Replies: 1 comment
-
|
Ok I think I found answer for decreased performance. However this is not stated in documentation -> #453 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Hello, I have 2 questions that are partially connected.
Recently we have updated our software from Physx 4.1.2 to 5.5.1.
In one of our scenarious, we have noticed worse performance since then.
Our app is using PxDefaultCpuDispatcher with 2 threads.
Scenario is this:
So far I can see around 10% drop in performance.
It seems that when I remove manual wakeUp call, performance gets better, drop is only around 5%. (And its not because actor goes to sleep, there is force applied that keeps him awake)
Then I tried to play with snippets and used SnippetHelloWorld to measure time needed for simulation using std::chrono::high_resolution_clock
I have changed snippet to contain only stacked boxes without ball hitting them so they all go to sleep soon.
From my measurements, it seems that Physx5 performs around 5-10% worse when there is small number of boxes (e.g. < 1000). while performing much better with large number of actors. At the same time, they perform same when scene is empty, or when I set dispatcher cores to 0,
So my questions are these:
EDIT: I just tried HelloWorld snippet in 5.6.0 where high priority tasks are available and it seems to perform better than 4.1.2. But my questions stay.
Beta Was this translation helpful? Give feedback.
All reactions