View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0005723 | Spring engine | General | public | 2017-08-22 15:34 | 2017-09-14 08:01 |
| Reporter | Google_Frog | Assigned To | |||
| Priority | normal | Severity | minor | Reproducibility | have not tried |
| Status | feedback | Resolution | reopened | ||
| Product Version | 103.0 +git | ||||
| Summary | 0005723: Possible performance issue in Sim::Script | ||||
| Description | I saw 50% CPU usage in Sim::Script in this replay: http://zero-k.info/Battles/Detail/466517. What is Sim::Script? Lua is also very high, around 70%, when it is usually much lower. The lua profilers don't say that there is any problem in luarules or luaUI. | ||||
| Tags | No tags attached. | ||||
| Attached Files | |||||
| Checked infolog.txt for Errors | |||||
|
|
Sim::Script is the timer for animation script updates. |
|
|
It is likely that something in ZK's unit scripts is munching perf. That explains both script and lua |
|
|
What is the timer for animation script updates? Would the cause be something in a thread? Something in a 'script.' in a LUS? |
|
|
The LUS gadget does not show high usage in widget profiler so I doubt the issue is just within lua. Sim::Script first starts to climb at 5:41. Prior to this it fluctuates and does not exceed 0.80%. Sim::Script goes back to normal if the Rocko in this screenshot is self destructed. This unit is stuck against the factory with a fight order towards the center of the map. Its legs are in the air which would indicate that it is spamming script operations. |
|
|
gadgethandler doesn't profile LUS correctly, so no, it doesn't indicate that. |
|
|
The widget profiler doesn't profile the LUS gadget correctly either... |
|
|
find out how many threads are being spawned by the StartMoving callin. |
|
|
Here is the script: https://github.com/ZeroK-RTS/Zero-K/blob/5be10353e84d013b9822bb4b166e0a4082496ded/scripts/cloakskirm.lua The script seems to assume that StartMoving cannot be called twice in a row without StopMoving being called. So perhaps the unit was spamming StartMoving while stuck. |
|
|
As a general rule of thumb, make sure that every thread function starts with Signal and SetSignalMask |
|
|
Yes, obviously. But there are two questions remaining. Is it the engines responsibility to not spam StartMove and is there going to be a fix for the Rocko which is permanently stuck trying to move around a factory? |
|
|
Fixing the second would render the first a non-issue but is not going to be as easy. Fixing just the first could be done by insertion of a StopMoving call but might have adverse effects. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2017-08-22 15:34 | Google_Frog | New Issue | |
| 2017-08-22 15:34 | Google_Frog | File Added: screen00503.jpg | |
| 2017-08-22 16:20 | Kloot | Note Added: 0018255 | |
| 2017-08-22 16:24 | Kloot | Note Edited: 0018255 | |
| 2017-08-23 14:39 | hokomoko | Note Added: 0018260 | |
| 2017-08-23 14:39 | hokomoko | Note Edited: 0018260 | |
| 2017-08-23 16:45 | Google_Frog | Note Added: 0018261 | |
| 2017-08-25 02:44 | Google_Frog | File Added: rockoSelfD.jpg | |
| 2017-08-25 02:44 | Google_Frog | Note Added: 0018269 | |
| 2017-08-25 07:59 | hokomoko | Note Added: 0018271 | |
| 2017-08-25 08:10 | silentwings | Note Added: 0018272 | |
| 2017-08-25 13:01 | Kloot | Note Added: 0018274 | |
| 2017-08-26 09:55 | Google_Frog | Note Added: 0018283 | |
| 2017-08-26 10:19 | hokomoko | Note Added: 0018284 | |
| 2017-08-26 10:24 | hokomoko | Assigned To | => hokomoko |
| 2017-08-26 10:24 | hokomoko | Status | new => closed |
| 2017-08-26 10:24 | hokomoko | Resolution | open => no change required |
| 2017-08-26 11:06 | Google_Frog | Status | closed => feedback |
| 2017-08-26 11:06 | Google_Frog | Resolution | no change required => reopened |
| 2017-08-26 11:06 | Google_Frog | Note Added: 0018285 | |
| 2017-08-26 13:34 | Kloot | Note Added: 0018286 | |
| 2017-08-26 13:37 | Kloot | Note Edited: 0018286 | |
| 2017-09-14 08:01 | hokomoko | Assigned To | hokomoko => |
