On Mon, 2008-10-20 at 12:37 -0500, Danny Miller wrote:
> Many beginners jump right into "floating point- yeah I need decimals".  
> But you'll find FP on a system with no Floating Point Unit is incredibly 
> slow.  Hundreds of times slower (or worse).  In fact there are plenty of 
> ways to work with decimals as "fixed point" that don't involve Floating 
> Point calcs.  Like if you need a single digit to the right of the 
> decimal, just store the number as 10x larger than it is.  13.8=>138.  
> 13.8 + 1 becomes 138+10.

One of the FIRST things I do when starting a new MCU project is defining
what kind of math I need.

Often this means figuring out the min and max values I'll have to deal
with, the resolution I'll need, and how fast things need to be done.

Then I go through all the formulas I'll be using, and hand optimize
them, keeping in mind the min/max values that might be hit in the
intermediate steps. I also introduce approximations in this step, if
appropriate. For example if the sensor I'm using is non linear, but I'm
only using it for a small range, I might approximate things with a
linear function.

It sounds like a lot of work, but in the end the code is SO much easier
to write, plus you can catch ALOT of bugs before even writing a single
line of code.

FWIW I've never ended up needing to use FP on an MCU, sometimes I can
get away with just integers, the rest of the time a "fixed point" type
of math has been appropriate.

TTYL

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist