GPU to CPU copy taking long time

[abhilashabraham]

I'm using IMX6Q and has no display. I'm doing a capture of Y16 format with 1040*768 pixels. I want to use the GPU for the processing. As a start point I configured the shaders in such a way to just pass the capture through GPU without any change in pixels. I have used glTexImage2D with the v4l2 buffer as input. The display I have configured is the framebuffer and I'm able to get the frames to framebuffer. I did a mmap to read the framebuffer to the application for streaming. I could find that the memcpy was taking 12ms for 1040*768*2 bytes. Is there any means to improve it? I'm using yocto with Linux 3.10.17.

[igorpadykov]

Hi Abhilash

I am not aware of such patch but one can try solutions below for improving

memcpy copy time

https://community.freescale.com/thread/340705

http://allontutorials.com/benchmarking-the-gpu-bandwidth-capability-based-on-opencl-memory-copy-oper...

Best regards

igor

-----------------------------------------------------------------------------------------------------------------------

Note: If this post answers your question, please click the Correct Answer button. Thank you!

-----------------------------------------------------------------------------------------------------------------------