Re: question for soc-camera driver

[Guennadi Liakhovetski <g.liakhovetski@xxxxxx>]

On Thu, 17 Apr 2008, ·ëöÎ wrote:

> 2008/4/14 Guennadi Liakhovetski <g.liakhovetski@xxxxxx>:
> >>  I found the problem with reversed frame order. The same problem led to
> >>  corrupted frames. Please try the patch below on the top of the previous
> >>  one. With it I have no more problems with the test application with any
> >>  number of buffers.
> 
> 2008/4/14 ·ëöÎ <fengxin215@xxxxxxxxx>:
> >  I test it and It is good.But some frames is dropped,like
> >  1,2,3,4,5,8,9,10,11,14,15,16,17,18,21,22.The frame 6,7,12,13,19,20 is
> >  dropped.I request 4 buffers now.
> 
> Will you improve soc-camera driver for the lost frames?

How exactly are you counting lost frames? I just tried several 
configurations - with different number of buffers, writing in tmpfs and 
over nfs, without load and under a ping flood. And I'm counting FIFO 
overrun interrupts. At most I'm getting 1-3 overruns with dropped frames 
in the beginning, and only if I write over NFS.

If your system is using drivers, that block interrupts for a considerable 
amount of time, of course DMA done interrupts will be missed, FIFO will 
overflow and frames will be dropped. I don't think you can avoid this 
under such conditions. As I suggested before - you can use more buffers, 
put the frame read-out in a separate thread. As a test try writing to 
RAM-based tmpfs and see if frames get dropped then too.

Thanks
Guennadi
---
Guennadi Liakhovetski

--
video4linux-list mailing list
Unsubscribe mailto:video4linux-list-request@xxxxxxxxxx?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/video4linux-list