah if that works

#

only on profiling

#

lucky

#

that you are running prof

#

usually you'd be running non-prof if you want more performance

#

profiling has worse performance though

#

Seems weird that you say that you want better performance, while you're running the less performance build

#

could also be memory fragmentation over time πŸ€”

#

well profiling is worse than non-profiling.. because of the added.. profiling...

#

k, then I recommend you don't use that anymore

#

no I didn't look

inland dew
#

so what could have been observed in past days, using profiling, to be able to take snapshots is:

  1. no clients/HCs connected -> no netcode activity? -> server FPS doesn't change
  2. 2 HCs connected + low FPS limit -> low netcode activity? -> server is in relatively good shape for like 8 hrs
  3. 2 HCs connected + high FPS limit -> higher netcode activity? -> server is in relatively good shape for like 4 hrs
  4. a lot of players + high FPS limit -> even higher netcode activity? -> server is in relatively good shape for like 2 hrs

so I assume that with higher FPS limit + more players/AI/armor/helis/jets, there is more stuff to synchronize and that something is reached much quicker and once it has reached this something, you run into a wall, resulting abruptly/instantly in very very poor performance

inland dew
#

so yes, -limitFPS doesn't harm in itself, but the higher the FPS, the quicker "the limit" is reached, due to much more frequent updating/synchronization

#

in my understanding at least

#

hope Dedemen will come up with a kinda solution for this, at least partially

frosty nexus
#

We ran the 64bit Linux server over the weekend for a few hours with 40 players or so without issue (just had to remember to rename 64 bit extensions as *_x64.so). Thanks Dedmen and BI team!

feral harness
#

but the higher the FPS, the quicker "the limit" is reached, due to much more frequent updating/synchronization
the sync is not tied to FPS

#

that would be retarded if it did

whole cloud
#

veryfast simulates per frame tho, as fast as possible

nocturne obsidian
#

how to download performance branch?

rain moth
#

Its included with the profiling branch

empty goblet
#

it's the default binary inside the profiling branch

cerulean dirge
#

How much of a difference do you guys notice with the performance branch

rain moth
#

battleye/beserver_x64.so is not included in the linux64.zip nor the linux binary zip file

whole cloud
#

is it missing on the steam branch though. I think I skipped that when checking

rain moth
#

testing

#

no battleye/beserver_x64.so on steam branch profiling

eternal crow
#

profiling branch crashes on startup when I'm trying to load ArmaScriptProfiler1.5.37 with it but only when I set the -profilerEnableEngine flag πŸ€”

rain moth
#

are you using the perf binary or the profiling binary?

eternal crow
#

arma3profiling_x64

whole cloud
#

ASP prolly broke then

eternal crow
#

fwiw , -profilerEnableInstruction works

whole cloud
#

yeah different stuff

#

but afaik nothing of the engine stuff changed

#

but dunnΓΆ

mellow lark
#

Gotta say this new server perf build works wonders

#

I was testing Antistasi with around 100 active AI, and server fps didn't dip below 100 (I had a 120fps limit)

#

AI feels more responsive too

#

This is on a dedicated server with 3 allocated cores ( 6 threads) with an Arma client on that server running alongside it. 3.5ghz i7 9750h

halcyon mural
#

hi just tried 64bit and got ./arma3server: error while loading shared libraries: libsteam_api.so: cannot open shared object file: No such file or directory

#

file is there

rain moth
#

did you install the linux64.zip?

#

x64 server loads libsteam_api.so from a separate directory linux64

bleak garden
#

@nocturne obsidian Make sure to drop the linux64 folder into your arma3folder. It should look like $HOME/arma3folder/linux64

halcyon mural
#

ah ok i will try that thanks

rose prawn
#

how would i run diag_captureFrame 24 command on a multiplayer server @whole cloud

#

oh wait

#

does the profiling .exe not automatically launch thru the arma3 launcher?

stark canyon
whole cloud
#

does the profiling .exe not automatically launch thru the arma3 launcher?
correct, it doesn't

#

on server, remoteExec

rose prawn
#

i don't know what that means

#

i play on an arma life 3 server, does launching the profiling executable enable some command window or console?

empty summit
#

I have a stupid question

#

how do I install both Global Mobilization and perf branch on the (linux) server?

#

GM is not on perf branch and perf binaries are not on GM branch

whole cloud
#

get gm folder from gm branch, then switch to perf and use the gm folder that you grabbed

empty summit
#

it gets deleted, I mean I know I can move it back and forth

#

but it's weird

#

well, removing it from the folder and putting it back at the end of the script works so I'll live with "weird"

rain moth
#

Fetch profiling via arma 3 app id, fetch gm from its depot instead

snow oxide
#

Hey, I seen that I can improve my performances by using Dwarden's things but idk how to use/configure the files

#

I'm french sorry for my english

whole cloud
#

just switch to the steam branch

snow oxide
#

ok thx I'm not very good at configuration installation xD

whole cloud
#

Hey scripting people.
I'll try some things and need some benchmark script. A preferably long SQF script that provides reliable, comparable benchmark results when checked via the debug console code performance thing.
I think execution time of about 1-2 milliseconds would be good for testing, and it needs to be a standalone "one piece" thing that I can just plug into debug console to test.
Would be nice if someone has something nice for this purpose, if not I'll have to clobber some random thing together.

Maybe something that does some complicated math calculation thing. shouldn't be really dependent on ingame state, for example no line intersects or ASLToAGL (queries terrain surface) because these will make the results unreliable

high wedge
#

Can you remind which aspect of VM are you trying to improve? πŸ‘€

#

So it must be long? Like how exactly long? Does a loop count as a long script?

private _i = 500;
while {_i > 0} do {
private _a = _i;
private _b = _i + 1;
private _c = _i - 1;
private _d = _a * _b / (_c + 3);
_i = _i - 1;
};

This one takes 1.55ms for instance 🀷

spiral pond
#

long in sense of much code (lines)? or just execution time?

whole cloud
#

number of instructions is my main target, yeah that loop should work fine... I wanted to get some complicated mumbo jumbo but a simple loop should be fine yeah :3

feral harness
#

This one takes 1.55ms for instance
what are you on super computer?

whole cloud
#

Okey lets go.
Baseline test on current retail build.
Gamestart -> Editor VR -> execute in debug console

3.0581ms
3.01205ms
3.05488ms
game restart
2.96736ms
3.04559ms
2.98806ms

pretty consistent

#

After my first set of changes, same procedure

2.99401ms
2.9528ms
2.96165ms
game restart
2.92128ms
2.99401ms
2.95858ms

well... its a tiny bit better, but doesn't match my results from 2 years ago thonk

spiral pond
#

how about some of the realistic artillery or missile guidance code as ref?

whole cloud
#

This one should be sufficient, lots of instructions due to the looping

whole cloud
#

2.7960
2.8413
2.9098
restart
2.9069
2.9269
2.9298

recheck baseline to make sure all is well

3.0180
2.9615
2.9732

thonk not fast nuff, need to find better fixes ┬─┬ γƒŽ( γ‚œ-γ‚œγƒŽ)

whole cloud
#

2.6162
2.6246
2.6595
2.6809
πŸ‘€
15% πŸ™ƒ
But majority of that change is not viable for multiplayer because of cheaters :/
can only do about 5-8% without opening the door to cheaters

high wedge
#

This one takes 1.55ms for instance
what are you on super computer?
Laptop with 8750H meowtrash
We should call it 'SQFMark 2020' or something

#

15% is a great improvement too πŸ‘

whole cloud
#

I have one more idea, we'll see

#

lol... 20 minutes in I notice that I'm stupid and there's a much simpler solution πŸ€¦β€β™‚οΈ

whole cloud
#

Welll that worked.

2.4838
2.4772
2.5799
game restart
2.4900
2.5214
2.5932

Compared to baseline
baseline was average of 3.021
new is average of 2.524

16.5% improvement, I need to reintroduce one thing I removed

silent nest
#

gj

whole cloud
#

2.8276
2.8385
2.8764

Readding the thing that I need to readd because of cheaters.... back down to 5.7% improvement :U

fickle geyser
whole cloud
#

maybe we can still do that. But not my decision :u

#

wait.. I relaunched the game and now I'm at 2.7 which makes 10% :U

#

That's why you do multiple runs on benchmarks peeps

#

2.7081
2.7805
2.7472
9.1%, fine by me

spiral pond
#

any chance for new dev branch still this week? πŸ™

whole cloud
#

Don't see why not

#

oh wait, dev branch day is today πŸ™ƒ

spiral pond
#

😒

still flare
#

2 questions:

  1. Is it possible to switch to the Performance/Profiling Branch with a Windows A3 server via SteamCMD?

  2. If the answer to #1 is "yes," does the official Steam version of the Perf/Prof Branch include Dwarden's glorious, long-awaited fix for the error message about the server not being able to transmit all data with a large mod pack, i.e. steamProtocolMaxDataSize = 1024;?

Thanks πŸ˜„

rain moth
#

yes to both

#

app_update 233800 -beta profiling -betapasssword CautionSpecialProfilingAndTestingBranchArma3

still flare
#

Magnifico! Thank you. Is there anything else that I ought to know? I am submitting a support ticket to my server provider, and I want to keep it short, sweet, and direct -- just like what you just wrote to me.

rain moth
#

don't forget to update the config with steamProtocolMaxDataSize

still flare
#

I have access to everything except the executable itself, hence the support ticket -- I won't forget that πŸ˜‰ Thanks, mate. Cheers.

#

P.S. What is the maximum for steamProtocolMaxDataSize = __; , while we are at it? Stick with 1024, as Dwarden recommends?

rain moth
#

avoid going over MTU size

still flare
#

What is that normally?

#

(P.S. I will be sure to post my results here after testing tonight. Feedback is the life blood of this sort of project πŸ˜„ )

rain moth
#

depends on hardware and software configs in network equipment between you and the server

still flare
#

Ah. So I am going to assume that 1024 is the recommended 'safe' value and probably solves most problems, anyway, right?

whole cloud
#
  1. Is it possible to switch to the Performance/Profiling Branch with a Windows A3 server via SteamCMD?
    yes
  2. If the answer to #1 is "yes," does the official Steam version of the Perf/Prof Branch include Dwarden's glorious, long-awaited fix for the error message about the server not being able to transmit all data with a large mod pack, i.e. steamProtocolMaxDataSize = 1024;?
    Dedmen's fix*
#

Stick with 1024, as Dwarden recommends?
Dwarden never said a single word about this feature in public.
avoid going over MTU size
Nonsense, having to go OVER MTU size is the whole reason for this setting to exist, the default value is 1024, and is already the MTU limit.
So I am going to assume that 1024 is the recommended 'safe' value and probably solves most problems, anyway, right?
Thats the default value, it won't solve anything if you're having problems.

#

What is the maximum for steamProtocolMaxDataSize = __; , while we are at it?
I think 32bit signed integer, so 3gig.

#

But asking about recommended or maximum with this doesn't make any sense, if its too small it won't work, and if its too big it won't matter as the packet will only be as big as it needs to be.

still flare
#

Thank you for the help, Dedmen and Dahlgren.

whole cloud
#

New profiling pushed 7 hours ago.
Script performance improvements (If you want, run benchmarks and compare stable to performance build) and some Bugfixes, real changelog soon, just woke up sorry :u

whole cloud
#

1.98.146597 new PROFILING branch with PERFORMANCE binaries, v07, server and client, windows 32/64-bit, linux server 32/64-bit

Fixed: CTD when a helicopter's HUD / cockpit digital clock reached a full hour - https://feedback.bistudio.com/T148480
Fixed: CT_XSLIDER was created with CT_SLIDER type
Fixed: Rare Linux server freeze when loading specific modded terrains
Tweaked: Improved general script execution performance
inland dew
#

should it increase FPS or take long for it to lower?

#

or has this more to do with RAM management/consumption?

#

what should we look for?

#

server already updated to this build

whole cloud
#

its improves the speed that scripts execute

#

not really much to do with fps or memory

bleak harbor
#

not on dropbox or google drive? Where can we get it from?

rain moth
#

steam profiling branch

bleak harbor
#

ah ok, thanks

slim gull
#

no perf07/146597 build on both dropbox and google drive

bleak harbor
#

Err..... ^^

whole cloud
#

Yeah @empty goblet does these manually

empty goblet
#

1.98 v07 profiling is available btw. also on dropbox and google drive, i was busy with family issues so wasn't able do it before weekend

inland dew
#

family goes first!

heady parcel
#

πŸ‘†

mellow lark
#

Mhm, google drive tells me they detected a virus on the latest server perf build, surely that isn't the case?

whole cloud
#

there was a new perf/prof pushed to steam yesterday, I think there should be no changes in there but need to check.
Didn't intend to push a new release πŸš’

#

Oh dslyecxi will like this one

#

1.98.146671 new PROFILING branch with PERFORMANCE binaries, v08, server and client, windows 32/64-bit, linux server 32/64-bit

Fixed: Missiles would bounce and fail to explode if scripted view distance is bigger than view distance in video options - https://feedback.bistudio.com/T150535
Updated: BattlEye files to be in sync with main branch
#

@inland dew ^ KK found it finally

inland dew
#

@rain hazel ^^^

inland dew
#

How does it work now then?

#

Adjusts overall in video to what's chosen by client while on server?

whole cloud
#

I think it takes the active view distance

inland dew
#

you mean objects distance?

#

or what do you mean by active?

whole cloud
#

it should take the distance that you can actually see

#

no matter if script or game options is the limiting factor

inland dew
#

so the one on server then

#

I suppose it's mandatory only for clients?

whole cloud
#

everything on perf/prof branch is independent from server

inland dew
#

have to test it now )

#

works!
have shot Titan AT at a house 2200m away from my pos with setViewDistance/setObjectViewDistance 12000m, but in game video settings overall only 500m and missile hit the target!

#

tested in MP on our server

#

many thanks from all the people that knew this problem and from the majority that suffered and continue to suffer, as they don't know the reason why and what to do )

#

very important fix, really

rain hazel
#

Nice!

woven loom
#

it should take the distance that you can actually see

so the one on server then
not necessarily, the view distance you see can is first set by the server (1600m) but then can be over-ridden by script

#

but good to see this fixed πŸ˜„

rapid cradle
#

The game crash with current performance branch for me if i try to get in zeus. Anyone else?

#

Jep @whole clouds ArmaZeusCache was the problem...
"Update: 10. Sep. um 17:41 Uhr - Fixed crash caused by yesterdays Profiling branch update"

whole cloud
#

Yep. I didn't know because I didn't know that we even updated perf branch

feral harness
#

@rain hazel your video helped a lot

rain hazel
#

The only reason that was possible was because of that initial bug report I linked in the description, so props to that person for it.

bleak harbor
#

@whole cloud in your email of yesterday about perf08, there's only up to perf07 available in the google drive and dropbox folders. Just thought I'd let you know

whole cloud
#

Dwarden uploads these with some delay πŸ™ƒ And I didn't even tell him yet

bleak harbor
#

Ok cool. Just thought you should know as it was in your email. No doubt a few have reuploaded v7 like I started doing πŸ˜‚

empty goblet
#

oh, i shall ? i thought it was bugged, considering your reaction

#

plays it wasn't my fault poker face ...

whole cloud
#

yeah wasn't intended ^^

#

really only need client builds

#

no serverside changes

oak nymph
#

Dedman, good to know, thanks.

whole cloud
#

πŸ‘€

whole cloud
#

1.98.146728 new PROFILING branch with PERFORMANCE binaries, v09, server and client, windows 32/64-bit, linux server 32/64-bit
Fixed: Rare server crash
Fixed: RPT spam when using the 'name' script command on a dead person (the command will now return the name of the dead person wherever applicable)
Fixed: Target indicators in the HUD of some Jets were jittering
Tweaked: Minor performance improvements in unscheduled scripts
Tweaked: Improved performance when stamina is disabled
Tweaked: VoN volume handling (check for potential issues)

Also available for manual download (Downloads still WIP, they'll be up soon):
https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm
https://drive.google.com/folderview?id=0B03-H4YIbhkFMUt5RzNqZjFlNGs

woven loom
#

VoN volume handling (check for potential issues)
what are the changes here / what kind of issues would one expect?

whole cloud
#

anything with volume/not hearing in VoN could be issue, its supposed to not change anything

#

also, any luck on the memory CTD stuff?
look at the ticket, its state is "New" and there aren't any comments about it on any progress made on it

rain moth
#

Fixed: RPT spam when using the 'name' script command on a dead person (the command will now return the name of the dead person wherever applicable)
❀️

inland dew
#

VoN thing, is this part of you work on sound handling?

whole cloud
#

that's part of the script command to mute players

woven loom
#

i know the state is new, i'm teh one that wrote it πŸ˜‰

#

also not having any comments about progress is not exactly uncommon on FT even when there is progress or resolution πŸ€·β€β™€οΈ

#

anyway, was just a casual question, but thanks

inland dew
#

I probably have missed something, since one can mute them already since like forever, in players on map and mute when click on a specific player

whole cloud
#

muting is different thing

inland dew
#

muting is not same as muting...

whole cloud
#

the script command is called "setPlayerVoNVolume" and can be used to mute players

inland dew
#

wow!!!

#

so one can lower VoN volume of certain/specific players!!!

#

this is incredible!

#

if only there was a slider in players list, instead of the loudspeaker icon, so one could adjust the VoN volume for each player without the need of any scripts
the slider all the way on the left = mute

empty goblet
#

client win and server win/linux files are now on both dropbox and google drive

inland dew
#

and also like it's now, where you can mute/unmute everybody with 1 click, if there was a bar to change VoN volume for each player
it would have been good to also be able to click on the loudspeaker not only to mute/unmute everybody at once, but also to reset the VoN volume for all the players to same/max volume, equal to max VoN set in audio settings
since it's difficult/impossible to remember for which player(s) you've changed it and to which values

sand karma
#

Ey guys since i have the profiling branch on my server

i sometimes gets a whole list of

11:50:40 Server: Network message XXXXXX is pending.

Is there anything i can check to solve this

i got around 23 Player online and 2 headless clients running

spiral pond
#

too many network message in the queue

#

in other words something is too slow for sending them (hardware or bandwidth overloaded)
or something is generating too many

sand karma
#

Oke so its server side I guess then not the clients

spiral pond
#

i think the message is about outgoing

#

so if its a client induced overloaded of the server, one should at least also see those on the client logs from what i can tell

sand karma
#

Hmm I’ll need to check those aswell then and figure things out

kindred radish
#

I solved all pending message problems by forcing minimum Client bandwidth via sockets class

bleak harbor
#

what value did you go for out of interest?

#

because I found anything over 8000 meant the potato players couldn't join

kindred radish
#
///////////////////////////////////////////////////////////////////////////////
//
// Bandwidth Tuning 
//
// 100MB * 1024 * 1024 = 104857600
//
// Converter: http://www.matisse.net/bitcalc
///////////////////////////////////////////////////////////////////////////////

MinBandwidth     = 943718400;           // Server total guaranteed bandwidth in bits per second!
//MaxBandwidth     = 1153433600;        // Slightly above total bandwidth in bits per second. (broken atm, dont use it !)

///////////////////////////////////////////////////////////////////////////////
// Network Tuning 
// 
// Usually something we developers set. Only change these if you really know
// what you are doing. Do something wrong here and your sever will burn in hell 
// and cause massive desync.
///////////////////////////////////////////////////////////////////////////////

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

///////////////////////////////////////////////////////////////////////////////
//
// Socket (Client) bandwidth settings in bytes
//
///////////////////////////////////////////////////////////////////////////////

class sockets
{
    maxPacketSize    = 1440;
    MinBandwidth     = 185000;         // 1.48 mbit/s     // minimum client bandwith in byte/s
    initBandwidth    = 640000;         // 10 mbit         // Average of client bandwidth in byte/s
    maxBandwidth     = 6553600;        // 50 mbit         // Total Server bandwidth in bytes/s / number of players (In theory but arma doesnt use it at all, so set it high for faster connection/mission download)
                     
};```
bleak harbor
#

oof! πŸ˜„

#

no moaning from player playing on potato internet?

kindred radish
#

nope

#

it's 2020. 1,48 mbit/s is nothing

bleak harbor
#

i know but, I have one guy in the outer reaches of scotland who's been playing for years and he barely scrapes up 400k

kindred radish
#

65 players and not a single pending message in 3 hours. worth it πŸ˜„

bleak harbor
#

but, that's my problem..

kindred radish
#

before we had between 1k and 15k pending messages

sand karma
#

Let me try that out cheers man

untold parrot
#

using profiling build rn and whenever i load up into any map it just instantly crashes

#

it worked fine yesterday though, and the performance then was amazing!!

whole cloud
#

do you get a crashdump? a .mdmp? and a RPT?

#

I can't look at the crashdump currently though, and next update will be out before I can

untold parrot
#

im gonna try again and see what error the crash reporter will give me

whole cloud
#

crash reporter usually just displays useless info

untold parrot
#

oh

#

thanks for the headsup haha

#

how can i check the error then?

whole cloud
#

you can't, I can with the .mdmp file (the report that the crash reporter lets you create), but not this week and not next week πŸ™ƒ

#

if you send me that report, or post it on the feedback tracker I can take a look afterwards

untold parrot
#

ok! ill go trigger the crash again and send the .mdmp file or do the feedback tracker thingie

#

rly appreciate the support!!

whole cloud
#

send me the ticket link when you're done so I can put it on my todo list

untold parrot
#

needed to make an account for the feedback tracker so im waiting for verification email

#

i got the zip with the .mdmp file though!

untold parrot
#

it wont let me open a ticket on the bug tracker as i havent received a verification email yet after trying multiple times. could i just send the .mdmp file or wld that be inconvenient?

whole cloud
#

you can send it to me but I might forget

untold parrot
#

i cant rly open a ticket on the forum bc the email verification isnt working so ill send it to you

#

and if you forget abt it, it's not the end of the world lol

#

ill probs manage not being able to use profiling build for a little while

whole cloud
#

FYI perf/prof update delayed to tomorrow, sorry πŸ™ƒ

wide sparrow
#

no problem

fickle spade
#

@whole cloud are there some changes since main branch update?

whole cloud
#

Just remerge if stuff that was in 1.98 prof.
New stuff maybe end next week dunno how long that takes

fickle spade
#

Sorry, maybe I say something wrong, but I meant - is there something what wasn't included to a latest public release?

whole cloud
#

Yeah

#

Linux freeze fix. And some script optimization I think

#

See the last 2 1.98 prof updates

fickle spade
#

Thanks for information. Will wait for perf branch update then.

whole cloud
#

Can someone confirm to me that new profiling went out on steam?

fickle spade
#

Can someone confirm to me that new profiling went out on steam?
@whole cloud

Update available
* Local build: 5676575
* Remote build: 5615998
#

Is it right version?

#

Because I'm switched to public branch yesterday and can't be sure

#

ArmA Linux Server, ver. 1.98.146728

#

Seemed not updated yet

rain moth
#

no, not updated

#

no changes to steam manifests

whole cloud
#

πŸ™ƒ delays

night mulch
#

perf branch when xd

#

i cant play without it anymore

whole cloud
#

Sorry, unable to say :/ More things went wrong πŸ™ƒ

night mulch
#

that's ok, keep up the good work! πŸ‘

silk summit
#

i cant play without it anymore
@night mulch There is really that much of a difference?

night mulch
#

well, is not day & night difference, but its definitely noticeable!

#

and come on man, even 1% improvement is welcome for this game πŸ˜‚

obtuse sinew
#

can i play 2.0 with profiling build

#

has the update been pushed out yet

whole cloud
#

No

#

We hope tomorrow, but there's no guarantee of no unforseen consequences. A resonance cascade might be imminent

frozen meadow
#

Guys I know that I'm probably asking the same question as many other but... Is there a chance that performance version would get an update by the mid day CET? [Excuse a very generic reason] Me and most of the mission makers (in my group) are unable to go on with the mission 😦 I know you're doing your best and we are grateful for that it's just this inconvenience of telling about 50 people that the mission needs to be cancelled redchain

inland dew
#

read Dedmen's message again

analog jolt
#

well, it's after midnight in Europe, so I doubt anything will happen till Monday

#

since Friday's are "no push Fridays" in general

frozen meadow
#

I know I'm just hoping to get an honest answer like "guys, we can't comfortably pull it off, just wait till monday" instead of it's delayed till tomorrow unless something bad will happen

analog jolt
#

temp fix is to run the stable server version

#

it is an honest answer... devs never know if something breaks while releasing something, even with proper QA

#

and even when they release it, it could still break servers and require a fix which could take days

frozen meadow
#

@whole cloud can you please have a word? Sorry for tagging you

inland dew
#

@frozen meadow never tag devs. read the rules

#

he's not your personal dev

frozen meadow
#

Never said he is... I'm just sitting here drunk, sorry for that!

#

Not sure what to do

inland dew
#

the answer was already given in his last message, like I already pointed it out

analog jolt
#

and a temp fix as well

#

since 2.00 should have the performance fixes (and not the profiling stuff)

frozen meadow
#

I'm sorry it's sometimes feels like I dedicated so much for my group I feel very individual about it 😦

fickle geyser
#

You do not need perf branch to be able to play the game.

#

Just switch to stable.

nocturne obsidian
#

It’s a bit of a wait, but should be up fairly soon. Even so, a temporary switch to stable might help if it’s 100% necessary for a group session for you Diego. Although, I understand the hesitation. I love my units, but I refuse to play without the perf_branch because I’m a performance/frames freak.

short fulcrum
#

The perf builds don't offer increased performance, just the latest tweaks and fixes.

whole cloud
#

@frozen meadow read my last message again. I already had a word you just have to read it.
Also... Don't ping me at 2:30 AM in the middle of the night, also.. don't ping me #rules
Thank you.

#

@short fulcrum it currently has performance improvements in it, that aren't in 2.00 stable

#

The latest "tweak" is a performance improvement

whole cloud
#

Sorry chaos and flames everywhere.
Might be issue on Steam side, we can't solve and nows weekend

#

Its already dark outside so maybe I should go home now too

spiral pond
#

sounds like any regular day at the office πŸ˜›

whole cloud
#

But you will get JUICY performance improvement next week ;)
Should be able to get that done for prof v1

#

Well maybe not as juicy for everyone, but still great

bleak harbor
#

Sounds terrific.... Why? What's the change?

inland dew
#

finally your sound thing?

#

decoupling/parallel

#

or something like this

bleak garden
#

i lurk this discord to hear all the juicy news from dedmen

runic sphinx
#

ooh, decoupling is always nice

whole cloud
#

there was no change

#

Oh no

#

what

#

you mean the fail or the cool thing

#

the cool thing I'll let be a surprise

inland dew
#

no, don't tell me I didn't work in the end...

#

I think I know what's your surprise )))
you've 100% fixed OBJECT setVariable NIL does not delete variables

whole cloud
#

leaves the room

inland dew
#

well, then I don't know what else it could be (((

stone patio
#

Howdy is there a eta on the perf build?

rain moth
whole cloud
#

diag_log "teeeeeest";

Before:
0.0297ms
0.0270ms
0.0303ms
After:
0.0044ms
0.0037ms
0.0049ms

tease

whole cloud
#

And remember the abominable game freeze everyone on the server has when someone assigns himself to curator? or some script calls addCuratorEditableAddons with all addons?

17:48:29 ###### Add curator old 0.152273 seconds
17:48:29 ###### Add curator new 0.009469 seconds

20 minutes of work πŸ€¦β€β™‚οΈ

inland dew
#

well, there are also freezes each time you lock a target when flying a jet

#

or when you lock/cycle through targets

rain moth
#

Oh god yes

worldly badge
#

Yep, when there're a LOT of targets

inland dew
#

or simply by pressing the button to lock something, even if there isn't anything that can be locked

whole cloud
#

Can you make me a ticket for that? with repro mission of some AI disabled helicopters to aim at or smth

#

if its bad enough that I need to fix it that is

inland dew
#

will see how I can do it.

whole cloud
#

if its not performance then I don't care πŸ‘€

fickle geyser
#

I recall having some issue with vehicle sensors querying all targets on all clients despite being local to server and causing clients to lag every few seconds.

#

I will try to do a repro. soonℒ️

#

It was something like:

  • empty cup helo on airfield
  • a lot of vehicles and ai around map
  • every N seconds lag (visible as sensor target query in profiler)
#

I think we even talked bout this on ace slack ages ago meowtrash

whole cloud
#

I think I remember such a problem yeah. Target query spamming, I tried to look into that before

#

but yeah, gimme repro and I'll put it on the list for next week

fickle geyser
#

It was really weird that the query was done on all clients.

whole cloud
#

interesting, would warrant a ticket

empty goblet
#

ye, please make the ticket

feral harness
#

It was really weird that the query was done on all clients.
@fickle geyser was there a script command to disable this?

bleak harbor
fickle geyser
#

was there a script command to disable this?
enableVehicleSensor?

feral harness
#

No, cant remember the name

fickle geyser
#

But the issue was that the targeting was done by Empty vehicles, not AI. I need to do the repro but I fear it might be reproductible only with bigger amount of people on dedicated server. Tried it few times already in editor and no dice. The issue was always on our community sessions (30+ ppl on dedicated)

whole cloud
#

perf prof is up now, changelog once I have time

rain moth
#

can confirm ❀️

pastel trail
#

Lovely news, downloading now - thanks!

whole cloud
#

can someone give me the build number of the prof binary that was pushed plox?

pastel trail
#

Point me in the right direction and I'll see if I can.

rain moth
#

Original output filename: Arma3Retail_DX11_x64
Exe timestamp: 2020/10/12 10:56:58
Current time: 2020/10/20 10:53:23

Type: Public
Build: Stable
Version: 2.00.146773

#

same as public? think

#

my profiling branch download was 0 byte when switching from public

#

so it's the exact same data?

pastel trail
#

I can confirm the 0 byte download.

whole cloud
#

do you have the profiling binaries?

rain moth
#

no

#

strangely its the exact same thing both for regular arma 3 (107410) and arma 3 server (233780)

whole cloud
#

but server has the right server binaries?

rain moth
#

0 byte diff from public, no profiling

whole cloud
#

but server doesnt have client binaries, its just same as server main branch?

rain moth
#

yep, same as main branch

#

Original output filename: Arma3Retail_Server_x64
Exe timestamp: 2020/10/12 10:56:58
Current time: 2020/10/20 11:00:53

Type: Public
Build: Stable
Version: 2.00.146773

#

tested both with regular steam client and custom client

whole cloud
rain moth
#

with depot downloader

#
Got AppInfo for 233780
Using app branch: 'profiling'.
Warning: Depot 233781 does not have branch named "profiling". Trying public branch.
Got depot key for 233781 result: OK
Warning: Depot 233782 does not have branch named "profiling". Trying public branch.
Got depot key for 233782 result: OK
Retrieved 1 beta keys with result: OK
Got depot key for 233784 result: OK
Warning: Depot 233787 does not have branch named "profiling". Trying public branch.
Depot 233787 (Arma 3 Server Creator DLC - GM) missing public subsection or manifest section.
Depot 233788 (Arma 3 Server Prague Project 2) is not available from this account.
Warning: Depot 233789 does not have branch named "profiling". Trying public branch.
Depot 233789 (Arma 3 Server Creator DLC - CSLA) missing public subsection or manifest section.
Downloading depot 233781 - Arma 3 Alpha Dedicated Server Content (internal)```
#

don't think I have a dump of the previous profiling branch sadly

whole cloud
#

welp. Atleast its usable now

rain moth
#

πŸ™‚

#

no one can complain about profiling not being 2.00 at least πŸ˜„

whole cloud
#

Well just try to push v1 right away and see if that fixes it

rain moth
#

it fetches a key for 233784 which is the profiler depot

#

I can maybe dig out the raw manifest and check

whole cloud
#

nah looks like we just didnt push really

feral harness
#

But the issue was that the targeting was done by Empty vehicles, not AI.
@fickle geyser Quite impossible, however if you have a repro please do give

fickle geyser
#

Quite impossible
I'll try. It was empty vics it was clear in profiler that it was sensors of CUP_Mi24 taking a lot of ms every few seconds. (causing a freeze). Said mission had only empty helis on a helipad as objective to destroy.

#

Sadly the profier dumps are long lost in ace slack history soo... blobshrug no repro no bug.

whole cloud
#

I saw the profiling capture of it happening

#

but without being able to debug couldnt guess what was going on

inland dew
whole cloud
#

Thanks, link it on the ticket or Ill forget

inland dew
#

don't know how to describe steps to repro this

#

or I just make a ticket only describing this and attach this video

#

it happens with any jet or combat heli when pressing the button to scan/lock targets

#

FPS drops and frametimes skyrocket the more you press the lock/scan button, especially if doing it quickly

whole cloud
#

make a repro mission

#

as you describe it, placing a few helicopters, getting in the gunner seat of something and pressing the lock target button should be sufficent

#

no repro, no fix. Your video doesnt help with that either if there is no good repro

inland dew
#

don't know if different sensors influence it, like only emitting or only receiving or emitting and receiving, their range and number, the range of sensors on your specific jet/heli, how far you're from targets, which targets etc

whole cloud
#

Well either you find out, or you want me to try around blindly for a day and waste my time, which... I will just never do because I dont want to waste my time

#

While I have better things I can do or spend my time on, no repro == no fix
And i have a full task list for probably a couple months

inland dew
#

will see how I can do this and if it's reproduceable in SP, without any other players present

whole cloud
#

@rain moth may I inquire about the current version again?
I would like 146796

rain moth
#

I'll check

fickle geyser
#

["Arma 3","Arma3",200,146796,"Stable",true,"Windows","x64"]

#

profiling branch normal exe

rain moth
#

Original output filename: Arma3Retail_Server_x64
Exe timestamp: 2020/10/17 03:52:28
Current time: 2020/10/20 13:37:10

Type: Public
Build: Stable
Version: 2.00.146796

#

I got the profiling binaries as well

#

Original output filename: Arma3Retail_DX11_x64
Exe timestamp: 2020/10/17 03:52:28
Current time: 2020/10/20 13:37:41

Type: Public
Build: Stable
Version: 2.00.146796

fickle geyser
#

["Arma 3","Arma3",200,146796,"Profile",true,"Windows","x64"]
profling branch profiling exe

rain moth
#

both apps 107410 and 233780 seems to be good πŸ™‚

whole cloud
#

nice thanks

#

Changelog in 30 minutes i think, we actually pushed two prof updates at once then

fickle geyser
#

ASP next? bloblurkinglenny
atm it instantly crashes the game notlikemeow

whole cloud
#

Zeus open freeze fix is next on my list yeah

#

not sure if Ill fix ASP first, probably.

#

ah no

#

I was thinking of AZC :U

fickle geyser
#

notlikemeow rip me

whole cloud
#

on that topic, anyone know if ArmaZeusCache still works on 2.00 prof branch? πŸ™ƒ

rain moth
#

I can check

#

it didn't crash πŸ˜›

#

I'll load a bunch of stuff and make sure it speedy boi too

#

yep, initial load with cup+cfp is super slow

#

but with AZC it's speedy boi after first load

whole cloud
#

ok, then Ill not touch that for now

whole cloud
#

2.00.146796 new PROFILING branch with PERFORMANCE binaries, v0, server and client, windows 32/64-bit, linux server 32/64-bit

- Sync with 2.00 main branch; remerged some changes slated for 2.02 that were also in 1.98 profiling
- Fixed: Rare Linux server freeze when loading specific modded terrains
- Tweaked: Improved general script execution performance

- Fixed: Passing empty texture to setObjectTextureGlobal would not remove texture
- Fixed: RPT spam when using `name` command on a dead person. The command will now return name of the dead person wherever applicable.
- Fixed: ctrlText didn't work correctly with picture buttons
- Tweaked: VoN Direct Chat volume moved from Effects to VON volume slider
- Tweaked: Minor optimisations (array+array operator, polyline rendering on Map)
- Tweaked: Increased CfgSoundCategories maximum "limit" from 4 to 64
rain moth
#

I'll test the weferlingen winter fix on windows linux then πŸ™‚

whole cloud
#

afaik it was linux only issue

rain moth
#

why did I type Windows think

#

x64 linux perf server crashed on weferlingen winter

#

I'll test the 32 bit as well

whole cloud
#

Ill do that instead then, fixing zeus open freeze is annoying and we already have ArmaZeusCache πŸ‘€

#

was there a ticket for that? did anyone want to make one?

rain moth
#

32bit stuck on loading

#

sorry πŸ˜›

vale shoal
#

am I stupid or is there no new download on your Dropbox @empty goblet ?

whole cloud
#

there is none

#

because its not worth the effort when we push another build hopefully tomorrow

nocturne obsidian
#

Apologies if I come across as dumb, but does that mean the current steam prof build is up to date till getting the new exe tomorrow?

whole cloud
#

its up to date now

#

But I want to do another update with more performance stuff soon
So better get ready by switching to the branch now πŸ˜‰

nocturne obsidian
#

Perfect. We appreciate all the work y’all do to bring such a great thing forward πŸ’―

vale shoal
#

I have a last question - I downloaded the profiling branch via steamcmd. there are two _x64 files: arma3server_x64, arma3serverprofiling_x64 - I assume right, that arma3server_x64 is the .exe for performance?

whole cloud
#

ye

empty goblet
#

@vale shoal but they on dropbox and google drive WasntMe

vale shoal
#

cheeky you

#

thanks for your work πŸ™‚

night mulch
#

because its not worth the effort when we push another build hopefully tomorrow
Dedmen, so are we getting new update today? blobcloseenjoy

whole cloud
#

build says 80 minutes remaining. Dunno how long boss will be in office today.

fickle geyser
#

boss has the big red "release" button? πŸ˜„

whole cloud
#

I think its just a small grey button

runic sphinx
#

I guess v1 isn't happening today? blobcatsad

whole cloud
#

no. We'll do tomorrow. That way I'm also there in case of issues, cuz I'll go home soon and if we push now and somethings wrong, bad

inland dew
#

you live there now?

#

thought you were working remotely from your country

whole cloud
#

Yes I now live in my office and sleep under the table

#

πŸ˜„

#

I moved on 2nd oct

inland dew
#

oh, ok

night mulch
#

diag_log "teeeeeest";

Before:
0.0297ms
0.0270ms
0.0303ms
After:
0.0044ms
0.0037ms
0.0049ms

tease
Dedmen, is this part of current build or tomorrow’s?

whole cloud
#

next one

night mulch
#

Oh yeah nice πŸ‘πŸ‘

whole cloud
#

That's why I was told to never promise anything :D even like "new prof tomorrow" lul

bleak harbor
#

Did you see my question about the server/client object not found spam? Just wondering if there was an update on that

whole cloud
#

Yes. No

bleak harbor
#

ok

rain moth
#

new perf/profiling update on steam happy_thonk

whole cloud
#

2.00.146824 new PROFILING branch with PERFORMANCE binaries, v1, server and client, windows 32/64-bit, linux server 32/64-bit

  • Tweaked: Improved startup time when the official server list has to be updated
  • Tweaked: Moved RPT logging into a separate thread to reduce performance impact when lots of logging is going on
  • Tweaked: Added better context for some RPT error messages
  • Tweaked: Mitigated freeze on every MP client when addCuratorAddons is called with many addons
  • Tweaked: Increased default shadow offset to fix shadow banding effect
inland dew
#

Direct chat is more usable now, finally

swift narwhal
#

Thank you Heldmen

runic sphinx
#

I downloaded the profiling and performance branch via steam and it says I'm on 2.00.146824, not 2.00.146818

whole cloud
#

Ahh

#

Ill fix after lunch, probably got one more fix than planned-ish

whole cloud
#

getMarkerPos "mk_2000"
0.0112ms -> 0.0006ms
meowsweats

whole cloud
#

Changelog updated, last entry was missing

vast siren
#

probably a dumb question, but are the performance and profiling builds two different things? or are they both together?

bleak garden
#

together

whole cloud
#

together, but different binaries

vast siren
#

gotcha, so I should be able to hop into my folder and one would be ArmA_ profiling

#

and the other would be ArmA_performance

#

or something like that

rain moth
#

the performance binary is not suffixed, it has the same name as the regular binary (when using the steam branch)

vast siren
#

that's why I didn't see it. Makes sense. Thank you!

whole cloud
#

Also the game is called Arma

short fulcrum
#

ETA on Dropbox/Google Drive?

empty goblet
#

@short fulcrum already in there

lilac fable
#

any chance to bring back say3D working with lip files?

stone patio
#

is the one on dropbox/google drive different to the steam performance profile?

#

or is it the exactly same.

whole cloud
#

same

hexed verge
#

Just to confirm I am not going crazy here. Does anyone else have issues with the latest client Perf x64 binary on windows?

I am experiencing crashes OR sqf errors when just loading altis in the editor and then hitting preview. No mods loaded, game files verified.
Binary version: 2.00.146824
binary name: arma3_x64.exe

if !(isDedicated>
11:34:36   Error Undefined variable in expression: bis_fnc_logformat
11:34:36 File A3\functions_f\initFunctions.sqf..., line 596
11:34:36 Error in expression <qf"],"bis_fnc_execvm",false,false] call bis_fnc_mp;
"initPlayerLocal.sqf" call b>
11:34:36   Error position: <bis_fnc_mp;
"initPlayerLocal.sqf" call b>
11:34:36   Error Undefined variable in expression: bis_fnc_mp
11:34:36 File A3\functions_f\initFunctions.sqf..., line 602
11:34:36 Error in expression <f" call bis_fnc_logFormat;
};
0.90 call bis_fnc_progressloadingscreen;
whole cloud
#

crashes? giv crash reports

hexed verge
#

okay via pm?

pallid remnant
#

pic

rain moth
#

I have no such issue with previewing altis

pallid remnant
rain moth
#

neither crash nor sqf error

hexed verge
#

I am unsure why this suddenly is happening, it has started today after running fine yesterday. Hence my confusion

#

it appears no vanilla sqf function also gets compiled which just adds to my confusion

whole cloud
#

also first report. but I assume perf users are in the dozens

#

maybe something broke initFunctions

#

i had that too yesterday when testing, but i was sure it was caused by intercept being loaded without intercept dll

#

oh your RPT logging is busy πŸ˜„

#

11:26:51 Warning Message: No entry 'bin\config.bin/CfgWorlds.tem_ihantalaw'.
11:26:51 /Context: [] L506 (A3\functions_f\initFunctions.sqf)
[] L515 (A3\functions_f\initFunctions.sqf)
[] L519 (A3\functions_f\initFunctions.sqf)
why does initFunctions look into CfgWorlds :u

#

would be interesting to see what it does there πŸ€”

rain moth
#

looks like the error popup in the "load mission" dialog for non loaded worls?

whole cloud
#

11:27:22 Attempt to override final function - bis_functions_list
your mission start suddenly starts overriding all vanilla cfgfunctinos entries, whut

hexed verge
#

I have loaded an sqm on game start up. I left that in

#

hence why cfg world's is being looked at

#

it also happened when I used a none existing path for the sqm as I tend to do that to "disable that feature". usually just by putting a 1 behind the sqm file extension in the start up param

empty goblet
#

orders v02

fickle geyser
#

blobthinkingeyes all fine on my end

hexed verge
#

i feel like my arma is just cursed now

whole cloud
#

well according to your RPT I have an off by one somewhere

hexed verge
#

i'll just purge the vanilla arma addon folders and binaries and let it re download everything

whole cloud
#

ill check the crashdump on monday

hexed verge
#

sure

#

i'll switch to public branch then.

fickle geyser
#

verify game files?

hexed verge
#

done that

fickle geyser
#

how exactly to repro the issue?

hexed verge
#

still the same. going the nuclear option at the moment.

fickle geyser
#

start the game right into editor?

hexed verge
#

load game with sqm load start param, go into editor, select altis, press preview

#

either either crashed or showed all vanilla sqf functions as undefined

fickle geyser
#

start "" "C:\GRY\Steam\steamapps\common\Arma 3\arma3profiling_x64.exe" ^
-noSplash ^
-enableHT ^
-hugePages ^
-showScriptErrors ^
-par=C:\_TOOLS\arma\dev_cba.txt ^
-config=C:\_TOOLS\arma\server.cfg ^
-name=veteran29 ^
-window ^
-filePatching ^
-noPause ^
-world=VR ^
 ^
-profilerEnableInstruction ^
 ^
-mod="P:\z\afm;P:\x\cba;P:\z\afft;%WS%\@CUP Weapons;%WS%\@CUP Units;%WS%\@CH View Distance" ^
"C:\Users\filip\Documents\Arma 3 - Other Profiles\veteran29\mpmissions\tempMissionMP.VR\mission.sqm"

like this?

whole cloud
#

load game with sqm load start param
never tried doing that, what param exactly

hexed verge
#
"D:\SteamLibrary\steamapps\common\Arma 3\arma3_x64.exe" -world=empty "C:\Users\diwak\Documents\Arma 3 - Other Profiles\Diwako\missions\coldsnap.tem_ihantalaw\mission.sqm" -name=Diwako -showScriptErrors -noPause -enableHT -hugepages -nosplash
#

it is not even a param

#

just a path to an sqm

fickle geyser
#

I'm doing that like 10 times a day

whole cloud
#

Cant think of a change that would mess that up though

fickle geyser
#

all werks fine

hexed verge
#

i'll see if it start working again once i have downloaded all that 60gb

hexed verge
#

Just tried again after the download. Nope, nuke app_data folder and my profile. It loaded into the mission just fine, but restarting or backing out into the editor crashes it. Truly bizarre. Awaiting monday

novel dune
#

i'm getting tons of similar problems loading missions today since swapping to prof branch, lemme get a report for you.

whole cloud
#

exact same 18:08:16 Attempt to override final function - bis_fnc_packstaticweapon

#

but why

novel dune
#

anything I can provide to help you figure it out

hexed verge
#

for some reason the crash is only an issue in 3den so far. I am currently playing an OP on the perf branch without a crash (except for jsrs realted crashes)

novel dune
#

I've noticed though that if I open something in eden, don't push through then jump on a server and load in the mission will crash

rain moth
#

@hexed verge Thanks for reminding me to fix that

hexed verge
#

gimme the pbo once your done πŸ˜„

rain moth
#

Will do, I’ll send it upstream aswell

whole cloud
#

would explain why majority of people dont have issue

hexed verge
#

the crashes i have so far were without jsrs

vast siren
#

I'll have you know at least 3 people in my unit use the perf branch.

rain moth
#

we have a couple of perf branch users at our community as well

vast siren
#

and as for issues, my game's been running peachy.

#

little bit of a boost in FPS and a hell of a lot smoother.

fickle geyser
#

just played ww2 joint ops with perf branch and had more crashes than usually (usually not crashing at all).

rain moth
#

One hour into GM event with zero issues ThinkM

whole cloud
#

Well thats why we do perf branch :3

#

just fire all those dumps into the FT ticket

#

And I'll have a busy monday

granite dock
#

I am noticing some pretty significant stuttering on the performance binary in areas with high amounts of assets.
Reverting to standard exe eliminates the stuttering entirely.

Is there anything specific I can provide?

violet sluice
#

Would you mind pinning the steam code to the channel? I can't seem to find it, i'll find it elsewhere on the internet, but just asking for future reference

#

It autorolled me back onto the normal branch

whole cloud
#

thanks for the reminder, I wanted to do that yesterday

ebon zodiac
#

...

#

Thx @whole cloud

#

The reason I want to use it when I say I Im speak for many of the combat freaks of Altis life as we can see through smoke much clearer

whole cloud
#

wat

inland dew
#

The reason I want to use it when I say I Im speak for many of the combat freaks of Altis life as we can see through smoke much clearer
@ebon zodiac what do you mean? explain/share the trick

ebon zodiac
#

I’ll find the video @inland dew

#

We all use it for this

inland dew
#

yes, please

edgy comet
whole cloud
#

But that's not profiling branch related

whole cloud
#

Uh.. first crashdump I look at is memory allocation fail, while trying to compile scripts. Looks to be related to the "unable to override final" spam in there

whole cloud
#

Uh... can't repro... πŸ™ƒ
tried
#perf_prof_branch message
with a vanilla mission.... but without -showScriptErrors! Maybe that

#

nope, negative

#

repro with the same binary as you are running, negative too...
maybe -world=empty :U all the parameters

#

-showScriptErrors -noPause -enableHT -hugepages -nosplash -world=empty
and missionfile direct to editor, negative too

whole cloud
#

Now starting to mission editor on Stratis, and then switching to altis...
nope ┬─┬ γƒŽ( γ‚œ-γ‚œγƒŽ)

whole cloud
#

Well dunno.. I can't repro with the steps you posted...
Maybe the problem is trying to start a mission on a terrain that you don't have?
Like how Diwako started ihantala winter mission, without having that mod loaded πŸ™ƒ

#

nope :U

#

Don't know what else to do. The crashdumps are simple stuff, but no explanation WHY all scripts try to recompile on mission start, which seems to be the issue. And I can't repro

#

Diwako did it without any mods. Why can I not repro with exact same game files and same binary and same start parameters (although i don't have the mission file, but shouldn't matter as the terrain mod is missing and it'll fail to open it anyway)

whole cloud
#

reproed as far as the

12:03:01 Warning Message: No entry 'bin\config.bin/CfgWorlds.tem_ihantala'.
exact same command line besides different missionfile, even the -name=Diwako.
No repro of either the crash, or the CfgFunctions recompiling everything

hexed verge
#

(β•―Β°β–‘Β°οΌ‰β•―οΈ΅ ┻━┻

#

maybe it is something in the profile? E.g. the vars one. I have there several cba settings in it.

whole cloud
#

can you send me that πŸ™ƒ

hexed verge
#

sure thing. pls no hack

#

now dedmen knows my debug console cache... oh no

whole cloud
#

and your server passwords πŸ‘€

hexed verge
#

oh fugg

hexed verge
#

BTW the crash also happens witohut the sqm path. for some reason i can preview the mission, but the compile fails, I then can back out of the mission and preview again, then all functions are compiled. Hitting restart crashes the game.
It is seemlingly at random, sometimes backing out of the preview into the editor triggers the crash.

whole cloud
#

Reproed :U

hexed verge
whole cloud
#

Well trying to join through into a mission with objects that i dont have, on a terrain that I dont have crashed me :U
Actually not sure if thats really the same crash as yours

hexed verge
#

As said, it happens with no mods loaded and no mission loaded via start up params as well.

whole cloud
#

so whats repro. Start 3den preview, go back preview again. try mission restart a couple times?

hexed verge
#

Yeah. It either happens right away or after backing out and previewing again or restarting.

whole cloud
#

Ah... I guess there is some memory corruption in there πŸ™ƒ

hexed verge
#

It is just weird that it sometimes fails on compiling the functions but lets you preview the mission. Then after previewing again it is all fine, then suddenly crashes. It is something on pre init (?) or when ever A3\functions_f\initFunctions.sqf runs

whole cloud
#

jup. memory corruption somewhere

#

compiling functions I would understand that it crashes there. But I don't understand why it even does that at all

#

mission start shouldnt try to recompile all CfgFunctions and overwriting final variables

#

preInit and postInit events come from initFunctions

#

Ah I got it too
14:04:57 Attempt to override final function - bis_fnc_priorityqueue_popitem I wonder if it won't happen without -showScriptErrors πŸ˜„

#

Yep. without showScriptErrors I can repro the
Attempt to override final function spam, but no crash. I know where the problem is then blobcloseenjoy

#

@hexed verge whats up with your profile.
arma3_x64.exe -name=Diwako -noPause -enableHT -hugepages -nosplash -world=empty
spams Attempt to override
arma3_x64.exe -noPause -enableHT -hugepages -nosplash -world=empty
is completely quiet

hexed verge
#

huh, what you mean?

#

I do start the game with those params

whole cloud
#

I mean. When I start your profile, I get the error spam. If I just use my own profile, all is fine

#

somehow your profile is causing the functions to be recompiled

hexed verge
#

... how?

whole cloud
#

Thats what I'm trying to ask you :D
I don't understand

hexed verge
#

me neither. I have none of these values set anywhere besides as start params

whole cloud
#

Well that'll be for someone else to investigate (future me is also someone else 🧻)

hexed verge
#

hold up hold up. I found this in my 3den config file

class CustomAttributes
{
    class Attribute0
    {
        property="RecompileFunctions";
        expression="true";
        class Value
        {
            class data
            {
                class type
                {
                    type[]=
                    {
                        "BOOL"
                    };
                };
                value=1;
            };
        };
    };
    nAttributes=1;
};

this is the check box in the eden preferences which says "Recompile functions"

whole cloud
#

ahhhh

#

what is a "3den config file" ?

hexed verge
#

Diwako.3den.Arma3Profile

whole cloud
#

AH

hexed verge
#

currently trying if it crashes without this checked

whole cloud
#

So I guess, this is not considered a "bug" then

#

or is it still?

hexed verge
#

well that is a vanilla checkbox

#

it... does not crash any more

whole cloud
#

quite sure its my callstack printing

#

it also has weird char at end

hexed verge
#

rpt spam is also gone

#

well almost

15:04:33 Warning Message: No entry '.gridNumbersOverLines'.
15:04:33 \/Context:     [] L506 (A3\functions_f\initFunctions.sqf)
    [] L515 (A3\functions_f\initFunctions.sqf)
    [] L519 (A3\functions_f\initFunctions.sqf)
S
15:04:33 Warning Message: '/' is not a value
15:04:33 \/Context:     [] L506 (A3\functions_f\initFunctions.sqf)
    [] L515 (A3\functions_f\initFunctions.sqf)
    [] L519 (A3\functions_f\initFunctions.sqf)
S
fickle geyser
#

oof yeah it might be callstack printing then. In case of my crashes which I've sent you it was always when there was some callstack printed

#

eg.

22:07:29 Ref to nonnetwork object 1894581: <no shape>
22:07:30 \/Context:     [] L1 ()
    [] L21 ()
    [] L17 ()
    [] L19 ()
    [] L145 (z\ace\addons\goggles\XEH_postInit.sqf)
    [] L43 (z\ace\addons\goggles\functions\fnc_applyRainEffect.sqf)
    [] L44 (z\ace\addons\goggles\functions\fnc_applyRainEffect.sqf)
    [] L18 (z\ace\addons\goggles\functions\fnc_removeRainEffect.sqf)
    [] L19 (z\ace\addons\goggles\functions\fnc_removeRainEffect.sqf)

22:07:30 Ref to nonnetwork object 1894582: <no shape>
=======================================================
-------------------------------------------------------
cpu:  AMD Ryzen 7 2700X Eight-Core Processor
graphics:  D3D11, Device: NVIDIA GeForce GTX 1080
resolution:  3440x1440x32
Addons:
  ace_medical_treatment in z\ace\addons\medical_treatment\
#
22:11:02 Reading cached action map data
22:11:04 \/Context:     [] L1 ()
    [] L21 ()
    [] L17 ()
    [] L19 ()
    [] L145 (z\ace\addons\goggles\XEH_postInit.sqf)
    [] L43 (z\ace\addons\goggles\functions\fnc_applyRainEffect.sqf)
    [] L44 (z\ace\addons\goggles\functions\fnc_applyRainEffect.sqf)
    [] L18 (z\ace\addons\goggles\functions\fnc_removeRainEffect.sqf)
    [] L19 (z\ace\addons\goggles\functions\fnc_removeRainEffect.sqf)

22:11:04 Ref to nonnetwork object 1840144: <no shape>
=======================================================
-------------------------------------------------------
Exception code: C0000005 ACCESS_VIOLATION at DA879860

etc.

hexed verge
#

yup with mods loaded, I get the same, just on another part. prints a notice --> gone

whole cloud
#

no repro in debug πŸ™ƒ

fickle geyser
#

Some of rpt logs are after crash callstack πŸ€” I guess that's due to async logger?

whole cloud
#

yeah think_turtle I guess

#

atleast the RPT spam doesn't lag you out now πŸ˜„

hexed verge
#

depends on your definition kappa

fickle geyser
#

yeah, it's just now crashing my game sporadically. bloblurkinglenny

whole cloud
#

disable -showScriptErrors and it should be fine :U

#

I'll try to get fixed till tomorrow.
I just cant repro in debug because final variables are ignored ┬─┬ γƒŽ( γ‚œ-γ‚œγƒŽ)

hexed verge
#

i'll try -nologs as well

whole cloud
#

AHHHHHHH notlikemeowcry
make that heresy go away, I made the faster logging so that people would stop doing that

rain moth
#

πŸ˜„

hexed verge
#

haha

#

just trying to make sure this is actually the cause really.

#

so far I hit the restart button 4 times with mods loaded also my mission likes to log as well a bunch when previewing.

whole cloud
#

also testing that my extra context printing gets properly disabled with logs disabled. I hope it does

#

Thanks for being my QA

hexed verge
#

that is fine, I am also the QA in my company despite being a programmer as i manage to break everything while just looking at it

fickle geyser
#

switched back to profling exe and I have no callstack anymore πŸ€”

#

should it work from debug console?

whole cloud
#

only with -showScriptErrors

#

its too expensive so I put it behind that

fickle geyser
#
== "C:\GRY\Steam\steamapps\common\Arma 3\arma3profiling_x64.exe"  -noSplash -enableHT -hugePages -showScriptErrors

Type: Public
Build: Profile
Version: 2.00.146824
whole cloud
#

πŸ™ƒ

fickle geyser
#
test_fnc = { test + 123 getPos asd };
onEachFrame {
    call test_fnc;    
}

just some random stuff to trigger errors.

#

hoped that repeated callstack print will bork the game

whole cloud
#

I'm not in the knowing group

fickle geyser
#

u wot? πŸ˜…

whole cloud
#

it should print if showScriptErrors on

#

If it doesn't then I'm not in the group that knows why

hexed verge
#

In related new, yes removing the script error flag does indeed help.

whole cloud
#

I probably will add a -developmentMode flag or something like that.
And then hide a bunch of RPT logging thats only relevant for mod developers behind that

fickle geyser
#

I'm confused I just switched branches back, starting the game via the same batch file than always... it's not there anymore πŸ˜„

whole cloud
#

its memory corruption, its random-ish

fickle geyser
#

I mean the callstack

whole cloud
#

ah you mean the callstack

fickle geyser
#

is it randomish too? πŸ˜›

whole cloud
#

ahh notlikemeow

rain moth
#

maybe it doesn't feel like it?

fickle geyser
hexed verge
#

just set the random seed to 0

fickle geyser
#

Gotta go back to work, will poke at it later.

whole cloud
#

Oh....
I think any rpt logging triggers callstack... I didn't try diag_log

#

that'd be not wanted, though I guess maybe its fine?

#

Ah no nvm, only error rpt not general

#

brain borked

fickle geyser
#

lolwut

#

I executed the code I've shown above

#

then onEachFrame {} to stop it from erroring.

#

now when I call call test_fnc

#

it prints:

15:33:56 Error in expression <test_fnc = { test + 123 getPos asd };
onEachFrame {
   }>
15:33:56   Error position: <getPos asd };
onEachFrame {
   }>
15:33:56   Error getpos: Type Number,Not a Number, expected Array,Object
15:33:56 Error in expression <test_fnc = { test + 123 getPos asd };
onEachFrame {
   }>
15:33:56   Error position: <getPos asd };
onEachFrame {
   }>
15:33:56   Error Generic error in expression

stuff stuck in buffer/memory issues?

hexed verge
#

oh ffs steam did not update the binary, have to retest the removal of the show script error flag.

restarted my mission 10 times with mods loaded. no crashes

whole cloud
#

oh my gawd what is happening.. WTF.
Now my logging in itself is crashing because whatever ┬─┬ γƒŽ( γ‚œ-γ‚œγƒŽ)

silent nest
#

Idea: Stahp breaking things!
This way, you could work more efficient.

#

(no need to thank me for this, i know this idea is great! You can freely use it!)

hexed verge
#

words to live by

silent nest
#

Yeah

#

(waiting for mute hammer meowtrash )

whole cloud
#

something is writing pointers to script instructions into my logging context on the stack.. WAT

silent nest
#

wut?

#

"overflow" somewhere?

fickle geyser
#

mondays

whole cloud
#

#QuickFix

  IDebugLogContext *_prev = nullptr;
  IDebugLogContext *_next = nullptr;
  char dontStackoverflowIntoMeYouDingus[256] {};
#

1> error C2220: warning treated as error - no 'object' file generated
1> warning C4351: new behavior: elements of array 'dontStackoverflowIntoMeYouDingus' will be default initialized
(β•―Β°β–‘Β°οΌ‰β•―οΈ΅ ┻━┻

silent nest
#

"dontStackoverflowIntoMeYouDingus"

#

πŸ˜‚

whole cloud
silent nest
#

I can assure you this:

#

It's borked.

whole cloud
#

I think one of my performance improvement fixes is cooperating with the better RPT logging context and they try to fight against me

whole cloud
#

okey one bug found, but that shouldn't be it

#

AHHHHHH

#

frick

#
    size_t finalSize = 0;
    for (const auto& it : *this)
    {
      finalSize += it.GetLength();
    }

    auto buffer = result.CreateBuffer(finalSize + 1);

    for (const auto &it : *this)
    {
      auto length = it.GetLength();
      memcpy(buffer, it.Data(), length);
      buffer += length;
    }

    buffer[finalSize] = '\0';

fellow C++ programmers can have a laugh now

rancid swallow
#

So if I have the performance branch on and a friend don’t and he starts a server can I join him?

fickle geyser
#

yes, perf branch is always compatible with current stable.

#

fellow C++ programmers can have a laugh now
care to explain the "joke" to st0pid web devs that do not understand anything from it? notlikemeow

#

also, inconsistent code formatting reee

for (const auto& it : *this)
for (const auto &it : *this)
whole cloud
#

one is auto formatting, the other is me overpowering the stupid auto formatting that doesn't do stuff I want

#

buffer[finalSize] doesn't work.
I already did buffer += length and advanced the pointer.
so buffer pointer now points to its end. but I go another finalSize offset further past the end

fickle geyser
#

thanks for explaination, whoopsie.

rancid swallow
#

Ok I think I tried joining a friends server but did not work it could have been a mod

whole cloud
#

2.00.146846 new PROFILING branch with PERFORMANCE binaries, v2, server and client, windows 32/64-bit, linux server 32/64-bit

  • Fixed: Crash when script callstrack was logged using -showScriptErrors
  • Fixed: Freeze when opening the Zeus interface with many mods active
  • Fixed: Linux server freeze / crash when loading Weferlingen Winter
  • Fixed: say3D start offset parameter not working in all cases
  • Tweaked: Improved speed of stringtable loading during game start
rain moth
#

10:27:57 /Context: gm\gm_core\gm_core_localization\stringtable.xml
10:27:57 Unsupported language English in stringtable
10:27:57 /Context: gm\gm_core\stringtable.csv
10:27:57 Unsupported language English in stringtable

#

on linux x64 server

whole cloud
#

A stringtable entry has no "English" entry, and the parser needs to fallback to "Original"

#

I don't (yet) want to logspam what specific keys are unsupported. I will do that later, need to do some command line parameters overhaul

rain moth
#

weferlingen winter still crashes on x64

#

let's try 32bit

whole cloud
#

@marsh echo could take a look at that, but its not important. Only downside is that log line

#

AHHH :U

rain moth
#

32 bit is burning CPU cycles πŸ˜›

whole cloud
#

aww cmon.. it worked

#

I get cpu load freeze on x64 :/

#

I swear! it worked.... I think

marsh echo
#

Oh thats an easy fix! (Stringtable to "Original")

whole cloud
#

Atleast now that it tells you atleast in what file the problem is πŸ˜„

fickle geyser
#

Does this mean that Zeus Cache will be obsolete? bongocat

whole cloud
#

only the zeus open part, not the search part yet

#

14:37:44 Warning Message: Include file KPLIB_debriefs.hpp not found.
meh...
14:37:44 \/Context: Preprocessing file: description.ext from mpmissions/kp_liberation.chernarus_winter.pbo at 187
14:37:44 Warning Message: Include file KPLIB_debriefs.hpp not found.
yeeey

#

Before:

14:48:22 ErrorMessage: Include file userconfig\doesnt_exist.hpp not found.

After:

14:48:22 \/Context: Preprocessing file: userconfig\file1.sqf at 2
Include Stack:
- userconfig\file1.sqf:2
- userconfig\file2.hpp:3
- userconfig\file3.hpp:4

14:48:22 ErrorMessage: Include file userconfig\doesnt_exist.hpp not found.
whole cloud
whole cloud
#

The weferlingen crash thing will have to be on halt for a while. I can't sink so much time into that

wanton lynx
#

I look forward to being able to identify which of my mission makers keeps forgetting semicolons in description.ext

empty goblet
#

the v02 is available as usual on dropbox/google drive

torpid river
#

wow great change on including the.... include... thanks for preparing the pre... thanks for the new errors.

whole cloud
#

Yey! Ive reached peak programmer, people thanking me for introducing errors

torpid river
#

Hey you got thanked for removing useless errors from logs as well

inland dew
#

hey, you can't be wrong if you don't do anything

#

learning by doing + fixing stuff that no-one else would have done otherwise

whole cloud
#

AH! I didn't think of that. Thanks! I shall apply that technique :3

#

goes to bed

inland dew
#

so you're fixing stuff and people report crashes/errors and the best thing is that you understand (even if not insta) where it can come from and fix it to how it was before, at some point, but still keeping your fixes/improvements

#

so win win

feral harness
#

Next profiling will have a few tweaks in regards to chat channel disabling and channel access, please check thoroughly

#
Global channel is always available (lobby, briefing, game) but for the admin and server host when disabled
Other channels UI is not shown when channel is disabled for chat and von and cannot be switched over to (exception when it is forced by script command. but text input not available even if channel UI shows)
Markers now cannot be placed on disabled channels. Only when channel is completely disabled for the client (von and chat) the markers cannot be placed on this channel with place marker UI. If current channel is disabled (forced by script for example) marker placing UI shows this channel disabled in combo box and OK button is disabled
Disabled icons color is now semitransparent red as it was the similar color to direct or global before and wasnt obvious
If we have chatline displayed and change channels, only channels with chat enabled can be selected
If we have von on or talk button pressed while changing channel, only channels with von enabled can be selected
If we have von on and chatline displayed while changing channel, only channels with von and chat enabled can be selected
Custom channels do not show if fully disabled or not active or player is not a participant
getPlayerChannel now returns Direct when person talks on direct (before -1)
channelEnabled now returns [false false] instead of [nil,nil] on invalid channel```
whole cloud
#

2.00.146886 new PROFILING branch with PERFORMANCE binaries, v3, server and client, windows 32/64-bit, linux server 32/64-bit
- Added: Better RPT context for preprocessor errors
- Added: -debug command line parameter for more verbose logging
- Added: "LogObjectNotFound" server config option
- Added: "SkipDescriptionParsing" server config option
- Tweaked: Moved RPT log context to below the actual error
- Tweaked: Further game startup time improvement
- Tweaked: channelEnabled would return [nil, nil] instead of [false, false] on non-existing channels
- Fixed: disabledChannels did not work in briefings
- Fixed: getPlayerChannel would return -1 when a player was talking on the Direct channel
- Fixed: Crash when opening Zeus

fickle geyser
#

Can we expectct -debug being added to launcher?

whole cloud
#

yeah

#

already done (dev branch, next one)

weak panther
inland dew
#

yes, players on our server running current perf/prof have same problem

#

and there might be another problem, but I have to test it first

whole cloud
#

lol what. How :U

feral harness
#

Some text in Russian disappeared after 2.00.146886
You mean there is no Russian text in new 2.00 stable?

#

yes, players on our server running current perf/prof have same problem
if it is not in 2.00 it obviously wont be in perf (unless fixed but we are not aware of this)

inland dew
#

with perf v2 everything was ok
perf v3 russian text disappeared like on the screenshot from Trager

feral harness
#

Still makes no sense if it affects 2.00 already

whole cloud
#

hes talking about perf, not stable 2.00. 2.00.146886 is prof branch

inland dew
#

we have several russian players playing with perf/prof on our sever and yesterday on v2 the text was there, but not there anymore in v3

feral harness
#

What have you broken this time? you are ded, men!

night mulch
#

thanks for the update

#
- Tweaked: Further game startup time improvement

does this only apply to game startup or also mission/island load startup?

whole cloud
#

game

fickle spade
#

@whole cloud Linux server stack while starting without -debug option on the last Profiling version. Sometimes restart helps, but sometimes not.

#
21:20:16 Unexpected stringtable format inside <Text ID="str_input_device_mouse_21:20:16 Cannot register unknown string STR_STEAM_UPDATE_COMPOSITION_QUESTION  21:20:16 Cannot register unknown string STR_STEAM_DELETE_COMPOSITION_QUESTION  21:20:16 Cannot register unknown string STR_PUBLISH_COMPOSITION_PROGRESS       21:20:16 Cannot register unknown string STR_DELETE_COMPOSITION_PROGRESS        21:20:16 Cannot register unknown string STR_COMPOSITION_WORKSHOP_AGREEMENT_FAIL21:20:16 Cannot register unknown string STR_COMPOSITION_STEAM_SLIDER_TOOLTIP_AD21:20:16 Cannot register unknown string STR_COMPOSITION_STEAM_SLIDER_TOOLTIP_FR21:20:16 Cannot register unknown string STR_VEHICLES_ONLY                      21:20:16 Initializing stats manager.                                           21:20:16 Stats config disabled.                                                21:20:16 sessionID: d81f18a205dea3d4c64a4540827a3a2f406c224b

And that's all in log

#

Nothing more

#

And yes, confirmed missed Cyrillic

whole cloud
#

"stack"? you mean stuck?

#

interesting πŸ™ƒ

whole cloud
#

Can anyone else confirm that same thing ^ ?

#

your string errors are definitely weird, they should be in 2.00

I wanted to touch stringtable stuff, but that didn't land in profiling

bleak harbor
#

2.00.146886 new PROFILING branch with PERFORMANCE binaries, v3, server and client, windows 32/64-bit, linux server 32/64-bit
...
- Added: "LogObjectNotFound" server config option
...
@whole cloud Thanks for this. How to enable it? Is it LogObjectNotFound = 0; in server.cfg?

tardy lava
#

server.cfg

class AdvancedOptions {
// Skip logging "Server: Object not found" messages
LogObjectNotFound = true; (default true, logging enabled)
// Skip loading of description.ext/mission.sqm will show pbo filename instead of configured missionName, overviewText and such won't work, but loading mission list in MP is ALOT faster when you have many missions
SkipDescriptionParsing = false; (default false, parse description.ext)
}
bleak harbor
#

Thank you

#

out of interest, where did you find that?

whole cloud
bleak harbor
#

ok, thanks. Needs to have the descriptions commented out. Just for feedback

#

(default true, logging enabled)

whole cloud
#

no it doesn't

#

thats not a copy paste

bleak harbor
#

ok

whole cloud
#

thats a info for the community wiki people on how to document it on the wiki. And they know how configs work and that just a hint

#

I actually overdid my optimization in ArmaZeusCache..
I actually made 3 optimizations while I reimplemented tree filtering there, I wasn't even aware of one of the optimizations I did by accident.

Turns out, the one optimization that i didn't even know I did, improves filtering speed by about 3000x (with my modpack), and the other two optimizations are so small that I might actually disable them again.

whole cloud
worldly badge
#

Uhh... 😳 is this a snek pek?

whole cloud
#

Depends on what u talking about

#

Japanese is broken too :u

worldly badge
#

IIRC Magyar isn't working in current public build?

whole cloud
#

I don't even know what that is

#

"qatest" probably doesn't either.
no its not a teaser

inland dew
whole cloud
#

windows or linux? if windows, did you notice performance difference too?
for both logging being async now, and less stuff logged?

inland dew
#

windows

#

can't tell you

limpid crane
#

also there is still some other object spam left

2020/11/04, 10:32:12 Server: Object info 44:104 not found.
2020/11/03, 20:53:51 Server: Object info 12:38 not found during Changing Owner
2020/11/04, 13:48:57 Client: Object 234:58 (type Type_92) not found.
abstract goblet
#

@limpid crane slightly off topic, but what are you using to get those graphs of the server fps?

limpid crane
#

rpt log parser + https://plotly.com, I just started working on it last week, far from finished.

inland dew
#

on the first graph, until 05:00 am of 3rd November there was perf v2 installed and starting from 05:00 am of 3rd November it was already perf v3

abstract goblet
#

It looks good and would be good to see historical FPS values over a session

inland dew
#

there are also 2 HCs, not represented here

whole cloud
#

also there is still some other object spam left
but the majority is disabled though?

limpid crane
#

yea, you can see it by the log size on the screenshot that Groove posted

whole cloud
#

Found the bug with russian/japanese strings...
We store our text as char, not unsigned char.
so the unicode letters actually turn into negative numbers, which messes up our "remove all trailing whitespace" code

inland dew
#

on your screen I see that there are at least 3 languages missing )
next perf soon then?

whole cloud
#

Same bullshit as how we store array lengths as signed int, because we wanna be prepared for the day that we have negative size arrays!!!! ducking hell mate..

#

not sure if I push a perf this week. Theoretically I can do tomorrow

inland dew
#

would be great for players concerned

#

I also speak russian, but since I always play Arma in english, I'm not concerned

bleak harbor
#

Server object spam has gone for me but, client object and error object remain which I guess was intentional. So, it helps reduce the size of the report but leaves client object spam in the server report which is still relatively sizeable

whole cloud
#

Some programmer wrote code 10 years ago that was wrong, but was undefined behaviour, and ended up doing literally nothing.
Linux compiler complained about that undefined behaviour wrongness, so I fixed it and did what programmer wrote there years ago.
So I fixed the programmers wrong code that wasn't working, and made it work in doing things wrong πŸ€¦β€β™‚οΈ

inland dew
#

Arma...

fickle spade
#

@whole cloud > "stack"? you mean stuck?
Yes, sure.. sorry 🀣

whole cloud
#

when did I say that?

#

or did you say that? what? whats gonig on :confuseddoggo:

runic sphinx
wispy shuttle
#

Hello
Will the perf/prof branch v3 be updated on Dropbox / Google Drive?

whole cloud
#

its broken, and v4 is incoming so probably not

wispy shuttle
#

Ok, thank you for your feedback!

whole cloud
#

2.00.146898 new PROFILING branch with PERFORMANCE binaries, v4, server and client, windows 32/64-bit, linux server 32/64-bit
Fixed: Optics Post Processing effects not updating when switching between optics modes - https://feedback.bistudio.com/T80695
Fixed: Brief game freeze when searching in Tree controls (including Zeus)
Fixed: Broken text in some languages

quaint flame
#

Fixed: Brief game freeze when searching in Tree controls (including Zeus)
"Brief" in some cases 20 seconds πŸ˜„

torpid river
#

speaking of controls, would there be any hope in getting a condition check for 3den category groups displaying in menus? We have a very limited series of simple expression type filters but no filter for non player entities/player entities

#

I do a workaround by hiding/minimizing the categories post load but it's not pretty, it doesn't resize the other controls

#

objectPlayable ?

#

for messy workaround I do things like control = QGVAR(AreaSelectAttributePlayerOnly);

#

with a onLoad onLoad = QUOTE(_this call EFUNC(Core,PlayerOnlyOnLoad));

quaint flame
#

Oh yes, please πŸ™‚

whole cloud
#

"Brief" in some cases 20 seconds πŸ˜„
@quaint flame
I wrote "short game freeze" :U

quaint flame
#

πŸ˜„

#

We have a very limited series of simple expression type filters but no filter for non player entities/player entities
Ideally we could put anything in there, also something like profileNamspace getVar.... dreaming

torpid river
#

that would be nice, but even just a few more general filters would be nice...

#

the whole "dont do complex conditions in simple expressions" thing...

quaint flame
#

Makes no sense in the editor if you ask me. Performance is no concern there.

torpid river
#

shouldn't be, but some of the attribute menus can get pretty slow to load

quaint flame
#

oh yes

#

The vanilla equipment ducking away

torpid river
#

heh

empty goblet
#

as usual the binaries are on the dropbox/google drive ... so i hope we get some feedback thru weekend