> -----Original Message----- > From: Jonathan Cameron <Jonathan.Cameron@xxxxxxxxxx> > Sent: Monday, March 8, 2021 12:52 PM > To: Sa, Nuno <Nuno.Sa@xxxxxxxxxx>; Jonathan Cameron > <jic23@xxxxxxxxxx> > Cc: Hennerich, Michael <Michael.Hennerich@xxxxxxxxxx>; > zzzzArdelean, zzzzAlexandru <alexandru.Ardelean@xxxxxxxxxx>; > linux-kernel@xxxxxxxxxxxxxxx; linux-iio@xxxxxxxxxxxxxxx; > lars@xxxxxxxxxx; Bogdan, Dragos <Dragos.Bogdan@xxxxxxxxxx> > Subject: Re: [PATCH v3 0/6] iio: Add output buffer support > > [External] > > On Mon, 8 Mar 2021 10:07:05 +0000 > "Sa, Nuno" <Nuno.Sa@xxxxxxxxxx> wrote: > > > > -----Original Message----- > > > From: Jonathan Cameron <jic23@xxxxxxxxxx> > > > Sent: Saturday, March 6, 2021 6:35 PM > > > To: Hennerich, Michael <Michael.Hennerich@xxxxxxxxxx> > > > Cc: zzzzArdelean, zzzzAlexandru > <alexandru.Ardelean@xxxxxxxxxx>; > > > linux-kernel@xxxxxxxxxxxxxxx; linux-iio@xxxxxxxxxxxxxxx; > > > lars@xxxxxxxxxx; Sa, Nuno <Nuno.Sa@xxxxxxxxxx>; Bogdan, > Dragos > > > <Dragos.Bogdan@xxxxxxxxxx> > > > Subject: Re: [PATCH v3 0/6] iio: Add output buffer support > > > > > > On Fri, 5 Mar 2021 08:57:08 +0000 > > > "Hennerich, Michael" <Michael.Hennerich@xxxxxxxxxx> wrote: > > > > > > > Hi Jonathan and others, > > > > > > > > With output/dac buffer support the semantics of the > scan_element > > > type may change. > > > > > > > > Today the Format is > [be|le]:[s|u]bits/storagebitsXrepeat[>>shift]. > > > > > > > > While shift (if specified) is the shift that needs to be applied prior > to > > > masking out unused bits. > > > > > > > > So far so good and it sounds universal. > > > > > > > > However, we use the right shift (operator) for that, which makes > > > sense for capture devices. > > > > For output devices the more logical operator would be the left > shift. > > > > > > > > I'm not proposing a new Format here. I just want to get some > > > agreement that for an output device > > > > > > > > le:s12/16>>4 > > > > > > > > is understood as a left shift of 4, since the unused bits are then > on > > > the LSB. > > > > > > Good question. Guess I wasn't thinking ahead when I came up with > > > that :) > > > > > > I'm not sure I'd mind if we did decide to define a new format for > > > output > > > buffers. Feels like it should be easy to do. > > > > > > What do others think? > > > > > > > I guess the most straight forward thing would be just to add a 'shift_l' > variable > > to 'struct scan_type'' and make sure either 'shift_l' or 'shift' is defined > and then > > properly export either ">>" or "<<" to userspace? > > Given we already know it's an output channel, can we not just use that > to make the decision? > > Jonathan I would argue that having two variables gives us more flexibility for whatever the future brings us :). But if we can sanely say that an output buffer will always use left shifts, then we could definitely use that information... I mean, we are already doing that assumption for input buffers and right shifts... - Nuno Sá
