capture_mem limitations in OMAP ISP

capture_mem limitations in OMAP ISP

[Michael Jones]

Hi Laurent & co.,

I'm looking at the memory limitations in the omap3isp driver.  'struct 
isp_video' contains member 'capture_mem', which is set separately for 
each of our v4l2 video device nodes.  The CCDC, for example, has 
capture_mem = 4096 * 4096 * 3 = 48MB, while the previewer and resizer 
each have twice that. Where do these numbers come from?

Is the CCDC incapable of DMA'ing more than 48MB into memory?  I know 
that ISP_VIDEO_MAX_BUFFERS also limits the # of buffers, but I assume 
this is basically an arbitrary number so we can have a finite array of 
isp_video_buffer's.  The 48MB, on the other hand, looks like it might 
have a good reason.

thanks,
Michael

MATRIX VISION GmbH, Talstrasse 16, DE-71570 Oppenweiler
Registergericht: Amtsgericht Stuttgart, HRB 271090
Geschaeftsfuehrer: Gerhard Thullner, Werner Armingeon, Uwe Furtner, Erhard Meier

Re: capture_mem limitations in OMAP ISP

[Laurent Pinchart]

Hi Michael,

On Tuesday 03 July 2012 11:46:33 Michael Jones wrote:
> Hi Laurent & co.,
> 
> I'm looking at the memory limitations in the omap3isp driver. 'struct
> isp_video' contains member 'capture_mem', which is set separately for each
> of our v4l2 video device nodes. The CCDC, for example, has capture_mem =
> 4096 * 4096 * 3 = 48MB, while the previewer and resizer each have twice
> that. Where do these numbers come from?

That's mostly historical. When developing the driver for the N900 we set a 
limit to avoid putting too much pressure on the system memory, and 3x8MP 
buffers was considered to be enough.

We could raise the limit, remove it completely, or implement a policy 
mechanism to let a privileged userspace application specify limits. The later 
might be interesting as a core V4L2 mechanism.

> Is the CCDC incapable of DMA'ing more than 48MB into memory? I know that
> ISP_VIDEO_MAX_BUFFERS also limits the # of buffers, but I assume this is
> basically an arbitrary number so we can have a finite array of
> isp_video_buffer's. The 48MB, on the other hand, looks like it might have a
> good reason.

-- 
Regards,

Laurent Pinchart