TDA4VM: tivxVpacMscScaleNode process crop cost long time

Hi,

But why do you want to crop in MSC? Cropping can be done with just pointer moving into the buffer. 

Regards,

Brijesh

If move pointer. I may add user kernel.

You can add user kernel or whichever node is expecting this resolution can move the pointer.. 

I tested one case, such as input image size is 3840*1680, output image size is 3840*840, crop cost 15ms; I change the output image size to 300*500, also cost 15ms. So the time depends on the input image size? Does it make sense? 

The time depends on the input size, because MSC is only for downscaling, so it all depends on the input size only.. 

But if you just want to do cropping operation, it should not need MSC. You could just move pointer into the buffer. That will save 15ms of latency..