Changes between Version 1 and Version 2 of Ticket #5422, comment 14


Ignore:
Timestamp:
Apr 7, 2019, 5:48:06 PM (5 years ago)
Author:
theotherhiveking

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #5422, comment 14

    v1 v2  
    11I'm not quite sure the rendering itself is slow. Profiler2 claims that the gui only takes about 3-4ms under the 'gpu_arb' section. But takes ~8ms in the main thread. I'm not quite sure I'm interpreting the output correctly. See the attached image.
    22
    3 So far, I have noticed that the slowdown seems to be independent of the game resolution, I also set the gui scaling to very low and very high values, that didn't affect performance either 
     3So far, I have noticed that the slowdown seems to be independent of the game resolution, I also set the gui scaling to very low and very high values, that didn't affect performance either
    44
    5 I realized that I could very easily add aditional buttons to the gui to test their performance impact by just adding some units to the training queue. So far each button seems to take close to 0.1ms 
     5I realized that I could very easily add aditional buttons to the gui to test their performance impact by just adding some units to the training queue. So far each button seems to take close to 0.1ms
    66
    77Maybe Callgrind will have the answer.
     8
     9> If you say that hiding the GUI doesn't change the performance, then it might be that it's either not the rendering itself, or that there is still renderer code being called if the objects are hidden.
     10
     11I meant the performance of the JS functions doesnt change.
     12Disabling the UI greatly improves performance.  It's just that at first I thought it was faster because the JS code wasn't run if the GUI was disabled. So the problem might be in C++ land.