new pathing rapes CPU, produces unplayably laggy games

[rene7705]

on many maps spring is now unplayable for me and many others.

it doesnt matter what game/mod, but map does matter.

some maps result in more units getting stuck, and when that happens cpu usage reaches 100%.

even the multithreaded spring does this, as this is what i run.

imo, this "upgrade" of the pathing has turned out to be a bit of a disaster.

[KaiserJ]

what map? what mod?

did you know MT spring is somewhat buggy and can conflict with certain widgets and gadgets; (causing some mods to break, and uncontrollable lag on others) have you tried with the official executable?

how do you know it's pathing lagging you out?

i just find it strange you're the only one to complain of CPU issues, after the latest spring version has been around for a week

you can always revert to an old spring version if you desire the old pathfinding.

[Beherith]

Hit alt+b ingame to see what is eating your CPU. And stop with this reverting nonsense. You should instead test the new builds and report your findings. And no, 'spring now sucks' does not constitute a report.

[rene7705]

what map? what mod?

did you know MT spring is somewhat buggy and can conflict with certain widgets and gadgets; (causing some mods to break, and uncontrollable lag on others) have you tried with the official executable?

several maps, most notably the ones with many height differences..

several mods, i play zero-k and tech-anni, both have this problem.

others are using the std exe and have the same problem

how do you know it's pathing lagging you out?

i just find it strange you're the only one to complain of CPU issues, after the latest spring version has been around for a week

i'm not the first to complain in-game or in the battle chat rooms.

tbh the other players have indicated it's the path finding, probably from experience.

i'll check tommorow with alt-b.

you can always revert to an old spring version if you desire the old pathfinding.

how?

[Floris]

Yeah I also think the new pathing is worse, (pathingwise and performancewise) and I'm using the normal singlethreaded client. But it isnt that much worse performancewise that it lags you out, or comes even close to that.

If I use the MT client nowadays.. I crash before even lagging out. But a few releases before I didnt crash but just lagged out if the game had lots of units. Your fps stays high but your pc can't actually handle it and will lag you out. With the single threaded client you get just like 5 fps but you dont lag out.

[luckywaldo7]

It has already been reported to mantis: http://springrts.com/mantis/view.php?id=2301

Its definitely new since .6 or .7. Probably less noticeable if you are using multithreaded.

If engine devs put out potential fixed versions to test I am sure I could find some people to test with me.

[KaiserJ]

reverting won't allow you to play online; im assuming you don't want that bc you mention multiplayer

pathfinding in the old versions was hideously broken; BA had been tweaked many times in order to compensate for it. there are other factors at play such as heatmapping and collision volumes that require tweaking for proper unit behavior in older mods. the new method is better (IMO) after having seen it working in several properly configured games. so... worse for BA, better for most others.

i would suggest not using MT exe for ZK. im no expert, but since MT focuses more on working with BA, you're sort of asking for trouble (wrong tool for the job, eat your spoon with a fork etc)

if i had to take a stab in the dark, i would say that if you're playing a game with a lot of units, especially stuck ones, that the lag you're experiencing is coming from LOS update mayhem as said stuck units spin around and try to climb hills... causal factor may be bad heatmap settings making units stick, but the issue really is with LOS update being the pig that it is

[ginekolog]

I also notice cpu spikes in big games. I reported here:

http://springrts.com/phpbb/viewtopic.php?f=11&t=25009

Good luck.

[Licho]

It is lagging pathing, we made mantis ticket already.

It happens especially when location is unreachable.

[Cheesecan]

Get a better CPU, I never have this problem.

[luckywaldo7]

Do you play online a lot? What is your lobby nick?

[hoijui]

you can also see this phenomenon when playing with AIs. for example: setup BA, TheRockFinal, 4v4 AIs (i used mixes of KAIK, RAI, AAI and E323AIs, but i guess KAIK only should do too), east vs west. after some time, when there are big unit groups around (getting stuck), it gets down to ~0.05 fps for me, on a pretty recent CPU. This happens in 0.82.7, but did not in 0.82.6.

[smoth]

and you guys wondered why I made my ragey post about these people..

[jK]

Anyone ever checked with ctrl+b if it really is the pathfinder???
Collision checking can be a much much much bigger hit in crowds ...

[KingRaptor]

This replay (attached) illustrates the problem quite well.

Despite the large number of units, no serious lag occurs until relatively small groups of chickens get stuck at a given location, e.g. trying to climb hill in top-left corner.
Killing or redirecting chickens solves lag almost instantly.

Also, this screenshot from a quick test game:

[Forboding Angel]

Well if these idiots would stop using the non-functional MT build of spring (seriously, what a stupid idea to include something that doesn't even work, with the release) we might actually get some decent bug reporting.

[Kloot]

Yeah I also think the new pathing is worse, (pathingwise ...)

Concrete evidence of regressions in 82.7 wrt. previous Spring versions is more useful than personal opinion.

If engine devs put out potential fixed versions to test I am sure I could find some people to test with me.

Here's your chance to prove that assertion: http://springrts.com/dl/buildbot/defaul ... 89799d.exe (everyone in this thread just volunteered, especially rene7705 who snipped a remark about TESTING THOROUGHLY from his first post, so glhf)

[Beherith]

wrt to build 1310:
Units like jeffies and fleas get pushed into impassable areas (feature footprints) on LLTAComplex_V2

Jeffies and flash get stuck on the (non impassable) edges of craters on Ganymede V1. Units push each other into the impassable craters.

[luckywaldo7]

I am a github noob, how can I get the source? I need to compile as I am on linux.

[jK]

http://springrts.com/wiki/Building_Spri ... s_from_Git