Loading...
  OR  Zero-K Name:    Password:   

Massive jam at start / memleaks on linux

18 posts, 319 views
Post comment
Filter:    Player:  
sort

PLrankrooks[vegan]
13 days ago
Since like 3 months ago, When game starts there is massive lag with placing labs, up to a point of rejoining menu popping up and joining up for like 3-5 minutes. It happens on linux ubuntu 64 chobby.exe

Also since ~1 month ago there is massive memleak somewhere. after couple of games the process occupies like 17gig of ram including swap. ending game to return to menu was instantenous in late winter, but now takes like 3 minutes. i got more mem to full 5 gig and it still does it. Previoulsy on 3 gig mem was instantenous.

And now it often quits during loading or loads very slow so i miss out games from it when it starts with catching up menu that waits for 5 minutes...

So like 3 issues.

Massive memleak after completed game
Long loading times that lead to long catching up time 5minute
Sometimes short catching up after normal start when many place factory at once.

Anyone noticed that?
+0 / -0

NLrankmalric
13 days ago
Happened to me today for the first time, after one played game extremely long loading time while HDD was used. Killed and rejoined worked without an issue. Third game happened the same (had to kill process and rejoin).

Specced like 5-6 games after but did not notice the problem (but was not always by the computer).

arch 64 bit, chobby
+0 / -0




AUrankAdminGoogleFrog
13 days ago
(edited 12 days ago)

Try doing '/cheat' then '/give all' in a skirmish game. See this ticket: https://github.com/ZeroK-RTS/Zero-K/issues/2301

Can either of you link to a bug report where you had this issue?

NLrankmalric is referring to this game: https://zero-k.info/Battles/Detail/464667
+0 / -0




AUrankAdminGoogleFrog
13 days ago
PLrankrooks[vegan] are you using 64 bit Spring?

Within the last few months Spring started doing fancier things with memory management, for LuaMenu support. Some of these things were quickly found to crash 32 bit Spring so were disabled. Everyone on windows uses 32 bit Spring due to some lingering issues with 64 bit Spring on windows. Perhaps some issues with memory management were not restricted to 32 bit Spring but the larger amount of memory and fewer uses has made the issues harder to find. Try my test with both 64 bit spring and 32 bit spring and watch memory.
+0 / -0



RUrankFirepluk
13 days ago
(edited 13 days ago)

AUrankAdminGoogleFrog is there any chance of getting 64bit spring for windows?
It could have some performance benefits for users:
16 x 64 bit general purpose registers and 16 x SSE registers instead of 8 x 32 bit general purpose registers and 8 x SSE registers

compilers will be able to use registers more effectively and as you know CPU register is the fastest way to store and access a variable. Also if Spring currently uses 64 bit calculations for ANYTHING when compiled to x86 this gonna be implemented thru ass and works much much slower than when compiled to x64

If we are looking for performance, this would be high priority target to solve
Yes, process would use 15-30% more memory - but memory is cheap and affordable nowdays, while CPUs stagnate in single thread performance
+0 / -0



RUrankFirepluk
13 days ago
(edited 13 days ago)

https://www.viva64.com/en/k/0003/

> Note also that when you launch 32-bit versions of software on 64-bit systems of the Windows family, old 32-bit applications are executed a bit slower because of the WoW64 subsystem which emulates the 32-bit environment. An average performance loss because of this WoW64 layer is 2-3%, although in some special cases it might be much more.
+0 / -0




AUrankAdminGoogleFrog
13 days ago
As far as I can tell win64 Spring basically works. Here is the current win64 blocker: https://springrts.com/mantis/view.php?id=5637

Testing it would be useful. To test it download this: https://springrts.com/dl/buildbot/default/develop/103.0.1-1253-g39b8cbf/win64/spring_%7bdevelop%7d103.0.1-1253-g39b8cbf_win64-minimal-portable.7z

Replace everything in your 'Spring\engine\win64\103.0.1-1253-g39b8cbf' folder with the contents of that zip, except for done.txt.
+1 / -0

PLrankrooks[vegan]
12 days ago
(edited 12 days ago)

I have it on ubuntu 64 14.04.4,
core2duo 2.33
Radeon HD 7350/8350 / R5 220 fglrx drivers
5 gig ram.
mono chobby

worked ok previously on 3 gig ram, now it does wierd memory things... the process sometimes has like 18gig of virtual mem including all physical, so maybe that new mem thing messes things up.

it also looks like from process tracker like the menu doesnt releasse mem alocated previously from spring engine... and with each game that number grows till it makes it unplayable due to hdd page swap lag... hard to hard reset computer few times
+0 / -0




AUrankAdminGoogleFrog
12 days ago
Submit a crash report or link to your infolog.
+0 / -0

PLrankrooks[vegan]
11 days ago
I wish i could submit any report. but the symptom is whole system goes to standstill because of page swap with hdd blocking everything so i have to hard reset...
+0 / -0



RUrankFirepluk
11 days ago
Duh, why in almost 2018 year people still use HDD? Those who do shall be punished with low performance, freezes and lags as they certainly deserve it for optioning out of buying SSD
+0 / -0

NLrankmalric
11 days ago
RUrankFirepluk: you should use only RAM, and have a nuclear plant (or fusion if you prefer :p) not to depend on the electricity grid!

HDD-s can be fine and spring loads like in ~5-10s for me, when it does not behave same as for rooks. And as I want to play I generally just kill it and rejoin... Should try once to check what happens.
+0 / -0




AUrankAdminGoogleFrog
11 days ago
There is potentially a fix in the engine. I'll update to it. https://springrts.com/mantis/view.php?id=5691
+0 / -0

GBrankLynx
9 days ago
(edited 9 days ago)

I think I may have been suffering from this too (Archlinux); I noticed huge memory usage spikes, my system ground to a halt and I had to hard reset a and rejoin match on a few occasions. At the time I thought it was due to an unrelated problem relating to colord: Archlinux bug 54205. But perhaps it was actually this.
+1 / -0


DErankAdminmojjj
9 days ago
32gb ram with ubuntu1404. did not have any issues -.-
+1 / -0

GBrankLynx
3 days ago
(edited 3 days ago)

I just experienced a hard crash [needed to reset computer] last night on loading a game (Archlinux). Nothing visible on journalctl and infolog got overwritten once I hurriedly rejoined. Perhaps it would be helpful if infolog or whatever log file we need kept an account of last few games rather than just last game. Or serious errors could be spewed out elsewhere. Or a separate longer infolog is kept for last few games.
+1 / -0




AUrankAdminGoogleFrog
3 days ago
Past infologs exist in Spring/logs/
+1 / -0

GBrankLynx
2 days ago
Doh; I'll take a look. Thanks.
+0 / -0