performance killer behaviour "SAFE"

[Wiper]

anyone noticed this too ?

i recently reduced an AI script down to simply one line which was the culprit for a 30% FPS drop:
_group setBehaviour "SAFE";
executed once for about 30 groups !

i did some further comparisons and finally it was clear that only behaviour "safe" does that much load on the Arma engine.
No issues with "combat", "slealth" or speedmode etc.

It definately looks like a glitch in the engine for me because, if ever, particularly "safe" behaviour should be less cpu intense than f.e. "combat" or even "stealth" behaviour but these modes don't seem to have any performance impact AT ALL.

Can anyone of u guys with a more direct link to BIS and/or the bug tracker (which i left because its too much hassle for me there meanwhile ) report this to the developers ?

gonna post it on the official BIS board shortly too ... or is that one already known and/or even  investigeated in ?
thanks

[Mandoble]

Might be there is a reson for that, in SAFE mode the units still watch for enemies, but, at the same time, vehicles try to follow roads in formation. Probably that road driving is what causes the drop in performance.

[Wiper]

uh - yes, good point.
looks like another argument pushing me forward looking for a new (multi core) system :-\
anyway, thanks for pointing this out Mandoble 

[JasonO]

A multi core system wouldn't do you much good. ArmA currently only runs on one CPU core.

[Nixer6]

I have a Dual Core Intel CPU (C2Duo 6600). I, and some others, did did a few tests.

While ArmA may not in itself be optimized for multi-core CPU's, they do improve my games performance.

Why?? 

I have an Nvidia 8800 GTX. It, and all graphics cards, still use lots of CPU cycles. While ArmA is hogging one core, my GPU and other stuff, sound etc is using the other one.

Sooooooo, if I only had one core, ALL this stuff would be running on that one core, and my games performance would take a hit accordingly. Don't take this as gospel....your mileage may vary. 

This is just how it works for me.

Link to ArmA hardware thread at my "other" hangout

[h-]

Awww, let's not get into the hardware troubleshooting/etc here as it's off-topic 

[Wiper]

lol,
that's why i braced (multi core). Although i'm shure Arma will be optimized fore multicore systems (if its not already) i'd get it just for some other applications and as im shure multi core is the way to go in the near future.

anyway, back to topic.

made some further tests and the results seem to contradict Mandoble's pathfinding theory:

the "setup":
- about 36 vehicle groups randomy placed on Sahrani on road positions
- about 25 infantry groups randomy placed all over  Sahrani

the result:
- putting just the 25 infantry groups into SAFE mode -> noticable FPS drop, about 25-35%
- SAFE mode just for the 36 vehicles (all with embarked crew !) -> no FPS drop at all

weired

[Mandoble]

25 infantry groups of how many units per group? You may try placing as many vehicles as units, and not already on the road so they must to look for roads first.

[Wiper]

the inf groups are randomly varying in size...

therfore i built some specific comparison tests/mission and placed
- 100 cars with one driver on 100 road positions and
- 100 single inf units on that same positions
alternatively

same 50 FPS for both cases - hm

after this i generated 5 man groups instead of singles (thus getting 500 men all over Sahrani) and varied between "SAFE" and "AWARE" as well as on road pos and some 50 meters besides the road. No movement command as this was my original testing condition:

30 FPS in *all* cases

sigh,
no idea what does my fps drop with safe inf in the original mission
anyway, will report back if i stumble upon some further findings