[batch] simulation duration increases steadily over time

[lesquoyb]

Is your request related to a problem? Please describe.
I noticed that the duration of the exact same simulation was constantly increasing as it is simulated again in a batch experiment.
For example with this model:

/**
* Name: keepsim
* Based on the internal empty template. 
* Author: baptiste
* Tags: 
*/


model keepsim

global{
	float seed <- 3.1415;
	float start;
	init {
		start <- gama.machine_time;
	}
	reflex d {
		int size <- 100;
		let m <- matrix_with({size,size}, rnd(0,255));
		m <- shuffle(m);
		loop i from:0 to:size-1 {
			loop j from:0 to:size-1 {
				m[i,j] <- rnd(255);
			}
		}
	}
	
	reflex s when:cycle=99{
		save [int(self), gama.machine_time - start] to:"logs/simulationDuration.csv" rewrite:false;
	}
	
}

experiment simulationBase type: batch repeat:500 parallel:1 keep_simulations:false until:cycle=100;

The duration looks like this:

There's a clear linear-like increase in time

Describe the improvement you'd like
The duration of a same simulation should be somewhat constant in a batch experiment

[AlexisDrogoul]

I have the same phenomenon, but it does not seem linear in my case.

[lesquoyb]

It's weird, even if I run 1000 simulations instead of 500 I still end up with something that looks linear:

Maybe the difference is cpu/os related ?

Also just to be sure it's a batch related problem I changed the model a bit to do the same thing but in one simulation and the duration is somewhat constant, as expected:

here is the modified model:

model longsim

global {
	
	float seed <- 3.1415;
	float last;
	
	
	init {
		gama.pref_csv_separator <- ";";
	}
	reflex d {
		int size <- 100;
		let m <- matrix_with({size,size}, rnd(0,255));
		m <- shuffle(m);
		loop i from:0 to:size-1 {
			loop j from:0 to:size-1 {
				m[i,j] <- rnd(255);
			}
		}
	}
	reflex f when:cycle=0{
		last <- gama.machine_time;
	}
	
	reflex s when:cycle mod 100 = 0 and cycle > 1 {
		float now <- gama.machine_time;
		save [int(self), int(now - last)] to:"logs/long sim.csv" rewrite:false;
		last <- now;
	}
	
	reflex stop when:cycle=50000{
		do die;
	}
		
}

experiment a;

[AlexisDrogoul]

You will be happy to know that, thanks to your oral comments and the use of YourKit, I've been able to spot a clear bug.

I've fixed it and the results seem ... quite good I think 😅

it is a one-line bug correction (I've also taken the occasion to comment some DEBUG.OUT statements that, surprisingly, take some time). I will try to make the whole PR thing to submit it.