> Aaron, thanks for the tip. I had never really looked at fixed point math
> before, and was surprised at the simplicity of your example. My
application
> is to read optical encoders mounted on the two axes of an astronomical
> telescope and convert the encoder readings to the celestial coordinates to
> which the telescope is pointing. The computational process involves a
> 3D transformation matrix and the usual trig manipulations of angles, so
> it could indeed be slow using traditional floating point. I need to worry
> a little bit about the precision though. I suppose I could use 2 bytes for
> the fractional part instead of one as you had done, since I don't think
> the magnitude of my numbers ever gets very large.

you can use

'long long'

and that'll give you a 64 bit integer *g* most likely, anything > 16bit
on palmos is actually done via software (integer operations) as the bus
is (correct me if i am wrong), 16 bit on the 68k CPU's

it will still be faster :) fixed point is fun - just depends on the
precision level you need :) using 8 bytes = 1/256 precision, that basically
means...

1/256 = 0.00390625

which, gives you at least two decimal points of precision

1/65536 = 0.0000152587890625

which, gives you four. up to you :) and, nothing stopping you from using
9, 10, 11 :) if you want 3 decimal points, your best bet would be

1/8192 = 0.0001220703125 or
1/4096 = 0.000244140625  or
1/2048 = 0.00048828125
1/1024 = 0.0009765626

i think you get the idea :)

---
Aaron Ardiri                           [EMAIL PROTECTED]
CEO - CTO                                              +46 70 656 1143
Mobile Wizardry                         http://www.mobilewizardry.com/


-- 
For information on using the Palm Developer Forums, or to unsubscribe, please see 
http://www.palmos.com/dev/support/forums/

Reply via email to