View Issue Details

Jump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0003612Spring engineGeneralpublic2013-03-23 22:092013-04-04 03:33
Reportersilentwings Assigned Tozerver  
PrioritynormalSeveritymajorReproducibilityalways
Status resolvedResolutionfixed 
Product Version94.1 
Summary0003612: self destructing units use huge amount of cpu in 94.0
Descriptionsee demo & screenshots. asking 200 corvamps to self d at the same time causes close to 100% cpu usage of simframe and unit::slowupdate while they count down to the explosion.

asking about >50 units to self d at once causes lag for me (with i7, good gfx card), with both st and mt. demo is with tabula v4 and ba 7.72.
TagsNo tags attached.
Attached Files
20130323_210549_Tabula-v4_94.sdf (Attachment missing)
screen00011.png (Attachment missing)
screen00012.png (Attachment missing)
screen00013.png (Attachment missing)
Checked infolog.txt for Errors

Relationships

duplicate of 0003664 closedKloot Ctrl-a ctrl-d massive fps drop 

Activities

silentwings

2013-03-23 22:14

reporter   ~0010170

tested same thing again with lua off, added another screenshot to show its not a widget/gadget error.

Kloot

2013-03-23 23:10

developer   ~0010171

doesn't reproduce for me (not with 200 vamps, not with 2000)

silentwings

2013-03-24 00:10

reporter   ~0010174

Yes, I asked a few others and it seems it only me...

abma

2013-03-24 05:23

administrator   ~0010177

infolog.txt maybe helps?

it could be maybe a problem with gfx drivers...

msafwan

2013-03-24 15:37

reporter   ~0010180

Last edited: 2013-03-24 15:38

 I tried same Corvamp self+d test using 2000 Corvamp. The debug info showed >10000 path-queue update request the moment those corvamps explode.

abma

2013-03-24 17:34

administrator   ~0010183

@msafwan: "100% cpu usage of simframe and unit::slowupdate while they count down to the explosion."

zerver

2013-03-24 18:44

reporter   ~0010184

I'm getting 20% CPU during countdown, which I agree is a bit high for just 200 units.

zerver

2013-03-24 23:20

reporter   ~0010187

This:
(changelog)
default LogFlush to true again

abma

2013-03-24 23:31

administrator   ~0010188

Last edited: 2013-03-24 23:39

 this questions are for windows only:
is a virus-scanner involved? has your harddrive the write cache disabled? (afaik it defaults to yes) maybe this as reference: http://www.sevenforums.com/tutorials/10392-write-caching-enable-disable.html

@silentwings:

can you try if setting LogFlush=0 has any effect?

silentwings

2013-03-25 01:57

reporter   ~0010190

Last edited: 2013-03-25 02:15

 Yes, setting LogFlush=0 fixes the problem!

(There is something a little odd - if I run spring.exe and use that to change LogFlush, it seemingly only takes effect after Spring.exe has been exited and restarted. I guess this is because the setting LogFlush is loaded before the point at which it can be changed.)

I don't use a virus scanner, other than what windows has built in and my harddrives both have write caching enabled. I tried watching the "queue length" (as its called in Window's Resource Monitor, I think this is the cache) for my hard drive while self d'ing 200 vamps and with LogFlush=1; the queue length is maxed out for all the time they are self destructing. If I do the same thing with LogFlush=0 then the queue length barely rises above 5%.

abma

2013-03-25 06:35

administrator   ~0010196

Last edited: 2013-03-25 06:36

 @zerver
https://github.com/spring/spring/blob/develop/rts/System/Log/FileSink.cpp#L202


removing the commit flag should make it behave as it should...

Issue History

Date Modified Username Field Change
2013-03-23 22:09 silentwings New Issue
2013-03-23 22:09 silentwings File Added: 20130323_210549_Tabula-v4_94.sdf
2013-03-23 22:10 silentwings File Added: screen00011.png
2013-03-23 22:11 silentwings File Added: screen00012.png
2013-03-23 22:14 silentwings Note Added: 0010170
2013-03-23 22:14 silentwings File Added: screen00013.png
2013-03-23 23:10 Kloot Note Added: 0010171
2013-03-24 00:10 silentwings Note Added: 0010174
2013-03-24 05:23 abma Note Added: 0010177
2013-03-24 15:37 msafwan Note Added: 0010180
2013-03-24 15:38 msafwan Note Edited: 0010180
2013-03-24 17:34 abma Note Added: 0010183
2013-03-24 18:44 zerver Note Added: 0010184
2013-03-24 18:57 zerver Assigned To => zerver
2013-03-24 18:57 zerver Status new => assigned
2013-03-24 23:20 zerver Note Added: 0010187
2013-03-24 23:31 abma Note Added: 0010188
2013-03-24 23:33 abma Note Edited: 0010188
2013-03-24 23:38 abma Note Edited: 0010188
2013-03-24 23:39 abma Note Edited: 0010188
2013-03-25 01:57 silentwings Note Added: 0010190
2013-03-25 02:00 silentwings Note Edited: 0010190
2013-03-25 02:07 silentwings Note Edited: 0010190
2013-03-25 02:15 silentwings Note Edited: 0010190
2013-03-25 06:35 abma Note Added: 0010196
2013-03-25 06:36 abma Note Edited: 0010196
2013-03-25 18:46 zerver Changeset attached => spring develop d6f4900b
2013-03-25 18:46 zerver Status assigned => resolved
2013-03-25 18:46 zerver Resolution open => fixed
2013-03-31 12:23 Kloot Relationship added duplicate of 0003664
2013-04-04 03:15 zerver Changeset attached => spring MTsim2 86f15a23
2013-04-04 03:33 zerver Changeset attached => spring MTsim3 e30546ce