2.2.3 Performance and Pop Jobs (Weights and Triggers)

[JanitorSecondClass]

Removed

 

[MechaThumper]

Or don't even bother with calculating job weights or job switching until the last couple of days of the month (like a week before) since most jobs only really produce resources on the first day of the month.

 

[icon41gimp]

Can you be confident that the checks are actually occurring every day for all jobs if nothing changes? Your experiment only shows that the system will recalculate some job assignments when a change in the population pool relevant to that job occurs.

Inferring the first is true from the second being true seems like a logical leap.

 

[Mithkabob]

Well with my devouring swarm game I can confirm that the game still ticks slow every day after eating the entire galaxy, with no ships moving, no ai doing anything, no sector ai turned on, and no trade happening, so this is a plausible explanation for the game running slow as time progresses since more jobs and pops are filling up the galaxy.

I'm not sure how helpful this speculation is, since the devs have much better tools available for checking exactly how long each part of the engine is running for each game day and I'm sure they are already working on solutions to the problem.

 

[RoverStorm]

Can you be confident that the checks are actually occurring every day for all jobs if nothing changes? Your experiment only shows that the system will recalculate some job assignments when a change in the population pool relevant to that job occurs.

I believe I can help answer this: yes. I was modding in a custom job, and I forgot to give it a weight modifier increase if the job was already held (e.g. 5 times weight to Admin if a pop is already an admin). This resulted in a lower pop (we'll call it Pop1) being promoted and replacing the current pop (Pop2) working that job. Then the next day, Pop2 (who was still higher than lower pops despite being unemployed) took back their job and kicked out Pop1. Pop1 then took back their job the next day, and this went on indefinitely. Every single day of the year, they replaced each other and kept taking the same job every single day.

 

[Ryika]

I believe I can help answer this: yes. I was modding in a custom job, and I forgot to give it a weight modifier increase if the job was already held (e.g. 5 times weight to Admin if a pop is already an admin). This resulted in a lower pop (we'll call it Pop1) being promoted and replacing the current pop (Pop2) working that job. Then the next day, Pop2 (who was still higher than lower pops despite being unemployed) took back their job and kicked out Pop1. Pop1 then took back their job the next day, and this went on indefinitely. Every single day of the year, they replaced each other and kept taking the same job every single day.

I also tested this intentionally a while ago (after seeing a post on reddit about how Glavius AI made pops jump around like crazy), with a very simple, modded setup:

Just by adding...

modifier = {
    factor = 0.1
    has_job = miner
}

...to the worker jobs, in this case the miner job (to make it so once a pop takes the job, it thinks it's ill-fitted for the job).

If this is done for all worker jobs and some worker jobs are available, the result is that worker pops jump in and out of their jobs every day even if nothing happens in the empire.

Unless I'm missing something, this should be direct and conclusive evidence that weights are recalculated on every day.

 

[Pointyearedgit]

Can you be confident that the checks are actually occurring every day for all jobs if nothing changes? Your experiment only shows that the system will recalculate some job assignments when a change in the population pool relevant to that job occurs.

Inferring the first is true from the second being true seems like a logical leap.

We know 2.2.2 reduced daily job calculations:

###################
# Performance
###################

* Fixed performance drop in the resettlement view
* Fixed one source of lag on daily ticks
* Reduced needless daily job related calculations

It seems reasonable to assume, even if we can't be certain. I have suspected this is the problem for some time, as lag seems to increase as the number of pops in the galaxy increases in quantity and variety, and is particularly bad with xenocompatibility, but seemingly unavoidable even with single species hive minds.

It's hard to say what a proper fix should be, the AI seems to have similar planets across the board, but a human player is far more likely to specialize planets (at least during initial development), so a change on one planet may mean that shifting jobs on another is necessary (i've done this manually in SP plenty of times)

 

[icon41gimp]

I believe I can help answer this: yes. I was modding in a custom job, and I forgot to give it a weight modifier increase if the job was already held (e.g. 5 times weight to Admin if a pop is already an admin). This resulted in a lower pop (we'll call it Pop1) being promoted and replacing the current pop (Pop2) working that job. Then the next day, Pop2 (who was still higher than lower pops despite being unemployed) took back their job and kicked out Pop1. Pop1 then took back their job the next day, and this went on indefinitely. Every single day of the year, they replaced each other and kept taking the same job every single day.

I'm not disputing that the process can get into loops like this. What I am saying is that in this scenario each day the job "environment" on a planet changes so the job calculation is likely forced to run again. In most in-game circumstances that environment only changes at specific times - new jobs are created, new pop is grown, resettlement, refugees arrive, etc. It's not necessarily clear that all the job calculations are being run each day when the system is in steady state.