Although already stated by Pete before in this thread it seems that I'm getting a slightly higher than expected overhead from the lightweight interceptor. The FPS seems to drop by around 3-4FPS and seems more jittery than when playing without Streamline capturing.
I'm seeing (with top) an average 60% CPU usage when gatord is capturing the data but also seen spikes of more than 200%. I'm launching the interceptor like follows:
python lwi_me.py --daemon 'C:\Program Files\Arm\Arm Mobile Studio 2020.2\streamline\bin\arm64\gatord' --lwi on --lwi-gles-layer-lib-path 'C:\Program Files\Arm\Arm Mobile Studio 2020.2\performance_advisor\lwi\target\android\arm\unrooted\arm64-v8a\libGLES_layer_lwi.so' --package com.mypackage.demo
I'm running this on a Samsung A20e but also seen similar behavior in a Samsung S10e. Is this expected, or are there any flags I should use/avoid using?
I should also add that our FPS counter matches what I'm seeing in Streamline.
Hi JPJ,
Thanks for reporting. We've recently found a problem in Streamline related to tracing threads that have exited - I'll get in touch offline with a gatord patch you can try.
Cheers, Pete