[lm-sensors] [RFC PATCH 09/10] lis3: Scale output values to mg

samu.p.onkalo at nokia.com samu.p.onkalo at nokia.com
Fri Nov 6 12:55:54 CET 2009


>-----Original Message-----
>From: ext Éric Piel [mailto:eric.piel at tremplin-utc.net] 
>Sent: 03 November, 2009 15:25
>To: Daniel Mack
>Cc: Onkalo Samu.P (Nokia-D/Tampere); kalhan.trisal at intel.com; 
>lm-sensors at lm-sensors.org
>Subject: Re: [RFC PATCH 09/10] lis3: Scale output values to mg
>Op 03-11-09 13:32, Daniel Mack schreef:
>> On Tue, Nov 03, 2009 at 01:30:07PM +0100, 
>samu.p.onkalo at nokia.com wrote:
>>>> On Tue, Nov 03, 2009 at 02:09:44PM +0200, Samu Onkalo wrote:
>>>>> Report output values as 1/1000 of earth gravity.
>>>> Without having tried that, I assume that changes the 
>values for all 
>>>> events reported? That would break all userspace clients of this 
>>>> device, right?
>>> Yes. That is one thing which need to agreed here. One 
>possibility is 
>>> to implement this as a configurable feature for example 
>using platform data.
>>> In that kind of solution default value must be the old way.
>> Yes please :)
>Actually we had already briefly discussed this with Samu. And 
>I had given my green light because:
> * I don't know any userspace program which would be affected 
>by this change (they only rely on the relative difference). 
>Arguably, I know only few programs.
> * Eventually, we should converge to a userspace API 
>compatible with any accelerometer device and exposed by all 
>the accelerometer drivers. So for now I consider the userspace 
>interface of the lis3 driver quite "soft", and any move toward 
>this goal good.
> * Having platform data to select between old unit/new unit 
>will add complexity to the driver, which should be avoided 
>whenever possible.
>Of course, if you come and tell me that you know some 
>userspace programs which will be affected by this change, this 
>blows away the first point.
>Then I'd be much more willing to have a configurable option. 
>But be aware that these programs would then blow away 
>themselves if, for example, they expect a LIS302DL and the 
>actual sensor is a LIS3LV02DL.
>So simply fixing these programs and moving to mG might be a 
>better option.

It is relatively simple to keep also non-scaled output. Scaling factor just need to
be set to a neutral value. Selection could be done via platform data.
Question is that what do we want to do? 


More information about the lm-sensors mailing list