Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

GAM percentage is too high when using VA i965 driver with help of Gstreamer Vaapi plugin for Decoding and Encoding. #5

Open
APrashant opened this issue Jan 11, 2022 · 1 comment

Comments

@APrashant
Copy link

Hi,

We are using Intel X86 architecture for our Project.
Build System : Yocto 2.1
VA_DRIVER = i965
We are using Gstreamer pipeline with VAAPI H264 plugins for Decoding, Encoding and Display ( vaapisink ), for 8 H264 input streaming processed parallelly.

We observed GAM % is too high for 8 input streams, when try to get performance using intel_gpu_top tool.

Render Busy = 47%
GAM = 97%
CS = 48%

Since our requirement is to use 16 input streams, we could not able to add more than 8 streams, due to GAM% is too high.

We understand that GAM is similar to TLB, and suspecting, if we increase PAGE_SIZE from 4K to 16K or 64K, then performance might be improved.

Also, we are liking to know the following information.

  1. Whether GAM and CPU is having common PAGE SIZE values.
  2. How to increase the PAGE SIZE values for GPU processing, for the above mentioned scenario.
  3. Whether there is common memory for both CPU and GPU, if yes, whether we can resize the memory, as we mainly using GPU for our purpose.

Is there any other factor will affect the GAM %, if yes, shall you please share the details and help us in resolving this issue.

Also, is there any other tool to get the memory details for GPU processing. Like /proc/meminfo for CPU processing.

@kraj
Copy link
Contributor

kraj commented Jan 13, 2022

perhaps its better to ask this question also at [email protected].

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants