Performance: Multiple Paralell Loops vs One big loop

[barbolani 198]

Hi!

 

As in my environment is difficult sometimes to make performace tests I ask the wise guys.

 

I have some soldier init script, which makes 1 loop to check if he's fleeing and if so, do some captive stuff.

 

My question is, what's better, one script per unit (lets say 200 scripts) with this?

 

waitUntil {sleep 1; fleeing _unit};

 

or one script server sided with this:

 

while {true} do

{

sleep 1;

{

if (fleeing _x) then {do stuff}

} forEach allunits;

};

 

Thanks in advance.

[Moerderhoschi 106]

in theorie for performace, one thread for all, in practice it don't matter.

maybe this thread contains some useful information for you: https://forums.bistudio.com/topic/184412-codeperformance-question-one-thread-per-unit-or-one-thread-for-all-units/

regards

moerderhoschi

[barbolani 198]

Very useful. Thanks.

 

Every day I spend on code optimisation I come up with the same conclusion in the end: Being the worst coder in the world isEqualTo  spawning 8 units :)

 

A fleeing eventhandler would be useful anyway. No one has asked for it?

[zapat 54]

There are some very minor differences:

foreach allunits goes through all units always, instead of the ones actually need to "flee". You need to check which units it applies too.

The one loop runs the check in one "batch", depending on the "do stuff" and whether it is called or spwned it may have some bottleneck buildup.

 

If it's only one check for a variable state, I'd say whichever is easier for you to handle .

 

In my mission I have one big while and that controls all such checks so that they don't interfere with each other. 

[bull_a 44]

forEach loops are less optimised than for loops. If the array you are checking is small, the difference is only points of a microsecond, however for large arrays the difference can increase exponentially.

I recommend using a for loop, for example:

while {true} do {
   _allUnits = allUnits;
   for "_i" from 0 to ((count _allUnits) - 1) step 1 do {
      _unit = _allUnits select _i;
      if (fleeing _unit) then {
         // do something
      };
   };
   // Delay
   sleep 1;
};

Code not tested

 

Hope this helps,

 

Bull

[zapat 54]

Bull, what do you mean under "forEach loops are less optimised than for loops", and where did you get this info? I am just curious, and asking.

[bull_a 44]

Bull, what do you mean under "forEach loops are less optimised than for loops", and where did you get this info? I am just curious, and asking.

 

There is very little in it but there is a slight difference (you can test it using in-game functions from the debug console that measure script execution time), there is information the code optimization page. I'm not entirely sure why it does this (guessing it has something to do with forEach executing in blocks), but from testing in the past (Arma 2 and 3) forEach has (for me) always been the slower of the two when looping through large arrays.

[barbolani 198]

Thanks all.

 

From my understanding the difference is very very small, not noticeable.

 

I am trying to get rid of any waitUntil, while etc.. loop in Antistasi and I was worried about spawning one everytime an enemy spawns. But it seems there is no big difference between both methods and the only (myabe) noticeable thing is to just use eventHandlers...

[fn_Quiksilver 1633]

All else being equal, less threads = less performance burden of your code.

[mindstorm 8]

All else being equal, less threads = less performance burden of your code.

 

Also there is the fact that each thread which takes to long to execute (for example because it uses a while with sleep 1 inside) get's called eah frame, and suspended if it takes more then x amount of time.

Therefor for example 300 threads require more engine work to be done then 1 main thread which loops everything.