Arma 3 STABLE Server 2.18 "profiling / performance binary" feedback

[TheMasterofBlubb 10]

Ohh man im getting old. Seems like they changed something since 8 Years ago. Let me correct this:

PhysX is working on all Nvidia card with CUDA (Compute Unified Device Architecture)

As arma is using the open source variant there is no use of the GPU.

Source Wikipedia.

THX Dwarden again learned something.

[Dwarden 1125]

1.56.134711 profiling & performance v1 server and client, windows/(linux not yet)

+ additional crash fixes , yet the hunt continue (-crashDiag still has use)

++ those crashes were from in new sound engine, scripting, engine init

https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[killzone_kid 1331]

Hi Dwarden,

 

is known any time when the callExtension Bug gets fixed ?

since 1.56 it feels that this happens a lot more than with 1.54.

 

Our players are crying because the get killed when they are in Helicopters when the Simulation Thread freezes. (Helicopters drops down like a stone)

Could you remind, what bug is it?

[torndeco 128]

Could you remind, what bug is it?

Believe they are referring to callExtension blocking the engine.

ALIVE recently switched over to c++, so the new developer prob wasn't aware of this.

[markussr1984 10]

Believe they are referring to callExtension blocking the engine.

 

Right, this is what i mean.

Dwarden tells a few posts before that the fix is not includet in the 1.56 release, but its a terrible bug.

So i hoped that dwarden is able to tell us when there come out a fix for this issue.

I think this bug is priorized nearby the left crash fixes.

[killzone_kid 1331]

Right, this is what i mean.

Dwarden tells a few posts before that the fix is not includet in the 1.56 release, but its a terrible bug.

So i hoped that dwarden is able to tell us when there come out a fix for this issue.

I think this bug is priorized nearby the left crash fixes.

 

callExtension just like normal call command expects return value, so it will hold the thread until it receives it. It is up to you how you want to manage this. It was always like this because this is how callExtension was designed in the first place. In order to minimise dll impact on the game, dll creators start a new thread inside dll and return immediately, then provide a method of retrieving the result asynchronously. There might be some different way of managing extensions in the future but this is a different matter.

[markussr1984 10]

note: we know about the ALIVE issue - beware if callExtension plugin does something extensive it can stall to freeze the simulation thread

 

Sorry my fault :(

 

i didnt realized that this is depending on ALIVE Addon.

 

and KK i had eaten your blog :P

i know how callExtension works  :rolleyes: 

 

OK, then i have to serach whare the freezes come from on my server  :ph34r:

[dvdbrewster 12]

Ohh man im getting old. Seems like they changed something since 8 Years ago. Let me correct this:

PhysX is working on all Nvidia card with CUDA (Compute Unified Device Architecture)

As arma is using the open source variant there is no use of the GPU.

Source Wikipedia.

THX Dwarden again learned something.

  I suppose you could force the setting to your GPU or your GPU, I think it is available in your display drivers control panel. :blink:

[TheMasterofBlubb 10]

  I suppose you could force the setting to your GPU or your GPU, I think it is available in your display drivers control panel. :blink:

Humm i quote Wikipedia this time(with the additional information that Arma uses the Open Source variant of PhysX(tell me if im wrong dwarden)):

German: Wikipedia Article PhysX "Im März 2015 kündigte Nvidia in Teilen die kostenlose Freigabe des PhysX-Quellcodes auf GitHub an, mit Beschränkung auf die CPU-berechnete Codeteile. Die auf CUDA aufbauende GPU-Variante blieb von dem Angebot ausgenommen."

As i could not find the same passage in the english one i just translate: In march 2015 Nvidia announced to share parts of the PhysX-source code for free, with the restriction on CPU- calculation only. The CUDA based variant is excluded from the offer.

[benitoll 16]

Hi, about the memory allocator in the same dropbox folder that the binaries are located, what is the practical difference between "withSSE2" and "withoutSSE2"? I have an i7 3960x. 

 

And, is it worth to use it on Windows 10? I ceased using Fred41's malloc when I read it was worthless if you had Win8 or superior.

Last but not least, please add information about it to the readme (specially if this' been asked, I searched with no result).


Regards and good job!

[Dwarden 1125]

1.56.134739 profiling & performance v2 server and client, windows/linux
+ fixes for issues on windows and linux with multiple users / acess-rights
https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[BL1P 35]

Just wondering are servers still crashing with long persistent missions due to memory and just plain old crash ?

 

Our community died because of the bug, so we no longer have a server on which to get feedback. as this bug was before Xmas maybe its been fixed ?

[Dwarden 1125]

no crashes with v1 or v2 ? remember to toss the crash-dumps and logs on me at end of weekend !

[markussr1984 10]

Sorry Dwarden,

i cant´t report any Crash.

 

just this one crash with 1.56_perf_v0 on HC with i reportet to you. but this was only this one.

sice this our server and hc runs without any crash.

 

Good job Dwarden :)

Time for Holydays for you :P

[Tankbuster 1746]

 toss ~ on me at end of weekend !

You might be putting people off making reports like this, Dwarden.

[Dwarden 1125]

You might be putting people off making reports like this, Dwarden.

 

well if someone want to hoard crashdumps then that's up to him but don't blame me/us the next time it crashes :rolleyes:

[DaViSFiT 21]

So it crashes if too many crashdumps are in the log dir? Ok, who should i blame beside you ? :P

[Tankbuster 1746]

well if someone want to hoard crashdumps then that's up to him but don't blame me/us the next time it crashes :rolleyes:

Just in case you're not with the finer point of your second language; "Toss" is English slang for both "throw" and "masturbate". :)

[markussr1984 10]

156_perf_v2 was just crashing on our Server.

 

Crashlogs was send via pn to dwarden

[dagg929 1]

Crashed on newest perf branch. Logs sent to Dwarden

[Kerc Kasha 102]

Where should I send crash reports now that the feedback tracker is down? getting a lot of crashes since the hitpoint changes, usually involves using mlods using outdated hitpoint selections

[manzarek 14]

Since last update1.56 when a lot a player (like 120-150) download the mission at the same time the server cannot handle it.

The ram used goes to =~ 3.5GO then it crash.

 

A lot of

 

Small mapped regions: 18, size 86016 B
Virtual memory total 4095 MB (4294836224 B)
Virtual memory free 18 MB (19197952 B)
Physical memory free 25895 MB (27153731584 B)
Page file free 28567 MB (29955076096 B)
Process working set 3471 MB (3640262656 B)
Process page file used 3770 MB (3953782784 B)
Longest free VM region: 3276800 B
VM busy 4275703808 B (reserved 205877248 B, committed 4069826560 B, mapped 53534720 B), free 19132416 B
Small mapped regions: 18, size 86016 B
 

 

in the rpt then crash

[Dwarden 1125]

Where should I send crash reports now that the feedback tracker is down? getting a lot of crashes since the hitpoint changes, usually involves using mlods using outdated hitpoint selections

take the rpt, mdmp and bidmp files from moment of crash,

then compress those files into archive

and upload e.g. to dropbox and post URL here (or PM me that if you want privacy)

[manzarek 14]

PM sent :)

Thank you

[Dwarden 1125]

Crashed on newest perf branch. Logs sent to Dwarden

need the RPT too !!! (see PM)