#LTN + alternatives testing
analog girder
Thank you
unique crane
You need Mod Creator role for thread permissions currently 🙂
analog girder
makes sense
Right, so i finished the preliminary set-up for all three mods on my test maps
analog girder
I still need to make sure that all the train limits/requests, etc are gonna be identical between them, and have the same starting-state.
however, i seem have the issue that RLD's priority system is different from Cybersyn & LTN.
It seems more like a priority splitter vs the other two's regular splitter close to beginning of the bus, if that makes sense
so basically all those buffers have to be filled up, even if the trains are available and stations are open for the lower priority stations for that same resource
would you suggest i just stick to testing without any priority?
theoretically they should end up equal after some time allotted to account for that difference
calling @lime dragon as well, since you wanted to be kept updated
austere orbit
I'm gonna be honest I don't think anyone else has used all 3 different mods extensively enough to be able to know everything you're talking about
or will be, I should say
analog girder
basically stations can have priority set on them, and they get served first if the same resources are requested
and this is on all three mods
austere orbit
nothing's stopping you from doing multiple tests
analog girder
That's a fair point
Would Follower Robot Count be the go-to research here as it uses all seven packs?
get as many trains moving as possible and all that
austere orbit
that or you could get the mod that has a tech for all of the sciences and an absurd cost
analog girder
ah yeah, creative mod does have that void technology
damn, looks like i'll need to do a lot of testing on the right request & limits
clearly lower after the mods 
been using a simple stated buffer * 1.5 for those
analog girder
Right, I've done my best at syncing up the settings between each 3 builds. RLD needs further manual setup because apparently it doesn't store any settings whatsoever in the blueprints yet, which is annoying as hell... at least copy pasting within maps still works i guess
unique crane
Modded blueprints are hard 😅
analog girder
currently thinking of doing testing with the following:
No train limits, no priority.
No train limits with priority.
train limits, no priority.
train limits and priority.
analog girder
Modded blueprints are hard 😅
eh, it's probably possible to do a workaround in signals mode which is more like other two mods, but i'm too lazy to switch now
that's about 4 rounds of testing 4 each, assuming i only do them once.
so any suggestions for how long i run the tests?
oh, and also i saw the 'Very Sleepy' profiler mentioned in one of the Alt + F4's. Do you think it would be better to use that?
This one, for C/C++
unique crane
I don't think you need that since you're only interested in lua execution time, not C++ right?
analog girder
yeah, im only looking at how the mods perform
i just joined the technical factorio discord earlier today actually. I should probably ask and see what they think too
random marsh
The game already has built-in tools for measuring save performance. Are they not enough?
analog girder
i was just suggested to use that on the technical factorio discord
i dunno, i went for the profiler cause that's what was suggested last time i asked:
#mod-dev-discussion message
¯_(ツ)_/¯
random marsh
That's if you want to debug mods.
Pretty sure I've told you about command line arguments when you were asking about comparing modded bases specifically.
analog girder
if you did, then i'm sorry, but i can't recall it coming up previously
in any case it sounds like --benchmark is the way to go
random marsh
#mod-dev-discussion message
I guess, I wasn't very clear on what I was talking about 🤔
Sorry about that.
analog girder
No worries 
still have a bit of finalizing to do, but hopefully can get started on actually doing some benchmarks tomorrow/day after
analog girder
yeah no testing today... just debugging blueprints.
I forgot a single power pole on one of them and that caused half the steel production to be treated as nonexistent 
at least once this phase is done the only thing i need change between saves is priority and limits cause everything else will work
analog girder
RLD with limits and priority for 108000 ticks (30 mins)
Average over the whole run:
Trains: 0.772 ms
Train Pathfinder: 0.087 ms
scriptUpdate (which is as close as i can get RLD alone): 0.106 ms
will have to do the other tests later
analog girder
Hmm maybe I should just do 60tick/1s average in future graphs
Bit hard to see at this scale
random marsh
When you get to comparing stuff, make sure the graphs you are comparing are on the same scale.
What's the black line?
analog girder
It says below. Moving avg of 60 tick ~ 1s
analog girder
When you get to comparing stuff, make sure the graphs you are comparing are on t...
Yeah I’ll be sure to do that
Obviously I took some off the top otherwise the intricacies at the bottom would be indiscernible
random marsh
It says below. Moving avg of 60 tick ~ 1s
Oh, so that's just a lot of spikes.
analog girder
Yeah
I mean I haven’t run the other two yet, but we’ll see if they’re similar
I expect LTN to take a bit more on the trains/pathfinder since they always have to return to the depot, but I’m not entirely sure
On the flip side both LTN and cybersyn seemed stabler and lower on the ups
random marsh
What's RLD, again? This? https://mods.factorio.com/mod/RailLogisticsDispatcher
analog girder
Yep
ms numbers seemed higher in actual game, but that’s probably cause I run on integrated graphics
So 1/12 tests done (assuming i don’t do any repeats down the line)
analog girder
RLD with limits and priority for 108000 ticks (30 mins)
Average over the whole r...
LTN test for the same:
Average over the whole run:
Trains: 1.010ms
Train Pathfinder: 0.139 ms
script Update: 0.075 ms
that's pretty interesting: LTN saves updates on the scripting/lua side, but loses all the gains and then some on the trains/pathfinding side. The trains needing to return to the depot likely play a big role there
on the plus side, it is significantly less 'spiky' on the ups
I'll have to change all the Cybersyn stations individually to not have an inactivity condition on the train since RLD didn't have the option for it and i opted for the same on LTN, so that'll be a bit later
Although that could also be that LTN goes round robin style for its providers and will happily grab from far-off outposts.
where RLD doesn't seem to
analog girder
Although I'm now thinking it's the latter reason far more than the former. So a better/more extensive use of requester/supplier priorities will likely allow LTN to catch back up.
I only did priorities on the oil and the input/output of the most central smelters and didn't think about doing anything on the outposts
analog girder
Cybersyn Test:
Average over 30mins:
Trains: 0.747 ms
Train Pathfinder: 0.077 ms
ScriptUpdate: 0.074 ms
Roughly equivalent to RLD on the Trains, while also being nearly identical to LTN on the scripts. Though it is noticeably less-consistent and spikier on the graphs.
So Cybersyn seems to be the clear winner for the first test round. (LTN could probably gain a lot ground back with proper utilization of Priority on the outposts)
random marsh
Interesting.
analog girder
i uploaded the saves and data if anyone wants to have a look
https://drive.google.com/drive/folders/1jHc0v9shLMcKrVo6urEbqd0_IZI1ppMY?usp=sharing
random marsh
Welcome to Windows 😄
analog girder
Welcome to Windows 😄
the OS that always picks the most inconvenient option possible
analog girder
Right, I'm ready to put the results out since viidi hasn't replied and optera is happy to just have it as a public discussion
is this the best spot to post it? https://forums.factorio.com/viewforum.php?f=190 [Mods]
Or is https://forums.factorio.com/viewforum.php?f=88 [questions, reviews, and ratings] better?
analog girder