[PATCH 46/55] staging: comedi: drivers: handle SDF_PACKED in comedi_inc_scan_progress()

[Ian Abbott]

On 24/10/14 17:34, Hartley Sweeten wrote:
> On Friday, October 24, 2014 1:49 AM, Ian Abbott wrote:
>> On 23/10/14 17:35, Hartley Sweeten wrote:
>>> On Thursday, October 23, 2014 4:28 AM, Ian Abbott wrote:
[snip]
>>>> Is this patch even necessary?  comedi_bytes_per_scan() assumes digital
>>>> samples will be packed into a whole number of "samples" and calculates
>>>> the scan length in bytes accordingly.
>>>
>>> I guess it's not really "necessary" but it adds clarification to what
>>> the SDF_PACKED flag does. Right now the only documentation
>>> for that flag is:
>>>
>>> #define SDF_PACKED	0x20000000	/* subdevice can do packed DIO */
>>>
>>> But nothing in the core uses it. I don't think comedilib
>>> references the flag either.
>>
>> Applications could see it, though they would probably ignore it as well,
>> as any application using async commands for digital I/O probably needs
>> to be hardware specific as the order of data in the scan data doesn't
>> match the order of channels in the chanlist for all drivers. (Some just
>> return all channels in channel number order.)
>
> Should this be fixed?

Probably best left alone for the sake of stability.  In particular, 
ni_pcidio is a reasonably high-speed device, so looping up to 32 times 
per scan to rearrange the bits may be a bit much.  It currently ignores 
the chanlist altogether!

-- 
-=( Ian Abbott @ MEV Ltd.    E-mail: <abbotti at mev.co.uk> )=-
-=(                          Web: http://www.mev.co.uk/  )=-