Give modders the tools to identify and avoid/solve performance bottlenecks

[.kju 3244]

Having a somewhat decent understanding of the engine, know a lot about modding arma,

I believe BI and most people miss out this important aspect.

While performance issues have a wide range of sources, there are two main problems:

1) There are no tools available to identify or narrow down the source of a given performance problem.

This also leads to generic, mostly useless reports for the developers.

2) It appears the mission design, especially the number of entities (AI, players, empty vehicles, objects) and even more in MP due to synchronization, is one key element for certain performance problems.

Hence modders (and players) need tools to be able to identify and cope with these limitations.

I hope you guys help to push this topic.

Vote here: http://feedback.arma3.com/view.php?id=12658

Discuss away to keep this topic on the radar. :cool:

Second highest vote ticket with over 1750 votes: Low CPU/GPU Utilization - need I say more :confused:

Edited August 12, 2013 by .kju [PvPscene]

[petek 62]

upvoted

[jumpinghubert 49]

good idea. Upvoted too.

[Tyl3r99 41]

ooooo +100000 votes (because im god)

[Rydygier 1309]

Something like this will be very useful indeed for better code/mission design. Ability to see, what (running script, unit/object on map, whatever else...) drains how much (value and/or %) of what (CPU load, scheduler load, GPU, memory...). Perhaps even ability to change amount of CPU power given eg. for scripts.

[doln 10]

This would likely require access to the source code, which is incredibly unlikely to happen

[.kju 3244]

No it would not

[Blue1 10]

100% agree!

Of course, this is needed for more advanced Missionmaking.

But this still could be called another "workaround" sadly, ... instead of finally make the ressources of an Year-2013-Server usable for the Engine

(i don't want to repeat the wellknown issues regarding CPU-Core-usages, Scheduling loads and balances... :j: )...

Edit* ...so i think, it would be more useful if BIS invests that time/manpower to fix those problems - instead of inventing monitoringtools (or whatever) to proof the already known problems over and over again ;)

best regards,

Blue1

Edited August 13, 2013 by Blue1

[mac81 0]

upvoted!

[.kju 3244]

@ Blue1

Fair points.

However those deep engine design changes or even optimization are nothing short-term, probably not even mid term (aka A4 at earliest).

And we need both - better tools and an improved engine (at some point).

[On_Sabbatical 11]

Man,thanks for making this thread !

I will add also,we badly need some real network monitoring tool,to allow to evaluate missions' requirements and tune the server's network settings to it !

[enex 11]

but I would imagine devs got some internal tool to monitor performance issues.

[boneboys 0]

This will only help in the continuing development of the game.

Problem identified, patch applied.

[.kju 3244]

I am curious if people see FPS problems with all large® scale MP mission or if its mainly about some specific missions/game modes?

As far as I can tell the 32vs32 Blitzkrieg of mine still had great fps - both for the server and clients.

Therefore I suspect it is mainly about the number of entities (objects, players, AI).

[Vigil Vindex 64]

This is obviously a great idea but it does make me wonder what the BIS developers use to test performance, you would hope they already have bench marking batch scripts in place already...

[[evo] dan 79]

voted on both kju

[fabio_chavez 103]

most mp missions really need optimization, this has my vote!

[sektor 2]

Upvote. The sooner the better.

[rebb 10]

+9001

[D_wolf 1]

If this sandbox game had a proper sandbox editor EVERYTHING would be easier. It feels like we are using technology from the stone age.

I will pay good money if BI releases a DLC of just GREAT modding tools.

http://abload.de/img/lol1ssrg.jpg (199 kB)

[windies 11]

I upvoted this. Kju, do you think that the performance issue's we see come from poor mission making, or is it a deeper engine issue? I can see how limiting the amount of active assets at a given time and the amount of scripts running at a given time can help to increase overall mission performance. What I don't see is how we can work around the limitations of the engine as far as AI, Objects and Scripting is concerned when trying to make a proper mission, especially larger scale coop missions.

[adanteh 58]

I upvoted this. Kju, do you think that the performance issue's we see come from poor mission making, or is it a deeper engine issue? I can see how limiting the amount of active assets at a given time and the amount of scripts running at a given time can help to increase overall mission performance. What I don't see is how we can work around the limitations of the engine as far as AI, Objects and Scripting is concerned when trying to make a proper mission, especially larger scale coop missions.

To me it just seems that we're limited to "small" missions and actually using all of Altis with lots of players and objects will be a no-go. It's just a shame to see we have such a nice map and a sandbox editor but still be plagued by the same old issues. The crazy locality and syncing issues just freak me out. I'd just love to make a custom game mode with near 120 players (or more even) and actually have it run without having 10fps. Even 60 players would be impossible right now though.

[windies 11]

To me it just seems that we're limited to "small" missions and actually using all of Altis with lots of players and objects will be a no-go. It's just a shame to see we have such a nice map and a sandbox editor but still be plagued by the same old issues. The crazy locality and syncing issues just freak me out. I'd just love to make a custom game mode with near 120 players (or more even) and actually have it run without having 10fps. Even 60 players would be impossible right now though.

It's like every iteration of the game, the performance constraints get tighter and tighter. I wonder how much of that is really the fault of mission makers, or if it's something that truly only BI can address. In ArmA I could have huge 100+ AI missions and it would run fine on a good period high spec machine. I also participated in scheduled pvp missions of 120+ players. There were some issue's here and there and the launch was rough, but I never saw terrible terrible performance issue's. ArmA 2 we started seeing the problems surfacing with lower AI capacity and player capacity in multiplayer.

Now we have ArmA 3 and I feel like it's just completely constricting. You can't play with more than 10 players or 20-30 AI without the game engine shitting itself basically. I support anything to help, I just don't see how tools for identifying performance problems are going to help if they just end up telling you that it's the engine at fault and not your mission, or rather your mission with 20 AI is too "Resource Intensive" for the engine.

[.kju 3244]

In the end I believe the engine design/architecture and netcode are the general limiting factor - especially in regards to number of entities.

Good mission design was and is always about intelligent resource usage though (spawn/create stuff only when required, de-spawn stuff, delete obsolete stuff at the basic level).

Now some/many missions out there and popular nowadays are not well coded/design in those regards.

Still it could be also engine issues/bugs relevant to A3 changes (physx).

In any case to learn more about the source of the issues, we need better tools/diagnostics. There is no way around that.

[whisper 0]

upvoted