Linux/AM5728: Appsink source code

xixi

Other Parts Discussed in Thread: AM5728

Hi all

Months ago, we found the am5728 appsink performance is very low and we send an post to e2e, the below is the website.

https://e2e.ti.com/support/arm/sitara_arm/f/791/t/589689

The e2e have told us the solution， which is modify the appsink’s source code and change memory allocation from omap_bo to CMEM.

Just now we don’t know how to modify the source code because we can’t find the memory allocation code.

Can you give us some advice.

Our Processor SDK RT3.3

Thanks

Regards

over 7 years ago

0 Biser Gatchev-XID over 7 years ago

TI__Guru**** 393215 points

You have already a thread opened on this subject. Your last post was moved to this thread. Please stop double-posting on this forum.

0 xixi over 7 years ago in reply to Biser Gatchev-XID

Expert 2990 points

OK , Thanks

We will pay attention。

But Why the status of post is TI Thinks Resolved.

Regards

0 Rogerio Almeida over 7 years ago in reply to xixi

TI__Mastermind 26205 points

because there is other post from you with same question and we are addressing the issues there, not here.

0 xixi over 7 years ago in reply to Rogerio Almeida

Expert 2990 points

thanks for your reply

But We can not edit the post "AM5728: GStreamer appsink performance issue"

It is locked.

Anyone can help us about this issue。

Thanks
e2e.ti.com/.../589689

Regards

0 xixi over 7 years ago in reply to Rogerio Almeida

Expert 2990 points

Hi ALL

Anyone can help us about this issue.

Because it is very important for us.

Just now we found Whether it's appsink or appsrc

They all have serious problems,their copy performance is very low.

We can not use the two element to development application.

We need modify the element's source code and change the memory allocation from omap_bo to CMEM.

Regards

0 Manisha Agrawal over 7 years ago in reply to xixi

TI__Mastermind 22946 points

The buffer for vpe is allocated inside the gstvpebuffer.c file in function gst_buffer_add_vpe_buffer_meta(). Look for omap_bo_new() function. As suggested in other post, you need to refer to dual-camera-demo example application to allocate buffers from CMEM. CPU based buffer copy also involves cache management. You will need to take care of that too in the plugin.

You can clone the vpe plugin code from following git repo -

Once modified the code, please follow below link to rebuild the plugin -

0 xixi over 7 years ago in reply to Manisha Agrawal

Expert 2990 points

HI manisha：
Thanks For your reply.

According to your opinion "The buffer for vpe is allocated inside the gstvpebuffer.c file "

So the appsink's buffer is from vpe's allocation.
Is that right?
Thanks

Regards

xixi

0 Manisha Agrawal over 7 years ago in reply to xixi

TI__Mastermind 22946 points

appsink is opensource plugin. I am not sure how it works. Based on the issue you reported, my understanding is that the appsink is doing a buffer copy from the upstream/source plugin. That source plugin is vpe. Now the vpe output buffer (input to appsink) is not cache enabled. This certainly is going to cost lot of CPU cycles to read from non cached buffer and hence my suggestion above.

Processors

Processors forum

Linux/AM5728: Appsink source code