On August 1, 2014 11:33:26 AM EDT, Antti Palosaari <crope@xxxxxx> wrote: > > >On 08/01/2014 06:19 PM, Antonio Ospite wrote: >> On Fri, 01 Aug 2014 12:14:15 +0200 >> Hans Verkuil <hverkuil@xxxxxxxxx> wrote: >> >>> On 07/31/2014 09:16 PM, Andy Walls wrote: >>>> On Wed, 2014-07-30 at 16:23 +0200, Hans Verkuil wrote: >>>>> From: Hans Verkuil <hans.verkuil@xxxxxxxxx> >>>>> >> [..] >>>>> +- Improve the sinus generation of the SDR radio. >>>> >>>> Maybe a lookup table, containing the first quarter wave of cos() >from 0 >>>> to pi/2 in pi/200 steps, and then linear interpolation for cos() of >>>> angles in between those steps. You could go with a larger lookup >table >>>> with finer grained steps to reduce the approximation errors. A >lookup >>>> table with linear interpolation, I would think, requires fewer >>>> mutliplies and divides than the current Taylor expansion >computation. >>> >>> Yeah, I had plans for that. There actually is a sine-table already >in vivid-tpg.c >>> since I'm using that to implement Hue support. >>> >> >> I don't know what your requirements are here but JFTR there is >already a >> simplistic implementation of fixed point operations in >> include/linux/fixp-arith.h I used them in >> drivers/media/usb/gspca/ov534.c for some hue calculation. > >I looked that too, but there was very small LUT => very bad resolution. > >So I ended up copying sin/cos from cx88 driver (Taylor method). > >regards >Antti I was thinking of implementing a fixed point sine based on the quintic polynomial approximation of sin (pi/2 * x) described here: http://www.coranac.com/2009/07/sines/ since I wanted to learn fixed point stuff anyway. But probably -ENOTIME . Regards, Andy -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
