View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0003867 | Spring engine | General | public | 2013-06-23 19:29 | 2013-08-02 05:12 |
| Reporter | abma | Assigned To | abma | ||
| Priority | normal | Severity | feature | Reproducibility | always |
| Status | resolved | Resolution | won't fix | ||
| Product Version | 94.1.1+git | ||||
| Summary | 0003867: allow partial repathing of one square across frames for games which have many different movetypes to avoid lags | ||||
| Description | spring heavily lags when stuff is reclaimed, one of this commits introduced it: https://github.com/spring/spring/commit/0dad8f81d0a4a8fd1ee7d6eea07645b9e3297b7e https://github.com/spring/spring/commit/8538573c6f10c15489015478ee9e29e879c3606f https://github.com/spring/spring/commit/376c4d4f28836e9b4c02e78ff204c606ae75d04b https://github.com/spring/spring/commit/8d0c2195faf0cb12c2fcfed54c8bbfd935cdff90 https://github.com/spring/spring/commit/de95c2000ff40fd83f5c7448756f1be7416a0343 it doesn't compile, so i can't get the exact commit. | ||||
| Tags | No tags attached. | ||||
| Attached Files | |||||
| Checked infolog.txt for Errors | |||||
|
|
maybe https://github.com/spring/spring/commit/0dad8f81d0a4a8fd1ee7d6eea07645b9e3297b7e (see the comment) |
|
|
possible related to 0003845 |
|
|
either the particles or the reclaimed feature (pathfinder update) could be the cause... |
|
|
Isnt it related with this? - http://springrts.com/mantis/view.php?id=3833 |
|
|
not sure, here it lags when ONE feature is reclaimed... (and some CEG's? are spawned) |
|
|
/give treetype1 already causes a lag, for building units the same, even with luaui / luarules disabled. creating / destroying a (single!) unit causes already a noticeable lag. not sure how to debug better. it looks a bit like creating / adding to the pathfinder queue causes the lag |
|
|
last comment is a different bug, see 0003873 (/give armcom causes no lag) |
|
|
happens on windows for me, too |
|
|
i dont see any slowdown, how do u detect the "lag" ? |
|
|
build a metal extractor and watch the animation. also, use a debug2 build... |
|
|
when the effect is spawned, fps drops from ~60fps to ~50fps for a short time. |
|
|
maybe as note: some recent changes reduced the lag, so i'm not sure if anything needs to be changed. |
|
|
hmm, as i created this bug report the fps drop was much more bigger, now its imo "ok" |
|
|
aww, the lag doesn't happend with xta... also the effect looks different. meh, sorry! |
|
|
@jk: an idea what could be the cause? http://springrts.com/phpbb/viewtopic.php?p=544753#p544753 (and the following posts) "it lags with /luaui disabled & /luarules disabled on ba v7.78. it doesn't lag with /luaui disabled & /luarules disabled in xta 9.729." |
|
|
DEBUG2 lag by themselves. Always test performance problems with REL* builds. |
|
|
you should also install perftools: http://en.wikipedia.org/wiki/Perf_(Linux) and maybe I should publish my oprofile script for spring. |
|
|
PS: one cause can be the different amount of movetypes, the thing is that when a square needs to get repathed _all_ movetypes need to be done at the same time -> the load cannot scattered across frames |
|
|
is there a mod-setting that makes the pathfinder ignore features? this would explain why xta doesn't lag... |
|
|
no |
|
|
ok, now im 100% sure its because of repathing. sorry, didn't know many things, this is why it took so long. xta seems to contain much less units, so repathing is much faster in comparison to BA. -> feature request to allow repathing of a square accross multiple frames to avoid micro lags. |
|
|
count of units doesn't matter, only count of movetypes does | -> feature request to allow repathing of a square accross multiple frames to avoid micro lags. that would massively increase memusage cause you need then a structure for each movetype to keep track if a square was updated or not. |
|
|
ok, that sounds not good. is there an easy way to delay the start of the pathing updates for a few frames? when a feature is destroyed (or created), mostly many events are triggered (sound, CEG effects, lua-events), which already creates a high load. delaying this would move the load to a time with possible fewer "load". also this would allow more updates (for example because of some chain explosions) to be merged into single updates. |
|
|
meh, i guess cleanest is to hack this into a thread... |
|
|
# of BA movedefs: 25 # of XTA movedefs: 23 |
|
|
- so generaly huge number of movetypes can cause the starting lag (increase the repathing time)? - i notice some lag after general pathing recalculation (i mean the one in loading screen, after massive heightmap change), too, which is running AFTER gadget terrain generator work is finished - so is this "starting lag pathing" included in genereal pathing recalculation (in loading screen) or this (this lag, issue) is some special update/event of pathing structures? |
|
|
(imo this can't be fixed like suggested, only multithreading / delaying this more would help) - so generaly huge number of movetypes can cause the starting lag (increase the repathing time)? yes, the more movetypes, the longer it will take on updates. - i notice some lag after general pathing recalculation (i mean the one in loading screen, after massive heightmap change), too, which is running AFTER gadget terrain generator work is finished - so is this "starting lag pathing" included in genereal pathing recalculation (in loading screen) or this (this lag, issue) is some special update/event of pathing structures? no, this is no special update event. when heightmap is changed, the pathing data needs to be updated. when the engine is started, the cached path is used, if a gadget changes heightmap, the pathing data has to be updated. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2013-06-23 19:29 | abma | New Issue | |
| 2013-06-23 19:30 | abma | Description Updated | |
| 2013-06-23 19:49 | abma | Note Added: 0010912 | |
| 2013-06-23 20:02 | abma | Summary | lags with particles, when stuff is reclaimed => lag when features are reclaimed |
| 2013-06-23 20:05 | abma | File Added: 20130623_200316_AI: Proving Grounds v01_94.1.1-655-gf217d9d develop.sdf | |
| 2013-06-23 20:05 | abma | File Added: infolog.txt | |
| 2013-06-23 20:06 | abma | Note Added: 0010915 | |
| 2013-06-23 21:22 | abma | Note Added: 0010916 | |
| 2013-06-24 01:34 | PepeAmpere | Note Added: 0010923 | |
| 2013-06-24 01:37 | abma | Note Added: 0010925 | |
| 2013-06-24 03:26 | abma | Note Added: 0010928 | |
| 2013-06-24 03:33 | abma | Note Edited: 0010928 | |
| 2013-06-24 03:41 | abma | Relationship added | related to 0003870 |
| 2013-06-24 03:44 | abma | Note Edited: 0010928 | |
| 2013-06-25 20:25 | abma | Note Added: 0010936 | |
| 2013-07-06 14:42 | abma | Note Added: 0010971 | |
| 2013-07-07 14:29 | cleanrock | Note Added: 0010979 | |
| 2013-07-07 14:47 | abma | Note Added: 0010980 | |
| 2013-07-07 14:50 | abma | Severity | major => minor |
| 2013-07-07 14:56 | abma | Note Added: 0010981 | |
| 2013-07-07 15:06 | abma | Note Added: 0010982 | |
| 2013-07-07 15:15 | abma | Note Added: 0010983 | |
| 2013-07-07 15:15 | abma | Status | new => resolved |
| 2013-07-07 15:15 | abma | Resolution | open => no change required |
| 2013-07-07 15:15 | abma | Assigned To | => abma |
| 2013-07-10 03:40 | abma | Relationship added | related to 0003845 |
| 2013-07-14 00:47 | abma | Note Added: 0010992 | |
| 2013-07-15 22:54 | abma | Note Added: 0011001 | |
| 2013-07-15 22:54 | abma | Assigned To | abma => jK |
| 2013-07-15 22:54 | abma | Status | resolved => feedback |
| 2013-07-15 22:54 | abma | Resolution | no change required => reopened |
| 2013-07-15 23:08 | jK | Note Added: 0011002 | |
| 2013-07-15 23:13 | jK | Note Added: 0011003 | |
| 2013-07-15 23:15 | jK | Note Added: 0011004 | |
| 2013-07-15 23:16 | jK | Note Edited: 0011004 | |
| 2013-07-15 23:19 | abma | Note Added: 0011005 | |
| 2013-07-15 23:19 | abma | Status | feedback => assigned |
| 2013-07-15 23:23 | jK | Note Added: 0011006 | |
| 2013-07-16 16:08 | abma | Assigned To | jK => |
| 2013-07-16 16:08 | abma | Severity | minor => feature |
| 2013-07-16 16:08 | abma | Status | assigned => new |
| 2013-07-16 16:08 | abma | Target Version | 95.0 => |
| 2013-07-16 16:08 | abma | Summary | lag when features are reclaimed => allow partial repathing of one square across frames for games which have many different movetypes to avoid lags |
| 2013-07-16 16:10 | abma | Note Added: 0011010 | |
| 2013-07-16 18:26 | jK | Note Added: 0011012 | |
| 2013-07-16 18:28 | jK | Note Edited: 0011012 | |
| 2013-07-16 19:34 | abma | Note Added: 0011013 | |
| 2013-07-16 20:12 | abma | Note Added: 0011014 | |
| 2013-07-17 00:10 | abma | Note Added: 0011018 | |
| 2013-07-18 00:23 | PepeAmpere | Note Added: 0011024 | |
| 2013-07-18 00:25 | PepeAmpere | Note Edited: 0011024 | |
| 2013-08-02 05:12 | abma | Note Added: 0011140 | |
| 2013-08-02 05:12 | abma | Status | new => resolved |
| 2013-08-02 05:12 | abma | Resolution | reopened => won't fix |
| 2013-08-02 05:12 | abma | Assigned To | => abma |
